diff --git a/exprements/PaddleClas/.clang_format.hook b/exprements/PaddleClas/.clang_format.hook new file mode 100644 index 00000000..1d928216 --- /dev/null +++ b/exprements/PaddleClas/.clang_format.hook @@ -0,0 +1,15 @@ +#!/bin/bash +set -e + +readonly VERSION="3.8" + +version=$(clang-format -version) + +if ! [[ $version == *"$VERSION"* ]]; then + echo "clang-format version check failed." + echo "a version contains '$VERSION' is needed, but get '$version'" + echo "you can install the right version, and make an soft-link to '\$PATH' env" + exit -1 +fi + +clang-format $@ diff --git a/exprements/PaddleClas/.gitignore b/exprements/PaddleClas/.gitignore new file mode 100644 index 00000000..d8f8bca6 --- /dev/null +++ b/exprements/PaddleClas/.gitignore @@ -0,0 +1,14 @@ +__pycache__/ +*.pyc +*.sw* +*/workerlog* +checkpoints/ +output*/ +pretrained/ +.ipynb_checkpoints/ +*.ipynb* +_build/ +build/ +log/ +nohup.out +.DS_Store diff --git a/exprements/PaddleClas/.pre-commit-config.yaml b/exprements/PaddleClas/.pre-commit-config.yaml new file mode 100644 index 00000000..1584bc76 --- /dev/null +++ b/exprements/PaddleClas/.pre-commit-config.yaml @@ -0,0 +1,35 @@ +- repo: https://github.com/PaddlePaddle/mirrors-yapf.git + sha: 0d79c0c469bab64f7229c9aca2b1186ef47f0e37 + hooks: + - id: yapf + files: \.py$ +- repo: https://github.com/pre-commit/pre-commit-hooks + sha: a11d9314b22d8f8c7556443875b731ef05965464 + hooks: + - id: check-merge-conflict + - id: check-symlinks + - id: detect-private-key + files: (?!.*paddle)^.*$ + - id: end-of-file-fixer + files: \.md$ + - id: trailing-whitespace + files: \.md$ +- repo: https://github.com/Lucas-C/pre-commit-hooks + sha: v1.0.1 + hooks: + - id: forbid-crlf + files: \.md$ + - id: remove-crlf + files: \.md$ + - id: forbid-tabs + files: \.md$ + - id: remove-tabs + files: \.md$ +- repo: local + hooks: + - id: clang-format + name: clang-format + description: Format files with ClangFormat + entry: bash .clang_format.hook -i + language: system + files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|cuh|proto)$ diff --git a/exprements/PaddleClas/LICENSE b/exprements/PaddleClas/LICENSE new file mode 100644 index 00000000..261eeb9e --- /dev/null +++ b/exprements/PaddleClas/LICENSE @@ -0,0 +1,201 @@ + Apache License + Version 2.0, January 2004 + http://www.apache.org/licenses/ + + TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION + + 1. Definitions. + + "License" shall mean the terms and conditions for use, reproduction, + and distribution as defined by Sections 1 through 9 of this document. + + "Licensor" shall mean the copyright owner or entity authorized by + the copyright owner that is granting the License. + + "Legal Entity" shall mean the union of the acting entity and all + other entities that control, are controlled by, or are under common + control with that entity. For the purposes of this definition, + "control" means (i) the power, direct or indirect, to cause the + direction or management of such entity, whether by contract or + otherwise, or (ii) ownership of fifty percent (50%) or more of the + outstanding shares, or (iii) beneficial ownership of such entity. + + "You" (or "Your") shall mean an individual or Legal Entity + exercising permissions granted by this License. + + "Source" form shall mean the preferred form for making modifications, + including but not limited to software source code, documentation + source, and configuration files. + + "Object" form shall mean any form resulting from mechanical + transformation or translation of a Source form, including but + not limited to compiled object code, generated documentation, + and conversions to other media types. + + "Work" shall mean the work of authorship, whether in Source or + Object form, made available under the License, as indicated by a + copyright notice that is included in or attached to the work + (an example is provided in the Appendix below). + + "Derivative Works" shall mean any work, whether in Source or Object + form, that is based on (or derived from) the Work and for which the + editorial revisions, annotations, elaborations, or other modifications + represent, as a whole, an original work of authorship. For the purposes + of this License, Derivative Works shall not include works that remain + separable from, or merely link (or bind by name) to the interfaces of, + the Work and Derivative Works thereof. + + "Contribution" shall mean any work of authorship, including + the original version of the Work and any modifications or additions + to that Work or Derivative Works thereof, that is intentionally + submitted to Licensor for inclusion in the Work by the copyright owner + or by an individual or Legal Entity authorized to submit on behalf of + the copyright owner. For the purposes of this definition, "submitted" + means any form of electronic, verbal, or written communication sent + to the Licensor or its representatives, including but not limited to + communication on electronic mailing lists, source code control systems, + and issue tracking systems that are managed by, or on behalf of, the + Licensor for the purpose of discussing and improving the Work, but + excluding communication that is conspicuously marked or otherwise + designated in writing by the copyright owner as "Not a Contribution." + + "Contributor" shall mean Licensor and any individual or Legal Entity + on behalf of whom a Contribution has been received by Licensor and + subsequently incorporated within the Work. + + 2. Grant of Copyright License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + copyright license to reproduce, prepare Derivative Works of, + publicly display, publicly perform, sublicense, and distribute the + Work and such Derivative Works in Source or Object form. + + 3. Grant of Patent License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + (except as stated in this section) patent license to make, have made, + use, offer to sell, sell, import, and otherwise transfer the Work, + where such license applies only to those patent claims licensable + by such Contributor that are necessarily infringed by their + Contribution(s) alone or by combination of their Contribution(s) + with the Work to which such Contribution(s) was submitted. If You + institute patent litigation against any entity (including a + cross-claim or counterclaim in a lawsuit) alleging that the Work + or a Contribution incorporated within the Work constitutes direct + or contributory patent infringement, then any patent licenses + granted to You under this License for that Work shall terminate + as of the date such litigation is filed. + + 4. Redistribution. You may reproduce and distribute copies of the + Work or Derivative Works thereof in any medium, with or without + modifications, and in Source or Object form, provided that You + meet the following conditions: + + (a) You must give any other recipients of the Work or + Derivative Works a copy of this License; and + + (b) You must cause any modified files to carry prominent notices + stating that You changed the files; and + + (c) You must retain, in the Source form of any Derivative Works + that You distribute, all copyright, patent, trademark, and + attribution notices from the Source form of the Work, + excluding those notices that do not pertain to any part of + the Derivative Works; and + + (d) If the Work includes a "NOTICE" text file as part of its + distribution, then any Derivative Works that You distribute must + include a readable copy of the attribution notices contained + within such NOTICE file, excluding those notices that do not + pertain to any part of the Derivative Works, in at least one + of the following places: within a NOTICE text file distributed + as part of the Derivative Works; within the Source form or + documentation, if provided along with the Derivative Works; or, + within a display generated by the Derivative Works, if and + wherever such third-party notices normally appear. The contents + of the NOTICE file are for informational purposes only and + do not modify the License. You may add Your own attribution + notices within Derivative Works that You distribute, alongside + or as an addendum to the NOTICE text from the Work, provided + that such additional attribution notices cannot be construed + as modifying the License. + + You may add Your own copyright statement to Your modifications and + may provide additional or different license terms and conditions + for use, reproduction, or distribution of Your modifications, or + for any such Derivative Works as a whole, provided Your use, + reproduction, and distribution of the Work otherwise complies with + the conditions stated in this License. + + 5. Submission of Contributions. Unless You explicitly state otherwise, + any Contribution intentionally submitted for inclusion in the Work + by You to the Licensor shall be under the terms and conditions of + this License, without any additional terms or conditions. + Notwithstanding the above, nothing herein shall supersede or modify + the terms of any separate license agreement you may have executed + with Licensor regarding such Contributions. + + 6. Trademarks. This License does not grant permission to use the trade + names, trademarks, service marks, or product names of the Licensor, + except as required for reasonable and customary use in describing the + origin of the Work and reproducing the content of the NOTICE file. + + 7. Disclaimer of Warranty. Unless required by applicable law or + agreed to in writing, Licensor provides the Work (and each + Contributor provides its Contributions) on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + implied, including, without limitation, any warranties or conditions + of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A + PARTICULAR PURPOSE. You are solely responsible for determining the + appropriateness of using or redistributing the Work and assume any + risks associated with Your exercise of permissions under this License. + + 8. Limitation of Liability. In no event and under no legal theory, + whether in tort (including negligence), contract, or otherwise, + unless required by applicable law (such as deliberate and grossly + negligent acts) or agreed to in writing, shall any Contributor be + liable to You for damages, including any direct, indirect, special, + incidental, or consequential damages of any character arising as a + result of this License or out of the use or inability to use the + Work (including but not limited to damages for loss of goodwill, + work stoppage, computer failure or malfunction, or any and all + other commercial damages or losses), even if such Contributor + has been advised of the possibility of such damages. + + 9. Accepting Warranty or Additional Liability. While redistributing + the Work or Derivative Works thereof, You may choose to offer, + and charge a fee for, acceptance of support, warranty, indemnity, + or other liability obligations and/or rights consistent with this + License. However, in accepting such obligations, You may act only + on Your own behalf and on Your sole responsibility, not on behalf + of any other Contributor, and only if You agree to indemnify, + defend, and hold each Contributor harmless for any liability + incurred by, or claims asserted against, such Contributor by reason + of your accepting any such warranty or additional liability. + + END OF TERMS AND CONDITIONS + + APPENDIX: How to apply the Apache License to your work. + + To apply the Apache License to your work, attach the following + boilerplate notice, with the fields enclosed by brackets "[]" + replaced with your own identifying information. (Don't include + the brackets!) The text should be enclosed in the appropriate + comment syntax for the file format. We also recommend that a + file or class name and description of purpose be included on the + same "printed page" as the copyright notice for easier + identification within third-party archives. + + Copyright [yyyy] [name of copyright owner] + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. diff --git a/exprements/PaddleClas/MANIFEST.in b/exprements/PaddleClas/MANIFEST.in new file mode 100644 index 00000000..b0a4f6dc --- /dev/null +++ b/exprements/PaddleClas/MANIFEST.in @@ -0,0 +1,7 @@ +include LICENSE.txt +include README.md +include docs/en/whl_en.md +recursive-include deploy/python predict_cls.py preprocess.py postprocess.py det_preprocess.py +recursive-include deploy/utils get_image_list.py config.py logger.py predictor.py + +recursive-include ppcls/ *.py *.txt \ No newline at end of file diff --git a/exprements/PaddleClas/README.md b/exprements/PaddleClas/README.md new file mode 100644 index 00000000..a80b5f53 --- /dev/null +++ b/exprements/PaddleClas/README.md @@ -0,0 +1 @@ +README_ch.md diff --git a/exprements/PaddleClas/README_ch.md b/exprements/PaddleClas/README_ch.md new file mode 100644 index 00000000..41c1d44f --- /dev/null +++ b/exprements/PaddleClas/README_ch.md @@ -0,0 +1,150 @@ +简体中文 | [English](README_en.md) + +# PaddleClas + +## 简介 + +飞桨图像识别套件PaddleClas是飞桨为工业界和学术界所准备的一个图像识别任务的工具集，助力使用者训练出更好的视觉模型和应用落地。 + +**近期更新** + +- 2022.1.27 全面升级文档；新增[PaddleServing C++ pipeline部署方式](./deploy/paddleserving)和[18M图像识别安卓部署Demo](./deploy/lite_shitu)。 +- 2021.11.1 发布[PP-ShiTu技术报告](https://arxiv.org/pdf/2111.00775.pdf)，新增饮料识别demo +- 2021.10.23 发布轻量级图像识别系统PP-ShiTu，CPU上0.2s即可完成在10w+库的图像识别。 +[点击这里](./docs/zh_CN/quick_start/quick_start_recognition.md)立即体验 +- 2021.09.17 发布PP-LCNet系列超轻量骨干网络模型, 在Intel CPU上，单张图像预测速度约5ms，ImageNet-1K数据集上Top1识别准确率达到80.82%，超越ResNet152的模型效果。PP-LCNet的介绍可以参考[论文](https://arxiv.org/pdf/2109.15099.pdf), 或者[PP-LCNet模型介绍](docs/zh_CN/models/PP-LCNet.md)，相关指标和预训练权重可以从 [这里](docs/zh_CN/algorithm_introduction/ImageNet_models.md)下载。 +- [more](./docs/zh_CN/others/update_history.md) + +## 特性 + +- PP-ShiTu轻量图像识别系统：集成了目标检测、特征学习、图像检索等模块，广泛适用于各类图像识别任务。cpu上0.2s即可完成在10w+库的图像识别。 + +- PP-LCNet轻量级CPU骨干网络：专门为CPU设备打造轻量级骨干网络，速度、精度均远超竞品。 + +- 丰富的预训练模型库：提供了36个系列共175个ImageNet预训练模型，其中7个精选系列模型支持结构快速修改。 + +- 全面易用的特征学习组件：集成arcmargin, triplet loss等12度量学习方法，通过配置文件即可随意组合切换。 + +- SSLD知识蒸馏：14个分类预训练模型，精度普遍提升3%以上；其中ResNet50_vd模型在ImageNet-1k数据集上的Top-1精度达到了84.0%， +Res2Net200_vd预训练模型Top-1精度高达85.1%。 + +
+ +
+ + +## 欢迎加入技术交流群 + +* 您可以扫描下面的微信群二维码， 加入PaddleClas 微信交流群。获得更高效的问题答疑，与各行各业开发者充分交流，期待您的加入。 + +
+ +
+ +## 快速体验 + +PP-ShiTu图像识别快速体验：[点击这里](./docs/zh_CN/quick_start/quick_start_recognition.md) + +## 文档教程 +- 安装说明 + - [安装Paddle](./docs/zh_CN/installation/install_paddle.md) + - [安装PaddleClas](./docs/zh_CN/installation/install_paddleclas.md) +- 快速体验 + - [PP-ShiTu图像识别快速体验](./docs/zh_CN/quick_start/quick_start_recognition.md) + - 图像分类快速体验 + - [尝鲜版](./docs/zh_CN/quick_start/quick_start_classification_new_user.md) + - [进阶版](./docs/zh_CN/quick_start/quick_start_classification_professional.md) + - [多标签分类](./docs/zh_CN/quick_start/quick_start_multilabel_classification.md) +- [PP-ShiTu图像识别系统介绍](#图像识别系统介绍) + - [主体检测](./docs/zh_CN/image_recognition_pipeline/mainbody_detection.md) + - [特征提取](./docs/zh_CN/image_recognition_pipeline/feature_extraction.md) + - [向量检索](./docs/zh_CN/image_recognition_pipeline/vector_search.md) +- [骨干网络和预训练模型库](./docs/zh_CN/algorithm_introduction/ImageNet_models.md) +- 数据准备 + - [图像分类数据集介绍](./docs/zh_CN/data_preparation/classification_dataset.md) + - [图像识别数据集介绍](./docs/zh_CN/data_preparation/recognition_dataset.md) +- 模型训练 + - [图像分类任务](./docs/zh_CN/models_training/classification.md) + - [图像识别任务](./docs/zh_CN/models_training/recognition.md) + - [训练参数调整策略](./docs/zh_CN/models_training/train_strategy.md) + - [配置文件说明](./docs/zh_CN/models_training/config_description.md) +- 模型预测部署 + - [模型导出](./docs/zh_CN/inference_deployment/export_model.md) + - Python/C++ 预测引擎 + - [基于Python预测引擎预测推理](./docs/zh_CN/inference_deployment/python_deploy.md) + - [基于C++分类预测引擎预测推理](./docs/zh_CN/inference_deployment/cpp_deploy.md)、[基于C++的PP-ShiTu预测引擎预测推理](deploy/cpp_shitu/readme.md) + - 服务化部署 + - [Paddle Serving服务化部署(推荐)](./docs/zh_CN/inference_deployment/paddle_serving_deploy.md) + - [Hub serving服务化部署](./docs/zh_CN/inference_deployment/paddle_hub_serving_deploy.md) + - [端侧部署](./deploy/lite/readme.md) + - [whl包预测](./docs/zh_CN/inference_deployment/whl_deploy.md) +- 算法介绍 + - [图像分类任务介绍](./docs/zh_CN/algorithm_introduction/image_classification.md) + - [度量学习介绍](./docs/zh_CN/algorithm_introduction/metric_learning.md) +- 高阶使用 + - [数据增广](./docs/zh_CN/advanced_tutorials/DataAugmentation.md) + - [模型量化](./docs/zh_CN/advanced_tutorials/model_prune_quantization.md) + - [知识蒸馏](./docs/zh_CN/advanced_tutorials/knowledge_distillation.md) + - [PaddleClas结构解析](./docs/zh_CN/advanced_tutorials/code_overview.md) + - [社区贡献指南](./docs/zh_CN/advanced_tutorials/how_to_contribute.md) +- FAQ + - [图像识别精选问题](docs/zh_CN/faq_series/faq_2021_s2.md) + - [图像分类精选问题](docs/zh_CN/faq_series/faq_selected_30.md) + - [图像分类FAQ第一季](docs/zh_CN/faq_series/faq_2020_s1.md) + - [图像分类FAQ第二季](docs/zh_CN/faq_series/faq_2021_s1.md) +- [许可证书](#许可证书) +- [贡献代码](#贡献代码) + + +## PP-ShiTu图像识别系统介绍 + +
+ +
+ +PP-ShiTu是一个实用的轻量级通用图像识别系统，主要由主体检测、特征学习和向量检索三个模块组成。该系统从骨干网络选择和调整、损失函数的选择、数据增强、学习率变换策略、正则化参数选择、预训练模型使用以及模型裁剪量化8个方面，采用多种策略，对各个模块的模型进行优化，最终得到在CPU上仅0.2s即可完成10w+库的图像识别的系统。更多细节请参考[PP-ShiTu技术方案](https://arxiv.org/pdf/2111.00775.pdf)。 + + + +## PP-ShiTu图像识别系统效果展示 +- 瓶装饮料识别 +
+ +
+ +- 商品识别 +
+ +
+ +- 动漫人物识别 +
+ +
+ +- logo识别 +
+ +
+ + +- 车辆识别 +
+ +
+ + + + +## 许可证书 +本项目的发布受Apache 2.0 license许可认证。 + + + +## 贡献代码 +我们非常欢迎你为PaddleClas贡献代码，也十分感谢你的反馈。 +如果想为PaddleCLas贡献代码，可以参考[贡献指南](./docs/zh_CN/advanced_tutorials/how_to_contribute.md)。 + +- 非常感谢[nblib](https://github.com/nblib)修正了PaddleClas中RandErasing的数据增广配置文件。 +- 非常感谢[chenpy228](https://github.com/chenpy228)修正了PaddleClas文档中的部分错别字。 +- 非常感谢[jm12138](https://github.com/jm12138)为PaddleClas添加ViT，DeiT系列模型和RepVGG系列模型。 diff --git a/exprements/PaddleClas/README_en.md b/exprements/PaddleClas/README_en.md new file mode 100644 index 00000000..e86ed3fe --- /dev/null +++ b/exprements/PaddleClas/README_en.md @@ -0,0 +1,132 @@ +[简体中文](README_ch.md) | English + +# PaddleClas + +## Introduction + +PaddleClas is an image recognition toolset for industry and academia, helping users train better computer vision models and apply them in real scenarios. + +**Recent updates** + +- 2021.09.17 Add PP-LCNet series model developed by PaddleClas, these models show strong competitiveness on Intel CPUs. +For the introduction of PP-LCNet, please refer to [paper](https://arxiv.org/pdf/2109.15099.pdf) or [PP-LCNet model introduction](docs/en/models/PP-LCNet_en.md). The metrics and pretrained model are available [here](docs/en/ImageNet_models_en.md). + +- 2021.06.29 Add Swin-transformer series model，Highest top1 acc on ImageNet1k dataset reaches 87.2%, training, evaluation and inference are all supported. Pretrained models can be downloaded [here](docs/en/models/models_intro_en.md). +- 2021.06.16 PaddleClas release/2.2. Add metric learning and vector search modules. Add product recognition, animation character recognition, vehicle recognition and logo recognition. Added 30 pretrained models of LeViT, Twins, TNT, DLA, HarDNet, and RedNet, and the accuracy is roughly the same as that of the paper. +- [more](./docs/en/update_history_en.md) + +## Features + +- A practical image recognition system consist of detection, feature learning and retrieval modules, widely applicable to all types of image recognition tasks. +Four sample solutions are provided, including product recognition, vehicle recognition, logo recognition and animation character recognition. + +- Rich library of pre-trained models: Provide a total of 164 ImageNet pre-trained models in 35 series, among which 6 selected series of models support fast structural modification. + +- Comprehensive and easy-to-use feature learning components: 12 metric learning methods are integrated and can be combined and switched at will through configuration files. + +- SSLD knowledge distillation: The 14 classification pre-training models generally improved their accuracy by more than 3%; among them, the ResNet50_vd model achieved a Top-1 accuracy of 84.0% on the Image-Net-1k dataset and the Res2Net200_vd pre-training model achieved a Top-1 accuracy of 85.1%. + +- Data augmentation: Provide 8 data augmentation algorithms such as AutoAugment, Cutout, Cutmix, etc. with detailed introduction, code replication and evaluation of effectiveness in a unified experimental environment. + + + + +
+ +
+ + +## Welcome to Join the Technical Exchange Group + +* You can also scan the QR code below to join the PaddleClas WeChat group to get more efficient answers to your questions and to communicate with developers from all walks of life. We look forward to hearing from you. + +
+ +
+ +## Quick Start +Quick experience of image recognition：[Link](./docs/en/tutorials/quick_start_recognition_en.md) + +## Tutorials + +- [Quick Installation](./docs/en/tutorials/install_en.md) +- [Quick Start of Recognition](./docs/en/tutorials/quick_start_recognition_en.md) +- [Introduction to Image Recognition Systems](#Introduction_to_Image_Recognition_Systems) +- [Demo images](#Demo_images) +- Algorithms Introduction + - [Backbone Network and Pre-trained Model Library](./docs/en/ImageNet_models_en.md) + - [Mainbody Detection](./docs/en/application/mainbody_detection_en.md) + - [Image Classification](./docs/en/tutorials/image_classification_en.md) + - [Feature Learning](./docs/en/application/feature_learning_en.md) + - [Product Recognition](./docs/en/application/product_recognition_en.md) + - [Vehicle Recognition](./docs/en/application/vehicle_recognition_en.md) + - [Logo Recognition](./docs/en/application/logo_recognition_en.md) + - [Animation Character Recognition](./docs/en/application/cartoon_character_recognition_en.md) + - [Vector Search](./deploy/vector_search/README.md) +- Models Training/Evaluation + - [Image Classification](./docs/en/tutorials/getting_started_en.md) + - [Feature Learning](./docs/en/tutorials/getting_started_retrieval_en.md) +- Inference Model Prediction + - [Python Inference](./docs/en/inference.md) + - [C++ Classfication Inference](./deploy/cpp/readme_en.md)， [C++ PP-ShiTu Inference](deploy/cpp_shitu/readme_en.md) +- Model Deploy (only support classification for now, recognition coming soon) + - [Hub Serving Deployment](./deploy/hubserving/readme_en.md) + - [Mobile Deployment](./deploy/lite/readme_en.md) + - [Inference Using whl](./docs/en/whl_en.md) +- Advanced Tutorial + - [Knowledge Distillation](./docs/en/advanced_tutorials/distillation/distillation_en.md) + - [Model Quantization](./docs/en/extension/paddle_quantization_en.md) + - [Data Augmentation](./docs/en/advanced_tutorials/image_augmentation/ImageAugment_en.md) +- [License](#License) +- [Contribution](#Contribution) + + +## Introduction to Image Recognition Systems + +
+ +
+ +Image recognition can be divided into three steps: +- （1）Identify region proposal for target objects through a detection model； +- （2）Extract features for each region proposal; +- （3）Search features in the retrieval database and output results; + +For a new unknown category, there is no need to retrain the model, just prepare images of new category, extract features and update retrieval database and the category can be recognised. + + +## Demo images [more](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.2/docs/images/recognition/more_demo_images) +- Product recognition +
+ +
+ +- Cartoon character recognition +
+ +
+ +- Logo recognition +
+ +
+ +- Car recognition +
+ +
+ + +## License +PaddleClas is released under the Apache 2.0 license Apache 2.0 license + + + +## Contribution +Contributions are highly welcomed and we would really appreciate your feedback!! + + +- Thank [nblib](https://github.com/nblib) to fix bug of RandErasing. +- Thank [chenpy228](https://github.com/chenpy228) to fix some typos PaddleClas. +- Thank [jm12138](https://github.com/jm12138) to add ViT, DeiT models and RepVGG models into PaddleClas. +- Thank [FutureSI](https://aistudio.baidu.com/aistudio/personalcenter/thirdview/76563) to parse and summarize the PaddleClas code. diff --git a/exprements/PaddleClas/__init__.py b/exprements/PaddleClas/__init__.py new file mode 100644 index 00000000..2128a6cc --- /dev/null +++ b/exprements/PaddleClas/__init__.py @@ -0,0 +1,17 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +__all__ = ['PaddleClas'] +from .paddleclas import PaddleClas +from ppcls.arch.backbone import * diff --git a/exprements/PaddleClas/benchmark/README.md b/exprements/PaddleClas/benchmark/README.md new file mode 100644 index 00000000..2e892d97 --- /dev/null +++ b/exprements/PaddleClas/benchmark/README.md @@ -0,0 +1,27 @@ +# benchmark使用说明 + +此目录所有shell脚本是为了测试PaddleClas中不同模型的速度指标，如单卡训练速度指标、多卡训练速度指标等。 + +## 相关脚本说明 + +一共有3个脚本： + +- `prepare_data.sh`: 下载相应的测试数据，并配置好数据路径 +- `run_benchmark.sh`: 执行单独一个训练测试的脚本，具体调用方式，可查看脚本注释 +- `run_all.sh`: 执行所有训练测试的入口脚本 + +## 使用说明 + +**注意**：为了跟PaddleClas中其他的模块的执行目录保持一致，此模块的执行目录为`PaddleClas`的根目录。 + +### 1.准备数据 + +```shell +bash benchmark/prepare_data.sh +``` + +### 2.执行所有模型的测试 + +```shell +bash benchmark/run_all.sh +``` diff --git a/exprements/PaddleClas/benchmark/prepare_data.sh b/exprements/PaddleClas/benchmark/prepare_data.sh new file mode 100644 index 00000000..411459cf --- /dev/null +++ b/exprements/PaddleClas/benchmark/prepare_data.sh @@ -0,0 +1,22 @@ +#!/bin/bash +dataset_url=$1 + +package_check_list=(imageio tqdm Cython pycocotools tb_paddle scipy pandas wget h5py sklearn opencv-python visualdl) +for package in ${package_check_list[@]}; do + if python -c "import ${package}" >/dev/null 2>&1; then + echo "${package} have already installed" + else + echo "${package} NOT FOUND" + pip install ${package} + echo "${package} installed" + fi +done + +cd dataset +rm -rf ILSVRC2012 +wget -nc ${dataset_url} +tar xf ILSVRC2012_val.tar +ln -s ILSVRC2012_val ILSVRC2012 +cd ILSVRC2012 +ln -s val_list.txt train_list.txt +cd ../../ diff --git a/exprements/PaddleClas/benchmark/run_all.sh b/exprements/PaddleClas/benchmark/run_all.sh new file mode 100644 index 00000000..a6d68841 --- /dev/null +++ b/exprements/PaddleClas/benchmark/run_all.sh @@ -0,0 +1,31 @@ +# 提供可稳定复现性能的脚本，默认在标准docker环境内py37执行： paddlepaddle/paddle:latest-gpu-cuda10.1-cudnn7 paddle=2.1.2 py=37 +# 执行目录：需说明 +# cd ** +# 1 安装该模型需要的依赖 (如需开启优化策略请注明) +# pip install ... +# 2 拷贝该模型需要数据、预训练模型 +# 3 批量运行（如不方便批量，1，2需放到单个模型中） +log_path=${LOG_PATH_INDEX_DIR:-$(pwd)} # LOG_PATH_INDEX_DIR 后续QA设置参数 +model_mode_list=(MobileNetV1 MobileNetV2 MobileNetV3_large_x1_0 ShuffleNetV2_x1_0 HRNet_W48_C SwinTransformer_tiny_patch4_window7_224 alt_gvt_base) # benchmark 监控模型列表 +#model_mode_list=(MobileNetV1 MobileNetV2 MobileNetV3_large_x1_0 EfficientNetB0 ShuffleNetV2_x1_0 DenseNet121 HRNet_W48_C SwinTransformer_tiny_patch4_window7_224 alt_gvt_base) # 该脚本支持列表 +fp_item_list=(fp32) +#bs_list=(32 64 96 128) +for model_mode in ${model_mode_list[@]}; do + for fp_item in ${fp_item_list[@]}; do + if [ ${model_mode} = MobileNetV3_large_x1_0 ] || [ ${model_mode} = ShuffleNetV2_x1_0 ]; then + bs_list=(256) + else + bs_list=(64) + fi + for bs_item in ${bs_list[@]};do + echo "index is speed, 1gpus, begin, ${model_name}" + run_mode=sp + CUDA_VISIBLE_DEVICES=0 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 1 ${model_mode} | tee ${log_path}/clas_${model_mode}_${run_mode}_bs${bs_item}_${fp_item}_1gpus 2>&1 # (5min) + sleep 10 + echo "index is speed, 8gpus, run_mode is multi_process, begin, ${model_name}" + run_mode=mp + CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 1 ${model_mode}| tee ${log_path}/clas_${model_mode}_${run_mode}_bs${bs_item}_${fp_item}_8gpus8p 2>&1 + sleep 10 + done + done +done diff --git a/exprements/PaddleClas/benchmark/run_benchmark.sh b/exprements/PaddleClas/benchmark/run_benchmark.sh new file mode 100644 index 00000000..8cee6fce --- /dev/null +++ b/exprements/PaddleClas/benchmark/run_benchmark.sh @@ -0,0 +1,68 @@ +#!/usr/bin/env bash +set -xe +# 运行示例：CUDA_VISIBLE_DEVICES=0 bash run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 500 ${model_mode} +# 参数说明 +function _set_params(){ + run_mode=${1:-"sp"} # 单卡sp|多卡mp + batch_size=${2:-"64"} + fp_item=${3:-"fp32"} # fp32|fp16 + epochs=${4:-"2"} # 可选，如果需要修改代码提前中断 + model_item=${5:-"model_item"} + run_log_path=${TRAIN_LOG_DIR:-$(pwd)} # TRAIN_LOG_DIR 后续QA设置该参数 + + index=1 + mission_name="图像分类" # 模型所属任务名称，具体可参考scripts/config.ini （必填） + direction_id=0 # 任务所属方向，0：CV，1：NLP，2：Rec。 (必填) + skip_steps=8 # 解析日志，有些模型前几个step耗时长，需要跳过 (必填) + keyword="ips:" # 解析日志，筛选出数据所在行的关键字 (必填) + keyword_loss="loss:" #选填 + model_mode=-1 # 解析日志，具体参考scripts/analysis.py. (必填) + ips_unit="images/s" + base_batch_size=$batch_size +# 以下不用修改 + device=${CUDA_VISIBLE_DEVICES//,/ } + arr=(${device}) + num_gpu_devices=${#arr[*]} + log_file=${run_log_path}/clas_${model_item}_${run_mode}_bs${batch_size}_${fp_item}_${num_gpu_devices} + model_name=${model_item}_bs${batch_size}_${fp_item} # model_item 用于yml匹配，model_name用于入库 +} +function _train(){ + echo "Train on ${num_gpu_devices} GPUs" + echo "current CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES, gpus=$num_gpu_devices, batch_size=$batch_size" + + if [ ${fp_item} = "fp32" ];then + model_config=`find ppcls/configs/ImageNet -name ${model_item}.yaml` + else + model_config=`find ppcls/configs/ImageNet -name ${model_item}_fp16.yaml` + fi + + train_cmd="-c ${model_config} -o DataLoader.Train.sampler.batch_size=${batch_size} -o Global.epochs=${epochs} -o Global.eval_during_train=False -o Global.print_batch_step=2" + case ${run_mode} in + sp) train_cmd="python -u tools/train.py ${train_cmd}" ;; + mp) + train_cmd="python -m paddle.distributed.launch --log_dir=./mylog --gpus=$CUDA_VISIBLE_DEVICES tools/train.py ${train_cmd}" + log_parse_file="mylog/workerlog.0" ;; + *) echo "choose run_mode(sp or mp)"; exit 1; + esac + rm -rf mylog +# 以下不用修改 + timeout 5m ${train_cmd} > ${log_file} 2>&1 + if [ $? -ne 0 ];then + echo -e "${model_name}, FAIL" + export job_fail_flag=1 + else + echo -e "${model_name}, SUCCESS" + export job_fail_flag=0 + fi + kill -9 `ps -ef|grep 'python'|awk '{print $2}'` + + if [ $run_mode = "mp" -a -d mylog ]; then + rm ${log_file} + cp mylog/workerlog.0 ${log_file} + fi +} + +source ${BENCHMARK_ROOT}/scripts/run_model.sh # 在该脚本中会对符合benchmark规范的log使用analysis.py 脚本进行性能数据解析;该脚本在连调时可从benchmark repo中下载https://github.com/PaddlePaddle/benchmark/blob/master/scripts/run_model.sh;如果不联调只想要产出训练log可以注掉本行,提交时需打开 +_set_params $@ +_run +#_train diff --git a/exprements/PaddleClas/dataset/.gitkeep b/exprements/PaddleClas/dataset/.gitkeep new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/deploy/__init__.py b/exprements/PaddleClas/deploy/__init__.py new file mode 100644 index 00000000..eb018c3f --- /dev/null +++ b/exprements/PaddleClas/deploy/__init__.py @@ -0,0 +1 @@ +from . import utils diff --git a/exprements/PaddleClas/deploy/configs/build_general.yaml b/exprements/PaddleClas/deploy/configs/build_general.yaml new file mode 100644 index 00000000..258b40a1 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/build_general.yaml @@ -0,0 +1,36 @@ +Global: + rec_inference_model_dir: "./models/general_PPLCNet_x2_5_lite_v1.0_infer" + batch_size: 32 + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +RecPreProcess: + transform_ops: + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null + +# indexing engine config +IndexProcess: + index_method: "HNSW32" # supported: HNSW32, IVF, Flat + image_root: "./drink_dataset_v1.0/gallery/" + index_dir: "./drink_dataset_v1.0/index" + data_file: "./drink_dataset_v1.0/gallery/drink_label.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: "\t" + dist_type: "IP" + embedding_size: 512 diff --git a/exprements/PaddleClas/deploy/configs/inference_cartoon.yaml b/exprements/PaddleClas/deploy/configs/inference_cartoon.yaml new file mode 100644 index 00000000..7d93d98c --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_cartoon.yaml @@ -0,0 +1,65 @@ +Global: + infer_imgs: "./recognition_demo_data_v1.1/test_cartoon" + det_inference_model_dir: "./models/ppyolov2_r50vd_dcn_mainbody_v1.0_infer/" + rec_inference_model_dir: "./models/cartoon_rec_ResNet50_iCartoon_v1.0_infer/" + rec_nms_thresold: 0.05 + + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 5 + labe_list: + - foreground + + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} + +DetPostProcess: {} + +RecPreProcess: + transform_ops: + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null + +# indexing engine config +IndexProcess: + index_method: "HNSW32" # supported: HNSW32, IVF, Flat + index_dir: "./recognition_demo_data_v1.1/gallery_cartoon/index/" + image_root: "./recognition_demo_data_v1.1/gallery_cartoon/" + data_file: "./recognition_demo_data_v1.1/gallery_cartoon/data_file.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: "\t" + dist_type: "IP" + embedding_size: 2048 + batch_size: 32 + return_k: 5 + score_thres: 0.5 \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_cls.yaml b/exprements/PaddleClas/deploy/configs/inference_cls.yaml new file mode 100644 index 00000000..fc0f0fe6 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_cls.yaml @@ -0,0 +1,35 @@ +Global: + infer_imgs: "./images/ILSVRC2012_val_00000010.jpeg" + inference_model_dir: "./models" + batch_size: 1 + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +PreProcess: + transform_ops: + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + channel_num: 3 + - ToCHWImage: + +PostProcess: + main_indicator: Topk + Topk: + topk: 5 + class_id_map_file: "../ppcls/utils/imagenet1k_label_list.txt" + SavePreLabel: + save_dir: ./pre_label/ \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_cls_ch4.yaml b/exprements/PaddleClas/deploy/configs/inference_cls_ch4.yaml new file mode 100644 index 00000000..9b740ed8 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_cls_ch4.yaml @@ -0,0 +1,35 @@ +Global: + infer_imgs: "./images/ILSVRC2012_val_00000010.jpeg" + inference_model_dir: "./models" + batch_size: 1 + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +PreProcess: + transform_ops: + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + channel_num: 4 + - ToCHWImage: + +PostProcess: + main_indicator: Topk + Topk: + topk: 5 + class_id_map_file: "../ppcls/utils/imagenet1k_label_list.txt" + SavePreLabel: + save_dir: ./pre_label/ \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_cls_multilabel.yaml b/exprements/PaddleClas/deploy/configs/inference_cls_multilabel.yaml new file mode 100644 index 00000000..78912d0a --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_cls_multilabel.yaml @@ -0,0 +1,35 @@ +Global: + infer_imgs: "./images/0517_2715693311.jpg" + inference_model_dir: "../inference/" + batch_size: 1 + use_gpu: True + enable_mkldnn: False + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +PreProcess: + transform_ops: + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + channel_num: 3 + - ToCHWImage: + +PostProcess: + main_indicator: MultiLabelTopk + MultiLabelTopk: + topk: 5 + class_id_map_file: None + SavePreLabel: + save_dir: ./pre_label/ \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_det.yaml b/exprements/PaddleClas/deploy/configs/inference_det.yaml new file mode 100644 index 00000000..c809a025 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_det.yaml @@ -0,0 +1,34 @@ +Global: + infer_imgs: "./images/wangzai.jpg" + det_inference_model_dir: "./models/ppyolov2_r50vd_dcn_mainbody_v1.0_infer" + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 1 + labe_list: + - foreground + + # inference engine config + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} + +DetPostProcess: {} \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_drink.yaml b/exprements/PaddleClas/deploy/configs/inference_drink.yaml new file mode 100644 index 00000000..61d116f9 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_drink.yaml @@ -0,0 +1,63 @@ +Global: + infer_imgs: "./drink_dataset_v1.0/test_images/hongniu_1.jpg" + det_inference_model_dir: "./models/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer" + rec_inference_model_dir: "./models/general_PPLCNet_x2_5_lite_v1.0_infer" + rec_nms_thresold: 0.05 + + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 5 + labe_list: + - foreground + + use_gpu: True + enable_mkldnn: False + cpu_num_threads: 10 + enable_benchmark: False + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} + +DetPostProcess: {} + +RecPreProcess: + transform_ops: + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null + +# indexing engine config +IndexProcess: + index_method: "HNSW32" # supported: HNSW32, IVF, Flat + index_dir: "./drink_dataset_v1.0/gallery" + image_root: "./drink_dataset_v1.0/index" + data_file: "./drink_dataset_v1.0/gallery/drink_label.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: " " + dist_type: "IP" + embedding_size: 512 + batch_size: 32 + return_k: 5 + score_thres: 0.4 \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_general.yaml b/exprements/PaddleClas/deploy/configs/inference_general.yaml new file mode 100644 index 00000000..6b397b50 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_general.yaml @@ -0,0 +1,63 @@ +Global: + infer_imgs: "./drink_dataset_v1.0/test_images/nongfu_spring.jpeg" + det_inference_model_dir: "./models/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer" + rec_inference_model_dir: "./models/general_PPLCNet_x2_5_lite_v1.0_infer" + rec_nms_thresold: 0.05 + + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 5 + labe_list: + - foreground + + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} + +DetPostProcess: {} + +RecPreProcess: + transform_ops: + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null + +# indexing engine config +IndexProcess: + index_method: "HNSW32" # supported: HNSW32, IVF, Flat + image_root: "./drink_dataset_v1.0/gallery/" + index_dir: "./drink_dataset_v1.0/index" + data_file: "./drink_dataset_v1.0/gallery/drink_label.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: "\t" + dist_type: "IP" + embedding_size: 512 + batch_size: 32 + return_k: 5 + score_thres: 0.5 \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_general_binary.yaml b/exprements/PaddleClas/deploy/configs/inference_general_binary.yaml new file mode 100644 index 00000000..d76dae8f --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_general_binary.yaml @@ -0,0 +1,65 @@ +Global: + infer_imgs: "./drink_dataset_v1.0/test_images/001.jpeg" + det_inference_model_dir: "./models/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer" + rec_inference_model_dir: "./models/general_PPLCNet_x2_5_lite_binary_v1.0_infer" + rec_nms_thresold: 0.05 + + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 5 + labe_list: + - foreground + + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} +DetPostProcess: {} + +RecPreProcess: + transform_ops: + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: + main_indicator: Binarize + Binarize: + method: "sign" + +# indexing engine config +IndexProcess: + index_method: "Flat" # supported: HNSW32, Flat + image_root: "./drink_dataset_v1.0/gallery/" + index_dir: "./drink_dataset_v1.0/index_bin" + data_file: "./drink_dataset_v1.0/gallery/drink_label.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: "\t" + dist_type: "hamming" + embedding_size: 512 + batch_size: 32 + return_k: 5 + hamming_radius: 100 diff --git a/exprements/PaddleClas/deploy/configs/inference_logo.yaml b/exprements/PaddleClas/deploy/configs/inference_logo.yaml new file mode 100644 index 00000000..f78ca25a --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_logo.yaml @@ -0,0 +1,63 @@ +Global: + infer_imgs: "./recognition_demo_data_v1.1/test_logo" + det_inference_model_dir: "./models/ppyolov2_r50vd_dcn_mainbody_v1.0_infer/" + rec_inference_model_dir: "./models/logo_rec_ResNet50_Logo3K_v1.0_infer/" + rec_nms_thresold: 0.05 + + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 5 + labe_list: + - foreground + + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} + +DetPostProcess: {} + +RecPreProcess: + transform_ops: + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null + +# indexing engine config +IndexProcess: + index_method: "HNSW32" # supported: HNSW32, IVF, Flat + index_dir: "./recognition_demo_data_v1.1/gallery_logo/index/" + image_root: "./recognition_demo_data_v1.1/gallery_logo/" + data_file: "./recognition_demo_data_v1.1/gallery_logo/data_file.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: "\t" + dist_type: "IP" + embedding_size: 512 + batch_size: 32 + return_k: 5 + score_thres: 0.5 \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_product.yaml b/exprements/PaddleClas/deploy/configs/inference_product.yaml new file mode 100644 index 00000000..e7b494c3 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_product.yaml @@ -0,0 +1,64 @@ +Global: + infer_imgs: "./recognition_demo_data_v1.1/test_product/daoxiangcunjinzhubing_6.jpg" + det_inference_model_dir: "./models/ppyolov2_r50vd_dcn_mainbody_v1.0_infer" + rec_inference_model_dir: "./models/product_ResNet50_vd_aliproduct_v1.0_infer" + rec_nms_thresold: 0.05 + + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 5 + labe_list: + - foreground + + # inference engine config + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} + +DetPostProcess: {} + +RecPreProcess: + transform_ops: + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null + +# indexing engine config +IndexProcess: + index_method: "HNSW32" # supported: HNSW32, IVF, Flat + index_dir: "./recognition_demo_data_v1.1/gallery_product/index" + image_root: "./recognition_demo_data_v1.1/gallery_product/" + data_file: "./recognition_demo_data_v1.1/gallery_product/data_file.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: "\t" + dist_type: "IP" + embedding_size: 512 + batch_size: 32 + return_k: 5 + score_thres: 0.5 \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_rec.yaml b/exprements/PaddleClas/deploy/configs/inference_rec.yaml new file mode 100644 index 00000000..e183ef07 --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_rec.yaml @@ -0,0 +1,28 @@ +Global: + infer_imgs: "./images/wangzai.jpg" + rec_inference_model_dir: "./models/product_ResNet50_vd_aliproduct_v1.0_infer" + batch_size: 1 + use_gpu: False + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +RecPreProcess: + transform_ops: + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/configs/inference_vehicle.yaml b/exprements/PaddleClas/deploy/configs/inference_vehicle.yaml new file mode 100644 index 00000000..d99f42ad --- /dev/null +++ b/exprements/PaddleClas/deploy/configs/inference_vehicle.yaml @@ -0,0 +1,63 @@ +Global: + infer_imgs: "./recognition_demo_data_v1.1/test_vehicle/" + det_inference_model_dir: "./models/ppyolov2_r50vd_dcn_mainbody_v1.0_infer/" + rec_inference_model_dir: "./models/vehicle_cls_ResNet50_CompCars_v1.0_infer/" + rec_nms_thresold: 0.05 + + batch_size: 1 + image_shape: [3, 640, 640] + threshold: 0.2 + max_det_results: 5 + labe_list: + - foreground + + use_gpu: True + enable_mkldnn: True + cpu_num_threads: 10 + enable_benchmark: True + use_fp16: False + ir_optim: True + use_tensorrt: False + gpu_mem: 8000 + enable_profile: False + +DetPreProcess: + transform_ops: + - DetResize: + interp: 2 + keep_ratio: false + target_size: [640, 640] + - DetNormalizeImage: + is_scale: true + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - DetPermute: {} + +DetPostProcess: {} + +RecPreProcess: + transform_ops: + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + +RecPostProcess: null + +# indexing engine config +IndexProcess: + index_method: "HNSW32" # supported: HNSW32, IVF, Flat + index_dir: "./recognition_demo_data_v1.1/gallery_vehicle/index/" + image_root: "./recognition_demo_data_v1.1/gallery_vehicle/" + data_file: "./recognition_demo_data_v1.1/gallery_vehicle/data_file.txt" + index_operation: "new" # suported: "append", "remove", "new" + delimiter: "\t" + dist_type: "IP" + embedding_size: 512 + batch_size: 32 + return_k: 5 + score_thres: 0.5 \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/cpp/CMakeLists.txt b/exprements/PaddleClas/deploy/cpp/CMakeLists.txt new file mode 100644 index 00000000..74c1c72f --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/CMakeLists.txt @@ -0,0 +1,225 @@ +project(clas_system CXX C) +cmake_minimum_required(VERSION 3.14) + +option(WITH_MKL "Compile demo with MKL/OpenBlas support, default use MKL." ON) +option(WITH_GPU "Compile demo with GPU/CPU, default use CPU." OFF) +option(WITH_STATIC_LIB "Compile demo with static/shared library, default use static." ON) +option(WITH_TENSORRT "Compile demo with TensorRT." OFF) + +SET(PADDLE_LIB "" CACHE PATH "Location of libraries") +SET(OPENCV_DIR "" CACHE PATH "Location of libraries") +SET(CUDA_LIB "" CACHE PATH "Location of libraries") +SET(CUDNN_LIB "" CACHE PATH "Location of libraries") +SET(TENSORRT_DIR "" CACHE PATH "Compile demo with TensorRT") + +set(DEMO_NAME "clas_system") + +include(external-cmake/yaml-cpp.cmake) +include_directories("${CMAKE_SOURCE_DIR}/") +include_directories("${CMAKE_CURRENT_BINARY_DIR}/ext/yaml-cpp/src/ext-yaml-cpp/include") +link_directories("${CMAKE_CURRENT_BINARY_DIR}/ext/yaml-cpp/lib") + +macro(safe_set_static_flag) + foreach(flag_var + CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_DEBUG CMAKE_CXX_FLAGS_RELEASE + CMAKE_CXX_FLAGS_MINSIZEREL CMAKE_CXX_FLAGS_RELWITHDEBINFO) + if(${flag_var} MATCHES "/MD") + string(REGEX REPLACE "/MD" "/MT" ${flag_var} "${${flag_var}}") + endif(${flag_var} MATCHES "/MD") + endforeach(flag_var) +endmacro() + +if (WITH_MKL) + ADD_DEFINITIONS(-DUSE_MKL) +endif() + +if(NOT DEFINED PADDLE_LIB) + message(FATAL_ERROR "please set PADDLE_LIB with -DPADDLE_LIB=/path/paddle/lib") +endif() + +if(NOT DEFINED OPENCV_DIR) + message(FATAL_ERROR "please set OPENCV_DIR with -DOPENCV_DIR=/path/opencv") +endif() + + +if (WIN32) + include_directories("${PADDLE_LIB}/paddle/fluid/inference") + include_directories("${PADDLE_LIB}/paddle/include") + link_directories("${PADDLE_LIB}/paddle/fluid/inference") + find_package(OpenCV REQUIRED PATHS ${OPENCV_DIR}/build/ NO_DEFAULT_PATH) + +else () + find_package(OpenCV REQUIRED PATHS ${OPENCV_DIR}/share/OpenCV NO_DEFAULT_PATH) + include_directories("${PADDLE_LIB}/paddle/include") + link_directories("${PADDLE_LIB}/paddle/lib") +endif () +include_directories(${OpenCV_INCLUDE_DIRS}) + +if (WIN32) + add_definitions("/DGOOGLE_GLOG_DLL_DECL=") + set(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} /bigobj /MTd") + set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} /bigobj /MT") + set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /bigobj /MTd") + set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /bigobj /MT") + if (WITH_STATIC_LIB) + safe_set_static_flag() + add_definitions(-DSTATIC_LIB) + endif() +else() + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -O3 -std=c++11") + set(CMAKE_STATIC_LIBRARY_PREFIX "") +endif() +message("flags" ${CMAKE_CXX_FLAGS}) + + +if (WITH_GPU) + if (NOT DEFINED CUDA_LIB OR ${CUDA_LIB} STREQUAL "") + message(FATAL_ERROR "please set CUDA_LIB with -DCUDA_LIB=/path/cuda-8.0/lib64") + endif() + if (NOT WIN32) + if (NOT DEFINED CUDNN_LIB) + message(FATAL_ERROR "please set CUDNN_LIB with -DCUDNN_LIB=/path/cudnn_v7.4/cuda/lib64") + endif() + endif(NOT WIN32) +endif() + +include_directories("${PADDLE_LIB}/third_party/install/protobuf/include") +include_directories("${PADDLE_LIB}/third_party/install/glog/include") +include_directories("${PADDLE_LIB}/third_party/install/gflags/include") +include_directories("${PADDLE_LIB}/third_party/install/xxhash/include") +include_directories("${PADDLE_LIB}/third_party/install/zlib/include") +include_directories("${PADDLE_LIB}/third_party/boost") +include_directories("${PADDLE_LIB}/third_party/eigen3") + +include_directories("${CMAKE_SOURCE_DIR}/") + +if (NOT WIN32) + if (WITH_TENSORRT AND WITH_GPU) + include_directories("${TENSORRT_DIR}/include") + link_directories("${TENSORRT_DIR}/lib") + endif() +endif(NOT WIN32) + +link_directories("${PADDLE_LIB}/third_party/install/zlib/lib") + +link_directories("${PADDLE_LIB}/third_party/install/protobuf/lib") +link_directories("${PADDLE_LIB}/third_party/install/glog/lib") +link_directories("${PADDLE_LIB}/third_party/install/gflags/lib") +link_directories("${PADDLE_LIB}/third_party/install/xxhash/lib") +link_directories("${PADDLE_LIB}/paddle/lib") + + +if(WITH_MKL) + include_directories("${PADDLE_LIB}/third_party/install/mklml/include") + if (WIN32) + set(MATH_LIB ${PADDLE_LIB}/third_party/install/mklml/lib/mklml.lib + ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5md.lib) + else () + set(MATH_LIB ${PADDLE_LIB}/third_party/install/mklml/lib/libmklml_intel${CMAKE_SHARED_LIBRARY_SUFFIX} + ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5${CMAKE_SHARED_LIBRARY_SUFFIX}) + execute_process(COMMAND cp -r ${PADDLE_LIB}/third_party/install/mklml/lib/libmklml_intel${CMAKE_SHARED_LIBRARY_SUFFIX} /usr/lib) + endif () + set(MKLDNN_PATH "${PADDLE_LIB}/third_party/install/mkldnn") + if(EXISTS ${MKLDNN_PATH}) + include_directories("${MKLDNN_PATH}/include") + if (WIN32) + set(MKLDNN_LIB ${MKLDNN_PATH}/lib/mkldnn.lib) + else () + set(MKLDNN_LIB ${MKLDNN_PATH}/lib/libmkldnn.so.0) + endif () + endif() +else() + if (WIN32) + set(MATH_LIB ${PADDLE_LIB}/third_party/install/openblas/lib/openblas${CMAKE_STATIC_LIBRARY_SUFFIX}) + else () + set(MATH_LIB ${PADDLE_LIB}/third_party/install/openblas/lib/libopenblas${CMAKE_STATIC_LIBRARY_SUFFIX}) + endif () +endif() + +# Note: libpaddle_inference_api.so/a must put before libpaddle_fluid.so/a +if(WITH_STATIC_LIB) + if(WIN32) + set(DEPS + ${PADDLE_LIB}/paddle/lib/paddle_inference${CMAKE_STATIC_LIBRARY_SUFFIX}) + else() + set(DEPS + ${PADDLE_LIB}/paddle/lib/libpaddle_inference${CMAKE_STATIC_LIBRARY_SUFFIX}) + endif() +else() + if(WIN32) + set(DEPS + ${PADDLE_LIB}/paddle/lib/paddle_inference${CMAKE_SHARED_LIBRARY_SUFFIX}) + else() + set(DEPS + ${PADDLE_LIB}/paddle/lib/libpaddle_inference${CMAKE_SHARED_LIBRARY_SUFFIX}) + endif() +endif(WITH_STATIC_LIB) + +if (NOT WIN32) + set(DEPS ${DEPS} + ${MATH_LIB} ${MKLDNN_LIB} + glog gflags protobuf z xxhash yaml-cpp + ) + if(EXISTS "${PADDLE_LIB}/third_party/install/snappystream/lib") + set(DEPS ${DEPS} snappystream) + endif() + if (EXISTS "${PADDLE_LIB}/third_party/install/snappy/lib") + set(DEPS ${DEPS} snappy) + endif() +else() + set(DEPS ${DEPS} + ${MATH_LIB} ${MKLDNN_LIB} + glog gflags_static libprotobuf xxhash libyaml-cppmt) + set(DEPS ${DEPS} libcmt shlwapi) + if (EXISTS "${PADDLE_LIB}/third_party/install/snappy/lib") + set(DEPS ${DEPS} snappy) + endif() + if(EXISTS "${PADDLE_LIB}/third_party/install/snappystream/lib") + set(DEPS ${DEPS} snappystream) + endif() +endif(NOT WIN32) + + +if(WITH_GPU) + if(NOT WIN32) + if (WITH_TENSORRT) + set(DEPS ${DEPS} ${TENSORRT_DIR}/lib/libnvinfer${CMAKE_SHARED_LIBRARY_SUFFIX}) + set(DEPS ${DEPS} ${TENSORRT_DIR}/lib/libnvinfer_plugin${CMAKE_SHARED_LIBRARY_SUFFIX}) + endif() + set(DEPS ${DEPS} ${CUDA_LIB}/libcudart${CMAKE_SHARED_LIBRARY_SUFFIX}) + set(DEPS ${DEPS} ${CUDNN_LIB}/libcudnn${CMAKE_SHARED_LIBRARY_SUFFIX}) + else() + set(DEPS ${DEPS} ${CUDA_LIB}/cudart${CMAKE_STATIC_LIBRARY_SUFFIX} ) + set(DEPS ${DEPS} ${CUDA_LIB}/cublas${CMAKE_STATIC_LIBRARY_SUFFIX} ) + set(DEPS ${DEPS} ${CUDNN_LIB}/cudnn${CMAKE_STATIC_LIBRARY_SUFFIX}) + endif() +endif() + + +if (NOT WIN32) + set(EXTERNAL_LIB "-ldl -lrt -lgomp -lz -lm -lpthread") + set(DEPS ${DEPS} ${EXTERNAL_LIB}) +endif() + +set(DEPS ${DEPS} ${OpenCV_LIBS}) + +include(FetchContent) +include(external-cmake/auto-log.cmake) +include_directories(${FETCHCONTENT_BASE_DIR}/extern_autolog-src) + +AUX_SOURCE_DIRECTORY(./src SRCS) +add_executable(${DEMO_NAME} ${SRCS}) +ADD_DEPENDENCIES(${DEMO_NAME} ext-yaml-cpp) + +target_link_libraries(${DEMO_NAME} ${DEPS}) + +if (WIN32 AND WITH_MKL) + add_custom_command(TARGET ${DEMO_NAME} POST_BUILD + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/mklml.dll ./mklml.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5md.dll ./libiomp5md.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mkldnn/lib/mkldnn.dll ./mkldnn.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/mklml.dll ./release/mklml.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5md.dll ./release/libiomp5md.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mkldnn/lib/mkldnn.dll ./release/mkldnn.dll + ) +endif() diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/ILSVRC2012_val_00000666.JPEG b/exprements/PaddleClas/deploy/cpp/docs/imgs/ILSVRC2012_val_00000666.JPEG new file mode 100644 index 00000000..ebb56252 Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/ILSVRC2012_val_00000666.JPEG differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/cpp_infer_result.png b/exprements/PaddleClas/deploy/cpp/docs/imgs/cpp_infer_result.png new file mode 100644 index 00000000..6a1adfe3 Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/cpp_infer_result.png differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step1.png b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step1.png new file mode 100644 index 00000000..58fdf2fd Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step1.png differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step2.png b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step2.png new file mode 100644 index 00000000..f1b5fd0a Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step2.png differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step3.png b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step3.png new file mode 100644 index 00000000..cb1b2aea Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step3.png differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step4.png b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step4.png new file mode 100644 index 00000000..5fc04082 Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step4.png differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step5.png b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step5.png new file mode 100644 index 00000000..fbb2e4ce Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step5.png differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step6.png b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step6.png new file mode 100644 index 00000000..86a8039c Binary files /dev/null and b/exprements/PaddleClas/deploy/cpp/docs/imgs/vs2019_step6.png differ diff --git a/exprements/PaddleClas/deploy/cpp/docs/windows_vs2019_build.md b/exprements/PaddleClas/deploy/cpp/docs/windows_vs2019_build.md new file mode 100644 index 00000000..6e2b3703 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/docs/windows_vs2019_build.md @@ -0,0 +1,119 @@ +# Visual Studio 2019 Community CMake 编译指南 + +PaddleClas在Windows 平台下基于`Visual Studio 2019 Community` 进行了测试。微软从`Visual Studio 2017`开始即支持直接管理`CMake`跨平台编译项目，但是直到`2019`才提供了稳定和完全的支持，所以如果你想使用CMake管理项目编译构建，我们推荐使用`Visual Studio 2019`。如果您希望通过生成`sln解决方案`的方式进行编译，可以参考该文档：[https://zhuanlan.zhihu.com/p/145446681](https://zhuanlan.zhihu.com/p/145446681)。 + + +## 前置条件 +* Visual Studio 2019 +* CUDA 9.0 / CUDA 10.0，cudnn 7.6+ （仅在使用GPU版本的预测库时需要） +* CMake 3.0+ + +请确保系统已经正确安装并配置好上述基本软件，其中： + * 在安装`Visual Studio 2019`时，`工作负载`需要勾选`使用C++的桌面开发`； + * CUDA需要正确安装并设置系统环境变量； + * CMake需要正确安装并将路径添加到系统环境变量中。 + +以下测试基于`Visual Studio 2019 Community`版本。 + +**下面所有示例以工作目录为 `D:\projects`演示**。 + +### Step1: 下载PaddlePaddle C++ 预测库 paddle_inference_install_dir + +PaddlePaddle C++ 预测库针对不同的`CPU`和`CUDA`版本提供了不同的预编译版本，请根据实际情况下载: [C++预测库下载列表](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/windows_cpp_inference.html)。 + +解压后`D:\projects\paddle_inference_install_dir`目录包含内容为： + +``` +paddle_inference_install_dir +├── paddle # paddle核心库和头文件 +| +├── third_party # 第三方依赖库和头文件 +| +└── version.txt # 版本和编译信息 +``` + +然后需要将`Paddle预测库`的路径`D:\projects\paddle_inference_install_dir\paddle\lib`添加到系统环境变量`Path`中。 + +### Step2: 安装配置OpenCV + +1. 在OpenCV官网下载适用于Windows平台的3.4.6版本， [下载地址](https://sourceforge.net/projects/opencvlibrary/files/3.4.6/opencv-3.4.6-vc14_vc15.exe/download) +2. 运行下载的可执行文件，将OpenCV解压至指定目录，如`D:\projects\opencv` +3. 配置环境变量，如下流程所示 + - 此电脑（我的电脑）-> 属性 -> 高级系统设置 -> 环境变量 + - 在系统变量中找到Path（如没有，自行创建），并双击编辑 + - 新建，将OpenCV路径填入并保存，如 `D:\projects\opencv\build\x64\vc14\bin` + +### Step3: 使用Visual Studio 2019直接编译CMake + +1. 打开Visual Studio 2019 Community，点击 `继续但无需代码` + +![step2](./imgs/vs2019_step1.png) + +2. 点击： `文件`->`打开`->`CMake` + +![step2.1](./imgs/vs2019_step2.png) + +选择项目代码所在路径，并打开`CMakeList.txt`： + +![step2.2](./imgs/vs2019_step3.png) + +3. 点击：`项目`->`cpp_inference_demo的CMake设置` + +![step3](./imgs/vs2019_step4.png) + +4. 请设置以下参数的值 + + +| 名称 | 值 | 保存到 JSON | +| ----------------------------- | ------------------ | ----------- | +| CMAKE_BACKWARDS_COMPATIBILITY | 3.17 | [√] | +| CMAKE_BUILD_TYPE | RelWithDebInfo | [√] | +| CUDA_LIB | CUDA的库路径 | [√] | +| CUDNN_LIB | CUDNN的库路径 | [√] | +| OpenCV_DIR | OpenCV的安装路径 | [√] | +| PADDLE_LIB | Paddle预测库的路径 | [√] | +| WITH_GPU | [√] | [√] | +| WITH_MKL | [√] | [√] | +| WITH_STATIC_LIB | [√] | [√] | + +**注意**： + +1. `CMAKE_BACKWARDS_COMPATIBILITY` 的值请根据自己 `cmake` 版本设置，`cmake` 版本可以通过命令：`cmake --version` 查询； +2. `CUDA_LIB` 、 `CUDNN_LIB` 的值仅需在使用**GPU版本**预测库时指定，其中CUDA库版本尽量对齐，**使用9.0、10.0版本，不使用9.2、10.1等版本CUDA库**； +3. 在设置 `CUDA_LIB`、`CUDNN_LIB`、`OPENCV_DIR`、`PADDLE_LIB` 时，点击 `浏览`，分别设置相应的路径； + * `CUDA_LIB`和`CUDNN_LIB`：该路径取决于CUDA与CUDNN的安装位置。 + * `OpenCV_DIR`：该路径下需要有`.cmake`文件，一般为`opencv/build/`； + * `PADDLE_LIB`：该路径下需要有`CMakeCache.txt`文件，一般为`paddle_inference_install_dir/`。 +4. 在使用 `CPU` 版预测库时，请不要勾选 `WITH_GPU` - `保存到 JSON`。 + +![step4](./imgs/vs2019_step5.png) + +**设置完成后**, 点击上图中 `保存并生成CMake缓存以加载变量` 。 + +5. 点击`生成`->`全部生成` + +![step6](./imgs/vs2019_step6.png) + + +### Step4: 预测及可视化 + +在完成上述操作后，`Visual Studio 2019` 编译产出的可执行文件 `clas_system.exe` 在 `out\build\x64-Release`目录下，打开`cmd`，并切换到该目录： + +``` +cd D:\projects\PaddleClas\deploy\cpp_infer\out\build\x64-Release +``` +可执行文件`clas_system.exe`即为编译产出的的预测程序，其使用方法如下： + +```shell +.\clas_system.exe D:\projects\PaddleClas\deploy\cpp_infer\tools\config.txt .\docs\ILSVRC2012_val_00008306.JPEG +``` + +上述命令中，第一个参数（`D:\projects\PaddleClas\deploy\cpp_infer\tools\config.txt`）为配置文件路径，第二个参数（`.\docs\ILSVRC2012_val_00008306.JPEG`）为需要预测的图片路径。 + +注意，需要在配置文件中正确设置预测参数，包括所用模型文件的路径（`cls_model_path`和`cls_params_path`）。 + + +### 注意 +* 在Windows下的终端中执行文件exe时，可能会发生乱码的现象，此时需要在终端中输入`CHCP 65001`，将终端的编码方式由GBK编码(默认)改为UTF-8编码，更加具体的解释可以参考这篇博客：[https://blog.csdn.net/qq_35038153/article/details/78430359](https://blog.csdn.net/qq_35038153/article/details/78430359)。 +* 如果需要使用CPU预测，PaddlePaddle在Windows上仅支持avx的CPU预测，目前不支持noavx的CPU预测。 +* 在使用生成的`clas_system.exe`进行预测时，如提示`由于找不到paddle_fluid.dll，无法继续执行代码。重新安装程序可能会解决此问题`，请检查是否将Paddle预测库路径添加到系统环境变量，详见[Step1: 下载PaddlePaddle C++ 预测库 paddle_inference_install_dir](#step1-下载paddlepaddle-c-预测库-paddle_inference_install_dir) diff --git a/exprements/PaddleClas/deploy/cpp/external-cmake/auto-log.cmake b/exprements/PaddleClas/deploy/cpp/external-cmake/auto-log.cmake new file mode 100644 index 00000000..9be9c2fb --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/external-cmake/auto-log.cmake @@ -0,0 +1,12 @@ +find_package(Git REQUIRED) +include(FetchContent) + +set(FETCHCONTENT_BASE_DIR "${CMAKE_CURRENT_BINARY_DIR}/third-party") + +FetchContent_Declare( + extern_Autolog + PREFIX autolog + GIT_REPOSITORY https://github.com/LDOUBLEV/AutoLog.git + GIT_TAG main +) +FetchContent_MakeAvailable(extern_Autolog) diff --git a/exprements/PaddleClas/deploy/cpp/external-cmake/yaml-cpp.cmake b/exprements/PaddleClas/deploy/cpp/external-cmake/yaml-cpp.cmake new file mode 100644 index 00000000..7bc7f34d --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/external-cmake/yaml-cpp.cmake @@ -0,0 +1,30 @@ + +find_package(Git REQUIRED) + +include(ExternalProject) + +message("${CMAKE_BUILD_TYPE}") + +ExternalProject_Add( + ext-yaml-cpp + URL https://bj.bcebos.com/paddlex/deploy/deps/yaml-cpp.zip + URL_MD5 9542d6de397d1fbd649ed468cb5850e6 + CMAKE_ARGS + -DYAML_CPP_BUILD_TESTS=OFF + -DYAML_CPP_BUILD_TOOLS=OFF + -DYAML_CPP_INSTALL=OFF + -DYAML_CPP_BUILD_CONTRIB=OFF + -DMSVC_SHARED_RT=OFF + -DBUILD_SHARED_LIBS=OFF + -DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE} + -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS} + -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG} + -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE} + -DCMAKE_LIBRARY_OUTPUT_DIRECTORY=${CMAKE_BINARY_DIR}/ext/yaml-cpp/lib + -DCMAKE_ARCHIVE_OUTPUT_DIRECTORY=${CMAKE_BINARY_DIR}/ext/yaml-cpp/lib + PREFIX "${CMAKE_BINARY_DIR}/ext/yaml-cpp" + # Disable install step + INSTALL_COMMAND "" + LOG_DOWNLOAD ON + LOG_BUILD 1 +) diff --git a/exprements/PaddleClas/deploy/cpp/include/cls.h b/exprements/PaddleClas/deploy/cpp/include/cls.h new file mode 100644 index 00000000..8456d108 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/include/cls.h @@ -0,0 +1,90 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include "paddle_inference_api.h" +#include +#include +#include +#include +#include + +#include +#include +#include + +#include "include/cls_config.h" +#include + +using namespace paddle_infer; + +namespace PaddleClas { + + class Classifier { + public: + explicit Classifier(const ClsConfig &config) { + this->use_gpu_ = config.use_gpu; + this->gpu_id_ = config.gpu_id; + this->gpu_mem_ = config.gpu_mem; + this->cpu_math_library_num_threads_ = config.cpu_threads; + this->use_fp16_ = config.use_fp16; + this->use_mkldnn_ = config.use_mkldnn; + this->use_tensorrt_ = config.use_tensorrt; + this->mean_ = config.mean; + this->std_ = config.std; + this->resize_short_size_ = config.resize_short_size; + this->scale_ = config.scale; + this->crop_size_ = config.crop_size; + this->ir_optim_ = config.ir_optim; + LoadModel(config.cls_model_path, config.cls_params_path); + } + + // Load Paddle inference model + void LoadModel(const std::string &model_path, const std::string ¶ms_path); + + // Run predictor + void Run(cv::Mat &img, std::vector &out_data, std::vector &idx, + std::vector ×); + + private: + std::shared_ptr predictor_; + + bool use_gpu_ = false; + int gpu_id_ = 0; + int gpu_mem_ = 4000; + int cpu_math_library_num_threads_ = 4; + bool use_mkldnn_ = false; + bool use_tensorrt_ = false; + bool use_fp16_ = false; + bool ir_optim_ = true; + + std::vector mean_ = {0.485f, 0.456f, 0.406f}; + std::vector std_ = {0.229f, 0.224f, 0.225f}; + float scale_ = 0.00392157; + + int resize_short_size_ = 256; + int crop_size_ = 224; + + // pre-process + ResizeImg resize_op_; + Normalize normalize_op_; + Permute permute_op_; + CenterCropImg crop_op_; + }; + +} // namespace PaddleClas diff --git a/exprements/PaddleClas/deploy/cpp/include/cls_config.h b/exprements/PaddleClas/deploy/cpp/include/cls_config.h new file mode 100644 index 00000000..225e728b --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/include/cls_config.h @@ -0,0 +1,133 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#ifdef WIN32 +#define OS_PATH_SEP "\\" +#else +#define OS_PATH_SEP "/" +#endif + +#include "include/utility.h" +#include "yaml-cpp/yaml.h" +#include +#include +#include +#include +#include +#include + +namespace PaddleClas { + + class ClsConfig { + public: + explicit ClsConfig(const std::string &path) { + ReadYamlConfig(path); + this->infer_imgs = + this->config_file["Global"]["infer_imgs"].as(); + this->batch_size = this->config_file["Global"]["batch_size"].as(); + this->use_gpu = this->config_file["Global"]["use_gpu"].as(); + if (this->config_file["Global"]["gpu_id"].IsDefined()) + this->gpu_id = this->config_file["Global"]["gpu_id"].as(); + else + this->gpu_id = 0; + this->gpu_mem = this->config_file["Global"]["gpu_mem"].as(); + this->cpu_threads = + this->config_file["Global"]["cpu_num_threads"].as(); + this->use_mkldnn = this->config_file["Global"]["enable_mkldnn"].as(); + this->use_tensorrt = this->config_file["Global"]["use_tensorrt"].as(); + this->use_fp16 = this->config_file["Global"]["use_fp16"].as(); + this->enable_benchmark = + this->config_file["Global"]["enable_benchmark"].as(); + this->ir_optim = this->config_file["Global"]["ir_optim"].as(); + this->enable_profile = + this->config_file["Global"]["enable_profile"].as(); + this->cls_model_path = + this->config_file["Global"]["inference_model_dir"].as() + + OS_PATH_SEP + "inference.pdmodel"; + this->cls_params_path = + this->config_file["Global"]["inference_model_dir"].as() + + OS_PATH_SEP + "inference.pdiparams"; + this->resize_short_size = + this->config_file["PreProcess"]["transform_ops"][0]["ResizeImage"] + ["resize_short"] + .as(); + this->crop_size = + this->config_file["PreProcess"]["transform_ops"][1]["CropImage"]["size"] + .as(); + this->scale = this->config_file["PreProcess"]["transform_ops"][2] + ["NormalizeImage"]["scale"] + .as(); + this->mean = this->config_file["PreProcess"]["transform_ops"][2] + ["NormalizeImage"]["mean"] + .as < std::vector < float >> (); + this->std = this->config_file["PreProcess"]["transform_ops"][2] + ["NormalizeImage"]["std"] + .as < std::vector < float >> (); + if (this->config_file["Global"]["benchmark"].IsDefined()) + this->benchmark = this->config_file["Global"]["benchmark"].as(); + else + this->benchmark = false; + + if (this->config_file["PostProcess"]["Topk"]["topk"].IsDefined()) + this->topk = this->config_file["PostProcess"]["Topk"]["topk"].as(); + if (this->config_file["PostProcess"]["Topk"]["class_id_map_file"] + .IsDefined()) + this->class_id_map_path = + this->config_file["PostProcess"]["Topk"]["class_id_map_file"] + .as(); + if (this->config_file["PostProcess"]["SavePreLabel"]["save_dir"] + .IsDefined()) + this->label_save_dir = + this->config_file["PostProcess"]["SavePreLabel"]["save_dir"] + .as(); + ReadLabelMap(); + } + + YAML::Node config_file; + bool use_gpu = false; + int gpu_id = 0; + int gpu_mem = 4000; + int cpu_threads = 1; + bool use_mkldnn = false; + bool use_tensorrt = false; + bool use_fp16 = false; + bool benchmark = false; + int batch_size = 1; + bool enable_benchmark = false; + bool ir_optim = true; + bool enable_profile = false; + + std::string cls_model_path; + std::string cls_params_path; + std::string infer_imgs; + + int resize_short_size = 256; + int crop_size = 224; + float scale = 0.00392157; + std::vector mean = {0.485, 0.456, 0.406}; + std::vector std = {0.229, 0.224, 0.225}; + int topk = 5; + std::string class_id_map_path; + std::map id_map; + std::string label_save_dir; + + void PrintConfigInfo(); + + void ReadLabelMap(); + + void ReadYamlConfig(const std::string &path); + }; +} // namespace PaddleClas diff --git a/exprements/PaddleClas/deploy/cpp/include/preprocess_op.h b/exprements/PaddleClas/deploy/cpp/include/preprocess_op.h new file mode 100644 index 00000000..48c04e47 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/include/preprocess_op.h @@ -0,0 +1,56 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); you may not +// use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include +#include +#include +#include +#include + +#include +#include +#include + +using namespace std; + +namespace PaddleClas { + + class Normalize { + public: + virtual void Run(cv::Mat *im, const std::vector &mean, + const std::vector &std, float &scale); + }; + +// RGB -> CHW + class Permute { + public: + virtual void Run(const cv::Mat *im, float *data); + }; + + class CenterCropImg { + public: + virtual void Run(cv::Mat &im, const int crop_size = 224); + }; + + class ResizeImg { + public: + virtual void Run(const cv::Mat &img, cv::Mat &resize_img, int max_size_len); + }; + +} // namespace PaddleClas diff --git a/exprements/PaddleClas/deploy/cpp/include/utility.h b/exprements/PaddleClas/deploy/cpp/include/utility.h new file mode 100644 index 00000000..73b594ef --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/include/utility.h @@ -0,0 +1,46 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" + +namespace PaddleClas { + + class Utility { + public: + static std::vector ReadDict(const std::string &path); + + // template + // inline static size_t argmax(ForwardIterator first, ForwardIterator last) + // { + // return std::distance(first, std::max_element(first, last)); + // } + }; + +} // namespace PaddleClas \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/cpp/readme.md b/exprements/PaddleClas/deploy/cpp/readme.md new file mode 100644 index 00000000..ec40f23d --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/readme.md @@ -0,0 +1,229 @@ +# 服务器端C++预测 + +本教程将介绍在服务器端部署PaddleClas模型的详细步骤。 + + +## 1. 准备环境 + +### 运行准备 +- Linux环境，推荐使用docker。 +- Windows环境，目前支持基于`Visual Studio 2019 Community`进行编译；此外，如果您希望通过生成`sln解决方案`的方式进行编译，可以参考该文档：[https://zhuanlan.zhihu.com/p/145446681](https://zhuanlan.zhihu.com/p/145446681) + +* 该文档主要介绍基于Linux环境下的PaddleClas C++预测流程，如果需要在Windows环境下使用预测库进行C++预测，具体编译方法请参考[Windows下编译教程](./docs/windows_vs2019_build.md)。 + +### 1.1 编译opencv库 + +* 首先需要从opencv官网上下载在Linux环境下源码编译的包，以3.4.7版本为例，下载及解压缩命令如下： + +``` +wget https://github.com/opencv/opencv/archive/3.4.7.tar.gz +tar -xvf 3.4.7.tar.gz +``` + +最终可以在当前目录下看到`opencv-3.4.7/`的文件夹。 + +* 编译opencv，首先设置opencv源码路径(`root_path`)以及安装路径(`install_path`)，`root_path`为下载的opencv源码路径，`install_path`为opencv的安装路径。在本例中，源码路径即为当前目录下的`opencv-3.4.7/`。 + +```shell +cd ./opencv-3.4.7 +export root_path=$PWD +export install_path=${root_path}/opencv3 +``` + +* 然后在opencv源码路径下，按照下面的方式进行编译。 + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DCMAKE_INSTALL_PREFIX=${install_path} \ + -DCMAKE_BUILD_TYPE=Release \ + -DBUILD_SHARED_LIBS=OFF \ + -DWITH_IPP=OFF \ + -DBUILD_IPP_IW=OFF \ + -DWITH_LAPACK=OFF \ + -DWITH_EIGEN=OFF \ + -DCMAKE_INSTALL_LIBDIR=lib64 \ + -DWITH_ZLIB=ON \ + -DBUILD_ZLIB=ON \ + -DWITH_JPEG=ON \ + -DBUILD_JPEG=ON \ + -DWITH_PNG=ON \ + -DBUILD_PNG=ON \ + -DWITH_TIFF=ON \ + -DBUILD_TIFF=ON + +make -j +make install +``` + +* `make install`完成之后，会在该文件夹下生成opencv头文件和库文件，用于后面的PaddleClas代码编译。 + +以opencv3.4.7版本为例，最终在安装路径下的文件结构如下所示。**注意**：不同的opencv版本，下述的文件结构可能不同。 + +``` +opencv3/ +|-- bin +|-- include +|-- lib64 +|-- share +``` + +### 1.2 下载或者编译Paddle预测库 + +* 有2种方式获取Paddle预测库，下面进行详细介绍。 + +#### 1.2.1 预测库源码编译 +* 如果希望获取最新预测库特性，可以从Paddle github上克隆最新代码，源码编译预测库。 +* 可以参考[Paddle预测库官网](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_and_install_lib_cn.html#id16)的说明，从github上获取Paddle代码，然后进行编译，生成最新的预测库。使用git获取代码方法如下。 + +```shell +git clone https://github.com/PaddlePaddle/Paddle.git +``` + +* 进入Paddle目录后，使用如下方法编译。 + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DWITH_CONTRIB=OFF \ + -DWITH_MKL=ON \ + -DWITH_MKLDNN=ON \ + -DWITH_TESTING=OFF \ + -DCMAKE_BUILD_TYPE=Release \ + -DWITH_INFERENCE_API_TEST=OFF \ + -DON_INFER=ON \ + -DWITH_PYTHON=ON +make -j +make inference_lib_dist +``` + +更多编译参数选项可以参考Paddle C++预测库官网：[https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_and_install_lib_cn.html#id16](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_and_install_lib_cn.html#id16)。 + + +* 编译完成之后，可以在`build/paddle_inference_install_dir/`文件下看到生成了以下文件及文件夹。 + +``` +build/paddle_inference_install_dir/ +|-- CMakeCache.txt +|-- paddle +|-- third_party +|-- version.txt +``` + +其中`paddle`就是之后进行C++预测时所需的Paddle库，`version.txt`中包含当前预测库的版本信息。 + +#### 1.2.2 直接下载安装 + +* [Paddle预测库官网](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_and_install_lib_cn.html#id1)上提供了不同cuda版本的Linux预测库，可以在官网查看并选择合适的预测库版本，注意必须选择`develop`版本。 + + 以`2.1.1-gpu-cuda10.2-cudnn8.1-mkl-gcc8.2`的`develop`版本为例，使用下述命令下载并解压： + + +```shell +wget https://paddle-inference-lib.bj.bcebos.com/2.1.1-gpu-cuda10.2-cudnn8.1-mkl-gcc8.2/paddle_inference.tgz +tar -xvf paddle_inference.tgz +``` + + +最终会在当前的文件夹中生成`paddle_inference/`的子文件夹。 + + +## 2 开始运行 + +### 2.1 将模型导出为inference model + +* 可以参考[模型导出](../../tools/export_model.py)，导出`inference model`，用于模型预测。得到预测模型后，假设模型文件放在`inference`目录下，则目录结构如下。 + +``` +inference/ +|--inference.pdmodel +|--inference.pdiparams +``` +**注意**：上述文件中，`inference.pdmodel`文件存储了模型结构信息，`inference.pdiparams`文件存储了模型参数信息。模型目录可以随意设置，但是模型名字不能修改。 + +### 2.2 编译PaddleClas C++预测demo + +* 编译命令如下，其中Paddle C++预测库、opencv等其他依赖库的地址需要换成自己机器上的实际地址。 + + +```shell +sh tools/build.sh +``` + +具体地，`tools/build.sh`中内容如下。 + +```shell +OPENCV_DIR=your_opencv_dir +LIB_DIR=your_paddle_inference_dir +CUDA_LIB_DIR=your_cuda_lib_dir +CUDNN_LIB_DIR=your_cudnn_lib_dir +TENSORRT_DIR=your_tensorrt_lib_dir + +BUILD_DIR=build +rm -rf ${BUILD_DIR} +mkdir ${BUILD_DIR} +cd ${BUILD_DIR} +cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DDEMO_NAME=clas_system \ + -DWITH_GPU=OFF \ + -DWITH_STATIC_LIB=OFF \ + -DWITH_TENSORRT=OFF \ + -DTENSORRT_DIR=${TENSORRT_DIR} \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + +make -j +cd .. +``` + +上述命令中， + +* `OPENCV_DIR`为opencv编译安装的地址（本例中为`opencv-3.4.7/opencv3`文件夹的路径）； + +* `LIB_DIR`为下载的Paddle预测库（`paddle_inference`文件夹），或编译生成的Paddle预测库（`build/paddle_inference_install_dir`文件夹）的路径； + +* `CUDA_LIB_DIR`为cuda库文件地址，在docker中为`/usr/local/cuda/lib64`； + +* `CUDNN_LIB_DIR`为cudnn库文件地址，在docker中为`/usr/lib/x86_64-linux-gnu/`。 + +* `TENSORRT_DIR`是tensorrt库文件地址，在dokcer中为`/usr/local/TensorRT6-cuda10.0-cudnn7/`，TensorRT需要结合GPU使用。 + +在执行上述命令，编译完成之后，会在当前路径下生成`build`文件夹，其中生成一个名为`clas_system`的可执行文件。 + + +### 2.3 运行demo +#### 2.3.1 设置配置文件 + +```shell +cp ../configs/inference_cls.yaml tools/ +``` + +根据[python预测推理](../../docs/zh_CN/inference_deployment/python_deploy.md)的`图像分类推理`部分修改好`tools`目录下`inference_cls.yaml`文件。`yaml`文件的参数说明详见[python预测推理](../../docs/zh_CN/inference_deployment/python_deploy.md)。 + +请根据实际存放文件，修改好`Global.infer_imgs`、`Global.inference_model_dir`、`PostProcess.Topk.topk`、`PostProcess.Topk.class_id_map_file`等参数。 + +#### 2.3.2 执行 + +```shell +./build/clas_system -c tools/inference_cls.yaml +# or +./build/clas_system -config tools/inference_cls.yaml +``` + +最终屏幕上会输出结果，如下图所示。 + +
+ +
+ + +其中`class id`表示置信度最高的类别对应的id，score表示图片属于该类别的概率。 diff --git a/exprements/PaddleClas/deploy/cpp/readme_en.md b/exprements/PaddleClas/deploy/cpp/readme_en.md new file mode 100644 index 00000000..fe5abe25 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/readme_en.md @@ -0,0 +1,240 @@ +# Server-side C++ inference + + +In this tutorial, we will introduce the detailed steps of deploying PaddleClas models on the server side. + + +## 1. Prepare the environment + +### Environment + +- Linux, docker is recommended. +- Windows, compilation based on `Visual Studio 2019 Community` is supported. In addition, you can refer to [How to use PaddleDetection to make a complete project](https://zhuanlan.zhihu.com/p/145446681) to compile by generating the `sln solution`. +- This document mainly introduces the compilation and inference of PaddleClas C++ in Linux environment. +- If you need to use the Inference Library in Windows environment, please refer to [The compilation tutorial in Windows](./docs/windows_vs2019_build.md) for detailed information. + + +### 1.1 Compile opencv + +* First of all, you need to download the source code compiled package in the Linux environment from the opencv official website. Taking opencv3.4.7 as an example, the download and uncompress command are as follows. + +``` +wget https://github.com/opencv/opencv/archive/3.4.7.tar.gz +tar -xf 3.4.7.tar.gz +``` + +Finally, you can see the folder of `opencv-3.4.7/` in the current directory. + +* Compile opencv, the opencv source path (`root_path`) and installation path (`install_path`) should be set by yourself. Among them, `root_path` is the downloaded opencv source code path, and `install_path` is the installation path of opencv. In this case, the opencv source is `./opencv-3.4.7`. + +```shell +cd ./opencv-3.4.7 +export root_path=$PWD +export install_path=${root_path}/opencv3 +``` + +* After entering the opencv source code path, you can compile it in the following way. + + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DCMAKE_INSTALL_PREFIX=${install_path} \ + -DCMAKE_BUILD_TYPE=Release \ + -DBUILD_SHARED_LIBS=OFF \ + -DWITH_IPP=OFF \ + -DBUILD_IPP_IW=OFF \ + -DWITH_LAPACK=OFF \ + -DWITH_EIGEN=OFF \ + -DCMAKE_INSTALL_LIBDIR=lib64 \ + -DWITH_ZLIB=ON \ + -DBUILD_ZLIB=ON \ + -DWITH_JPEG=ON \ + -DBUILD_JPEG=ON \ + -DWITH_PNG=ON \ + -DBUILD_PNG=ON \ + -DWITH_TIFF=ON \ + -DBUILD_TIFF=ON + +make -j +make install +``` + +* After `make install` is completed, the opencv header file and library file will be generated in this folder for later PaddleClas source code compilation. + +Take opencv3.4.7 for example, the final file structure under the opencv installation path is as follows. **NOTICE**:The following file structure may be different for different Versions of Opencv. + +``` +opencv3/ +|-- bin +|-- include +|-- lib64 +|-- share +``` + +### 1.2 Compile or download the Paddle Inference Library + +* There are 2 ways to obtain the Paddle Inference Library, described in detail below. + + +#### 1.2.1 Compile from the source code +* If you want to get the latest Paddle Inference Library features, you can download the latest code from Paddle GitHub repository and compile the inference library from the source code. +* You can refer to [Paddle Inference Library](https://www.paddlepaddle.org.cn/documentation/docs/en/develop/guides/05_inference_deployment/inference/build_and_install_lib_en.html#build-from-source-code) to get the Paddle source code from github, and then compile To generate the latest inference library. The method of using git to access the code is as follows. + + +```shell +git clone https://github.com/PaddlePaddle/Paddle.git +``` + +* After entering the Paddle directory, the compilation method is as follows. + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DWITH_CONTRIB=OFF \ + -DWITH_MKL=ON \ + -DWITH_MKLDNN=ON \ + -DWITH_TESTING=OFF \ + -DCMAKE_BUILD_TYPE=Release \ + -DWITH_INFERENCE_API_TEST=OFF \ + -DON_INFER=ON \ + -DWITH_PYTHON=ON +make -j +make inference_lib_dist +``` + +For more compilation parameter options, please refer to the official website of the Paddle C++ inference library:[https://www.paddlepaddle.org.cn/documentation/docs/en/develop/guides/05_inference_deployment/inference/build_and_install_lib_en.html#build-from-source-code](https://www.paddlepaddle.org.cn/documentation/docs/en/develop/guides/05_inference_deployment/inference/build_and_install_lib_en.html#build-from-source-code). + + +* After the compilation process, you can see the following files in the folder of `build/paddle_inference_install_dir/`. + +``` +build/paddle_inference_install_dir/ +|-- CMakeCache.txt +|-- paddle +|-- third_party +|-- version.txt +``` + +Among them, `paddle` is the Paddle library required for C++ prediction later, and `version.txt` contains the version information of the current inference library. + + + +#### 1.2.2 Direct download and installation + +* Different cuda versions of the Linux inference library (based on GCC 4.8.2) are provided on the +[Paddle Inference Library official website](https://www.paddlepaddle.org.cn/documentation/docs/en/develop/guides/05_inference_deployment/inference/build_and_install_lib_en.html). You can view and select the appropriate version of the inference library on the official website. + +* Please select the `develop` version. + +* After downloading, use the following method to uncompress. + +``` +tar -xf paddle_inference.tgz +``` + +Finally you can see the following files in the folder of `paddle_inference/`. + + +## 2. Compile and run the demo + +### 2.1 Export the inference model + +* You can refer to [Model inference](../../tools/export_model.py)，export the inference model. After the model is exported, assuming it is placed in the `inference` directory, the directory structure is as follows. + +``` +inference/ +|--cls_infer.pdmodel +|--cls_infer.pdiparams +``` + +**NOTICE**: Among them, `cls_infer.pdmodel` file stores the model structure information and the `cls_infer.pdiparams` file stores the model parameter information.The paths of the two files need to correspond to the parameters of `cls_model_path` and `cls_params_path` in the configuration file `tools/config.txt`. + +### 2.2 Compile PaddleClas C++ inference demo + + +* The compilation commands are as follows. The addresses of Paddle C++ inference library, opencv and other Dependencies need to be replaced with the actual addresses on your own machines. + +```shell +sh tools/build.sh +``` + +Specifically, the content in `tools/build.sh` is as follows. + +```shell +OPENCV_DIR=your_opencv_dir +LIB_DIR=your_paddle_inference_dir +CUDA_LIB_DIR=your_cuda_lib_dir +CUDNN_LIB_DIR=your_cudnn_lib_dir +TENSORRT_DIR=your_tensorrt_lib_dir + +BUILD_DIR=build +rm -rf ${BUILD_DIR} +mkdir ${BUILD_DIR} +cd ${BUILD_DIR} +cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DDEMO_NAME=clas_system \ + -DWITH_GPU=OFF \ + -DWITH_STATIC_LIB=OFF \ + -DWITH_TENSORRT=OFF \ + -DTENSORRT_DIR=${TENSORRT_DIR} \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + +make -j +``` + +In the above parameters of command: + +* `OPENCV_DIR` is the opencv installation path; + +* `LIB_DIR` is the download (`paddle_inference` folder) or the generated Paddle Inference Library path (`build/paddle_inference_install_dir` folder); + +* `CUDA_LIB_DIR` is the cuda library file path, in docker; it is `/usr/local/cuda/lib64`; + +* `CUDNN_LIB_DIR` is the cudnn library file path, in docker it is `/usr/lib/x86_64-linux-gnu/`. + +* `TENSORRT_DIR` is the tensorrt library file path，in dokcer it is `/usr/local/TensorRT6-cuda10.0-cudnn7/`，TensorRT is just enabled for GPU. + +After the compilation is completed, an executable file named `clas_system` will be generated in the `build` folder. + + +### Run the demo +* First, please modify the `tools/config.txt` and `tools/run.sh`. + +* Some key words in `tools/config.txt` is as follows. + * use_gpu: Whether to use GPU. + * gpu_id: GPU id. + * gpu_mem：GPU memory. + * cpu_math_library_num_threads：Number of thread for math library acceleration. + * use_mkldnn：Whether to use mkldnn. + * use_tensorrt: Whether to use tensorRT. + * use_fp16：Whether to use Float16 (half precision), it is just enabled when use_tensorrt is set as 1. + * cls_model_path: Model path of inference model. + * cls_params_path: Params path of inference model. + * resize_short_size：Short side length of the image after resize. + * crop_size：Image size after center crop. + + +* Then execute the following command to complete the classification of an image. + +```shell +sh tools/run.sh +``` + +* The detection results will be shown on the screen, which is as follows. + +
+ +
+ +* In the above results,`class id` represents the id corresponding to the category with the highest confidence, and `score` represents the probability that the image belongs to that category. diff --git a/exprements/PaddleClas/deploy/cpp/src/cls.cpp b/exprements/PaddleClas/deploy/cpp/src/cls.cpp new file mode 100644 index 00000000..63d8c5f7 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/src/cls.cpp @@ -0,0 +1,116 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include +#include + +namespace PaddleClas { + + void Classifier::LoadModel(const std::string &model_path, + const std::string ¶ms_path) { + paddle_infer::Config config; + config.SetModel(model_path, params_path); + + if (this->use_gpu_) { + config.EnableUseGpu(this->gpu_mem_, this->gpu_id_); + if (this->use_tensorrt_) { + config.EnableTensorRtEngine( + 1 << 20, 1, 3, + this->use_fp16_ ? paddle_infer::Config::Precision::kHalf + : paddle_infer::Config::Precision::kFloat32, + false, false); + } + } else { + config.DisableGpu(); + if (this->use_mkldnn_) { + config.EnableMKLDNN(); + // cache 10 different shapes for mkldnn to avoid memory leak + config.SetMkldnnCacheCapacity(10); + } + config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_); + } + + config.SwitchUseFeedFetchOps(false); + // true for multiple input + config.SwitchSpecifyInputNames(true); + + config.SwitchIrOptim(this->ir_optim_); + + config.EnableMemoryOptim(); + config.DisableGlogInfo(); + + this->predictor_ = CreatePredictor(config); + } + + void Classifier::Run(cv::Mat &img, std::vector &out_data, + std::vector &idx, std::vector ×) { + cv::Mat srcimg; + cv::Mat resize_img; + img.copyTo(srcimg); + + auto preprocess_start = std::chrono::system_clock::now(); + this->resize_op_.Run(img, resize_img, this->resize_short_size_); + + this->crop_op_.Run(resize_img, this->crop_size_); + + this->normalize_op_.Run(&resize_img, this->mean_, this->std_, this->scale_); + std::vector input(1 * 3 * resize_img.rows * resize_img.cols, 0.0f); + this->permute_op_.Run(&resize_img, input.data()); + + auto input_names = this->predictor_->GetInputNames(); + auto input_t = this->predictor_->GetInputHandle(input_names[0]); + input_t->Reshape({1, 3, resize_img.rows, resize_img.cols}); + auto preprocess_end = std::chrono::system_clock::now(); + + auto infer_start = std::chrono::system_clock::now(); + input_t->CopyFromCpu(input.data()); + this->predictor_->Run(); + + auto output_names = this->predictor_->GetOutputNames(); + auto output_t = this->predictor_->GetOutputHandle(output_names[0]); + std::vector output_shape = output_t->shape(); + int out_num = std::accumulate(output_shape.begin(), output_shape.end(), 1, + std::multiplies()); + + out_data.resize(out_num); + idx.resize(out_num); + output_t->CopyToCpu(out_data.data()); + auto infer_end = std::chrono::system_clock::now(); + + auto postprocess_start = std::chrono::system_clock::now(); + // int maxPosition = + // max_element(out_data.begin(), out_data.end()) - out_data.begin(); + iota(idx.begin(), idx.end(), 0); + stable_sort(idx.begin(), idx.end(), [&out_data](int i1, int i2) { + return out_data[i1] > out_data[i2]; + }); + auto postprocess_end = std::chrono::system_clock::now(); + + std::chrono::duration preprocess_diff = + preprocess_end - preprocess_start; + times[0] = double(preprocess_diff.count() * 1000); + std::chrono::duration inference_diff = infer_end - infer_start; + double inference_cost_time = double(inference_diff.count() * 1000); + times[1] = inference_cost_time; + std::chrono::duration postprocess_diff = + postprocess_end - postprocess_start; + times[2] = double(postprocess_diff.count() * 1000); + + /* std::cout << "result: " << std::endl; */ + /* std::cout << "\tclass id: " << maxPosition << std::endl; */ + /* std::cout << std::fixed << std::setprecision(10) */ + /* << "\tscore: " << double(out_data[maxPosition]) << std::endl; */ + } +} // namespace PaddleClas diff --git a/exprements/PaddleClas/deploy/cpp/src/cls_config.cpp b/exprements/PaddleClas/deploy/cpp/src/cls_config.cpp new file mode 100644 index 00000000..ec3a44f1 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/src/cls_config.cpp @@ -0,0 +1,52 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include + +namespace PaddleClas { + + void ClsConfig::PrintConfigInfo() { + std::cout << "=======Paddle Class inference config======" << std::endl; + std::cout << this->config_file << std::endl; + std::cout << "=======End of Paddle Class inference config======" << std::endl; + } + + void ClsConfig::ReadYamlConfig(const std::string &path) { + + try { + this->config_file = YAML::LoadFile(path); + } catch (YAML::BadFile &e) { + std::cout << "Something wrong in yaml file, please check yaml file" + << std::endl; + exit(1); + } + } + + void ClsConfig::ReadLabelMap() { + if (this->class_id_map_path.empty()) { + std::cout << "The Class Label file dose not input" << std::endl; + return; + } + std::ifstream in(this->class_id_map_path); + std::string line; + if (in) { + while (getline(in, line)) { + int split_flag = line.find_first_of(" "); + this->id_map[std::stoi(line.substr(0, split_flag))] = + line.substr(split_flag + 1, line.size()); + } + } + } +}; // namespace PaddleClas diff --git a/exprements/PaddleClas/deploy/cpp/src/main.cpp b/exprements/PaddleClas/deploy/cpp/src/main.cpp new file mode 100644 index 00000000..120c4a46 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/src/main.cpp @@ -0,0 +1,137 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +#include +#include +#include +#include + +using namespace std; +using namespace cv; +using namespace PaddleClas; + +DEFINE_string(config, +"", "Path of yaml file"); +DEFINE_string(c, +"", "Path of yaml file"); + +int main(int argc, char **argv) { + google::ParseCommandLineFlags(&argc, &argv, true); + std::string yaml_path = ""; + if (FLAGS_config == "" && FLAGS_c == "") { + std::cerr << "[ERROR] usage: " << std::endl + << argv[0] << " -c $yaml_path" << std::endl + << "or:" << std::endl + << argv[0] << " -config $yaml_path" << std::endl; + exit(1); + } else if (FLAGS_config != "") { + yaml_path = FLAGS_config; + } else { + yaml_path = FLAGS_c; + } + ClsConfig config(yaml_path); + config.PrintConfigInfo(); + + std::string path(config.infer_imgs); + + std::vector img_files_list; + if (cv::utils::fs::isDirectory(path)) { + std::vector filenames; + cv::glob(path, filenames); + for (auto f : filenames) { + img_files_list.push_back(f); + } + } else { + img_files_list.push_back(path); + } + + std::cout << "img_file_list length: " << img_files_list.size() << std::endl; + + Classifier classifier(config); + + std::vector cls_times = {0, 0, 0}; + std::vector cls_times_total = {0, 0, 0}; + double infer_time; + std::vector out_data; + std::vector result_index; + int warmup_iter = 5; + bool label_output_equal_flag = true; + for (int idx = 0; idx < img_files_list.size(); ++idx) { + std::string img_path = img_files_list[idx]; + cv::Mat srcimg = cv::imread(img_path, cv::IMREAD_COLOR); + if (!srcimg.data) { + std::cerr << "[ERROR] image read failed! image path: " << img_path + << "\n"; + exit(-1); + } + + cv::cvtColor(srcimg, srcimg, cv::COLOR_BGR2RGB); + classifier.Run(srcimg, out_data, result_index, cls_times); + if (label_output_equal_flag and out_data.size() != config.id_map.size()) { + std::cout << "Warning: the label size is not equal to output size!" + << std::endl; + label_output_equal_flag = false; + } + + int max_len = std::min(config.topk, int(out_data.size())); + std::cout << "Current image path: " << img_path << std::endl; + infer_time = cls_times[0] + cls_times[1] + cls_times[2]; + std::cout << "Current total inferen time cost: " << infer_time << " ms." + << std::endl; + for (int i = 0; i < max_len; ++i) { + printf("\tTop%d: class_id: %d, score: %.4f, ", i + 1, result_index[i], + out_data[result_index[i]]); + if (label_output_equal_flag) + printf("label: %s\n", config.id_map[result_index[i]].c_str()); + } + if (idx >= warmup_iter) { + for (int i = 0; i < cls_times.size(); ++i) + cls_times_total[i] += cls_times[i]; + } + } + if (img_files_list.size() > warmup_iter) { + + infer_time = cls_times_total[0] + cls_times_total[1] + cls_times_total[2]; + std::cout << "average time cost in all: " + << infer_time / (img_files_list.size() - warmup_iter) << " ms." + << std::endl; + } + + std::string presion = "fp32"; + if (config.use_fp16) + presion = "fp16"; + if (config.benchmark) { + AutoLogger autolog("Classification", config.use_gpu, config.use_tensorrt, + config.use_mkldnn, config.cpu_threads, 1, + "1, 3, 224, 224", presion, cls_times_total, + img_files_list.size()); + autolog.report(); + } + return 0; +} diff --git a/exprements/PaddleClas/deploy/cpp/src/preprocess_op.cpp b/exprements/PaddleClas/deploy/cpp/src/preprocess_op.cpp new file mode 100644 index 00000000..7e863081 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/src/preprocess_op.cpp @@ -0,0 +1,88 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include "paddle_api.h" +#include "paddle_inference_api.h" +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +#include + +namespace PaddleClas { + + void Permute::Run(const cv::Mat *im, float *data) { + int rh = im->rows; + int rw = im->cols; + int rc = im->channels(); + for (int i = 0; i < rc; ++i) { + cv::extractChannel(*im, cv::Mat(rh, rw, CV_32FC1, data + i * rh * rw), i); + } + } + + void Normalize::Run(cv::Mat *im, const std::vector &mean, + const std::vector &std, float &scale) { + if (scale) { + (*im).convertTo(*im, CV_32FC3, scale); + } + for (int h = 0; h < im->rows; h++) { + for (int w = 0; w < im->cols; w++) { + im->at(h, w)[0] = + (im->at(h, w)[0] - mean[0]) / std[0]; + im->at(h, w)[1] = + (im->at(h, w)[1] - mean[1]) / std[1]; + im->at(h, w)[2] = + (im->at(h, w)[2] - mean[2]) / std[2]; + } + } + } + + void CenterCropImg::Run(cv::Mat &img, const int crop_size) { + int resize_w = img.cols; + int resize_h = img.rows; + int w_start = int((resize_w - crop_size) / 2); + int h_start = int((resize_h - crop_size) / 2); + cv::Rect rect(w_start, h_start, crop_size, crop_size); + img = img(rect); + } + + void ResizeImg::Run(const cv::Mat &img, cv::Mat &resize_img, + int resize_short_size) { + int w = img.cols; + int h = img.rows; + + float ratio = 1.f; + if (h < w) { + ratio = float(resize_short_size) / float(h); + } else { + ratio = float(resize_short_size) / float(w); + } + + int resize_h = round(float(h) * ratio); + int resize_w = round(float(w) * ratio); + + cv::resize(img, resize_img, cv::Size(resize_w, resize_h)); + } + +} // namespace PaddleClas diff --git a/exprements/PaddleClas/deploy/cpp/src/utility.cpp b/exprements/PaddleClas/deploy/cpp/src/utility.cpp new file mode 100644 index 00000000..dbb4549b --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/src/utility.cpp @@ -0,0 +1,39 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include +#include + +#include + +namespace PaddleClas { + + std::vector Utility::ReadDict(const std::string &path) { + std::ifstream in(path); + std::string line; + std::vector m_vec; + if (in) { + while (getline(in, line)) { + m_vec.push_back(line); + } + } else { + std::cout << "no such label file: " << path << ", exit the program..." + << std::endl; + exit(1); + } + return m_vec; + } + +} // namespace PaddleClas \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/cpp/tools/build.sh b/exprements/PaddleClas/deploy/cpp/tools/build.sh new file mode 100644 index 00000000..72447eca --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp/tools/build.sh @@ -0,0 +1,21 @@ +OPENCV_DIR=/work/project/project/test/opencv-3.4.7/opencv3 +LIB_DIR=/work/project/project/test/paddle_inference/ +CUDA_LIB_DIR=/usr/local/cuda/lib64 +CUDNN_LIB_DIR=/usr/lib/x86_64-linux-gnu/ + +BUILD_DIR=build +rm -rf ${BUILD_DIR} +mkdir ${BUILD_DIR} +cd ${BUILD_DIR} +cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DWITH_GPU=OFF \ + -DWITH_STATIC_LIB=OFF \ + -DUSE_TENSORRT=OFF \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + +make -j +cd .. diff --git a/exprements/PaddleClas/deploy/cpp_shitu/CMakeLists.txt b/exprements/PaddleClas/deploy/cpp_shitu/CMakeLists.txt new file mode 100644 index 00000000..f412654d --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/CMakeLists.txt @@ -0,0 +1,249 @@ +project(pp_shitu CXX C) +cmake_minimum_required(VERSION 3.14) + +option(WITH_MKL "Compile demo with MKL/OpenBlas support, default use MKL." ON) +option(WITH_GPU "Compile demo with GPU/CPU, default use CPU." OFF) +option(WITH_STATIC_LIB "Compile demo with static/shared library, default use static." ON) +option(WITH_TENSORRT "Compile demo with TensorRT." OFF) +option(FAISS_WITH_MKL "Faiss Compile demo with MKL." OFF) + +SET(PADDLE_LIB "" CACHE PATH "Location of libraries") +SET(OPENCV_DIR "" CACHE PATH "Location of libraries") +SET(CUDA_LIB "" CACHE PATH "Location of libraries") +SET(CUDNN_LIB "" CACHE PATH "Location of libraries") +SET(TENSORRT_DIR "" CACHE PATH "Compile demo with TensorRT") +SET(FAISS_DIR "" CACHE PATH "Location of libraries") + +set(DEMO_NAME "pp_shitu") + +if (FAISS_WITH_MKL) + SET(BLAS_NAME "mklml_intel") +else() + SET(BLAS_NAME "openblas") +endif() + +include(external-cmake/yaml-cpp.cmake) +include_directories("${CMAKE_SOURCE_DIR}/") +include_directories("${CMAKE_CURRENT_BINARY_DIR}/ext/yaml-cpp/src/ext-yaml-cpp/include") +link_directories("${CMAKE_CURRENT_BINARY_DIR}/ext/yaml-cpp/lib") + +macro(safe_set_static_flag) + foreach(flag_var + CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_DEBUG CMAKE_CXX_FLAGS_RELEASE + CMAKE_CXX_FLAGS_MINSIZEREL CMAKE_CXX_FLAGS_RELWITHDEBINFO) + if(${flag_var} MATCHES "/MD") + string(REGEX REPLACE "/MD" "/MT" ${flag_var} "${${flag_var}}") + endif(${flag_var} MATCHES "/MD") + endforeach(flag_var) +endmacro() + + +if (WITH_MKL) + ADD_DEFINITIONS(-DUSE_MKL) +endif() + +if(NOT DEFINED FAISS_DIR) + message(FATAL_ERROR "please set FAISS_DIR with -DFAISS_DIR=/path/faiss") +endif() + +if(NOT DEFINED PADDLE_LIB) + message(FATAL_ERROR "please set PADDLE_LIB with -DPADDLE_LIB=/path/paddle/lib") +endif() + +if(NOT DEFINED OPENCV_DIR) + message(FATAL_ERROR "please set OPENCV_DIR with -DOPENCV_DIR=/path/opencv") +endif() + +if (WIN32) + include_directories("${FAISS_DIR}/include") + link_directories("${FAISS_DIR}/lib") + find_package(faiss REQUIRED PATHS ${FAISS_DIR}/share/faiss/ NO_DEFAULT_PATH) +else() + find_package(faiss REQUIRED PATHS ${FAISS_DIR}/share/faiss NO_DEFAULT_PATH) + include_directories("${FAISS_DIR}/include") + link_directories("${FAISS_DIR}/lib") +endif() + +if (WIN32) + include_directories("${PADDLE_LIB}/paddle/fluid/inference") + include_directories("${PADDLE_LIB}/paddle/include") + link_directories("${PADDLE_LIB}/paddle/fluid/inference") + find_package(OpenCV REQUIRED PATHS ${OPENCV_DIR}/build/ NO_DEFAULT_PATH) + +else () + find_package(OpenCV REQUIRED PATHS ${OPENCV_DIR}/share/OpenCV NO_DEFAULT_PATH) + include_directories("${PADDLE_LIB}/paddle/include") + link_directories("${PADDLE_LIB}/paddle/lib") +endif () +include_directories(${OpenCV_INCLUDE_DIRS}) + +if (WIN32) + add_definitions("/DGOOGLE_GLOG_DLL_DECL=") + set(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} /bigobj /MTd") + set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} /bigobj /MT") + set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /bigobj /MTd") + set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /bigobj /MT") + if (WITH_STATIC_LIB) + safe_set_static_flag() + add_definitions(-DSTATIC_LIB) + endif() +else() + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -O3 -std=c++11") + set(CMAKE_STATIC_LIBRARY_PREFIX "") +endif() +message("flags" ${CMAKE_CXX_FLAGS}) + + +if (WITH_GPU) + if (NOT DEFINED CUDA_LIB OR ${CUDA_LIB} STREQUAL "") + message(FATAL_ERROR "please set CUDA_LIB with -DCUDA_LIB=/path/cuda-8.0/lib64") + endif() + if (NOT WIN32) + if (NOT DEFINED CUDNN_LIB) + message(FATAL_ERROR "please set CUDNN_LIB with -DCUDNN_LIB=/path/cudnn_v7.4/cuda/lib64") + endif() + endif(NOT WIN32) +endif() + +include_directories("${PADDLE_LIB}/third_party/install/protobuf/include") +include_directories("${PADDLE_LIB}/third_party/install/glog/include") +include_directories("${PADDLE_LIB}/third_party/install/gflags/include") +include_directories("${PADDLE_LIB}/third_party/install/xxhash/include") +include_directories("${PADDLE_LIB}/third_party/install/zlib/include") +include_directories("${PADDLE_LIB}/third_party/boost") +include_directories("${PADDLE_LIB}/third_party/eigen3") + +include_directories("${CMAKE_SOURCE_DIR}/") + +if (NOT WIN32) + if (WITH_TENSORRT AND WITH_GPU) + include_directories("${TENSORRT_DIR}/include") + link_directories("${TENSORRT_DIR}/lib") + endif() +endif(NOT WIN32) + +link_directories("${PADDLE_LIB}/third_party/install/zlib/lib") + +link_directories("${PADDLE_LIB}/third_party/install/protobuf/lib") +link_directories("${PADDLE_LIB}/third_party/install/glog/lib") +link_directories("${PADDLE_LIB}/third_party/install/gflags/lib") +link_directories("${PADDLE_LIB}/third_party/install/xxhash/lib") +link_directories("${PADDLE_LIB}/paddle/lib") + + +if(WITH_MKL) + include_directories("${PADDLE_LIB}/third_party/install/mklml/include") + if (WIN32) + set(MATH_LIB ${PADDLE_LIB}/third_party/install/mklml/lib/mklml.lib + ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5md.lib) + else () + set(MATH_LIB ${PADDLE_LIB}/third_party/install/mklml/lib/libmklml_intel${CMAKE_SHARED_LIBRARY_SUFFIX} + ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5${CMAKE_SHARED_LIBRARY_SUFFIX}) + execute_process(COMMAND cp -r ${PADDLE_LIB}/third_party/install/mklml/lib/libmklml_intel${CMAKE_SHARED_LIBRARY_SUFFIX} /usr/lib) + endif () + set(MKLDNN_PATH "${PADDLE_LIB}/third_party/install/mkldnn") + if(EXISTS ${MKLDNN_PATH}) + include_directories("${MKLDNN_PATH}/include") + if (WIN32) + set(MKLDNN_LIB ${MKLDNN_PATH}/lib/mkldnn.lib) + else () + set(MKLDNN_LIB ${MKLDNN_PATH}/lib/libmkldnn.so.0) + endif () + endif() +else() + if (WIN32) + set(MATH_LIB ${PADDLE_LIB}/third_party/install/openblas/lib/openblas${CMAKE_STATIC_LIBRARY_SUFFIX}) + else () + set(MATH_LIB ${PADDLE_LIB}/third_party/install/openblas/lib/libopenblas${CMAKE_STATIC_LIBRARY_SUFFIX}) + endif () +endif() + +# Note: libpaddle_inference_api.so/a must put before libpaddle_fluid.so/a +if(WITH_STATIC_LIB) + if(WIN32) + set(DEPS + ${PADDLE_LIB}/paddle/lib/paddle_inference${CMAKE_STATIC_LIBRARY_SUFFIX}) + else() + set(DEPS + ${PADDLE_LIB}/paddle/lib/libpaddle_inference${CMAKE_STATIC_LIBRARY_SUFFIX}) + endif() +else() + if(WIN32) + set(DEPS + ${PADDLE_LIB}/paddle/lib/paddle_inference${CMAKE_SHARED_LIBRARY_SUFFIX}) + else() + set(DEPS + ${PADDLE_LIB}/paddle/lib/libpaddle_inference${CMAKE_SHARED_LIBRARY_SUFFIX}) + endif() +endif(WITH_STATIC_LIB) + +if (NOT WIN32) + set(DEPS ${DEPS} + ${MATH_LIB} ${MKLDNN_LIB} + glog gflags protobuf z xxhash yaml-cpp + ) + if(EXISTS "${PADDLE_LIB}/third_party/install/snappystream/lib") + set(DEPS ${DEPS} snappystream) + endif() + if (EXISTS "${PADDLE_LIB}/third_party/install/snappy/lib") + set(DEPS ${DEPS} snappy) + endif() +else() + set(DEPS ${DEPS} + ${MATH_LIB} ${MKLDNN_LIB} + glog gflags_static libprotobuf xxhash libyaml-cppmt) + set(DEPS ${DEPS} libcmt shlwapi) + if (EXISTS "${PADDLE_LIB}/third_party/install/snappy/lib") + set(DEPS ${DEPS} snappy) + endif() + if(EXISTS "${PADDLE_LIB}/third_party/install/snappystream/lib") + set(DEPS ${DEPS} snappystream) + endif() +endif(NOT WIN32) + + +if(WITH_GPU) + if(NOT WIN32) + if (WITH_TENSORRT) + set(DEPS ${DEPS} ${TENSORRT_DIR}/lib/libnvinfer${CMAKE_SHARED_LIBRARY_SUFFIX}) + set(DEPS ${DEPS} ${TENSORRT_DIR}/lib/libnvinfer_plugin${CMAKE_SHARED_LIBRARY_SUFFIX}) + endif() + set(DEPS ${DEPS} ${CUDA_LIB}/libcudart${CMAKE_SHARED_LIBRARY_SUFFIX}) + set(DEPS ${DEPS} ${CUDNN_LIB}/libcudnn${CMAKE_SHARED_LIBRARY_SUFFIX}) + else() + set(DEPS ${DEPS} ${CUDA_LIB}/cudart${CMAKE_STATIC_LIBRARY_SUFFIX} ) + set(DEPS ${DEPS} ${CUDA_LIB}/cublas${CMAKE_STATIC_LIBRARY_SUFFIX} ) + set(DEPS ${DEPS} ${CUDNN_LIB}/cudnn${CMAKE_STATIC_LIBRARY_SUFFIX}) + endif() +endif() + + +if (NOT WIN32) + set(EXTERNAL_LIB "-ldl -lrt -lgomp -lz -lm -lpthread") + set(DEPS ${DEPS} ${EXTERNAL_LIB}) +endif() + +set(DEPS ${DEPS} ${OpenCV_LIBS}) + +include(FetchContent) +include(external-cmake/auto-log.cmake) +include_directories(${FETCHCONTENT_BASE_DIR}/extern_autolog-src) + + +AUX_SOURCE_DIRECTORY(./src SRCS) +add_executable(${DEMO_NAME} ${SRCS}) +ADD_DEPENDENCIES(${DEMO_NAME} ext-yaml-cpp) +target_link_libraries(${DEMO_NAME} ${DEPS}) +target_link_libraries(${DEMO_NAME} ${FAISS_DIR}/lib/libfaiss${CMAKE_STATIC_LIBRARY_SUFFIX}) +target_link_libraries(${DEMO_NAME} ${BLAS_NAME}${CMAKE_SHARED_LIBRARY_SUFFIX}) + +if (WIN32 AND WITH_MKL) + add_custom_command(TARGET ${DEMO_NAME} POST_BUILD + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/mklml.dll ./mklml.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5md.dll ./libiomp5md.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mkldnn/lib/mkldnn.dll ./mkldnn.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/mklml.dll ./release/mklml.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mklml/lib/libiomp5md.dll ./release/libiomp5md.dll + COMMAND ${CMAKE_COMMAND} -E copy_if_different ${PADDLE_LIB}/third_party/install/mkldnn/lib/mkldnn.dll ./release/mkldnn.dll + ) +endif() diff --git a/exprements/PaddleClas/deploy/cpp_shitu/external-cmake/auto-log.cmake b/exprements/PaddleClas/deploy/cpp_shitu/external-cmake/auto-log.cmake new file mode 100644 index 00000000..9be9c2fb --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/external-cmake/auto-log.cmake @@ -0,0 +1,12 @@ +find_package(Git REQUIRED) +include(FetchContent) + +set(FETCHCONTENT_BASE_DIR "${CMAKE_CURRENT_BINARY_DIR}/third-party") + +FetchContent_Declare( + extern_Autolog + PREFIX autolog + GIT_REPOSITORY https://github.com/LDOUBLEV/AutoLog.git + GIT_TAG main +) +FetchContent_MakeAvailable(extern_Autolog) diff --git a/exprements/PaddleClas/deploy/cpp_shitu/external-cmake/yaml-cpp.cmake b/exprements/PaddleClas/deploy/cpp_shitu/external-cmake/yaml-cpp.cmake new file mode 100644 index 00000000..7bc7f34d --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/external-cmake/yaml-cpp.cmake @@ -0,0 +1,30 @@ + +find_package(Git REQUIRED) + +include(ExternalProject) + +message("${CMAKE_BUILD_TYPE}") + +ExternalProject_Add( + ext-yaml-cpp + URL https://bj.bcebos.com/paddlex/deploy/deps/yaml-cpp.zip + URL_MD5 9542d6de397d1fbd649ed468cb5850e6 + CMAKE_ARGS + -DYAML_CPP_BUILD_TESTS=OFF + -DYAML_CPP_BUILD_TOOLS=OFF + -DYAML_CPP_INSTALL=OFF + -DYAML_CPP_BUILD_CONTRIB=OFF + -DMSVC_SHARED_RT=OFF + -DBUILD_SHARED_LIBS=OFF + -DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE} + -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS} + -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG} + -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE} + -DCMAKE_LIBRARY_OUTPUT_DIRECTORY=${CMAKE_BINARY_DIR}/ext/yaml-cpp/lib + -DCMAKE_ARCHIVE_OUTPUT_DIRECTORY=${CMAKE_BINARY_DIR}/ext/yaml-cpp/lib + PREFIX "${CMAKE_BINARY_DIR}/ext/yaml-cpp" + # Disable install step + INSTALL_COMMAND "" + LOG_DOWNLOAD ON + LOG_BUILD 1 +) diff --git a/exprements/PaddleClas/deploy/cpp_shitu/include/feature_extracter.h b/exprements/PaddleClas/deploy/cpp_shitu/include/feature_extracter.h new file mode 100644 index 00000000..e83caee7 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/include/feature_extracter.h @@ -0,0 +1,110 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include "paddle_inference_api.h" +#include +#include +#include +#include +#include + +#include +#include +#include + +#include +#include + +using namespace paddle_infer; + +namespace Feature { + + class FeatureExtracter { + public: + explicit FeatureExtracter(const YAML::Node &config_file) { + this->use_gpu_ = config_file["Global"]["use_gpu"].as(); + if (config_file["Global"]["gpu_id"].IsDefined()) + this->gpu_id_ = config_file["Global"]["gpu_id"].as(); + else + this->gpu_id_ = 0; + this->gpu_mem_ = config_file["Global"]["gpu_mem"].as(); + this->cpu_math_library_num_threads_ = + config_file["Global"]["cpu_num_threads"].as(); + this->use_mkldnn_ = config_file["Global"]["enable_mkldnn"].as(); + this->use_tensorrt_ = config_file["Global"]["use_tensorrt"].as(); + this->use_fp16_ = config_file["Global"]["use_fp16"].as(); + + this->cls_model_path_ = + config_file["Global"]["rec_inference_model_dir"].as() + + OS_PATH_SEP + "inference.pdmodel"; + this->cls_params_path_ = + config_file["Global"]["rec_inference_model_dir"].as() + + OS_PATH_SEP + "inference.pdiparams"; + this->resize_size_ = + config_file["RecPreProcess"]["transform_ops"][0]["ResizeImage"]["size"] + .as(); + this->scale_ = config_file["RecPreProcess"]["transform_ops"][1]["NormalizeImage"]["scale"].as(); + this->mean_ = config_file["RecPreProcess"]["transform_ops"][1] + ["NormalizeImage"]["mean"] + .as < std::vector < float >> (); + this->std_ = config_file["RecPreProcess"]["transform_ops"][1] + ["NormalizeImage"]["std"] + .as < std::vector < float >> (); + if (config_file["Global"]["rec_feature_normlize"].IsDefined()) + this->feature_norm = + config_file["Global"]["rec_feature_normlize"].as(); + + LoadModel(cls_model_path_, cls_params_path_); + } + + // Load Paddle inference model + void LoadModel(const std::string &model_path, const std::string ¶ms_path); + + // Run predictor + void Run(cv::Mat &img, std::vector &out_data, + std::vector ×); + + void FeatureNorm(std::vector &feature); + + std::shared_ptr predictor_; + + private: + bool use_gpu_ = false; + int gpu_id_ = 0; + int gpu_mem_ = 4000; + int cpu_math_library_num_threads_ = 4; + bool use_mkldnn_ = false; + bool use_tensorrt_ = false; + bool feature_norm = true; + bool use_fp16_ = false; + std::vector mean_ = {0.485f, 0.456f, 0.406f}; + std::vector std_ = {0.229f, 0.224f, 0.225f}; + float scale_ = 0.00392157; + int resize_size_ = 224; + int resize_short_ = 224; + std::string cls_model_path_; + std::string cls_params_path_; + + // pre-process + ResizeImg resize_op_; + Normalize normalize_op_; + Permute permute_op_; + }; + +} // namespace Feature diff --git a/exprements/PaddleClas/deploy/cpp_shitu/include/nms.h b/exprements/PaddleClas/deploy/cpp_shitu/include/nms.h new file mode 100644 index 00000000..02956246 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/include/nms.h @@ -0,0 +1,85 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// This code is adpated from opencv(https://github.com/opencv/opencv) + +#include +#include + +template +static inline bool SortScorePairDescend(const std::pair &pair1, + const std::pair &pair2) { + return pair1.first > pair2.first; +} + +float RectOverlap(const Detection::ObjectResult &a, + const Detection::ObjectResult &b) { + float Aa = (a.rect[2] - a.rect[0] + 1) * (a.rect[3] - a.rect[1] + 1); + float Ab = (b.rect[2] - b.rect[0] + 1) * (b.rect[3] - b.rect[1] + 1); + + int iou_w = max(min(a.rect[2], b.rect[2]) - max(a.rect[0], b.rect[0]) + 1, 0); + int iou_h = max(min(a.rect[3], b.rect[3]) - max(a.rect[1], b.rect[1]) + 1, 0); + float Aab = iou_w * iou_h; + return Aab / (Aa + Ab - Aab); +} + +// Get max scores with corresponding indices. +// scores: a set of scores. +// threshold: only consider scores higher than the threshold. +// top_k: if -1, keep all; otherwise, keep at most top_k. +// score_index_vec: store the sorted (score, index) pair. +inline void +GetMaxScoreIndex(const std::vector &det_result, + const float threshold, + std::vector > &score_index_vec) { + // Generate index score pairs. + for (size_t i = 0; i < det_result.size(); ++i) { + if (det_result[i].confidence > threshold) { + score_index_vec.push_back(std::make_pair(det_result[i].confidence, i)); + } + } + + // Sort the score pair according to the scores in descending order + std::stable_sort(score_index_vec.begin(), score_index_vec.end(), + SortScorePairDescend); + + // // Keep top_k scores if needed. + // if (top_k > 0 && top_k < (int)score_index_vec.size()) + // { + // score_index_vec.resize(top_k); + // } +} + +void NMSBoxes(const std::vector det_result, + const float score_threshold, const float nms_threshold, + std::vector &indices) { + int a = 1; + // Get top_k scores (with corresponding indices). + std::vector > score_index_vec; + GetMaxScoreIndex(det_result, score_threshold, score_index_vec); + + // Do nms + indices.clear(); + for (size_t i = 0; i < score_index_vec.size(); ++i) { + const int idx = score_index_vec[i].second; + bool keep = true; + for (int k = 0; k < (int) indices.size() && keep; ++k) { + const int kept_idx = indices[k]; + float overlap = RectOverlap(det_result[idx], det_result[kept_idx]); + keep = overlap <= nms_threshold; + } + if (keep) + indices.push_back(idx); + } +} diff --git a/exprements/PaddleClas/deploy/cpp_shitu/include/object_detector.h b/exprements/PaddleClas/deploy/cpp_shitu/include/object_detector.h new file mode 100644 index 00000000..5bfc5625 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/include/object_detector.h @@ -0,0 +1,138 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include +#include +#include +#include +#include + +#include +#include +#include + +#include "paddle_inference_api.h" // NOLINT + +#include "include/preprocess_op_det.h" +#include "include/yaml_config.h" + +using namespace paddle_infer; + +namespace Detection { +// Object Detection Result + struct ObjectResult { + // Rectangle coordinates of detected object: left, right, top, down + std::vector rect; + // Class id of detected object + int class_id; + // Confidence of detected object + float confidence; + }; + +// Generate visualization colormap for each class + std::vector GenerateColorMap(int num_class); + +// Visualiztion Detection Result + cv::Mat VisualizeResult(const cv::Mat &img, + const std::vector &results, + const std::vector &lables, + const std::vector &colormap, const bool is_rbox); + + class ObjectDetector { + public: + explicit ObjectDetector(const YAML::Node &config_file) { + this->use_gpu_ = config_file["Global"]["use_gpu"].as(); + if (config_file["Global"]["gpu_id"].IsDefined()) + this->gpu_id_ = config_file["Global"]["gpu_id"].as(); + this->gpu_mem_ = config_file["Global"]["gpu_mem"].as(); + this->cpu_math_library_num_threads_ = + config_file["Global"]["cpu_num_threads"].as(); + this->use_mkldnn_ = config_file["Global"]["enable_mkldnn"].as(); + this->use_tensorrt_ = config_file["Global"]["use_tensorrt"].as(); + this->use_fp16_ = config_file["Global"]["use_fp16"].as(); + this->model_dir_ = + config_file["Global"]["det_inference_model_dir"].as(); + this->threshold_ = config_file["Global"]["threshold"].as(); + this->max_det_results_ = config_file["Global"]["max_det_results"].as(); + this->image_shape_ = + config_file["Global"]["image_shape"].as < std::vector < int >> (); + this->label_list_ = + config_file["Global"]["labe_list"].as < std::vector < std::string >> (); + this->ir_optim_ = config_file["Global"]["ir_optim"].as(); + this->batch_size_ = config_file["Global"]["batch_size"].as(); + + preprocessor_.Init(config_file["DetPreProcess"]["transform_ops"]); + LoadModel(model_dir_, batch_size_, run_mode); + } + + // Load Paddle inference model + void LoadModel(const std::string &model_dir, const int batch_size = 1, + const std::string &run_mode = "fluid"); + + // Run predictor + void Predict(const std::vector imgs, const int warmup = 0, + const int repeats = 1, + std::vector *result = nullptr, + std::vector *bbox_num = nullptr, + std::vector *times = nullptr); + + const std::vector &GetLabelList() const { + return this->label_list_; + } + + const float &GetThreshold() const { return this->threshold_; } + + private: + bool use_gpu_ = true; + int gpu_id_ = 0; + int gpu_mem_ = 800; + int cpu_math_library_num_threads_ = 6; + std::string run_mode = "fluid"; + bool use_mkldnn_ = false; + bool use_tensorrt_ = false; + bool batch_size_ = 1; + bool use_fp16_ = false; + std::string model_dir_; + float threshold_ = 0.5; + float max_det_results_ = 5; + std::vector image_shape_ = {3, 640, 640}; + std::vector label_list_; + bool ir_optim_ = true; + bool det_permute_ = true; + bool det_postprocess_ = true; + int min_subgraph_size_ = 30; + bool use_dynamic_shape_ = false; + int trt_min_shape_ = 1; + int trt_max_shape_ = 1280; + int trt_opt_shape_ = 640; + bool trt_calib_mode_ = false; + + // Preprocess image and copy data to input buffer + void Preprocess(const cv::Mat &image_mat); + + // Postprocess result + void Postprocess(const std::vector mats, + std::vector *result, std::vector bbox_num, + bool is_rbox); + + std::shared_ptr predictor_; + Preprocessor preprocessor_; + ImageBlob inputs_; + std::vector output_data_; + std::vector out_bbox_num_data_; + }; + +} // namespace Detection diff --git a/exprements/PaddleClas/deploy/cpp_shitu/include/preprocess_op.h b/exprements/PaddleClas/deploy/cpp_shitu/include/preprocess_op.h new file mode 100644 index 00000000..db4c6b2c --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/include/preprocess_op.h @@ -0,0 +1,57 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include +#include +#include +#include +#include + +#include +#include +#include + +using namespace std; + +namespace Feature { + + class Normalize { + public: + virtual void Run(cv::Mat *im, const std::vector &mean, + const std::vector &std, float scale); + }; + +// RGB -> CHW + class Permute { + public: + virtual void Run(const cv::Mat *im, float *data); + }; + + class CenterCropImg { + public: + virtual void Run(cv::Mat &im, const int crop_size = 224); + }; + + class ResizeImg { + public: + virtual void Run(const cv::Mat &img, cv::Mat &resize_img, int max_size_len, + int size = 0); + }; + +} // namespace Feature diff --git a/exprements/PaddleClas/deploy/cpp_shitu/include/preprocess_op_det.h b/exprements/PaddleClas/deploy/cpp_shitu/include/preprocess_op_det.h new file mode 100644 index 00000000..cd663409 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/include/preprocess_op_det.h @@ -0,0 +1,158 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include +#include + +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +namespace Detection { + +// Object for storing all preprocessed data + class ImageBlob { + public: + // image width and height + std::vector im_shape_; + // Buffer for image data after preprocessing + std::vector im_data_; + // in net data shape(after pad) + std::vector in_net_shape_; + // Evaluation image width and height + // std::vector eval_im_size_f_; + // Scale factor for image size to origin image size + std::vector scale_factor_; + }; + +// Abstraction of preprocessing opration class + class PreprocessOp { + public: + virtual void Init(const YAML::Node &item) = 0; + + virtual void Run(cv::Mat *im, ImageBlob *data) = 0; + }; + + class InitInfo : public PreprocessOp { + public: + virtual void Init(const YAML::Node &item) {} + + virtual void Run(cv::Mat *im, ImageBlob *data); + }; + + class NormalizeImage : public PreprocessOp { + public: + virtual void Init(const YAML::Node &item) { + mean_ = item["mean"].as < std::vector < float >> (); + scale_ = item["std"].as < std::vector < float >> (); + is_scale_ = item["is_scale"].as(); + } + + virtual void Run(cv::Mat *im, ImageBlob *data); + + private: + // CHW or HWC + std::vector mean_; + std::vector scale_; + bool is_scale_; + }; + + class Permute : public PreprocessOp { + public: + virtual void Init(const YAML::Node &item) {} + + virtual void Run(cv::Mat *im, ImageBlob *data); + }; + + class Resize : public PreprocessOp { + public: + virtual void Init(const YAML::Node &item) { + interp_ = item["interp"].as(); + // max_size_ = item["target_size"].as(); + keep_ratio_ = item["keep_ratio"].as(); + target_size_ = item["target_size"].as < std::vector < int >> (); + } + + // Compute best resize scale for x-dimension, y-dimension + std::pair GenerateScale(const cv::Mat &im); + + virtual void Run(cv::Mat *im, ImageBlob *data); + + private: + int interp_ = 2; + bool keep_ratio_; + std::vector target_size_; + std::vector in_net_shape_; + }; + +// Models with FPN need input shape % stride == 0 + class PadStride : public PreprocessOp { + public: + virtual void Init(const YAML::Node &item) { + stride_ = item["stride"].as(); + } + + virtual void Run(cv::Mat *im, ImageBlob *data); + + private: + int stride_; + }; + + class Preprocessor { + public: + void Init(const YAML::Node &config_node) { + // initialize image info at first + ops_["InitInfo"] = std::make_shared(); + for (int i = 0; i < config_node.size(); ++i) { + if (config_node[i]["DetResize"].IsDefined()) { + ops_["Resize"] = std::make_shared(); + ops_["Resize"]->Init(config_node[i]["DetResize"]); + } + + if (config_node[i]["DetNormalizeImage"].IsDefined()) { + ops_["NormalizeImage"] = std::make_shared(); + ops_["NormalizeImage"]->Init(config_node[i]["DetNormalizeImage"]); + } + + if (config_node[i]["DetPermute"].IsDefined()) { + ops_["Permute"] = std::make_shared(); + ops_["Permute"]->Init(config_node[i]["DetPermute"]); + } + + if (config_node[i]["DetPadStrid"].IsDefined()) { + ops_["PadStride"] = std::make_shared(); + ops_["PadStride"]->Init(config_node[i]["DetPadStrid"]); + } + } + } + + void Run(cv::Mat *im, ImageBlob *data); + + public: + static const std::vector RUN_ORDER; + + private: + std::unordered_map > ops_; + }; + +} // namespace Detection diff --git a/exprements/PaddleClas/deploy/cpp_shitu/include/vector_search.h b/exprements/PaddleClas/deploy/cpp_shitu/include/vector_search.h new file mode 100644 index 00000000..4b354f78 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/include/vector_search.h @@ -0,0 +1,70 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once +#ifdef WIN32 +#define OS_PATH_SEP "\\" +#else +#define OS_PATH_SEP "/" +#endif + +#include "yaml-cpp/yaml.h" +#include +#include +#include +#include + +struct SearchResult { + std::vector I; + std::vector D; + int return_k; +}; + +class VectorSearch { +public: + explicit VectorSearch(const YAML::Node &config_file) { + // IndexProcess + this->index_dir = + config_file["IndexProcess"]["index_dir"].as(); + this->return_k = config_file["IndexProcess"]["return_k"].as(); + this->score_thres = config_file["IndexProcess"]["score_thres"].as(); + this->max_query_number = + config_file["Global"]["max_det_results"].as() + 1; + LoadIdMap(); + LoadIndexFile(); + this->I.resize(this->return_k * this->max_query_number); + this->D.resize(this->return_k * this->max_query_number); + }; + + void LoadIdMap(); + + void LoadIndexFile(); + + const SearchResult &Search(float *feature, int query_number); + + const std::string &GetLabel(faiss::Index::idx_t ind); + + const float &GetThreshold() { return this->score_thres; } + +private: + std::string index_dir; + int return_k = 5; + float score_thres = 0.5; + std::map id_map; + faiss::Index *index; + int max_query_number = 6; + std::vector D; + std::vector I; + SearchResult sr; +}; diff --git a/exprements/PaddleClas/deploy/cpp_shitu/include/yaml_config.h b/exprements/PaddleClas/deploy/cpp_shitu/include/yaml_config.h new file mode 100644 index 00000000..78243e2f --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/include/yaml_config.h @@ -0,0 +1,58 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#ifdef WIN32 +#define OS_PATH_SEP "\\" +#else +#define OS_PATH_SEP "/" +#endif + +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include "yaml-cpp/yaml.h" + + +class YamlConfig { +public: + explicit YamlConfig(const std::string &path) { + config_file = ReadYamlConfig(path); + } + + static std::vector ReadDict(const std::string &path); + + static std::map ReadIndexId(const std::string &path); + + static YAML::Node ReadYamlConfig(const std::string &path); + + void PrintConfigInfo(); + + YAML::Node config_file; +}; diff --git a/exprements/PaddleClas/deploy/cpp_shitu/readme.md b/exprements/PaddleClas/deploy/cpp_shitu/readme.md new file mode 100644 index 00000000..97315ec3 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/readme.md @@ -0,0 +1,340 @@ +# 服务器端C++预测 + +本教程将介绍在服务器端部署PP-ShiTU的详细步骤。 + +## 目录 + +- [1.准备环境](#1) + - [1.1 升级cmake](#1.1) + - [1.2 编译opencv库](#1.2) + - [1.3 下载或者编译Paddle预测库](#1.3) + - [1.3.1 预测库源码编译](#1.3.1) + - [1.3.2 直接下载安装](#1.3.2) + - [1.4 安装faiss库](#1.4) +- [2.代码编译](#2) +- [3.运行demo](#3) +- [4.使用自己模型](#4) + + + +## 1. 准备环境 + +### 运行准备 +- Linux环境，推荐使用ubuntu docker。 + + + +### 1.1 升级cmake + +由于依赖库编译需要较高版本的cmake，因此，第一步首先将cmake升级。 + +- 下载最新版本cmake + +```shell +# 当前版本最新为3.22.0，根据实际情况自行下载，建议最新版本 +wget https://github.com/Kitware/CMake/releases/download/v3.22.0/cmake-3.22.0.tar.gz +tar xf cmake-3.22.0.tar.gz +``` + +最终可以在当前目录下看到`cmake-3.22.0/`的文件夹。 + +- 编译cmake，首先设置came源码路径(`root_path`)以及安装路径(`install_path`)，`root_path`为下载的came源码路径，`install_path`为came的安装路径。在本例中，源码路径即为当前目录下的`cmake-3.22.0/`。 + +```shell +cd ./cmake-3.22.0 +export root_path=$PWD +export install_path=${root_path}/cmake +``` + +- 然后在cmake源码路径下，按照下面的方式进行编译 + +```shell +./bootstrap --prefix=${install_path} +make -j +make install +``` + +- 设置环境变量 + +```shell +export PATH=${install_path}/bin:$PATH +#检查是否正常使用 +cmake --version +``` + +此时，cmake就可以使用了 + + + +### 1.2 编译opencv库 + +* 首先需要从opencv官网上下载在Linux环境下源码编译的包，以3.4.7版本为例，下载及解压缩命令如下： + +``` +wget https://github.com/opencv/opencv/archive/3.4.7.tar.gz +tar -xvf 3.4.7.tar.gz +``` + +最终可以在当前目录下看到`opencv-3.4.7/`的文件夹。 + +* 编译opencv，首先设置opencv源码路径(`root_path`)以及安装路径(`install_path`)，`root_path`为下载的opencv源码路径，`install_path`为opencv的安装路径。在本例中，源码路径即为当前目录下的`opencv-3.4.7/`。 + +```shell +cd ./opencv-3.4.7 +export root_path=$PWD +export install_path=${root_path}/opencv3 +``` + +* 然后在opencv源码路径下，按照下面的方式进行编译。 + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DCMAKE_INSTALL_PREFIX=${install_path} \ + -DCMAKE_BUILD_TYPE=Release \ + -DBUILD_SHARED_LIBS=OFF \ + -DWITH_IPP=OFF \ + -DBUILD_IPP_IW=OFF \ + -DWITH_LAPACK=OFF \ + -DWITH_EIGEN=OFF \ + -DCMAKE_INSTALL_LIBDIR=lib64 \ + -DWITH_ZLIB=ON \ + -DBUILD_ZLIB=ON \ + -DWITH_JPEG=ON \ + -DBUILD_JPEG=ON \ + -DWITH_PNG=ON \ + -DBUILD_PNG=ON \ + -DWITH_TIFF=ON \ + -DBUILD_TIFF=ON + +make -j +make install +``` + +* `make install`完成之后，会在该文件夹下生成opencv头文件和库文件，用于后面的PaddleClas代码编译。 + +以opencv3.4.7版本为例，最终在安装路径下的文件结构如下所示。**注意**：不同的opencv版本，下述的文件结构可能不同。 + +``` +opencv3/ +|-- bin +|-- include +|-- lib64 +|-- share +``` + + + +### 1.3 下载或者编译Paddle预测库 + +* 有2种方式获取Paddle预测库，下面进行详细介绍。 + + + +#### 1.3.1 预测库源码编译 + +* 如果希望获取最新预测库特性，可以从Paddle github上克隆最新代码，源码编译预测库。 +* 可以参考[Paddle预测库官网](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_and_install_lib_cn.html#id16)的说明，从github上获取Paddle代码，然后进行编译，生成最新的预测库。使用git获取代码方法如下。 + +```shell +git clone https://github.com/PaddlePaddle/Paddle.git +``` + +* 进入Paddle目录后，使用如下方法编译。 + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DWITH_CONTRIB=OFF \ + -DWITH_MKL=ON \ + -DWITH_MKLDNN=ON \ + -DWITH_TESTING=OFF \ + -DCMAKE_BUILD_TYPE=Release \ + -DWITH_INFERENCE_API_TEST=OFF \ + -DON_INFER=ON \ + -DWITH_PYTHON=ON +make -j +make inference_lib_dist +``` + +更多编译参数选项可以参考[Paddle C++预测库官网](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_and_install_lib_cn.html#id16)。 + + +* 编译完成之后，可以在`build/paddle_inference_install_dir/`文件下看到生成了以下文件及文件夹。 + +``` +build/paddle_inference_install_dir/ +|-- CMakeCache.txt +|-- paddle +|-- third_party +|-- version.txt +``` + +其中`paddle`就是之后进行C++预测时所需的Paddle库，`version.txt`中包含当前预测库的版本信息。 + + + +#### 1.3.2 直接下载安装 + +* [Paddle预测库官网](https://paddle-inference.readthedocs.io/en/latest/user_guides/download_lib.html)上提供了不同cuda版本的Linux预测库，可以在官网查看并选择合适的预测库版本，注意必须选择`develop`版本。 + + 以`https://paddle-inference-lib.bj.bcebos.com/2.1.1-gpu-cuda10.2-cudnn8.1-mkl-gcc8.2/paddle_inference.tgz`的`develop`版本为例，使用下述命令下载并解压： + + +```shell +wget https://paddle-inference-lib.bj.bcebos.com/2.1.1-gpu-cuda10.2-cudnn8.1-mkl-gcc8.2/paddle_inference.tgz + +tar -xvf paddle_inference.tgz +``` + + +最终会在当前的文件夹中生成`paddle_inference/`的子文件夹。 + + + +### 1.4 安装faiss库 + +```shell + # 下载 faiss + git clone https://github.com/facebookresearch/faiss.git + cd faiss + export faiss_install_path=$PWD/faiss_install + cmake -B build . -DFAISS_ENABLE_PYTHON=OFF -DCMAKE_INSTALL_PREFIX=${faiss_install_path} + make -C build -j faiss + make -C build install +``` + +在安装`faiss`前，请安装`openblas`，`ubuntu`系统中安装命令如下： + +```shell +apt-get install libopenblas-dev +``` + +注意本教程以安装faiss cpu版本为例，安装时请参考[faiss](https://github.com/facebookresearch/faiss)官网文档，根据需求自行安装。 + + + +## 2. 代码编译 + +编译命令如下，其中Paddle C++预测库、opencv等其他依赖库的地址需要换成自己机器上的实际地址。同时，编译过程中需要下载编译`yaml-cpp`等C++库，请保持联网环境。 + + +```shell +sh tools/build.sh +``` + +具体地，`tools/build.sh`中内容如下，请根据具体路径修改。 + +```shell +OPENCV_DIR=${opencv_install_dir} +LIB_DIR=${paddle_inference_dir} +CUDA_LIB_DIR=/usr/local/cuda/lib64 +CUDNN_LIB_DIR=/usr/lib/x86_64-linux-gnu/ +FAISS_DIR=${faiss_install_dir} +FAISS_WITH_MKL=OFF + +BUILD_DIR=build +rm -rf ${BUILD_DIR} +mkdir ${BUILD_DIR} +cd ${BUILD_DIR} +cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DWITH_GPU=OFF \ + -DWITH_STATIC_LIB=OFF \ + -DUSE_TENSORRT=OFF \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + -DFAISS_DIR=${FAISS_DIR} \ + -DFAISS_WITH_MKL=${FAISS_WITH_MKL} + +make -j +cd .. +``` + +上述命令中， + +* `OPENCV_DIR`为opencv编译安装的地址（本例中为`opencv-3.4.7/opencv3`文件夹的路径）； +* `LIB_DIR`为下载的Paddle预测库（`paddle_inference`文件夹），或编译生成的Paddle预测库（`build/paddle_inference_install_dir`文件夹）的路径； +* `CUDA_LIB_DIR`为cuda库文件地址，在docker中为`/usr/local/cuda/lib64`； +* `CUDNN_LIB_DIR`为cudnn库文件地址，在docker中为`/usr/lib/x86_64-linux-gnu/`。 +* `TENSORRT_DIR`是tensorrt库文件地址，在dokcer中为`/usr/local/TensorRT6-cuda10.0-cudnn7/`，TensorRT需要结合GPU使用。 +* `FAISS_DIR`是faiss的安装地址 +* `FAISS_WITH_MKL`是指在编译faiss的过程中，是否使用了mkldnn，本文档中编译faiss，没有使用，而使用了openblas，故设置为`OFF`，若使用了mkldnn，则为`ON`. + + +在执行上述命令，编译完成之后，会在当前路径下生成`build`文件夹，其中生成一个名为`pp_shitu`的可执行文件。 + + + +## 3. 运行demo + +- 请参考[识别快速开始文档](../../docs/zh_CN/quick_start/quick_start_recognition.md)，下载好相应的 轻量级通用主体检测模型、轻量级通用识别模型及瓶装饮料测试数据并解压。 + + ```shell + mkdir models + cd models + wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer.tar + tar -xf picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer.tar + wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/general_PPLCNet_x2_5_lite_v1.0_infer.tar + tar -xf general_PPLCNet_x2_5_lite_v1.0_infer.tar + cd .. + + mkdir data + cd data + wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/data/drink_dataset_v1.0.tar + tar -xf drink_dataset_v1.0.tar + cd .. + ``` + +- 将相应的yaml文件拷到当前文件夹下 + + ```shell + cp ../configs/inference_drink.yaml . + ``` + +- 将`inference_drink.yaml`中的相对路径，改成基于本目录的路径或者绝对路径。涉及到的参数有 + + - Global.infer_imgs ：此参数可以是具体的图像地址，也可以是图像集所在的目录 + - Global.det_inference_model_dir ： 检测模型存储目录 + - Global.rec_inference_model_dir ： 识别模型存储目录 + - IndexProcess.index_dir ： 检索库的存储目录，在示例中，检索库在下载的demo数据中。 + +- 字典转换 + + 由于python的检索库的字典，使用`pickle`进行的序列化存储，导致C++不方便读取，因此进行转换 + + ```shell + python tools/transform_id_map.py -c inference_drink.yaml + ``` + + 转换成功后，在`IndexProcess.index_dir`目录下生成`id_map.txt`，方便c++ 读取。 + +- 执行程序 + + ```shell + ./build/pp_shitu -c inference_drink.yaml + # or + ./build/pp_shitu -config inference_drink.yaml + ``` + + 若对图像集进行检索，则可能得到，如下结果。注意，此结果只做展示，具体以实际运行结果为准。 + + 同时，需注意的是，由于opencv 版本问题，会导致图像在预处理的过程中，resize产生细微差别，导致python 和c++结果，轻微不同，如bbox相差几个像素，检索结果小数点后3位diff等。但不会改变最终检索label。 + + ![](../../docs/images/quick_start/shitu_c++_result.png) + + + +## 4. 使用自己模型 + +使用自己训练的模型，可以参考[模型导出](../../docs/zh_CN/inference_deployment/export_model.md)，导出`inference model`，用于模型预测。 + +同时注意修改`yaml`文件中具体参数。 diff --git a/exprements/PaddleClas/deploy/cpp_shitu/readme_en.md b/exprements/PaddleClas/deploy/cpp_shitu/readme_en.md new file mode 100644 index 00000000..76867c73 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/readme_en.md @@ -0,0 +1,336 @@ +# C++ Inference on Server + +This tutorial details the steps to deploy PP-ShiTU on the server side. + +## Catalogue + +- [1.Prepare the Environment](#1) + - [1.1 Update cmake](#1.1) + - [1.2 Compile opencv Library](#1.2) + - [1.3 Download or Compile Paddle Inference Library](#1.3) + - [1.3.1 Compile the Source of Inference Library](#1.3.1) + - [1.3.2 Direct Download and Installation](#1.3.2) + - [1.4 Install faiss Library](#1.4) +- [2.Code Compilation](#2) +- [3.Run the demo](#3) +- [4.Use Your Own Model](#4) + + + +## 1. Prepare the Environment + +### Environment Preparation + +- Linux environment, ubuntu docker recommended. + + + +### 1.1 Update cmake + +The first step is to upgrade `cmake` considering the requirements of the dependency library compilation. + +- Download the latest version of cmake + +``` +# The latest version is 3.22.0, please download the appropriate one, the latest recommended. +wget https://github.com/Kitware/CMake/releases/download/v3.22.0/cmake-3.22.0.tar.gz +tar xf cmake-3.22.0.tar.gz +``` + +You can find `cmake-3.22.0/` folder in the current directory. + +- To compile cmake, first set the source path of `cmake` (`root_path`) and installation path (`install_path`). In this example, the source path is `cmake-3.22.0/` in the current directory. + +```shell +cd ./cmake-3.22.0 +export root_path=$PWD +export install_path=${root_path}/cmake +``` + +- Then compile under the source path as follows: + +```shell +./bootstrap --prefix=${install_path} +make -j +make install +``` + +- Set environment variables + +```shell +export PATH=${install_path}/bin:$PATH +#Check its well functioning +cmake --version +``` + +cmake is now ready for use. + + + +### 1.2 Compile opencv Library + +- First, download the package for source compilation in Linux environment from the official website of opencv. Taking version 3.4.7 as an example, follow the command below to download and unzip it: + +```shell +wget https://github.com/opencv/opencv/archive/3.4.7.tar.gz +tar -xvf 3.4.7.tar.gz +``` + +You can find`opencv-3.4.7/`folder in the current directory. + +- To compile opencv, first set the source path of opencv(`root_path`) and installation path (`install_path`). In this example, the source path is`opencv-3.4.7/`in the current directory. + +``` +cd ./opencv-3.4.7 +export root_path=$PWD +export install_path=${root_path}/opencv3 +``` + +- Then compile under the source path as follows: + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DCMAKE_INSTALL_PREFIX=${install_path} \ + -DCMAKE_BUILD_TYPE=Release \ + -DBUILD_SHARED_LIBS=OFF \ + -DWITH_IPP=OFF \ + -DBUILD_IPP_IW=OFF \ + -DWITH_LAPACK=OFF \ + -DWITH_EIGEN=OFF \ + -DCMAKE_INSTALL_LIBDIR=lib64 \ + -DWITH_ZLIB=ON \ + -DBUILD_ZLIB=ON \ + -DWITH_JPEG=ON \ + -DBUILD_JPEG=ON \ + -DWITH_PNG=ON \ + -DBUILD_PNG=ON \ + -DWITH_TIFF=ON \ + -DBUILD_TIFF=ON + +make -j +make install +``` + +- After `make install` is done, opencv header and library files will be generated in this folder for later compilation of PaddleClas code. + +For opencv version 3.4.7, the final file structure under the installation path is shown below. **Note**: The following file structure may vary for different opencv versions. + +``` +opencv3/ +|-- bin +|-- include +|-- lib64 +|-- share +``` + + + +### 1.3 Download or Compile Paddle Inference Library + +- Here we detail 2 ways to obtain Paddle inference library. + + + +#### 1.3.1 Compile the Source of Inference Library + +- To obtain the latest features of the inference library, you can clone the latest code from Paddle github and compile the source code of the library. +- Please refer to the website of [Paddle Inference Library](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_ and_install_lib_cn.html#id16) to get Paddle code from github and then compile it to generate the latest inference library. The method to obtain the code using git is as follows. + +```shell +git clone https://github.com/PaddlePaddle/Paddle.git +``` + +- Adopt the following method to compile after entering Paddle directory. + +```shell +rm -rf build +mkdir build +cd build + +cmake .. \ + -DWITH_CONTRIB=OFF \ + -DWITH_MKL=ON \ + -DWITH_MKLDNN=ON \ + -DWITH_TESTING=OFF \ + -DCMAKE_BUILD_TYPE=Release \ + -DWITH_INFERENCE_API_TEST=OFF \ + -DON_INFER=ON \ + -DWITH_PYTHON=ON +make -j +make inference_lib_dist +``` + +See the official website of [Paddle C++ Inference Library](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/05_inference_deployment/inference/build_and_install_lib_cn.html#id16) for more compilation parameters. + +- The following files and folders can be found generated under `build/paddle_inference_install_dir/` after compilation. + +``` +build/paddle_inference_install_dir/ +|-- CMakeCache.txt +|-- paddle +|-- third_party +|-- version.txt +``` + + `paddle` is the Paddle library needed for later C++ inference, and `version.txt` contains the version information of the current inference library. + + + +#### 1.3.2 Direct Download and Installation + +- The Linux inference library of different cuda versions are available on the official website of [Paddle Inference Library ](https://paddle-inference.readthedocs.io/en/latest/user_guides/download_lib.html), where you can choose the appropriate version. Note that you must select the `develop` version. + + For the `develop` version of `https://paddle-inference-lib.bj.bcebos.com/2.1.1-gpu-cuda10.2-cudnn8.1-mkl-gcc8.2/paddle_inference.tgz`, use the following command to download and unzip it: + +```shell +wget https://paddle-inference-lib.bj.bcebos.com/2.1.1-gpu-cuda10.2-cudnn8.1-mkl-gcc8.2/paddle_inference.tgz + +tar -xvf paddle_inference.tgz +``` + +The subfolder `paddle_inference/` will finally be created in the current folder. + + + +### 1.4 Install faiss Library + +```shell + # Download faiss + git clone https://github.com/facebookresearch/faiss.git + cd faiss + export faiss_install_path=$PWD/faiss_install + cmake -B build . -DFAISS_ENABLE_PYTHON=OFF -DCMAKE_INSTALL_PREFIX=${faiss_install_path} + make -C build -j faiss + make -C build install +``` + +Please install `openblas` before `faiss`, the installation command in `ubuntu` system is as follows: + +``` +apt-get install libopenblas-dev +``` + +Note that this tutorial installs the cpu version of faiss as an example, please install it as your need by referring to the official documents of [faiss](https://github.com/facebookresearch/faiss). + + + +## 2. Code Compilation + +The command is as follows, where the address of Paddle C++ inference library, opencv and other dependency libraries need to be replaced with the actual address on your own machine. Also, you need to download and compile `yaml-cpp` and other C++ libraries during the compilation, so please keep the network unblocked. + +```shell +sh tools/build.sh +``` + +Specifically, the contents of `tools/build.sh` are as follows, please modify according to the specific path. + +```shell +OPENCV_DIR=${opencv_install_dir} +LIB_DIR=${paddle_inference_dir} +CUDA_LIB_DIR=/usr/local/cuda/lib64 +CUDNN_LIB_DIR=/usr/lib/x86_64-linux-gnu/ +FAISS_DIR=${faiss_install_dir} +FAISS_WITH_MKL=OFF + +BUILD_DIR=build +rm -rf ${BUILD_DIR} +mkdir ${BUILD_DIR} +cd ${BUILD_DIR} +cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DWITH_GPU=OFF \ + -DWITH_STATIC_LIB=OFF \ + -DUSE_TENSORRT=OFF \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + -DFAISS_DIR=${FAISS_DIR} \ + -DFAISS_WITH_MKL=${FAISS_WITH_MKL} + +make -j +cd .. +``` + +In the above commands: + +- `OPENCV_DIR` is the address of the opencv compilation and installation (in this case, the path of the `opencv-3.4.7/opencv3` folder). +- `LIB_DIR` is the path of the downloaded Paddle inference library (`paddle_inference` folder), or the generated Paddle inference library after compilation (`build/paddle_inference_install_dir` folder). +- `CUDA_LIB_DIR` is path of the cuda library file, which in docker is `/usr/local/cuda/lib64`. +- `CUDNN_LIB_DIR` is the path of the cudnn library file, which in docker is `/usr/lib/x86_64-linux-gnu/` . +- `TENSORRT_DIR` is the path of the tensorrt library file, which in docker is `/usr/local/TensorRT6-cuda10.0-cudnn7/`. TensorRT needs to be used in combination with GPU. +- `FAISS_DIR` is the installation path of faiss. +- `FAISS_WITH_MKL` means whether mkldnn is used during the compilation of faiss. The compilation in this document employs openbals instead of mkldnn, so it is set to `OFF`, otherwise it is `ON`. + +A `build` folder will be created in the current path after the compilation, which generates an executable file named `pp_shitu`. + + + +## 3. Run the demo + +- Please refer to the [Quick Start of Recognition](../../docs/en/quick_start/quick_start_recognition_en.md), download the corresponding Lightweight Generic Mainbody Detection Model, Lightweight Generic Recognition Model, and the beverage test data and unzip them. + + ```shell + mkdir models + cd models + wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer.tar + tar -xf picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer.tar + wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/general_PPLCNet_x2_5_lite_v1.0_infer.tar + tar -xf general_PPLCNet_x2_5_lite_v1.0_infer.tar + cd .. + + mkdir data + cd data + wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/data/drink_dataset_v1.0.tar + tar -xf drink_dataset_v1.0.tar + cd .. + ``` + +- Copy the corresponding yaml file to the current folder + + ``` + cp ../configs/inference_drink.yaml . + ``` + +- Change the relative path in `inference_drink.yaml` to a path based on this directory or an absolute path. The parameters involved are: + + - Global.infer_imgs: This parameter can be a specific image path or the directory where the image set is located + - Global.det_inference_model_dir: The directory where detection models are saved. + - Global.rec_inference_model_dir: The directory where recognition models are saved. + - IndexProcess.index_dir: The storage directory of the retrieval library, in the exmple, the retrieval library is in the downloaded demo data. + +- Transform the ID-Label Map Dictionary + + The id-label map dictionary in python is serialized using `pickle`, which make it hard to read for C++, so the transformation is needed: + + ``` + python tools/transform_id_map.py -c inference_drink.yaml + ``` + + `id_map.txt` is generated in `IndexProcess.index_dir` directory for convenience of C++ reading. + +- Execute the program + + ```shell + ./build/pp_shitu -c inference_drink.yaml + # or + ./build/pp_shitu -config inference_drink.yaml + ``` + + The following results can be obtained after searching the image set. + + At the same time, it should be noticed that a slight difference may occur during the pre-processing of the image due to the version of opencv, resulting in a minor discrepancy in python and c++ results, such as a few pixels for bbox, 3 decimal places for retrieval results, etc. But it has no impact on the final search label. + + [![img](../../docs/images/quick_start/shitu_c++_result.png)](https://github.com/PaddlePaddle/PaddleClas/blob/develop/docs/images/quick_start/shitu_c++_result.png) + + + +## 4. Use Your Own Model + +You can also use your self-trained models. Please refer to [model export](../../docs/en/inference_deployment/export_model_en.md) to export ` inference model` for model inference. + +Mind modifying the specific parameters in the `yaml` file. diff --git a/exprements/PaddleClas/deploy/cpp_shitu/src/feature_extracter.cpp b/exprements/PaddleClas/deploy/cpp_shitu/src/feature_extracter.cpp new file mode 100644 index 00000000..37c7590f --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/src/feature_extracter.cpp @@ -0,0 +1,119 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include +#include + +namespace Feature { + + void FeatureExtracter::LoadModel(const std::string &model_path, + const std::string ¶ms_path) { + paddle_infer::Config config; + config.SetModel(model_path, params_path); + + if (this->use_gpu_) { + config.EnableUseGpu(this->gpu_mem_, this->gpu_id_); + if (this->use_tensorrt_) { + config.EnableTensorRtEngine( + 1 << 20, 1, 3, + this->use_fp16_ ? paddle_infer::Config::Precision::kHalf + : paddle_infer::Config::Precision::kFloat32, + false, false); + } + } else { + config.DisableGpu(); + if (this->use_mkldnn_) { + config.EnableMKLDNN(); + // cache 10 different shapes for mkldnn to avoid memory leak + config.SetMkldnnCacheCapacity(10); + } + config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_); + } + + config.SwitchUseFeedFetchOps(false); + // true for multiple input + config.SwitchSpecifyInputNames(true); + + config.SwitchIrOptim(true); + + config.EnableMemoryOptim(); + config.DisableGlogInfo(); + + this->predictor_ = CreatePredictor(config); + } + + void FeatureExtracter::Run(cv::Mat &img, std::vector &out_data, + std::vector ×) { + cv::Mat resize_img; + std::vector time; + + auto preprocess_start = std::chrono::steady_clock::now(); + this->resize_op_.Run(img, resize_img, this->resize_short_, + this->resize_size_); + + this->normalize_op_.Run(&resize_img, this->mean_, this->std_, this->scale_); + std::vector input(1 * 3 * resize_img.rows * resize_img.cols, 0.0f); + this->permute_op_.Run(&resize_img, input.data()); + + auto input_names = this->predictor_->GetInputNames(); + auto input_t = this->predictor_->GetInputHandle(input_names[0]); + input_t->Reshape({1, 3, resize_img.rows, resize_img.cols}); + auto preprocess_end = std::chrono::steady_clock::now(); + + auto infer_start = std::chrono::steady_clock::now(); + input_t->CopyFromCpu(input.data()); + this->predictor_->Run(); + + auto output_names = this->predictor_->GetOutputNames(); + auto output_t = this->predictor_->GetOutputHandle(output_names[0]); + std::vector output_shape = output_t->shape(); + int out_num = std::accumulate(output_shape.begin(), output_shape.end(), 1, + std::multiplies()); + + out_data.resize(out_num); + output_t->CopyToCpu(out_data.data()); + auto infer_end = std::chrono::steady_clock::now(); + + auto postprocess_start = std::chrono::steady_clock::now(); + if (this->feature_norm) + FeatureNorm(out_data); + auto postprocess_end = std::chrono::steady_clock::now(); + + std::chrono::duration preprocess_diff = + preprocess_end - preprocess_start; + time.push_back(double(preprocess_diff.count()) * 1000); + std::chrono::duration inference_diff = infer_end - infer_start; + double inference_cost_time = double(inference_diff.count() * 1000); + time.push_back(inference_cost_time); + // std::chrono::duration postprocess_diff = + // postprocess_end - postprocess_start; + time.push_back(0); + + // std::cout << "result: " << std::endl; + // std::cout << "\tclass id: " << maxPosition << std::endl; + // std::cout << std::fixed << std::setprecision(10) + // << "\tscore: " << double(out_data[maxPosition]) << std::endl; + times[0] += time[0]; + times[1] += time[1]; + times[2] += time[2]; + } + + void FeatureExtracter::FeatureNorm(std::vector &featuer) { + float featuer_sqrt = std::sqrt(std::inner_product( + featuer.begin(), featuer.end(), featuer.begin(), 0.0f)); + for (int i = 0; i < featuer.size(); ++i) + featuer[i] /= featuer_sqrt; + } +} // namespace Feature diff --git a/exprements/PaddleClas/deploy/cpp_shitu/src/main.cpp b/exprements/PaddleClas/deploy/cpp_shitu/src/main.cpp new file mode 100644 index 00000000..be37d3af --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/src/main.cpp @@ -0,0 +1,342 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include + +using namespace std; +using namespace cv; + +DEFINE_string(config, +"", "Path of yaml file"); +DEFINE_string(c, +"", "Path of yaml file"); + +void DetPredictImage(const std::vector &batch_imgs, + const std::vector &all_img_paths, + const int batch_size, Detection::ObjectDetector *det, + std::vector &im_result, + std::vector &im_bbox_num, std::vector &det_t, + const bool visual_det = false, + const bool run_benchmark = false, + const std::string &output_dir = "output") { + int steps = ceil(float(all_img_paths.size()) / batch_size); + // printf("total images = %d, batch_size = %d, total steps = %d\n", + // all_img_paths.size(), batch_size, steps); + for (int idx = 0; idx < steps; idx++) { + int left_image_cnt = all_img_paths.size() - idx * batch_size; + if (left_image_cnt > batch_size) { + left_image_cnt = batch_size; + } + // for (int bs = 0; bs < left_image_cnt; bs++) { + // std::string image_file_path = all_img_paths.at(idx * batch_size+bs); + // cv::Mat im = cv::imread(image_file_path, 1); + // batch_imgs.insert(batch_imgs.end(), im); + // } + + // Store all detected result + std::vector result; + std::vector bbox_num; + std::vector det_times; + bool is_rbox = false; + if (run_benchmark) { + det->Predict(batch_imgs, 10, 10, &result, &bbox_num, &det_times); + } else { + det->Predict(batch_imgs, 0, 1, &result, &bbox_num, &det_times); + // get labels and colormap + auto labels = det->GetLabelList(); + auto colormap = Detection::GenerateColorMap(labels.size()); + + int item_start_idx = 0; + for (int i = 0; i < left_image_cnt; i++) { + cv::Mat im = batch_imgs[i]; + int detect_num = 0; + + for (int j = 0; j < bbox_num[i]; j++) { + Detection::ObjectResult item = result[item_start_idx + j]; + if (item.confidence < det->GetThreshold() || item.class_id == -1) { + continue; + } + detect_num += 1; + im_result.push_back(item); + if (visual_det) { + if (item.rect.size() > 6) { + is_rbox = true; + printf( + "class=%d confidence=%.4f rect=[%d %d %d %d %d %d %d %d]\n", + item.class_id, item.confidence, item.rect[0], item.rect[1], + item.rect[2], item.rect[3], item.rect[4], item.rect[5], + item.rect[6], item.rect[7]); + } else { + printf("class=%d confidence=%.4f rect=[%d %d %d %d]\n", + item.class_id, item.confidence, item.rect[0], item.rect[1], + item.rect[2], item.rect[3]); + } + } + } + im_bbox_num.push_back(detect_num); + item_start_idx = item_start_idx + bbox_num[i]; + + // Visualization result + if (visual_det) { + std::cout << all_img_paths.at(idx * batch_size + i) + << " The number of detected box: " << detect_num + << std::endl; + cv::Mat vis_img = Detection::VisualizeResult(im, im_result, labels, + colormap, is_rbox); + std::vector compression_params; + compression_params.push_back(CV_IMWRITE_JPEG_QUALITY); + compression_params.push_back(95); + std::string output_path(output_dir); + if (output_dir.rfind(OS_PATH_SEP) != output_dir.size() - 1) { + output_path += OS_PATH_SEP; + } + std::string image_file_path = all_img_paths.at(idx * batch_size + i); + output_path += + image_file_path.substr(image_file_path.find_last_of('/') + 1); + cv::imwrite(output_path, vis_img, compression_params); + printf("Visualized output saved as %s\n", output_path.c_str()); + } + } + } + det_t[0] += det_times[0]; + det_t[1] += det_times[1]; + det_t[2] += det_times[2]; + } +} + +void PrintResult(std::string &img_path, + std::vector &det_result, + std::vector &indeices, VectorSearch &vector_search, + SearchResult &search_result) { + printf("%s:\n", img_path.c_str()); + for (int i = 0; i < indeices.size(); ++i) { + int t = indeices[i]; + printf("\tresult%d: bbox[%d, %d, %d, %d], score: %f, label: %s\n", i, + det_result[t].rect[0], det_result[t].rect[1], det_result[t].rect[2], + det_result[t].rect[3], det_result[t].confidence, + vector_search.GetLabel(search_result.I[search_result.return_k * t]) + .c_str()); + } +} + +int main(int argc, char **argv) { + google::ParseCommandLineFlags(&argc, &argv, true); + std::string yaml_path = ""; + if (FLAGS_config == "" && FLAGS_c == "") { + std::cerr << "[ERROR] usage: " << std::endl + << argv[0] << " -c $yaml_path" << std::endl + << "or:" << std::endl + << argv[0] << " -config $yaml_path" << std::endl; + exit(1); + } else if (FLAGS_config != "") { + yaml_path = FLAGS_config; + } else { + yaml_path = FLAGS_c; + } + + YamlConfig config(yaml_path); + config.PrintConfigInfo(); + + // initialize detector, rec_Model, vector_search + Feature::FeatureExtracter feature_extracter(config.config_file); + Detection::ObjectDetector detector(config.config_file); + VectorSearch searcher(config.config_file); + + // config + const int batch_size = config.config_file["Global"]["batch_size"].as(); + bool visual_det = false; + if (config.config_file["Global"]["visual_det"].IsDefined()) { + visual_det = config.config_file["Global"]["visual_det"].as(); + } + bool benchmark = false; + if (config.config_file["Global"]["benchmark"].IsDefined()) { + benchmark = config.config_file["Global"]["benchmark"].as(); + } + int max_det_results = 5; + if (config.config_file["Global"]["max_det_results"].IsDefined()) { + max_det_results = config.config_file["Global"]["max_det_results"].as(); + } + float rec_nms_thresold = 0.05; + if (config.config_file["Global"]["rec_nms_thresold"].IsDefined()) { + rec_nms_thresold = + config.config_file["Global"]["rec_nms_thresold"].as(); + } + + // load image_file_path + std::string path = + config.config_file["Global"]["infer_imgs"].as(); + std::vector img_files_list; + if (cv::utils::fs::isDirectory(path)) { + std::vector filenames; + cv::glob(path, filenames); + for (auto f : filenames) { + img_files_list.push_back(f); + } + } else { + img_files_list.push_back(path); + } + std::cout << "img_file_list length: " << img_files_list.size() << std::endl; + // for time log + std::vector cls_times = {0, 0, 0}; + std::vector det_times = {0, 0, 0}; + std::vector search_times = {0, 0, 0}; + int instance_num = 0; + // for read images + std::vector batch_imgs; + std::vector img_paths; + // for detection + std::vector det_result; + std::vector det_bbox_num; + // for vector search + std::vector features; + std::vector feature; + // for nms + std::vector indeices; + + int warmup_iter = img_files_list.size() > 5 ? 5 : img_files_list.size(); + if (benchmark) { + img_files_list.insert(img_files_list.begin(), img_files_list.begin(), + img_files_list.begin() + warmup_iter); + } + + for (int idx = 0; idx < img_files_list.size(); ++idx) { + std::string img_path = img_files_list[idx]; + cv::Mat srcimg = cv::imread(img_path, cv::IMREAD_COLOR); + if (!srcimg.data) { + std::cerr << "[ERROR] image read failed! image path: " << img_path + << "\n"; + exit(-1); + } + cv::cvtColor(srcimg, srcimg, cv::COLOR_BGR2RGB); + + batch_imgs.push_back(srcimg); + img_paths.push_back(img_path); + + // step1: get all detection results + DetPredictImage(batch_imgs, img_paths, batch_size, &detector, det_result, + det_bbox_num, det_times, visual_det, false); + + // select max_det_results bbox + if (det_result.size() > max_det_results) { + det_result.resize(max_det_results); + } + instance_num += det_result.size(); + + // step2: add the whole image for recognition to improve recall + Detection::ObjectResult result_whole_img = { + {0, 0, srcimg.cols - 1, srcimg.rows - 1}, 0, 1.0}; + det_result.push_back(result_whole_img); + det_bbox_num[0] = det_result.size() + 1; + + // step3: extract feature for all boxes in an inmage + SearchResult search_result; + for (int j = 0; j < det_result.size(); ++j) { + int w = det_result[j].rect[2] - det_result[j].rect[0]; + int h = det_result[j].rect[3] - det_result[j].rect[1]; + cv::Rect rect(det_result[j].rect[0], det_result[j].rect[1], w, h); + cv::Mat crop_img = srcimg(rect); + feature_extracter.Run(crop_img, feature, cls_times); + features.insert(features.end(), feature.begin(), feature.end()); + } + + // step4: get search result + auto search_start = std::chrono::steady_clock::now(); + search_result = searcher.Search(features.data(), det_result.size()); + auto search_end = std::chrono::steady_clock::now(); + + // nms for search result + for (int i = 0; i < det_result.size(); ++i) { + det_result[i].confidence = search_result.D[search_result.return_k * i]; + } + NMSBoxes(det_result, searcher.GetThreshold(), rec_nms_thresold, indeices); + auto nms_end = std::chrono::steady_clock::now(); + std::chrono::duration search_diff = search_end - search_start; + search_times[1] += double(search_diff.count() * 1000); + + std::chrono::duration nms_diff = nms_end - search_end; + search_times[2] += double(nms_diff.count() * 1000); + + // print result + if (not benchmark or (benchmark and idx >= warmup_iter)) + PrintResult(img_path, det_result, indeices, searcher, search_result); + + // for postprocess + batch_imgs.clear(); + img_paths.clear(); + det_bbox_num.clear(); + det_result.clear(); + feature.clear(); + features.clear(); + indeices.clear(); + if (benchmark and warmup_iter == idx + 1) { + det_times = {0, 0, 0}; + cls_times = {0, 0, 0}; + search_times = {0, 0, 0}; + instance_num = 0; + } + } + + if (benchmark) { + std::string presion = "fp32"; + if (config.config_file["Global"]["use_fp16"].IsDefined() and + config.config_file["Global"]["use_fp16"].as()) + presion = "fp16"; + bool use_gpu = config.config_file["Global"]["use_gpu"].as(); + bool use_tensorrt = config.config_file["Global"]["use_tensorrt"].as(); + bool enable_mkldnn = + config.config_file["Global"]["enable_mkldnn"].as(); + int cpu_num_threads = + config.config_file["Global"]["cpu_num_threads"].as(); + int batch_size = config.config_file["Global"]["batch_size"].as(); + std::vector shape = + config.config_file["Global"]["image_shape"].as < std::vector < int >> (); + std::string det_shape = std::to_string(shape[0]); + for (int i = 1; i < shape.size(); ++i) + det_shape = det_shape + ", " + std::to_string(shape[i]); + + AutoLogger autolog_det("Det", use_gpu, use_tensorrt, enable_mkldnn, + cpu_num_threads, batch_size, det_shape, presion, + det_times, img_files_list.size() - warmup_iter); + autolog_det.report(); + AutoLogger autolog_rec("Rec", use_gpu, use_tensorrt, enable_mkldnn, + cpu_num_threads, batch_size, "3, 224, 224", presion, + cls_times, instance_num); + autolog_rec.report(); + AutoLogger autolog_search("Search", false, use_tensorrt, enable_mkldnn, + cpu_num_threads, batch_size, "dynamic", presion, + search_times, instance_num); + autolog_search.report(); + } + return 0; +} diff --git a/exprements/PaddleClas/deploy/cpp_shitu/src/object_detector.cpp b/exprements/PaddleClas/deploy/cpp_shitu/src/object_detector.cpp new file mode 100644 index 00000000..03b02b97 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/src/object_detector.cpp @@ -0,0 +1,365 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +#include +// for setprecision +#include "include/object_detector.h" +#include +#include + +using namespace paddle_infer; + +namespace Detection { + +// Load Model and create model predictor + void ObjectDetector::LoadModel(const std::string &model_dir, + const int batch_size, + const std::string &run_mode) { + paddle_infer::Config config; + std::string prog_file = model_dir + OS_PATH_SEP + "inference.pdmodel"; + std::string params_file = model_dir + OS_PATH_SEP + "inference.pdiparams"; + config.SetModel(prog_file, params_file); + if (this->use_gpu_) { + config.EnableUseGpu(this->gpu_mem_, this->gpu_id_); + config.SwitchIrOptim(this->ir_optim_); + // // use tensorrt + // if (run_mode != "fluid") { + // auto precision = paddle_infer::Config::Precision::kFloat32; + // if (run_mode == "trt_fp32") { + // precision = paddle_infer::Config::Precision::kFloat32; + // } + // else if (run_mode == "trt_fp16") { + // precision = paddle_infer::Config::Precision::kHalf; + // } + // else if (run_mode == "trt_int8") { + // precision = paddle_infer::Config::Precision::kInt8; + // } else { + // printf("run_mode should be 'fluid', 'trt_fp32', 'trt_fp16' or + // 'trt_int8'"); + // } + // set tensorrt + + if (this->use_tensorrt_) { + config.EnableTensorRtEngine( + 1 << 30, batch_size, this->min_subgraph_size_, + this->use_fp16_ ? paddle_infer::Config::Precision::kHalf + : paddle_infer::Config::Precision::kFloat32, + false, this->trt_calib_mode_); + // set use dynamic shape + if (this->use_dynamic_shape_) { + // set DynamicShsape for image tensor + const std::vector min_input_shape = {1, 3, this->trt_min_shape_, + this->trt_min_shape_}; + const std::vector max_input_shape = {1, 3, this->trt_max_shape_, + this->trt_max_shape_}; + const std::vector opt_input_shape = {1, 3, this->trt_opt_shape_, + this->trt_opt_shape_}; + const std::map > map_min_input_shape = { + {"image", min_input_shape}}; + const std::map > map_max_input_shape = { + {"image", max_input_shape}}; + const std::map > map_opt_input_shape = { + {"image", opt_input_shape}}; + + config.SetTRTDynamicShapeInfo(map_min_input_shape, map_max_input_shape, + map_opt_input_shape); + std::cout << "TensorRT dynamic shape enabled" << std::endl; + } + } + + // } else if (this->device_ == "XPU"){ + // config.EnableXpu(10*1024*1024); + } else { + config.DisableGpu(); + if (this->use_mkldnn_) { + config.EnableMKLDNN(); + // cache 10 different shapes for mkldnn to avoid memory leak + config.SetMkldnnCacheCapacity(10); + } + config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_); + } + config.SwitchUseFeedFetchOps(false); + config.SwitchIrOptim(this->ir_optim_); + config.DisableGlogInfo(); + // Memory optimization + config.EnableMemoryOptim(); + predictor_ = std::move(CreatePredictor(config)); + } + +// Visualiztion MaskDetector results + cv::Mat VisualizeResult(const cv::Mat &img, + const std::vector &results, + const std::vector &lables, + const std::vector &colormap, + const bool is_rbox = false) { + cv::Mat vis_img = img.clone(); + for (int i = 0; i < results.size(); ++i) { + // Configure color and text size + std::ostringstream oss; + oss << std::setiosflags(std::ios::fixed) << std::setprecision(4); + oss << lables[results[i].class_id] << " "; + oss << results[i].confidence; + std::string text = oss.str(); + int c1 = colormap[3 * results[i].class_id + 0]; + int c2 = colormap[3 * results[i].class_id + 1]; + int c3 = colormap[3 * results[i].class_id + 2]; + cv::Scalar roi_color = cv::Scalar(c1, c2, c3); + int font_face = cv::FONT_HERSHEY_COMPLEX_SMALL; + double font_scale = 0.5f; + float thickness = 0.5; + cv::Size text_size = + cv::getTextSize(text, font_face, font_scale, thickness, nullptr); + cv::Point origin; + + if (is_rbox) { + // Draw object, text, and background + for (int k = 0; k < 4; k++) { + cv::Point pt1 = cv::Point(results[i].rect[(k * 2) % 8], + results[i].rect[(k * 2 + 1) % 8]); + cv::Point pt2 = cv::Point(results[i].rect[(k * 2 + 2) % 8], + results[i].rect[(k * 2 + 3) % 8]); + cv::line(vis_img, pt1, pt2, roi_color, 2); + } + } else { + int w = results[i].rect[2] - results[i].rect[0]; + int h = results[i].rect[3] - results[i].rect[1]; + cv::Rect roi = cv::Rect(results[i].rect[0], results[i].rect[1], w, h); + // Draw roi object, text, and background + cv::rectangle(vis_img, roi, roi_color, 2); + } + + origin.x = results[i].rect[0]; + origin.y = results[i].rect[1]; + + // Configure text background + cv::Rect text_back = + cv::Rect(results[i].rect[0], results[i].rect[1] - text_size.height, + text_size.width, text_size.height); + // Draw text, and background + cv::rectangle(vis_img, text_back, roi_color, -1); + cv::putText(vis_img, text, origin, font_face, font_scale, + cv::Scalar(255, 255, 255), thickness); + } + return vis_img; + } + + void ObjectDetector::Preprocess(const cv::Mat &ori_im) { + // Clone the image : keep the original mat for postprocess + cv::Mat im = ori_im.clone(); + cv::cvtColor(im, im, cv::COLOR_BGR2RGB); + preprocessor_.Run(&im, &inputs_); + } + + void ObjectDetector::Postprocess(const std::vector mats, + std::vector *result, + std::vector bbox_num, + bool is_rbox = false) { + result->clear(); + int start_idx = 0; + for (int im_id = 0; im_id < mats.size(); im_id++) { + cv::Mat raw_mat = mats[im_id]; + int rh = 1; + int rw = 1; + // if (config_.arch_ == "Face") { + // rh = raw_mat.rows; + // rw = raw_mat.cols; + // } + for (int j = start_idx; j < start_idx + bbox_num[im_id]; j++) { + if (is_rbox) { + // Class id + int class_id = static_cast(round(output_data_[0 + j * 10])); + // Confidence score + float score = output_data_[1 + j * 10]; + int x1 = (output_data_[2 + j * 10] * rw); + int y1 = (output_data_[3 + j * 10] * rh); + int x2 = (output_data_[4 + j * 10] * rw); + int y2 = (output_data_[5 + j * 10] * rh); + int x3 = (output_data_[6 + j * 10] * rw); + int y3 = (output_data_[7 + j * 10] * rh); + int x4 = (output_data_[8 + j * 10] * rw); + int y4 = (output_data_[9 + j * 10] * rh); + + ObjectResult result_item; + result_item.rect = {x1, y1, x2, y2, x3, y3, x4, y4}; + result_item.class_id = class_id; + result_item.confidence = score; + result->push_back(result_item); + } else { + // Class id + int class_id = static_cast(round(output_data_[0 + j * 6])); + // Confidence score + float score = output_data_[1 + j * 6]; + int xmin = (output_data_[2 + j * 6] * rw); + int ymin = (output_data_[3 + j * 6] * rh); + int xmax = (output_data_[4 + j * 6] * rw); + int ymax = (output_data_[5 + j * 6] * rh); + int wd = xmax - xmin; + int hd = ymax - ymin; + + ObjectResult result_item; + result_item.rect = {xmin, ymin, xmax, ymax}; + result_item.class_id = class_id; + result_item.confidence = score; + result->push_back(result_item); + } + } + start_idx += bbox_num[im_id]; + } + } + + void ObjectDetector::Predict(const std::vector imgs, const int warmup, + const int repeats, + std::vector *result, + std::vector *bbox_num, + std::vector *times) { + auto preprocess_start = std::chrono::steady_clock::now(); + int batch_size = imgs.size(); + + // in_data_batch + std::vector in_data_all; + std::vector im_shape_all(batch_size * 2); + std::vector scale_factor_all(batch_size * 2); + + // Preprocess image + for (int bs_idx = 0; bs_idx < batch_size; bs_idx++) { + cv::Mat im = imgs.at(bs_idx); + Preprocess(im); + im_shape_all[bs_idx * 2] = inputs_.im_shape_[0]; + im_shape_all[bs_idx * 2 + 1] = inputs_.im_shape_[1]; + + scale_factor_all[bs_idx * 2] = inputs_.scale_factor_[0]; + scale_factor_all[bs_idx * 2 + 1] = inputs_.scale_factor_[1]; + + // TODO: reduce cost time + in_data_all.insert(in_data_all.end(), inputs_.im_data_.begin(), + inputs_.im_data_.end()); + } + + // Prepare input tensor + auto input_names = predictor_->GetInputNames(); + for (const auto &tensor_name : input_names) { + auto in_tensor = predictor_->GetInputHandle(tensor_name); + if (tensor_name == "image") { + int rh = inputs_.in_net_shape_[0]; + int rw = inputs_.in_net_shape_[1]; + in_tensor->Reshape({batch_size, 3, rh, rw}); + in_tensor->CopyFromCpu(in_data_all.data()); + } else if (tensor_name == "im_shape") { + in_tensor->Reshape({batch_size, 2}); + in_tensor->CopyFromCpu(im_shape_all.data()); + } else if (tensor_name == "scale_factor") { + in_tensor->Reshape({batch_size, 2}); + in_tensor->CopyFromCpu(scale_factor_all.data()); + } + } + + auto preprocess_end = std::chrono::steady_clock::now(); + // Run predictor + // warmup + for (int i = 0; i < warmup; i++) { + predictor_->Run(); + // Get output tensor + auto output_names = predictor_->GetOutputNames(); + auto out_tensor = predictor_->GetOutputHandle(output_names[0]); + std::vector output_shape = out_tensor->shape(); + auto out_bbox_num = predictor_->GetOutputHandle(output_names[1]); + std::vector out_bbox_num_shape = out_bbox_num->shape(); + // Calculate output length + int output_size = 1; + for (int j = 0; j < output_shape.size(); ++j) { + output_size *= output_shape[j]; + } + + if (output_size < 6) { + std::cerr << "[WARNING] No object detected." << std::endl; + } + output_data_.resize(output_size); + out_tensor->CopyToCpu(output_data_.data()); + + int out_bbox_num_size = 1; + for (int j = 0; j < out_bbox_num_shape.size(); ++j) { + out_bbox_num_size *= out_bbox_num_shape[j]; + } + out_bbox_num_data_.resize(out_bbox_num_size); + out_bbox_num->CopyToCpu(out_bbox_num_data_.data()); + } + + bool is_rbox = false; + auto inference_start = std::chrono::steady_clock::now(); + for (int i = 0; i < repeats; i++) { + predictor_->Run(); + // Get output tensor + auto output_names = predictor_->GetOutputNames(); + auto out_tensor = predictor_->GetOutputHandle(output_names[0]); + std::vector output_shape = out_tensor->shape(); + auto out_bbox_num = predictor_->GetOutputHandle(output_names[1]); + std::vector out_bbox_num_shape = out_bbox_num->shape(); + // Calculate output length + int output_size = 1; + for (int j = 0; j < output_shape.size(); ++j) { + output_size *= output_shape[j]; + } + is_rbox = output_shape[output_shape.size() - 1] % 10 == 0; + + if (output_size < 6) { + std::cerr << "[WARNING] No object detected." << std::endl; + } + output_data_.resize(output_size); + out_tensor->CopyToCpu(output_data_.data()); + + int out_bbox_num_size = 1; + for (int j = 0; j < out_bbox_num_shape.size(); ++j) { + out_bbox_num_size *= out_bbox_num_shape[j]; + } + out_bbox_num_data_.resize(out_bbox_num_size); + out_bbox_num->CopyToCpu(out_bbox_num_data_.data()); + } + auto inference_end = std::chrono::steady_clock::now(); + auto postprocess_start = std::chrono::steady_clock::now(); + // Postprocessing result + result->clear(); + Postprocess(imgs, result, out_bbox_num_data_, is_rbox); + bbox_num->clear(); + for (int k = 0; k < out_bbox_num_data_.size(); k++) { + int tmp = out_bbox_num_data_[k]; + bbox_num->push_back(tmp); + } + auto postprocess_end = std::chrono::steady_clock::now(); + + std::chrono::duration preprocess_diff = + preprocess_end - preprocess_start; + times->push_back(double(preprocess_diff.count() * 1000)); + std::chrono::duration inference_diff = inference_end - inference_start; + times->push_back(double(inference_diff.count() / repeats * 1000)); + std::chrono::duration postprocess_diff = + postprocess_end - postprocess_start; + times->push_back(double(postprocess_diff.count() * 1000)); + } + + std::vector GenerateColorMap(int num_class) { + auto colormap = std::vector(3 * num_class, 0); + for (int i = 0; i < num_class; ++i) { + int j = 0; + int lab = i; + while (lab) { + colormap[i * 3] |= (((lab >> 0) & 1) << (7 - j)); + colormap[i * 3 + 1] |= (((lab >> 1) & 1) << (7 - j)); + colormap[i * 3 + 2] |= (((lab >> 2) & 1) << (7 - j)); + ++j; + lab >>= 3; + } + } + return colormap; + } + +} // namespace Detection diff --git a/exprements/PaddleClas/deploy/cpp_shitu/src/preprocess_op.cpp b/exprements/PaddleClas/deploy/cpp_shitu/src/preprocess_op.cpp new file mode 100644 index 00000000..6bf1ec91 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/src/preprocess_op.cpp @@ -0,0 +1,91 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "opencv2/core.hpp" +#include "opencv2/imgcodecs.hpp" +#include "opencv2/imgproc.hpp" +#include "paddle_api.h" +#include "paddle_inference_api.h" +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +#include + +namespace Feature { + + void Permute::Run(const cv::Mat *im, float *data) { + int rh = im->rows; + int rw = im->cols; + int rc = im->channels(); + for (int i = 0; i < rc; ++i) { + cv::extractChannel(*im, cv::Mat(rh, rw, CV_32FC1, data + i * rh * rw), i); + } + } + + void Normalize::Run(cv::Mat *im, const std::vector &mean, + const std::vector &std, float scale) { + (*im).convertTo(*im, CV_32FC3, scale); + for (int h = 0; h < im->rows; h++) { + for (int w = 0; w < im->cols; w++) { + im->at(h, w)[0] = + (im->at(h, w)[0] - mean[0]) / std[0]; + im->at(h, w)[1] = + (im->at(h, w)[1] - mean[1]) / std[1]; + im->at(h, w)[2] = + (im->at(h, w)[2] - mean[2]) / std[2]; + } + } + } + + void CenterCropImg::Run(cv::Mat &img, const int crop_size) { + int resize_w = img.cols; + int resize_h = img.rows; + int w_start = int((resize_w - crop_size) / 2); + int h_start = int((resize_h - crop_size) / 2); + cv::Rect rect(w_start, h_start, crop_size, crop_size); + img = img(rect); + } + + void ResizeImg::Run(const cv::Mat &img, cv::Mat &resize_img, + int resize_short_size, int size) { + int resize_h = 0; + int resize_w = 0; + if (size > 0) { + resize_h = size; + resize_w = size; + } else { + int w = img.cols; + int h = img.rows; + + float ratio = 1.f; + if (h < w) { + ratio = float(resize_short_size) / float(h); + } else { + ratio = float(resize_short_size) / float(w); + } + resize_h = round(float(h) * ratio); + resize_w = round(float(w) * ratio); + } + cv::resize(img, resize_img, cv::Size(resize_w, resize_h)); + } + +} // namespace Feature diff --git a/exprements/PaddleClas/deploy/cpp_shitu/src/preprocess_op_det.cpp b/exprements/PaddleClas/deploy/cpp_shitu/src/preprocess_op_det.cpp new file mode 100644 index 00000000..e32b03a4 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/src/preprocess_op_det.cpp @@ -0,0 +1,130 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include + +#include "include/preprocess_op_det.h" + +namespace Detection { + + void InitInfo::Run(cv::Mat *im, ImageBlob *data) { + data->im_shape_ = {static_cast(im->rows), + static_cast(im->cols)}; + data->scale_factor_ = {1., 1.}; + data->in_net_shape_ = {static_cast(im->rows), + static_cast(im->cols)}; + } + + void NormalizeImage::Run(cv::Mat *im, ImageBlob *data) { + double e = 1.0; + if (is_scale_) { + e /= 255.0; + } + (*im).convertTo(*im, CV_32FC3, e); + for (int h = 0; h < im->rows; h++) { + for (int w = 0; w < im->cols; w++) { + im->at(h, w)[0] = + (im->at(h, w)[0] - mean_[0]) / scale_[0]; + im->at(h, w)[1] = + (im->at(h, w)[1] - mean_[1]) / scale_[1]; + im->at(h, w)[2] = + (im->at(h, w)[2] - mean_[2]) / scale_[2]; + } + } + } + + void Permute::Run(cv::Mat *im, ImageBlob *data) { + int rh = im->rows; + int rw = im->cols; + int rc = im->channels(); + (data->im_data_).resize(rc * rh * rw); + float *base = (data->im_data_).data(); + for (int i = 0; i < rc; ++i) { + cv::extractChannel(*im, cv::Mat(rh, rw, CV_32FC1, base + i * rh * rw), i); + } + } + + void Resize::Run(cv::Mat *im, ImageBlob *data) { + auto resize_scale = GenerateScale(*im); + data->im_shape_ = {static_cast(im->cols * resize_scale.first), + static_cast(im->rows * resize_scale.second)}; + data->in_net_shape_ = {static_cast(im->cols * resize_scale.first), + static_cast(im->rows * resize_scale.second)}; + cv::resize(*im, *im, cv::Size(), resize_scale.first, resize_scale.second, + interp_); + data->im_shape_ = { + static_cast(im->rows), static_cast(im->cols), + }; + data->scale_factor_ = { + resize_scale.second, resize_scale.first, + }; + } + + std::pair Resize::GenerateScale(const cv::Mat &im) { + std::pair resize_scale; + int origin_w = im.cols; + int origin_h = im.rows; + + if (keep_ratio_) { + int im_size_max = std::max(origin_w, origin_h); + int im_size_min = std::min(origin_w, origin_h); + int target_size_max = + *std::max_element(target_size_.begin(), target_size_.end()); + int target_size_min = + *std::min_element(target_size_.begin(), target_size_.end()); + double scale_min = + static_cast(target_size_min) / static_cast(im_size_min); + double scale_max = + static_cast(target_size_max) / static_cast(im_size_max); + double scale_ratio = std::min(scale_min, scale_max); + resize_scale = {scale_ratio, scale_ratio}; + } else { + resize_scale.first = + static_cast(target_size_[1]) / static_cast(origin_w); + resize_scale.second = + static_cast(target_size_[0]) / static_cast(origin_h); + } + return resize_scale; + } + + void PadStride::Run(cv::Mat *im, ImageBlob *data) { + if (stride_ <= 0) { + return; + } + int rc = im->channels(); + int rh = im->rows; + int rw = im->cols; + int nh = (rh / stride_) * stride_ + (rh % stride_ != 0) * stride_; + int nw = (rw / stride_) * stride_ + (rw % stride_ != 0) * stride_; + cv::copyMakeBorder(*im, *im, 0, nh - rh, 0, nw - rw, cv::BORDER_CONSTANT, + cv::Scalar(0)); + data->in_net_shape_ = { + static_cast(im->rows), static_cast(im->cols), + }; + } + +// Preprocessor op running order + const std::vector Preprocessor::RUN_ORDER = { + "InitInfo", "Resize", "NormalizeImage", "PadStride", "Permute"}; + + void Preprocessor::Run(cv::Mat *im, ImageBlob *data) { + for (const auto &name : RUN_ORDER) { + if (ops_.find(name) != ops_.end()) { + ops_[name]->Run(im, data); + } + } + } + +} // namespace Detection diff --git a/exprements/PaddleClas/deploy/cpp_shitu/src/vector_search.cpp b/exprements/PaddleClas/deploy/cpp_shitu/src/vector_search.cpp new file mode 100644 index 00000000..85c487a7 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/src/vector_search.cpp @@ -0,0 +1,62 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +#include "include/vector_search.h" +#include +#include +#include +#include +#include +#include + +void VectorSearch::LoadIndexFile() { + std::string file_path = this->index_dir + OS_PATH_SEP + "vector.index"; + const char *fname = file_path.c_str(); + this->index = faiss::read_index(fname, 0); +} + +void VectorSearch::LoadIdMap() { + std::string file_path = this->index_dir + OS_PATH_SEP + "id_map.txt"; + std::ifstream in(file_path); + std::string line; + std::vector m_vec; + if (in) { + while (getline(in, line)) { + std::regex ws_re("\\s+"); + std::vector v( + std::sregex_token_iterator(line.begin(), line.end(), ws_re, -1), + std::sregex_token_iterator()); + if (v.size() != 2) { + std::cout << "The number of element for each line in : " << file_path + << "must be 2, exit the program..." << std::endl; + exit(1); + } else + this->id_map.insert(std::pair( + std::stol(v[0], nullptr, 10), v[1])); + } + } +} + +const SearchResult &VectorSearch::Search(float *feature, int query_number) { + this->D.resize(this->return_k * query_number); + this->I.resize(this->return_k * query_number); + this->index->search(query_number, feature, return_k, D.data(), I.data()); + this->sr.return_k = this->return_k; + this->sr.D = this->D; + this->sr.I = this->I; + return this->sr; +} + +const std::string &VectorSearch::GetLabel(faiss::Index::idx_t ind) { + return this->id_map.at(ind); +} diff --git a/exprements/PaddleClas/deploy/cpp_shitu/src/yaml_config.cpp b/exprements/PaddleClas/deploy/cpp_shitu/src/yaml_config.cpp new file mode 100644 index 00000000..b20fcde9 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/src/yaml_config.cpp @@ -0,0 +1,78 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include +#include + +#include + + +std::vector YamlConfig::ReadDict(const std::string &path) { + std::ifstream in(path); + std::string line; + std::vector m_vec; + if (in) { + while (getline(in, line)) { + m_vec.push_back(line); + } + } else { + std::cout << "no such label file: " << path << ", exit the program..." + << std::endl; + exit(1); + } + return m_vec; +} + +std::map YamlConfig::ReadIndexId(const std::string &path) { + std::ifstream in(path); + std::string line; + std::map m_vec; + if (in) { + while (getline(in, line)) { + std::regex ws_re("\\s+"); + std::vector v( + std::sregex_token_iterator(line.begin(), line.end(), ws_re, -1), + std::sregex_token_iterator()); + if (v.size() != 3) { + std::cout << "The number of element for each line in : " << path + << "must be 3, exit the program..." << std::endl; + exit(1); + } else + m_vec.insert(std::pair(stoi(v[0]), v[2])); + } + } + return m_vec; +} + +YAML::Node YamlConfig::ReadYamlConfig(const std::string &path) { + YAML::Node config; + try { + config = YAML::LoadFile(path); + } catch (YAML::BadFile &e) { + std::cout << "Something wrong in yaml file, please check yaml file" + << std::endl; + exit(1); + } + return config; +} + +void YamlConfig::PrintConfigInfo() { + std::cout << this->config_file << std::endl; + // for (YAML::const_iterator + // it=config_file.begin();it!=config_file.end();++it) + // { + // std::cout << it->as() << "\n"; + // } +} diff --git a/exprements/PaddleClas/deploy/cpp_shitu/tools/build.sh b/exprements/PaddleClas/deploy/cpp_shitu/tools/build.sh new file mode 100644 index 00000000..a9e2f922 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/tools/build.sh @@ -0,0 +1,25 @@ +OPENCV_DIR=${opencv_install_dir} +LIB_DIR=${paddle_inference_dir} +CUDA_LIB_DIR=/usr/local/cuda/lib64 +CUDNN_LIB_DIR=/usr/lib/x86_64-linux-gnu/ +FAISS_DIR=${faiss_install_dir} +FAISS_WITH_MKL=OFF + +BUILD_DIR=build +rm -rf ${BUILD_DIR} +mkdir ${BUILD_DIR} +cd ${BUILD_DIR} +cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DWITH_GPU=OFF \ + -DWITH_STATIC_LIB=OFF \ + -DUSE_TENSORRT=OFF \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + -DFAISS_DIR=${FAISS_DIR} \ + -DFAISS_WITH_MKL=${FAISS_WITH_MKL} + +make -j +cd .. diff --git a/exprements/PaddleClas/deploy/cpp_shitu/tools/transform_id_map.py b/exprements/PaddleClas/deploy/cpp_shitu/tools/transform_id_map.py new file mode 100644 index 00000000..3ad6b6f9 --- /dev/null +++ b/exprements/PaddleClas/deploy/cpp_shitu/tools/transform_id_map.py @@ -0,0 +1,39 @@ +import argparse +import os +import pickle + +import yaml + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument('-c', '--config', type=str, required=True) + args = parser.parse_args() + return args + + +def main(): + args = parse_args() + with open(args.config) as fd: + config = yaml.load(fd.read(), yaml.FullLoader) + index_dir = "" + try: + index_dir = config["IndexProcess"]["index_dir"] + except Exception as e: + print("The IndexProcess.index_dir in config_file dose not exist") + exit(1) + id_map_path = os.path.join(index_dir, "id_map.pkl") + assert os.path.exists( + id_map_path), "The id_map file dose not exist: {}".format(id_map_path) + + with open(id_map_path, "rb") as fd: + ids = pickle.load(fd) + with open(os.path.join(index_dir, "id_map.txt"), "w") as fd: + for k, v in ids.items(): + v = v.split("\t")[1] + fd.write(str(k) + " " + v + "\n") + print('Transform id_map sucess') + + +if __name__ == "__main__": + main() diff --git a/exprements/PaddleClas/deploy/hubserving/ILSVRC2012_val_00006666.JPEG b/exprements/PaddleClas/deploy/hubserving/ILSVRC2012_val_00006666.JPEG new file mode 100644 index 00000000..f2845224 Binary files /dev/null and b/exprements/PaddleClas/deploy/hubserving/ILSVRC2012_val_00006666.JPEG differ diff --git a/exprements/PaddleClas/deploy/hubserving/clas/__init__.py b/exprements/PaddleClas/deploy/hubserving/clas/__init__.py new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/deploy/hubserving/clas/config.json b/exprements/PaddleClas/deploy/hubserving/clas/config.json new file mode 100644 index 00000000..647036f6 --- /dev/null +++ b/exprements/PaddleClas/deploy/hubserving/clas/config.json @@ -0,0 +1,16 @@ +{ + "modules_info": { + "clas_system": { + "init_args": { + "version": "1.0.0", + "use_gpu": true, + "enable_mkldnn": false + }, + "predict_args": { + } + } + }, + "port": 8866, + "use_multiprocess": false, + "workers": 2 +} diff --git a/exprements/PaddleClas/deploy/hubserving/clas/module.py b/exprements/PaddleClas/deploy/hubserving/clas/module.py new file mode 100644 index 00000000..98ec1d90 --- /dev/null +++ b/exprements/PaddleClas/deploy/hubserving/clas/module.py @@ -0,0 +1,100 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import sys +sys.path.insert(0, ".") + +import time + +import numpy as np +import paddle.nn as nn +from paddlehub.module.module import moduleinfo, serving + +from hubserving.clas.params import get_default_confg +from python.predict_cls import ClsPredictor +from utils import config +from utils.encode_decode import b64_to_np + + +@moduleinfo( + name="clas_system", + version="1.0.0", + summary="class system service", + author="paddle-dev", + author_email="paddle-dev@baidu.com", + type="cv/class") +class ClasSystem(nn.Layer): + def __init__(self, use_gpu=None, enable_mkldnn=None): + """ + initialize with the necessary elements + """ + self._config = self._load_config( + use_gpu=use_gpu, enable_mkldnn=enable_mkldnn) + self.cls_predictor = ClsPredictor(self._config) + + def _load_config(self, use_gpu=None, enable_mkldnn=None): + cfg = get_default_confg() + cfg = config.AttrDict(cfg) + config.create_attr_dict(cfg) + if use_gpu is not None: + cfg.Global.use_gpu = use_gpu + if enable_mkldnn is not None: + cfg.Global.enable_mkldnn = enable_mkldnn + cfg.enable_benchmark = False + if cfg.Global.use_gpu: + try: + _places = os.environ["CUDA_VISIBLE_DEVICES"] + int(_places[0]) + print("Use GPU, GPU Memery:{}".format(cfg.Global.gpu_mem)) + print("CUDA_VISIBLE_DEVICES: ", _places) + except: + raise RuntimeError( + "Environment Variable CUDA_VISIBLE_DEVICES is not set correctly. If you wanna use gpu, please set CUDA_VISIBLE_DEVICES via export CUDA_VISIBLE_DEVICES=cuda_device_id." + ) + else: + print("Use CPU") + print("Enable MKL-DNN") if enable_mkldnn else None + return cfg + + def predict(self, inputs): + if not isinstance(inputs, list): + raise Exception( + "The input data is inconsistent with expectations.") + + starttime = time.time() + outputs = self.cls_predictor.predict(inputs) + elapse = time.time() - starttime + return {"prediction": outputs, "elapse": elapse} + + @serving + def serving_method(self, images, revert_params): + """ + Run as a service. + """ + input_data = b64_to_np(images, revert_params) + results = self.predict(inputs=list(input_data)) + return results + + +if __name__ == "__main__": + import cv2 + import paddlehub as hub + + module = hub.Module(name="clas_system") + img_path = "./hubserving/ILSVRC2012_val_00006666.JPEG" + img = cv2.imread(img_path)[:, :, ::-1] + img = cv2.resize(img, (224, 224)).transpose((2, 0, 1)) + res = module.predict([img.astype(np.float32)]) + print("The returned result of {}: {}".format(img_path, res)) diff --git a/exprements/PaddleClas/deploy/hubserving/clas/params.py b/exprements/PaddleClas/deploy/hubserving/clas/params.py new file mode 100644 index 00000000..9d9c0dda --- /dev/null +++ b/exprements/PaddleClas/deploy/hubserving/clas/params.py @@ -0,0 +1,42 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + + +def get_default_confg(): + return { + 'Global': { + "inference_model_dir": "../inference/", + "batch_size": 1, + 'use_gpu': False, + 'use_fp16': False, + 'enable_mkldnn': False, + 'cpu_num_threads': 1, + 'use_tensorrt': False, + 'ir_optim': False, + "gpu_mem": 8000, + 'enable_profile': False, + "enable_benchmark": False + }, + 'PostProcess': { + 'main_indicator': 'Topk', + 'Topk': { + 'topk': 5, + 'class_id_map_file': './utils/imagenet1k_label_list.txt' + } + } + } diff --git a/exprements/PaddleClas/deploy/hubserving/readme.md b/exprements/PaddleClas/deploy/hubserving/readme.md new file mode 100644 index 00000000..6b2b2dd4 --- /dev/null +++ b/exprements/PaddleClas/deploy/hubserving/readme.md @@ -0,0 +1,188 @@ +[English](readme_en.md) | 简体中文 + +# 基于PaddleHub Serving的服务部署 + +hubserving服务部署配置服务包`clas`下包含3个必选文件，目录如下： +``` +hubserving/clas/ + └─ __init__.py 空文件，必选 + └─ config.json 配置文件，可选，使用配置启动服务时作为参数传入 + └─ module.py 主模块，必选，包含服务的完整逻辑 + └─ params.py 参数文件，必选，包含模型路径、前后处理参数等参数 +``` + +## 快速启动服务 +### 1. 准备环境 +```shell +# 安装paddlehub,请安装2.0版本 +pip3 install paddlehub==2.1.0 --upgrade -i https://pypi.tuna.tsinghua.edu.cn/simple +``` + +### 2. 下载推理模型 +安装服务模块前，需要准备推理模型并放到正确路径，默认模型路径为： +``` +分类推理模型结构文件：PaddleClas/inference/inference.pdmodel +分类推理模型权重文件：PaddleClas/inference/inference.pdiparams +``` + +**注意**： +* 模型文件路径可在`PaddleClas/deploy/hubserving/clas/params.py`中查看和修改： + ```python + "inference_model_dir": "../inference/" + ``` + 需要注意，模型文件（包括.pdmodel与.pdiparams）名称必须为`inference`。 +* 我们也提供了大量基于ImageNet-1k数据集的预训练模型，模型列表及下载地址详见[模型库概览](../../docs/zh_CN/models/models_intro.md)，也可以使用自己训练转换好的模型。 + +### 3. 安装服务模块 +针对Linux环境和Windows环境，安装命令如下。 + +* 在Linux环境下，安装示例如下： +```shell +cd PaddleClas/deploy +# 安装服务模块： +hub install hubserving/clas/ +``` + +* 在Windows环境下(文件夹的分隔符为`\`)，安装示例如下： + +```shell +cd PaddleClas\deploy +# 安装服务模块： +hub install hubserving\clas\ +``` + +### 4. 启动服务 +#### 方式1. 命令行命令启动（仅支持CPU） +**启动命令：** +```shell +$ hub serving start --modules Module1==Version1 \ + --port XXXX \ + --use_multiprocess \ + --workers \ +``` + +**参数：** +|参数|用途| +|-|-| +|--modules/-m| [**必选**] PaddleHub Serving预安装模型，以多个Module==Version键值对的形式列出
*`当不指定Version时，默认选择最新版本`*| +|--port/-p| [**可选**] 服务端口，默认为8866| +|--use_multiprocess| [**可选**] 是否启用并发方式，默认为单进程方式，推荐多核CPU机器使用此方式
*`Windows操作系统只支持单进程方式`*| +|--workers| [**可选**] 在并发方式下指定的并发任务数，默认为`2*cpu_count-1`，其中`cpu_count`为CPU核数| + +如按默认参数启动服务： ```hub serving start -m clas_system``` + +这样就完成了一个服务化API的部署，使用默认端口号8866。 + +#### 方式2. 配置文件启动（支持CPU、GPU） +**启动命令：** +```hub serving start -c config.json``` + +其中，`config.json`格式如下： +```json +{ + "modules_info": { + "clas_system": { + "init_args": { + "version": "1.0.0", + "use_gpu": true, + "enable_mkldnn": false + }, + "predict_args": { + } + } + }, + "port": 8866, + "use_multiprocess": false, + "workers": 2 +} +``` + +- `init_args`中的可配参数与`module.py`中的`_initialize`函数接口一致。其中， + - 当`use_gpu`为`true`时，表示使用GPU启动服务。 + - 当`enable_mkldnn`为`true`时，表示使用MKL-DNN加速。 +- `predict_args`中的可配参数与`module.py`中的`predict`函数接口一致。 + +**注意:** +- 使用配置文件启动服务时，其他参数会被忽略。 +- 如果使用GPU预测(即，`use_gpu`置为`true`)，则需要在启动服务之前，设置CUDA_VISIBLE_DEVICES环境变量，如：```export CUDA_VISIBLE_DEVICES=0```，否则不用设置。 +- **`use_gpu`不可与`use_multiprocess`同时为`true`**。 +- **`use_gpu`与`enable_mkldnn`同时为`true`时，将忽略`enable_mkldnn`，而使用GPU**。 + +如，使用GPU 3号卡启动串联服务： +```shell +cd PaddleClas/deploy +export CUDA_VISIBLE_DEVICES=3 +hub serving start -c hubserving/clas/config.json +``` + +## 发送预测请求 +配置好服务端，可使用以下命令发送预测请求，获取预测结果: + +```shell +cd PaddleClas/deploy +python hubserving/test_hubserving.py server_url image_path +``` + +需要给脚本传递2个必须参数： +- **server_url**：服务地址，格式为 +`http://[ip_address]:[port]/predict/[module_name]` +- **image_path**：测试图像路径，可以是单张图片路径，也可以是图像集合目录路径。 +- **batch_size**：[**可选**] 以`batch_size`大小为单位进行预测，默认为`1`。 +- **resize_short**：[**可选**] 预处理时，按短边调整大小，默认为`256`。 +- **crop_size**：[**可选**] 预处理时，居中裁剪的大小，默认为`224`。 +- **normalize**：[**可选**] 预处理时，是否进行`normalize`，默认为`True`。 +- **to_chw**：[**可选**] 预处理时，是否调整为`CHW`顺序，默认为`True`。 + +**注意**：如果使用`Transformer`系列模型，如`DeiT_***_384`, `ViT_***_384`等，请注意模型的输入数据尺寸，需要指定`--resize_short=384 --crop_size=384`。 + + +访问示例： +```shell +python hubserving/test_hubserving.py --server_url http://127.0.0.1:8866/predict/clas_system --image_file ./hubserving/ILSVRC2012_val_00006666.JPEG --batch_size 8 +``` + +### 返回结果格式说明 +返回结果为列表（list），包含top-k个分类结果，以及对应的得分，还有此图片预测耗时，具体如下： +``` +list: 返回结果 +└─ list: 第一张图片结果 + └─ list: 前k个分类结果，依score递减排序 + └─ list: 前k个分类结果对应的score，依score递减排序 + └─ float: 该图分类耗时，单位秒 +``` + +**说明：** 如果需要增加、删除、修改返回字段，可对相应模块进行修改，完整流程参考下一节自定义修改服务模块。 + +## 自定义修改服务模块 +如果需要修改服务逻辑，你一般需要操作以下步骤： + +- 1、 停止服务 +```hub serving stop --port/-p XXXX``` + +- 2、 到相应的`module.py`和`params.py`等文件中根据实际需求修改代码。`module.py`修改后需要重新安装（`hub install hubserving/clas/`）并部署。在进行部署前，可通过`python hubserving/clas/module.py`测试已安装服务模块。 + +- 3、 卸载旧服务包 +```hub uninstall clas_system``` + +- 4、 安装修改后的新服务包 +```hub install hubserving/clas/``` + +- 5、重新启动服务 +```hub serving start -m clas_system``` + +**注意**： +常用参数可在[params.py](./clas/params.py)中修改： + * 更换模型，需要修改模型文件路径参数: + ```python + "inference_model_dir": + ``` + * 更改后处理时返回的`top-k`结果数量： + ```python + 'topk': + ``` + * 更改后处理时的lable与class id对应映射文件： + ```python + 'class_id_map_file': + ``` + +为了避免不必要的延时以及能够以batch_size进行预测，数据预处理逻辑（包括resize、crop等操作）在客户端完成，因此需要在[test_hubserving.py](./test_hubserving.py#L35-L52)中修改。 diff --git a/exprements/PaddleClas/deploy/hubserving/readme_en.md b/exprements/PaddleClas/deploy/hubserving/readme_en.md new file mode 100644 index 00000000..bb0ddbd2 --- /dev/null +++ b/exprements/PaddleClas/deploy/hubserving/readme_en.md @@ -0,0 +1,199 @@ +English | [简体中文](readme.md) + +# Service deployment based on PaddleHub Serving + +HubServing service pack contains 3 files, the directory is as follows: +``` +hubserving/clas/ + └─ __init__.py Empty file, required + └─ config.json Configuration file, optional, passed in as a parameter when using configuration to start the service + └─ module.py Main module file, required, contains the complete logic of the service + └─ params.py Parameter file, required, including parameters such as model path, pre- and post-processing parameters +``` + +## Quick start service +### 1. Prepare the environment +```shell +# Install version 2.0 of PaddleHub +pip3 install paddlehub==2.1.0 --upgrade -i https://pypi.tuna.tsinghua.edu.cn/simple +``` + +### 2. Download inference model +Before installing the service module, you need to prepare the inference model and put it in the correct path. The default model path is: + +``` +Model structure file: PaddleClas/inference/inference.pdmodel +Model parameters file: PaddleClas/inference/inference.pdiparams +``` + +* The model file path can be viewed and modified in `PaddleClas/deploy/hubserving/clas/params.py`. + + It should be noted that the prefix of model structure file and model parameters file must be `inference`. + +* More models provided by PaddleClas can be obtained from the [model library](../../docs/en/models/models_intro_en.md). You can also use models trained by yourself. + +### 3. Install Service Module + +* On Linux platform, the examples are as follows. +```shell +cd PaddleClas/deploy +hub install hubserving/clas/ +``` + +* On Windows platform, the examples are as follows. +```shell +cd PaddleClas\deploy +hub install hubserving\clas\ +``` + +### 4. Start service +#### Way 1. Start with command line parameters (CPU only) + +**start command：** +```shell +$ hub serving start --modules Module1==Version1 \ + --port XXXX \ + --use_multiprocess \ + --workers \ +``` +**parameters：** + +|parameters|usage| +|-|-| +|--modules/-m|PaddleHub Serving pre-installed model, listed in the form of multiple Module==Version key-value pairs
*`When Version is not specified, the latest version is selected by default`*| +|--port/-p|Service port, default is 8866| +|--use_multiprocess|Enable concurrent mode, the default is single-process mode, this mode is recommended for multi-core CPU machines
*`Windows operating system only supports single-process mode`*| +|--workers|The number of concurrent tasks specified in concurrent mode, the default is `2*cpu_count-1`, where `cpu_count` is the number of CPU cores| + +For example, start the 2-stage series service: +```shell +hub serving start -m clas_system +``` + +This completes the deployment of a service API, using the default port number 8866. + +#### Way 2. Start with configuration file（CPU、GPU） +**start command：** +```shell +hub serving start --config/-c config.json +``` +Wherein, the format of `config.json` is as follows: +```json +{ + "modules_info": { + "clas_system": { + "init_args": { + "version": "1.0.0", + "use_gpu": true, + "enable_mkldnn": false + }, + "predict_args": { + } + } + }, + "port": 8866, + "use_multiprocess": false, + "workers": 2 +} +``` +- The configurable parameters in `init_args` are consistent with the `_initialize` function interface in `module.py`. Among them, + - when `use_gpu` is `true`, it means that the GPU is used to start the service. + - when `enable_mkldnn` is `true`, it means that use MKL-DNN to accelerate. +- The configurable parameters in `predict_args` are consistent with the `predict` function interface in `module.py`. + +**Note:** +- When using the configuration file to start the service, other parameters will be ignored. +- If you use GPU prediction (that is, `use_gpu` is set to `true`), you need to set the environment variable CUDA_VISIBLE_DEVICES before starting the service, such as: ```export CUDA_VISIBLE_DEVICES=0```, otherwise you do not need to set it. +- **`use_gpu` and `use_multiprocess` cannot be `true` at the same time.** +- **When both `use_gpu` and `enable_mkldnn` are set to `true` at the same time, GPU is used to run and `enable_mkldnn` will be ignored.** + +For example, use GPU card No. 3 to start the 2-stage series service: +```shell +cd PaddleClas/deploy +export CUDA_VISIBLE_DEVICES=3 +hub serving start -c hubserving/clas/config.json +``` + +## Send prediction requests +After the service starts, you can use the following command to send a prediction request to obtain the prediction result: +```shell +cd PaddleClas/deploy +python hubserving/test_hubserving.py server_url image_path +``` + +Two required parameters need to be passed to the script: +- **server_url**: service address，format of which is +`http://[ip_address]:[port]/predict/[module_name]` +- **image_path**: Test image path, can be a single image path or an image directory path +- **batch_size**: [**Optional**] batch_size. Default by `1`. +- **resize_short**: [**Optional**] In preprocessing, resize according to short size. Default by `256`。 +- **crop_size**: [**Optional**] In preprocessing, centor crop size. Default by `224`。 +- **normalize**: [**Optional**] In preprocessing, whether to do `normalize`. Default by `True`。 +- **to_chw**: [**Optional**] In preprocessing, whether to transpose to `CHW`. Default by `True`。 + +**Notice**: +If you want to use `Transformer series models`, such as `DeiT_***_384`, `ViT_***_384`, etc., please pay attention to the input size of model, and need to set `--resize_short=384`, `--crop_size=384`. + +**Eg.** +```shell +python hubserving/test_hubserving.py --server_url http://127.0.0.1:8866/predict/clas_system --image_file ./hubserving/ILSVRC2012_val_00006666.JPEG --batch_size 8 +``` + +### Returned result format +The returned result is a list, including the `top_k`'s classification results, corresponding scores and the time cost of prediction, details as follows. + +``` +list: The returned results +└─ list: The result of first picture + └─ list: The top-k classification results, sorted in descending order of score + └─ list: The scores corresponding to the top-k classification results, sorted in descending order of score + └─ float: The time cost of predicting the picture, unit second +``` + +**Note：** If you need to add, delete or modify the returned fields, you can modify the corresponding module. For the details, refer to the user-defined modification service module in the next section. + +## User defined service module modification +If you need to modify the service logic, the following steps are generally required: + +1. Stop service +```shell +hub serving stop --port/-p XXXX +``` + +2. Modify the code in the corresponding files, like `module.py` and `params.py`, according to the actual needs. You need re-install(hub install hubserving/clas/) and re-deploy after modifing `module.py`. +After modifying and installing and before deploying, you can use `python hubserving/clas/module.py` to test the installed service module. + +For example, if you need to replace the model used by the deployed service, you need to modify model path parameters `cfg.model_file` and `cfg.params_file` in `params.py`. Of course, other related parameters may need to be modified at the same time. Please modify and debug according to the actual situation. + +3. Uninstall old service module +```shell +hub uninstall clas_system +``` + +4. Install modified service module +```shell +hub install hubserving/clas/ +``` + +5. Restart service +```shell +hub serving start -m clas_system +``` + +**Note**: + +Common parameters can be modified in params.py: +* Directory of model files(include model structure file and model parameters file): + ```python + "inference_model_dir": + ``` +* The number of Top-k results returned during post-processing: + ```python + 'topk': + ``` +* Mapping file corresponding to label and class ID during post-processing: + ```python + 'class_id_map_file': + ``` + +In order to avoid unnecessary delay and be able to predict in batch, the preprocessing (include resize, crop and other) is completed in the client, so modify [test_hubserving.py](./test_hubserving.py#L35-L52) if necessary. diff --git a/exprements/PaddleClas/deploy/hubserving/test_hubserving.py b/exprements/PaddleClas/deploy/hubserving/test_hubserving.py new file mode 100644 index 00000000..b778a828 --- /dev/null +++ b/exprements/PaddleClas/deploy/hubserving/test_hubserving.py @@ -0,0 +1,166 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +import time +import requests +import json +import base64 +import argparse + +import numpy as np +import cv2 + +from utils import logger +from utils.get_image_list import get_image_list +from utils import config +from utils.encode_decode import np_to_b64 +from python.preprocess import create_operators + + +def get_args(): + def str2bool(v): + return v.lower() in ("true", "t", "1") + + parser = argparse.ArgumentParser() + parser.add_argument("--server_url", type=str) + parser.add_argument("--image_file", type=str) + parser.add_argument("--batch_size", type=int, default=1) + parser.add_argument("--resize_short", type=int, default=256) + parser.add_argument("--crop_size", type=int, default=224) + parser.add_argument("--normalize", type=str2bool, default=True) + parser.add_argument("--to_chw", type=str2bool, default=True) + return parser.parse_args() + + +class PreprocessConfig(object): + def __init__(self, + resize_short=256, + crop_size=224, + normalize=True, + to_chw=True): + self.config = [{ + 'ResizeImage': { + 'resize_short': resize_short + } + }, { + 'CropImage': { + 'size': crop_size + } + }] + if normalize: + self.config.append({ + 'NormalizeImage': { + 'scale': 0.00392157, + 'mean': [0.485, 0.456, 0.406], + 'std': [0.229, 0.224, 0.225], + 'order': '' + } + }) + if to_chw: + self.config.append({'ToCHWImage': None}) + + def __call__(self): + return self.config + + +def main(args): + image_path_list = get_image_list(args.image_file) + headers = {"Content-type": "application/json"} + preprocess_ops = create_operators( + PreprocessConfig(args.resize_short, args.crop_size, args.normalize, + args.to_chw)()) + + cnt = 0 + predict_time = 0 + all_score = 0.0 + start_time = time.time() + + img_data_list = [] + img_name_list = [] + cnt = 0 + for idx, img_path in enumerate(image_path_list): + img = cv2.imread(img_path) + if img is None: + logger.warning( + f"Image file failed to read and has been skipped. The path: {img_path}" + ) + continue + else: + img = img[:, :, ::-1] + for ops in preprocess_ops: + img = ops(img) + img = np.array(img) + img_data_list.append(img) + + img_name = img_path.split('/')[-1] + img_name_list.append(img_name) + cnt += 1 + if cnt % args.batch_size == 0 or (idx + 1) == len(image_path_list): + inputs = np.array(img_data_list) + b64str, revert_shape = np_to_b64(inputs) + data = { + "images": b64str, + "revert_params": { + "shape": revert_shape, + "dtype": str(inputs.dtype) + } + } + try: + r = requests.post( + url=args.server_url, + headers=headers, + data=json.dumps(data)) + r.raise_for_status + if r.json()["status"] != "000": + msg = r.json()["msg"] + raise Exception(msg) + except Exception as e: + logger.error(f"{e}, in file(s): {img_name_list[0]} etc.") + continue + else: + results = r.json()["results"] + preds = results["prediction"] + elapse = results["elapse"] + + cnt += len(preds) + predict_time += elapse + + for number, result_list in enumerate(preds): + all_score += result_list["scores"][0] + pred_str = ", ".join( + [f"{k}: {result_list[k]}" for k in result_list]) + logger.info( + f"File:{img_name_list[number]}, The result(s): {pred_str}" + ) + + finally: + img_data_list = [] + img_name_list = [] + + total_time = time.time() - start_time + logger.info("The average time of prediction cost: {:.3f} s/image".format( + predict_time / cnt)) + logger.info("The average time cost: {:.3f} s/image".format(total_time / + cnt)) + logger.info("The average top-1 score: {:.3f}".format(all_score / cnt)) + + +if __name__ == '__main__': + args = get_args() + main(args) diff --git a/exprements/PaddleClas/deploy/images/0517_2715693311.jpg b/exprements/PaddleClas/deploy/images/0517_2715693311.jpg new file mode 100644 index 00000000..bd9d2f63 Binary files /dev/null and b/exprements/PaddleClas/deploy/images/0517_2715693311.jpg differ diff --git a/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00000010.jpeg b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00000010.jpeg new file mode 100644 index 00000000..6fcafb4d Binary files /dev/null and b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00000010.jpeg differ diff --git a/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00010010.jpeg b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00010010.jpeg new file mode 100644 index 00000000..4908ac44 Binary files /dev/null and b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00010010.jpeg differ diff --git a/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00020010.jpeg b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00020010.jpeg new file mode 100644 index 00000000..9c6f8779 Binary files /dev/null and b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00020010.jpeg differ diff --git a/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00030010.jpeg b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00030010.jpeg new file mode 100644 index 00000000..0a073fe4 Binary files /dev/null and b/exprements/PaddleClas/deploy/images/ILSVRC2012_val_00030010.jpeg differ diff --git a/exprements/PaddleClas/deploy/images/anmuxi.jpg b/exprements/PaddleClas/deploy/images/anmuxi.jpg new file mode 100644 index 00000000..735ea5a4 Binary files /dev/null and b/exprements/PaddleClas/deploy/images/anmuxi.jpg differ diff --git a/exprements/PaddleClas/deploy/images/coco_000000570688.jpg b/exprements/PaddleClas/deploy/images/coco_000000570688.jpg new file mode 100644 index 00000000..cb304bd5 Binary files /dev/null and b/exprements/PaddleClas/deploy/images/coco_000000570688.jpg differ diff --git a/exprements/PaddleClas/deploy/images/wangzai.jpg b/exprements/PaddleClas/deploy/images/wangzai.jpg new file mode 100644 index 00000000..fe7aa77d Binary files /dev/null and b/exprements/PaddleClas/deploy/images/wangzai.jpg differ diff --git a/exprements/PaddleClas/deploy/lite/Makefile b/exprements/PaddleClas/deploy/lite/Makefile new file mode 100644 index 00000000..f18864aa --- /dev/null +++ b/exprements/PaddleClas/deploy/lite/Makefile @@ -0,0 +1,68 @@ +ARM_ABI = arm8 +export ARM_ABI + +include ../Makefile.def + +LITE_ROOT=../../../ + +THIRD_PARTY_DIR=${LITE_ROOT}/third_party + +OPENCV_VERSION=opencv4.1.0 + +ifeq (${ARM_ABI}, arm8) + ARM_PATH=arm64-v8a +endif +ifeq (${ARM_ABI}, arm7) + ARM_PATH=armeabi-v7a +endif + +OPENCV_LIBS = ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/libs/libopencv_imgcodecs.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/libs/libopencv_imgproc.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/libs/libopencv_core.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/libtegra_hal.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/liblibjpeg-turbo.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/liblibwebp.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/liblibpng.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/liblibjasper.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/liblibtiff.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/libIlmImf.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/libtbb.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/libcpufeatures.a + +OPENCV_INCLUDE = -I../../../third_party/${OPENCV_VERSION}/${ARM_PATH}/include + +CXX_INCLUDES = $(INCLUDES) ${OPENCV_INCLUDE} -I$(LITE_ROOT)/cxx/include + +CXX_LIBS = ${OPENCV_LIBS} -L$(LITE_ROOT)/cxx/lib/ -lpaddle_light_api_shared $(SYSTEM_LIBS) + +############################################################### +# How to use one of static libaray: # +# `libpaddle_api_full_bundled.a` # +# `libpaddle_api_light_bundled.a` # +############################################################### +# Note: default use lite's shared library. # +############################################################### +# 1. Comment above line using `libpaddle_light_api_shared.so` +# 2. Undo comment below line using `libpaddle_api_light_bundled.a` + +#CXX_LIBS = $(LITE_ROOT)/cxx/lib/libpaddle_api_light_bundled.a $(SYSTEM_LIBS) + +clas_system: fetch_opencv clas_system.o + $(CC) $(SYSROOT_LINK) $(CXXFLAGS_LINK) clas_system.o -o clas_system $(CXX_LIBS) $(LDFLAGS) + +clas_system.o: image_classfication.cpp + $(CC) $(SYSROOT_COMPLILE) $(CXX_DEFINES) $(CXX_INCLUDES) $(CXX_FLAGS) -o clas_system.o -c image_classfication.cpp + +fetch_opencv: + @ test -d ${THIRD_PARTY_DIR} || mkdir ${THIRD_PARTY_DIR} + @ test -e ${THIRD_PARTY_DIR}/${OPENCV_VERSION}.tar.gz || \ + (echo "fetch opencv libs" && \ + wget -P ${THIRD_PARTY_DIR} https://paddle-inference-dist.bj.bcebos.com/${OPENCV_VERSION}.tar.gz) + @ test -d ${THIRD_PARTY_DIR}/${OPENCV_VERSION} || \ + tar -zxvf ${THIRD_PARTY_DIR}/${OPENCV_VERSION}.tar.gz -C ${THIRD_PARTY_DIR} + + +.PHONY: clean +clean: + rm -f clas_system.o + rm -f clas_system diff --git a/exprements/PaddleClas/deploy/lite/benchmark/benchmark.sh b/exprements/PaddleClas/deploy/lite/benchmark/benchmark.sh new file mode 100644 index 00000000..591331e4 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite/benchmark/benchmark.sh @@ -0,0 +1,73 @@ +#!/bin/bash +# ref1: https://github.com/PaddlePaddle/Paddle-Lite/blob/58b2d7dd89/lite/api/benchmark.cc +# ref2: https://paddle-inference-dist.bj.bcebos.com/PaddleLite/benchmark_0/benchmark.sh + +set -e + +# Check input +if [ $# -lt 3 ]; +then + echo "Input error" + echo "Usage:" + echo " sh benchmark.sh " + echo " sh benchmark.sh " + exit +fi + +# Set benchmark params +ANDROID_DIR=/data/local/tmp +BENCHMARK_BIN=$1 +MODELS_DIR=$2 +RESULT_FILENAME=$3 + +WARMUP=10 +REPEATS=30 +IS_RUN_MODEL_OPTIMIZE=false +IS_RUN_QUANTIZED_MODEL=false +NUM_THREADS_LIST=(1 2 4) +MODELS_LIST=$(ls $MODELS_DIR) + +# Check input +if [ $# -gt 3 ]; +then + IS_RUN_MODEL_OPTIMIZE=$4 +fi + +# Adb push benchmark_bin, models +adb push $BENCHMARK_BIN $ANDROID_DIR/benchmark_bin +adb shell chmod +x $ANDROID_DIR/benchmark_bin +adb push $MODELS_DIR $ANDROID_DIR + +# Run benchmark +adb shell "echo 'PaddleLite Benchmark' > $ANDROID_DIR/$RESULT_FILENAME" +for threads in ${NUM_THREADS_LIST[@]}; do + adb shell "echo Threads=$threads Warmup=$WARMUP Repeats=$REPEATS >> $ANDROID_DIR/$RESULT_FILENAME" + for model_name in ${MODELS_LIST[@]}; do + echo "Model=$model_name Threads=$threads" + if [ "$IS_RUN_MODEL_OPTIMIZE" = true ]; + then + adb shell "$ANDROID_DIR/benchmark_bin \ + --model_dir=$ANDROID_DIR/${MODELS_DIR}/$model_name \ + --model_filename=model \ + --param_filename=params \ + --warmup=$WARMUP \ + --repeats=$REPEATS \ + --threads=$threads \ + --result_filename=$ANDROID_DIR/$RESULT_FILENAME" + else + adb shell "$ANDROID_DIR/benchmark_bin \ + --optimized_model_path=$ANDROID_DIR/${MODELS_DIR}/$model_name \ + --warmup=$WARMUP \ + --repeats=$REPEATS \ + --threads=$threads \ + --result_filename=$ANDROID_DIR/$RESULT_FILENAME" + fi + done + adb shell "echo >> $ANDROID_DIR/$RESULT_FILENAME" +done + +# Adb pull benchmark result, show result +adb pull $ANDROID_DIR/$RESULT_FILENAME . +echo "\n--------------------------------------" +cat $RESULT_FILENAME +echo "--------------------------------------" diff --git a/exprements/PaddleClas/deploy/lite/config.txt b/exprements/PaddleClas/deploy/lite/config.txt new file mode 100644 index 00000000..08cee3db --- /dev/null +++ b/exprements/PaddleClas/deploy/lite/config.txt @@ -0,0 +1,6 @@ +clas_model_file ./MobileNetV3_large_x1_0.nb +label_path ./imagenet1k_label_list.txt +resize_short_size 256 +crop_size 224 +visualize 0 +enable_benchmark 0 diff --git a/exprements/PaddleClas/deploy/lite/image_classfication.cpp b/exprements/PaddleClas/deploy/lite/image_classfication.cpp new file mode 100644 index 00000000..535f3c86 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite/image_classfication.cpp @@ -0,0 +1,344 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "paddle_api.h" // NOLINT +#include +#include +#include +#include +#include +#include +#include +#include + +using namespace paddle::lite_api; // NOLINT +using namespace std; + +struct RESULT { + std::string class_name; + int class_id; + float score; +}; + +std::vector PostProcess(const float *output_data, int output_size, + const std::vector &word_labels, + cv::Mat &output_image) { + const int TOPK = 5; + int max_indices[TOPK]; + double max_scores[TOPK]; + for (int i = 0; i < TOPK; i++) { + max_indices[i] = 0; + max_scores[i] = 0; + } + for (int i = 0; i < output_size; i++) { + float score = output_data[i]; + int index = i; + for (int j = 0; j < TOPK; j++) { + if (score > max_scores[j]) { + index += max_indices[j]; + max_indices[j] = index - max_indices[j]; + index -= max_indices[j]; + score += max_scores[j]; + max_scores[j] = score - max_scores[j]; + score -= max_scores[j]; + } + } + } + + std::vector results(TOPK); + for (int i = 0; i < results.size(); i++) { + results[i].class_name = "Unknown"; + if (max_indices[i] >= 0 && max_indices[i] < word_labels.size()) { + results[i].class_name = word_labels[max_indices[i]]; + } + results[i].score = max_scores[i]; + results[i].class_id = max_indices[i]; + cv::putText(output_image, + "Top" + std::to_string(i + 1) + "." + results[i].class_name + + ":" + std::to_string(results[i].score), + cv::Point2d(5, i * 18 + 20), cv::FONT_HERSHEY_PLAIN, 1, + cv::Scalar(51, 255, 255)); + } + return results; +} + +// fill tensor with mean and scale and trans layout: nhwc -> nchw, neon speed up +void NeonMeanScale(const float *din, float *dout, int size, + const std::vector mean, + const std::vector scale) { + if (mean.size() != 3 || scale.size() != 3) { + std::cerr << "[ERROR] mean or scale size must equal to 3\n"; + exit(1); + } + float32x4_t vmean0 = vdupq_n_f32(mean[0]); + float32x4_t vmean1 = vdupq_n_f32(mean[1]); + float32x4_t vmean2 = vdupq_n_f32(mean[2]); + float32x4_t vscale0 = vdupq_n_f32(scale[0]); + float32x4_t vscale1 = vdupq_n_f32(scale[1]); + float32x4_t vscale2 = vdupq_n_f32(scale[2]); + + float *dout_c0 = dout; + float *dout_c1 = dout + size; + float *dout_c2 = dout + size * 2; + + int i = 0; + for (; i < size - 3; i += 4) { + float32x4x3_t vin3 = vld3q_f32(din); + float32x4_t vsub0 = vsubq_f32(vin3.val[0], vmean0); + float32x4_t vsub1 = vsubq_f32(vin3.val[1], vmean1); + float32x4_t vsub2 = vsubq_f32(vin3.val[2], vmean2); + float32x4_t vs0 = vmulq_f32(vsub0, vscale0); + float32x4_t vs1 = vmulq_f32(vsub1, vscale1); + float32x4_t vs2 = vmulq_f32(vsub2, vscale2); + vst1q_f32(dout_c0, vs0); + vst1q_f32(dout_c1, vs1); + vst1q_f32(dout_c2, vs2); + + din += 12; + dout_c0 += 4; + dout_c1 += 4; + dout_c2 += 4; + } + for (; i < size; i++) { + *(dout_c0++) = (*(din++) - mean[0]) * scale[0]; + *(dout_c1++) = (*(din++) - mean[1]) * scale[1]; + *(dout_c2++) = (*(din++) - mean[2]) * scale[2]; + } +} + +cv::Mat ResizeImage(const cv::Mat &img, const int &resize_short_size) { + int w = img.cols; + int h = img.rows; + + cv::Mat resize_img; + + float ratio = 1.f; + if (h < w) { + ratio = float(resize_short_size) / float(h); + } else { + ratio = float(resize_short_size) / float(w); + } + int resize_h = round(float(h) * ratio); + int resize_w = round(float(w) * ratio); + + cv::resize(img, resize_img, cv::Size(resize_w, resize_h)); + return resize_img; +} + +cv::Mat CenterCropImg(const cv::Mat &img, const int &crop_size) { + int resize_w = img.cols; + int resize_h = img.rows; + int w_start = int((resize_w - crop_size) / 2); + int h_start = int((resize_h - crop_size) / 2); + cv::Rect rect(w_start, h_start, crop_size, crop_size); + cv::Mat crop_img = img(rect); + return crop_img; +} + +std::vector +RunClasModel(std::shared_ptr predictor, const cv::Mat &img, + const std::map &config, + const std::vector &word_labels, double &cost_time) { + // Read img + int resize_short_size = stoi(config.at("resize_short_size")); + int crop_size = stoi(config.at("crop_size")); + int visualize = stoi(config.at("visualize")); + + cv::Mat resize_image = ResizeImage(img, resize_short_size); + + cv::Mat crop_image = CenterCropImg(resize_image, crop_size); + + cv::Mat img_fp; + double e = 1.0 / 255.0; + crop_image.convertTo(img_fp, CV_32FC3, e); + + // Prepare input data from image + std::unique_ptr input_tensor(std::move(predictor->GetInput(0))); + input_tensor->Resize({1, 3, img_fp.rows, img_fp.cols}); + auto *data0 = input_tensor->mutable_data(); + + std::vector mean = {0.485f, 0.456f, 0.406f}; + std::vector scale = {1 / 0.229f, 1 / 0.224f, 1 / 0.225f}; + const float *dimg = reinterpret_cast(img_fp.data); + NeonMeanScale(dimg, data0, img_fp.rows * img_fp.cols, mean, scale); + + auto start = std::chrono::system_clock::now(); + // Run predictor + predictor->Run(); + + // Get output and post process + std::unique_ptr output_tensor( + std::move(predictor->GetOutput(0))); + auto *output_data = output_tensor->data(); + auto end = std::chrono::system_clock::now(); + auto duration = + std::chrono::duration_cast(end - start); + cost_time = double(duration.count()) * + std::chrono::microseconds::period::num / + std::chrono::microseconds::period::den; + + int output_size = 1; + for (auto dim : output_tensor->shape()) { + output_size *= dim; + } + + cv::Mat output_image; + auto results = + PostProcess(output_data, output_size, word_labels, output_image); + + if (visualize) { + std::string output_image_path = "./clas_result.png"; + cv::imwrite(output_image_path, output_image); + std::cout << "save output image into " << output_image_path << std::endl; + } + + return results; +} + +std::shared_ptr LoadModel(std::string model_file) { + MobileConfig config; + config.set_model_from_file(model_file); + + std::shared_ptr predictor = + CreatePaddlePredictor(config); + return predictor; +} + +std::vector split(const std::string &str, + const std::string &delim) { + std::vector res; + if ("" == str) + return res; + char *strs = new char[str.length() + 1]; + std::strcpy(strs, str.c_str()); + + char *d = new char[delim.length() + 1]; + std::strcpy(d, delim.c_str()); + + char *p = std::strtok(strs, d); + while (p) { + string s = p; + res.push_back(s); + p = std::strtok(NULL, d); + } + + return res; +} + +std::vector ReadDict(std::string path) { + std::ifstream in(path); + std::string filename; + std::string line; + std::vector m_vec; + if (in) { + while (getline(in, line)) { + m_vec.push_back(line); + } + } else { + std::cout << "no such file" << std::endl; + } + return m_vec; +} + +std::map LoadConfigTxt(std::string config_path) { + auto config = ReadDict(config_path); + + std::map dict; + for (int i = 0; i < config.size(); i++) { + std::vector res = split(config[i], " "); + dict[res[0]] = res[1]; + } + return dict; +} + +void PrintConfig(const std::map &config) { + std::cout << "=======PaddleClas lite demo config======" << std::endl; + for (auto iter = config.begin(); iter != config.end(); iter++) { + std::cout << iter->first << " : " << iter->second << std::endl; + } + std::cout << "=======End of PaddleClas lite demo config======" << std::endl; +} + +std::vector LoadLabels(const std::string &path) { + std::ifstream file; + std::vector labels; + file.open(path); + while (file) { + std::string line; + std::getline(file, line); + std::string::size_type pos = line.find(" "); + if (pos != std::string::npos) { + line = line.substr(pos); + } + labels.push_back(line); + } + file.clear(); + file.close(); + return labels; +} + +int main(int argc, char **argv) { + if (argc < 3) { + std::cerr << "[ERROR] usage: " << argv[0] << " config_path img_path\n"; + exit(1); + } + + std::string config_path = argv[1]; + std::string img_path = argv[2]; + + // load config + auto config = LoadConfigTxt(config_path); + PrintConfig(config); + + double elapsed_time = 0.0; + int warmup_iter = 10; + + bool enable_benchmark = bool(stoi(config.at("enable_benchmark"))); + int total_cnt = enable_benchmark ? 1000 : 1; + + std::string clas_model_file = config.at("clas_model_file"); + std::string label_path = config.at("label_path"); + + // Load Labels + std::vector word_labels = LoadLabels(label_path); + + auto clas_predictor = LoadModel(clas_model_file); + for (int j = 0; j < total_cnt; ++j) { + cv::Mat srcimg = cv::imread(img_path, cv::IMREAD_COLOR); + cv::cvtColor(srcimg, srcimg, cv::COLOR_BGR2RGB); + + double run_time = 0; + std::vector results = + RunClasModel(clas_predictor, srcimg, config, word_labels, run_time); + + std::cout << "===clas result for image: " << img_path << "===" << std::endl; + for (int i = 0; i < results.size(); i++) { + std::cout << "\t" + << "Top-" << i + 1 << ", class_id: " << results[i].class_id + << ", class_name: " << results[i].class_name + << ", score: " << results[i].score << std::endl; + } + if (j >= warmup_iter) { + elapsed_time += run_time; + std::cout << "Current image path: " << img_path << std::endl; + std::cout << "Current time cost: " << run_time << " s, " + << "average time cost in all: " + << elapsed_time / (j + 1 - warmup_iter) << " s." << std::endl; + } else { + std::cout << "Current time cost: " << run_time << " s." << std::endl; + } + } + + return 0; +} diff --git a/exprements/PaddleClas/deploy/lite/imgs/lite_demo_result.png b/exprements/PaddleClas/deploy/lite/imgs/lite_demo_result.png new file mode 100644 index 00000000..b778f158 Binary files /dev/null and b/exprements/PaddleClas/deploy/lite/imgs/lite_demo_result.png differ diff --git a/exprements/PaddleClas/deploy/lite/imgs/tabby_cat.jpg b/exprements/PaddleClas/deploy/lite/imgs/tabby_cat.jpg new file mode 100644 index 00000000..6eb5b714 Binary files /dev/null and b/exprements/PaddleClas/deploy/lite/imgs/tabby_cat.jpg differ diff --git a/exprements/PaddleClas/deploy/lite/prepare.sh b/exprements/PaddleClas/deploy/lite/prepare.sh new file mode 100644 index 00000000..ed1f13a1 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite/prepare.sh @@ -0,0 +1,14 @@ +#!/bin/bash + +if [ $# != 1 ] ; then +echo "USAGE: $0 your_inference_lite_lib_path" +exit 1; +fi + +mkdir -p $1/demo/cxx/clas/debug/ +cp ../../ppcls/utils/imagenet1k_label_list.txt $1/demo/cxx/clas/debug/ +cp -r ./* $1/demo/cxx/clas/ +cp ./config.txt $1/demo/cxx/clas/debug/ +cp ./imgs/tabby_cat.jpg $1/demo/cxx/clas/debug/ + +echo "Prepare Done" diff --git a/exprements/PaddleClas/deploy/lite/readme.md b/exprements/PaddleClas/deploy/lite/readme.md new file mode 100644 index 00000000..176e62ce --- /dev/null +++ b/exprements/PaddleClas/deploy/lite/readme.md @@ -0,0 +1,265 @@ +# 端侧部署 + +本教程将介绍基于[Paddle Lite](https://github.com/PaddlePaddle/Paddle-Lite) 在移动端部署PaddleClas分类模型的详细步骤。 + +Paddle Lite是飞桨轻量化推理引擎，为手机、IOT端提供高效推理能力，并广泛整合跨平台硬件，为端侧部署及应用落地问题提供轻量化的部署方案。如果希望直接测试速度，可以参考[Paddle-Lite移动端benchmark测试教程](../../docs/zh_CN/extension/paddle_mobile_inference.md)。 + + +## 1. 准备环境 + +### 运行准备 +- 电脑（编译Paddle Lite） +- 安卓手机（armv7或armv8） + +### 1.1 准备交叉编译环境 +交叉编译环境用于编译 Paddle Lite 和 PaddleClas 的C++ demo。 +支持多种开发环境，不同开发环境的编译流程请参考对应文档。 + +1. [Docker](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#docker) +2. [Linux](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#linux) +3. [MAC OS](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#mac-os) + +### 1.2 准备预测库 + +预测库有两种获取方式： +1. [建议]直接下载，预测库下载链接如下： + |平台|预测库下载链接| + |-|-| + |Android|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv7.gcc.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv8.gcc.c++_static.with_extra.with_cv.tar.gz)| + |iOS|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| + + **注**： + 1. 如果是从 Paddle-Lite [官方文档](https://paddle-lite.readthedocs.io/zh/latest/quick_start/release_lib.html#android-toolchain-gcc)下载的预测库， + 注意选择`with_extra=ON，with_cv=ON`的下载链接。 + 2. 如果使用量化的模型部署在端侧，建议使用Paddle-Lite develop分支编译预测库。 + +2. 编译Paddle-Lite得到预测库，Paddle-Lite的编译方式如下： +```shell +git clone https://github.com/PaddlePaddle/Paddle-Lite.git +cd Paddle-Lite +# 如果使用编译方式，建议使用develop分支编译预测库 +git checkout develop +./lite/tools/build_android.sh --arch=armv8 --with_cv=ON --with_extra=ON +``` + +**注意**：编译Paddle-Lite获得预测库时，需要打开`--with_cv=ON --with_extra=ON`两个选项，`--arch`表示`arm`版本，这里指定为armv8，更多编译命令介绍请参考[链接](https://paddle-lite.readthedocs.io/zh/latest/user_guides/Compile/Android.html#id2)。 + +直接下载预测库并解压后，可以得到`inference_lite_lib.android.armv8/`文件夹，通过编译Paddle-Lite得到的预测库位于`Paddle-Lite/build.lite.android.armv8.gcc/inference_lite_lib.android.armv8/`文件夹下。 +预测库的文件目录如下： + +``` +inference_lite_lib.android.armv8/ +|-- cxx C++ 预测库和头文件 +| |-- include C++ 头文件 +| | |-- paddle_api.h +| | |-- paddle_image_preprocess.h +| | |-- paddle_lite_factory_helper.h +| | |-- paddle_place.h +| | |-- paddle_use_kernels.h +| | |-- paddle_use_ops.h +| | `-- paddle_use_passes.h +| `-- lib C++预测库 +| |-- libpaddle_api_light_bundled.a C++静态库 +| `-- libpaddle_light_api_shared.so C++动态库 +|-- java Java预测库 +| |-- jar +| | `-- PaddlePredictor.jar +| |-- so +| | `-- libpaddle_lite_jni.so +| `-- src +|-- demo C++和Java示例代码 +| |-- cxx C++ 预测库demo +| `-- java Java 预测库demo +``` + +## 2 开始运行 + +### 2.1 模型优化 + +Paddle-Lite 提供了多种策略来自动优化原始的模型，其中包括量化、子图融合、混合调度、Kernel优选等方法，使用Paddle-Lite的`opt`工具可以自动对inference模型进行优化，目前支持两种优化方式，优化后的模型更轻量，模型运行速度更快。 + +**注意**：如果已经准备好了 `.nb` 结尾的模型文件，可以跳过此步骤。 + +#### 2.1.1 [建议]pip安装paddlelite并进行转换 + +Python下安装 `paddlelite`，目前最高支持`Python3.7`。 +**注意**：`paddlelite`whl包版本必须和预测库版本对应。 + +```shell +pip install paddlelite==2.8 +``` + +之后使用`paddle_lite_opt`工具可以进行inference模型的转换。`paddle_lite_opt`的部分参数如下 + +|选项|说明| +|-|-| +|--model_dir|待优化的PaddlePaddle模型（非combined形式）的路径| +|--model_file|待优化的PaddlePaddle模型（combined形式）的网络结构文件路径| +|--param_file|待优化的PaddlePaddle模型（combined形式）的权重文件路径| +|--optimize_out_type|输出模型类型，目前支持两种类型：protobuf和naive_buffer，其中naive_buffer是一种更轻量级的序列化/反序列化实现。若您需要在mobile端执行模型预测，请将此选项设置为naive_buffer。默认为protobuf| +|--optimize_out|优化模型的输出路径| +|--valid_targets|指定模型可执行的backend，默认为arm。目前可支持x86、arm、opencl、npu、xpu，可以同时指定多个backend(以空格分隔)，Model Optimize Tool将会自动选择最佳方式。如果需要支持华为NPU（Kirin 810/990 Soc搭载的达芬奇架构NPU），应当设置为npu, arm| +|--record_tailoring_info|当使用 根据模型裁剪库文件 功能时，则设置该选项为true，以记录优化后模型含有的kernel和OP信息，默认为false| + +`--model_file`表示inference模型的model文件地址，`--param_file`表示inference模型的param文件地址；`optimize_out`用于指定输出文件的名称（不需要添加`.nb`的后缀）。直接在命令行中运行`paddle_lite_opt`，也可以查看所有参数及其说明。 + + +#### 2.1.2 源码编译Paddle-Lite生成opt工具 + +模型优化需要Paddle-Lite的`opt`可执行文件，可以通过编译Paddle-Lite源码获得，编译步骤如下： +```shell +# 如果准备环境时已经clone了Paddle-Lite，则不用重新clone Paddle-Lite +git clone https://github.com/PaddlePaddle/Paddle-Lite.git +cd Paddle-Lite +git checkout develop +# 启动编译 +./lite/tools/build.sh build_optimize_tool +``` + +编译完成后，`opt`文件位于`build.opt/lite/api/`下，可通过如下方式查看`opt`的运行选项和使用方式； +```shell +cd build.opt/lite/api/ +./opt +``` + +`opt`的使用方式与参数与上面的`paddle_lite_opt`完全一致。 + + + +#### 2.1.3 转换示例 + +下面以PaddleClas的 `MobileNetV3_large_x1_0` 模型为例，介绍使用`paddle_lite_opt`完成预训练模型到inference模型，再到Paddle-Lite优化模型的转换。 + +```shell +# 进入PaddleClas根目录 +cd PaddleClas_root_path + +# 下载并解压inference模型 +wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileNetV3_large_x1_0_infer.tar +tar -xf MobileNetV3_large_x1_0_infer.tar + +# 将inference模型转化为Paddle-Lite优化模型 +paddle_lite_opt --model_file=./MobileNetV3_large_x1_0_infer/inference.pdmodel --param_file=./MobileNetV3_large_x1_0_infer/inference.pdiparams --optimize_out=./MobileNetV3_large_x1_0 +``` + +最终在当前文件夹下生成`MobileNetV3_large_x1_0.nb`的文件。 + +**注意**：`--optimize_out` 参数为优化后模型的保存路径，无需加后缀`.nb`；`--model_file` 参数为模型结构信息文件的路径，`--param_file` 参数为模型权重信息文件的路径，请注意文件名。 + + +### 2.2 与手机联调 + +首先需要进行一些准备工作。 +1. 准备一台arm8的安卓手机，如果编译的预测库和opt文件是armv7，则需要arm7的手机，并修改Makefile中`ARM_ABI = arm7`。 +2. 电脑上安装ADB工具，用于调试。 ADB安装方式如下： + + 3.1. MAC电脑安装ADB: + + ```shell + brew cask install android-platform-tools + ``` + 3.2. Linux安装ADB + ```shell + sudo apt update + sudo apt install -y wget adb + ``` + 3.3. Window安装ADB + + win上安装需要去谷歌的安卓平台下载ADB软件包进行安装：[链接](https://developer.android.com/studio) + +4. 手机连接电脑后，开启手机`USB调试`选项，选择`文件传输`模式，在电脑终端中输入： + +```shell +adb devices +``` +如果有device输出，则表示安装成功，如下所示： +``` +List of devices attached +744be294 device +``` + +5. 准备优化后的模型、预测库文件、测试图像和类别映射文件。 + +```shell +cd PaddleClas_root_path +cd deploy/lite/ + +# 运行prepare.sh +# prepare.sh 会将预测库文件、测试图像和使用的字典文件放置在预测库中的demo/cxx/clas文件夹下 +sh prepare.sh /{lite prediction library path}/inference_lite_lib.android.armv8 + +# 进入lite demo的工作目录 +cd /{lite prediction library path}/inference_lite_lib.android.armv8/ +cd demo/cxx/clas/ + +# 将C++预测动态库so文件复制到debug文件夹中 +cp ../../../cxx/lib/libpaddle_light_api_shared.so ./debug/ +``` + +`prepare.sh` 以 `PaddleClas/deploy/lite/imgs/tabby_cat.jpg` 作为测试图像，将测试图像复制到`demo/cxx/clas/debug/` 文件夹下。 +将 `paddle_lite_opt` 工具优化后的模型文件放置到 `/{lite prediction library path}/inference_lite_lib.android.armv8/demo/cxx/clas/debug/` 文件夹下。本例中，使用[2.1.3](#2.1.3)生成的 `MobileNetV3_large_x1_0.nb` 模型文件。 + +执行完成后，clas文件夹下将有如下文件格式： + +``` +demo/cxx/clas/ +|-- debug/ +| |--MobileNetV3_large_x1_0.nb 优化后的分类器模型文件 +| |--tabby_cat.jpg 待测试图像 +| |--imagenet1k_label_list.txt 类别映射文件 +| |--libpaddle_light_api_shared.so C++预测库文件 +| |--config.txt 分类预测超参数配置 +|-- config.txt 分类预测超参数配置 +|-- image_classfication.cpp 图像分类代码文件 +|-- Makefile 编译文件 +``` + +#### 注意： +* 上述文件中，`imagenet1k_label_list.txt` 是ImageNet1k数据集的类别映射文件，如果使用自定义的类别，需要更换该类别映射文件。 + +* `config.txt` 包含了分类器的超参数，如下： + +```shell +clas_model_file ./MobileNetV3_large_x1_0.nb # 模型文件地址 +label_path ./imagenet1k_label_list.txt # 类别映射文本文件 +resize_short_size 256 # resize之后的短边边长 +crop_size 224 # 裁剪后用于预测的边长 +visualize 0 # 是否进行可视化，如果选择的话，会在当前文件夹下生成名为clas_result.png的图像文件。 +``` + +5. 启动调试，上述步骤完成后就可以使用ADB将文件夹 `debug/` push到手机上运行，步骤如下： + +```shell +# 执行编译，得到可执行文件clas_system +make -j + +# 将编译得到的可执行文件移动到debug文件夹中 +mv clas_system ./debug/ + +# 将上述debug文件夹push到手机上 +adb push debug /data/local/tmp/ + +adb shell +cd /data/local/tmp/debug +export LD_LIBRARY_PATH=/data/local/tmp/debug:$LD_LIBRARY_PATH + +# clas_system可执行文件的使用方式为: +# ./clas_system 配置文件路径 测试图像路径 +./clas_system ./config.txt ./tabby_cat.jpg +``` + +如果对代码做了修改，则需要重新编译并push到手机上。 + +运行效果如下： + +
+ +
+ + +## FAQ +Q1：如果想更换模型怎么办，需要重新按照流程走一遍吗？ +A1：如果已经走通了上述步骤，更换模型只需要替换 `.nb` 模型文件即可，同时要注意修改下配置文件中的 `.nb` 文件路径以及类别映射文件（如有必要）。 + +Q2：换一个图测试怎么做？ +A2：替换 debug 下的测试图像为你想要测试的图像，使用 ADB 再次 push 到手机上即可。 diff --git a/exprements/PaddleClas/deploy/lite/readme_en.md b/exprements/PaddleClas/deploy/lite/readme_en.md new file mode 100644 index 00000000..1027c781 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite/readme_en.md @@ -0,0 +1,257 @@ + +# Tutorial of PaddleClas Mobile Deployment + +This tutorial will introduce how to use [Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) to deploy PaddleClas models on mobile phones. + +Paddle-Lite is a lightweight inference engine for PaddlePaddle. It provides efficient inference capabilities for mobile phones and IoTs, and extensively integrates cross-platform hardware to provide lightweight deployment solutions for mobile-side deployment issues. + +If you only want to test speed, please refer to [The tutorial of Paddle-Lite mobile-side benchmark test](../../docs/zh_CN/extension/paddle_mobile_inference.md). + +## 1. Preparation + +- Computer (for compiling Paddle-Lite) +- Mobile phone (arm7 or arm8) + +## 2. Build Paddle-Lite library + +The cross-compilation environment is used to compile the C++ demos of Paddle-Lite and PaddleClas. + +For the detailed compilation directions of different development environments, please refer to the corresponding documents. + +1. [Docker](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#docker) +2. [Linux](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#linux) +3. [macOS](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#mac-os) + +## 3. Download inference library for Android or iOS + +|Platform|Inference Library Download Link| +|-|-| +|Android|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv7.gcc.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv8.gcc.c++_static.with_extra.with_cv.tar.gz)| +|iOS|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| + +**NOTE**: + +1. If you download the inference library from [Paddle-Lite official document](https://paddle-lite.readthedocs.io/zh/latest/quick_start/release_lib.html#android-toolchain-gcc), please choose `with_extra=ON` , `with_cv=ON` . + +2. It is recommended to build inference library using [Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) develop branch if you want to deploy the [quantitative](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/deploy/slim/quantization/README_en.md) model to mobile phones. Please refer to the [link](https://paddle-lite.readthedocs.io/zh/latest/user_guides/Compile/Android.html#id2) for more detailed information about compiling. + + +The structure of the inference library is as follows: + +``` +inference_lite_lib.android.armv8/ +|-- cxx C++ inference library and header files +| |-- include C++ header files +| | |-- paddle_api.h +| | |-- paddle_image_preprocess.h +| | |-- paddle_lite_factory_helper.h +| | |-- paddle_place.h +| | |-- paddle_use_kernels.h +| | |-- paddle_use_ops.h +| | `-- paddle_use_passes.h +| `-- lib C++ inference library +| |-- libpaddle_api_light_bundled.a C++ static library +| `-- libpaddle_light_api_shared.so C++ dynamic library +|-- java Java inference library +| |-- jar +| | `-- PaddlePredictor.jar +| |-- so +| | `-- libpaddle_lite_jni.so +| `-- src +|-- demo C++ and java demos +| |-- cxx C++ demos +| `-- java Java demos +``` + + + +## 4. Inference Model Optimization + +Paddle-Lite provides a variety of strategies to automatically optimize the original training model, including quantization, sub-graph fusion, hybrid scheduling, Kernel optimization and so on. In order to make the optimization process more convenient and easy to use, Paddle-Lite provides `opt` tool to automatically complete the optimization steps and output a lightweight, optimal executable model. + +**NOTE**: If you have already got the `.nb` file, you can skip this step. + + + +### 4.1 [RECOMMEND] Use `pip` to install Paddle-Lite and optimize model + +* Use pip to install Paddle-Lite. The following command uses `pip3.7` . + +```shell +pip install paddlelite==2.8 +``` +**Note**：The version of `paddlelite`'s wheel must match that of inference lib. + +* Use `paddle_lite_opt` to optimize inference model, the parameters of `paddle_lite_opt` are as follows: + +| Parameters | Explanation | +| ----------------------- | ------------------------------------------------------------ | +| --model_dir | Path to the PaddlePaddle model (no-combined) file to be optimized. | +| --model_file | Path to the net structure file of PaddlePaddle model (combined) to be optimized. | +| --param_file | Path to the net weight files of PaddlePaddle model (combined) to be optimized. | +| --optimize_out_type | Type of output model, `protobuf` by default. Supports `protobuf` and `naive_buffer` . Compared with `protobuf`, you can use`naive_buffer` to get a more lightweight serialization/deserialization model. If you need to predict on the mobile-side, please set it to `naive_buffer`. | +| --optimize_out | Path to output model, not needed to add `.nb` suffix. | +| --valid_targets | The executable backend of the model, `arm` by default. Supports one or some of `x86` , `arm` , `opencl` , `npu` , `xpu`. If set more than one, please separate the options by space, and the `opt` tool will choose the best way automatically. If need to support Huawei NPU (DaVinci core carried by Kirin 810/990 SoC), please set it to `npu arm` . | +| --record_tailoring_info | Whether to enable `Cut the Library Files According To the Model` , `false` by default. If need to record kernel and OP infos of optimized model, please set it to `true`. | + +In addition, you can run `paddle_lite_opt` to get more detailed information about how to use. + +### 4.2 Compile Paddle-Lite to generate `opt` tool + +Optimizing model requires Paddle-Lite's `opt` executable file, which can be obtained by compiling the Paddle-Lite. The steps are as follows: + +```shell +# get the Paddle-Lite source code, if have gotten , please skip +git clone https://github.com/PaddlePaddle/Paddle-Lite.git +cd Paddle-Lite +git checkout develop +# compile +./lite/tools/build.sh build_optimize_tool +``` + +After the compilation is complete, the `opt` file is located under `build.opt/lite/api/`. + +`opt` tool is used in the same way as `paddle_lite_opt` , please refer to [4.1](#4.1). + + + +### 4.3 Demo of get the optimized model + +Taking the `MobileNetV3_large_x1_0` model of PaddleClas as an example, we will introduce how to use `paddle_lite_opt` to complete the conversion from the pre-trained model to the inference model, and then to the Paddle-Lite optimized model. + +```shell +# enter PaddleClas root directory +cd PaddleClas_root_path + +# download and uncompress the inference model +wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileNetV3_large_x1_0_infer.tar +tar -xf MobileNetV3_large_x1_0_infer.tar + + +# convert inference model to Paddle-Lite optimized model +paddle_lite_opt --model_file=./MobileNetV3_large_x1_0_infer/inference.pdmodel --param_file=./MobileNetV3_large_x1_0_infer/inference.pdiparams --optimize_out=./MobileNetV3_large_x1_0 +``` + +When the above code command is completed, there will be ``MobileNetV3_large_x1_0.nb` in the current directory, which is the converted model file. + +## 5. Run optimized model on Phone + +1. Prepare an Android phone with `arm8`. If the compiled inference library and `opt` file are `armv7`, you need an `arm7` phone and modify `ARM_ABI = arm7` in the Makefile. + +2. Install the ADB tool on the computer. + + * Install ADB for MAC + + Recommend use homebrew to install. + + ```shell + brew cask install android-platform-tools + ``` + * Install ADB for Linux + + ```shell + sudo apt update + sudo apt install -y wget adb + ``` + * Install ADB for windows + If install ADB fo Windows, you need to download from Google's Android platform: [Download Link](https://developer.android.com/studio). + + First, make sure the phone is connected to the computer, turn on the `USB debugging` option of the phone, and select the `file transfer` mode. Verify whether ADB is installed successfully as follows: + + ```shell + $ adb devices + + List of devices attached + 744be294 device + ``` + + If there is `device` output like the above, it means the installation was successful. + +4. Prepare optimized model, inference library files, test image and dictionary file used. + +```shell +cd PaddleClas_root_path +cd deploy/lite/ + +# prepare.sh will put the inference library files, the test image and the dictionary files in demo/cxx/clas +sh prepare.sh /{lite inference library path}/inference_lite_lib.android.armv8 + +# enter the working directory of lite demo +cd /{lite inference library path}/inference_lite_lib.android.armv8/ +cd demo/cxx/clas/ + +# copy the C++ inference dynamic library file （ie. .so) to the debug folder +cp ../../../cxx/lib/libpaddle_light_api_shared.so ./debug/ +``` + +The `prepare.sh` take `PaddleClas/deploy/lite/imgs/tabby_cat.jpg` as the test image, and copy it to the `demo/cxx/clas/debug/` directory. + +You should put the model that optimized by `paddle_lite_opt` under the `demo/cxx/clas/debug/` directory. In this example, use `MobileNetV3_large_x1_0.nb` model file generated in [2.1.3](#4.3). + +The structure of the clas demo is as follows after the above command is completed: + +``` +demo/cxx/clas/ +|-- debug/ +| |--MobileNetV3_large_x1_0.nb class model +| |--tabby_cat.jpg test image +| |--imagenet1k_label_list.txt dictionary file +| |--libpaddle_light_api_shared.so C++ .so file +| |--config.txt config file +|-- config.txt config file +|-- image_classfication.cpp source code +|-- Makefile compile file +``` + +**NOTE**: + +* `Imagenet1k_label_list.txt` is the category mapping file of the `ImageNet1k` dataset. If use a custom category, you need to replace the category mapping file. +* `config.txt` contains the hyperparameters, as follows: + +```shell +clas_model_file ./MobileNetV3_large_x1_0.nb # path of model file +label_path ./imagenet1k_label_list.txt # path of category mapping file +resize_short_size 256 # the short side length after resize +crop_size 224 # side length used for inference after cropping + +visualize 0 # whether to visualize. If you set it to 1, an image file named 'clas_result.png' will be generated in the current directory. +``` + +5. Run Model on Phone + +```shell +# run compile to get the executable file 'clas_system' +make -j + +# move the compiled executable file to the debug folder +mv clas_system ./debug/ + +# push the debug folder to Phone +adb push debug /data/local/tmp/ + +adb shell +cd /data/local/tmp/debug +export LD_LIBRARY_PATH=/data/local/tmp/debug:$LD_LIBRARY_PATH + +# the usage of clas_system is as follows: +# ./clas_system "path of config file" "path of test image" +./clas_system ./config.txt ./tabby_cat.jpg +``` + +**NOTE**: If you make changes to the code, you need to recompile and repush the `debug ` folder to the phone. + +The result is as follows: + +
+ +
+ + + +## FAQ + +Q1：If I want to change the model, do I need to go through the all process again? +A1：If you have completed the above steps, you only need to replace the `.nb` model file after replacing the model. At the same time, you may need to modify the path of `.nb` file in the config file and change the category mapping file to be compatible the model . + +Q2：How to change the test picture? +A2：Replace the test image under debug folder with the image you want to test，and then repush to the Phone again. diff --git a/exprements/PaddleClas/deploy/lite_shitu/Makefile b/exprements/PaddleClas/deploy/lite_shitu/Makefile new file mode 100644 index 00000000..64ee6aed --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/Makefile @@ -0,0 +1,90 @@ +ARM_ABI = arm8#[arm7/arm8] +export ARM_ABI + +ifeq ($(ARM_ABI), arm8) + ARM_PLAT=arm64-v8a +else + ARM_PLAT=armeabi-v7a +endif +${info ARM_ABI: ${ARM_ABI}} +${info ARM_PLAT: ${ARM_PLAT}; option[arm7/arm8]} + +include ../Makefile.def + +LITE_ROOT=../../../ +${info LITE_ROOT: $(abspath ${LITE_ROOT})} + +THIRD_PARTY_DIR=third_party +${info THIRD_PARTY_DIR: $(abspath ${THIRD_PARTY_DIR})} + + +OPENCV_VERSION=opencv4.1.0 +OPENCV_LIBS = ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/libs/libopencv_imgcodecs.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/libs/libopencv_imgproc.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/libs/libopencv_core.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/libtegra_hal.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/liblibjpeg-turbo.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/liblibwebp.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/liblibpng.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/liblibjasper.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/liblibtiff.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/libIlmImf.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/libtbb.a \ + ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/libcpufeatures.a + + +LITE_LIBS = -L${LITE_ROOT}/cxx/lib/ -lpaddle_light_api_shared +############################################################### +# How to use one of static libaray: # +# `libpaddle_api_full_bundled.a` # +# `libpaddle_api_light_bundled.a` # +############################################################### +# Note: default use lite's shared library. # +############################################################### +# 1. Comment above line using `libpaddle_light_api_shared.so` +# 2. Undo comment below line using `libpaddle_api_light_bundled.a` +# LITE_LIBS = ${LITE_ROOT}/cxx/lib/libpaddle_api_light_bundled.a + +CXX_LIBS = $(LITE_LIBS) ${OPENCV_LIBS} $(SYSTEM_LIBS) + +LOCAL_DIRSRCS=$(wildcard src/*.cc) +LOCAL_SRCS=$(notdir $(LOCAL_DIRSRCS)) +LOCAL_OBJS=$(patsubst %.cpp, %.o, $(patsubst %.cc, %.o, $(LOCAL_SRCS))) + +JSON_OBJS = json_reader.o json_value.o json_writer.o + +pp_shitu: $(LOCAL_OBJS) $(JSON_OBJS) fetch_opencv + $(CC) $(SYSROOT_LINK) $(CXXFLAGS_LINK) $(LOCAL_OBJS) $(JSON_OBJS) -o pp_shitu $(CXX_LIBS) $(LDFLAGS) + +fetch_opencv: + @ test -d ${THIRD_PARTY_DIR} || mkdir ${THIRD_PARTY_DIR} + @ test -e ${THIRD_PARTY_DIR}/${OPENCV_VERSION}.tar.gz || \ + (echo "fetch opencv libs" && \ + wget -P ${THIRD_PARTY_DIR} https://paddle-inference-dist.bj.bcebos.com/${OPENCV_VERSION}.tar.gz) + @ test -d ${THIRD_PARTY_DIR}/${OPENCV_VERSION} || \ + tar -zxf ${THIRD_PARTY_DIR}/${OPENCV_VERSION}.tar.gz -C ${THIRD_PARTY_DIR} + +fetch_json_code: + @ test -d ${THIRD_PARTY_DIR} || mkdir ${THIRD_PARTY_DIR} + @ test -e ${THIRD_PARTY_DIR}/jsoncpp_code.tar.gz || \ + (echo "fetch jsoncpp_code.tar.gz" && \ + wget -P ${THIRD_PARTY_DIR} https://bj.bcebos.com/v1/paddledet/deploy/jsoncpp_code.tar.gz ) + @ test -d ${THIRD_PARTY_DIR}/jsoncpp_code || \ + tar -zxf ${THIRD_PARTY_DIR}/jsoncpp_code.tar.gz -C ${THIRD_PARTY_DIR} + +LOCAL_INCLUDES = -I./ -Iinclude +OPENCV_INCLUDE = -I${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/include +JSON_INCLUDE = -I${THIRD_PARTY_DIR}/jsoncpp_code/include +CXX_INCLUDES = ${LOCAL_INCLUDES} ${INCLUDES} ${OPENCV_INCLUDE} ${JSON_INCLUDE} -I$(LITE_ROOT)/cxx/include + + +$(LOCAL_OBJS): %.o: src/%.cc fetch_opencv fetch_json_code + $(CC) $(SYSROOT_COMPLILE) $(CXX_DEFINES) $(CXX_INCLUDES) $(CXX_FLAGS) -c $< -o $@ + +$(JSON_OBJS): %.o: ${THIRD_PARTY_DIR}/jsoncpp_code/%.cpp fetch_json_code + $(CC) $(SYSROOT_COMPLILE) $(CXX_DEFINES) $(CXX_INCLUDES) $(CXX_FLAGS) -c $< -o $@ + +.PHONY: clean fetch_opencv fetch_json_code +clean: + rm -rf $(LOCAL_OBJS) $(JSON_OBJS) + rm -f pp_shitu diff --git a/exprements/PaddleClas/deploy/lite_shitu/README.md b/exprements/PaddleClas/deploy/lite_shitu/README.md new file mode 100644 index 00000000..8f5462f6 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/README.md @@ -0,0 +1,218 @@ +# PP-ShiTu在Paddle-Lite端侧部署 + +本教程将介绍基于[Paddle Lite](https://github.com/PaddlePaddle/Paddle-Lite) 在移动端部署PaddleClas PP-ShiTu模型的详细步骤。 + +Paddle Lite是飞桨轻量化推理引擎，为手机、IOT端提供高效推理能力，并广泛整合跨平台硬件，为端侧部署及应用落地问题提供轻量化的部署方案。 + +## 1. 准备环境 + +### 运行准备 +- 电脑（编译Paddle Lite） +- 安卓手机（armv7或armv8） + +### 1.1 准备交叉编译环境 +交叉编译环境用于编译 Paddle Lite 和 PaddleClas 的PP-ShiTu Lite demo。 +支持多种开发环境，不同开发环境的编译流程请参考对应文档，请确保安装完成Java jdk、Android NDK(R17以上)。 + +1. [Docker](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#docker) +2. [Linux](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#linux) +3. [MAC OS](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#mac-os) + +```shell +# 配置完成交叉编译环境后，更新环境变量 +# for docker、Linux +source ~/.bashrc +# for Mac OS +source ~/.bash_profile +``` + +### 1.2 准备预测库 + +预测库有两种获取方式： +1. [**建议**]直接下载，预测库下载链接如下： + |平台| 架构 | 预测库下载链接| + |-|-|-| + |Android| arm7 | [inference_lite_lib](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10-rc/inference_lite_lib.android.armv7.clang.c++_static.with_extra.with_cv.tar.gz) | + | Android | arm8 | [inference_lite_lib](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10-rc/inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv.tar.gz) | + | Android | arm8(FP16) | [inference_lite_lib](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10-rc/inference_lite_lib.android.armv8_clang_c++_static_with_extra_with_cv_with_fp16.tiny_publish_427e46.zip) | + +**注意**：1. 如果是从 Paddle-Lite [官方文档](https://paddle-lite.readthedocs.io/zh/latest/quick_start/release_lib.html#android-toolchain-gcc)下载的预测库，注意选择`with_extra=ON，with_cv=ON`的下载链接。2. 目前只提供Android端demo，IOS端demo可以参考[Paddle-Lite IOS demo](https://github.com/PaddlePaddle/Paddle-Lite-Demo/tree/master/PaddleLite-ios-demo) + + +2. 编译Paddle-Lite得到预测库，Paddle-Lite的编译方式如下： +```shell +git clone https://github.com/PaddlePaddle/Paddle-Lite.git +cd Paddle-Lite +# 如果使用编译方式，建议使用develop分支编译预测库 +git checkout develop +# FP32 +./lite/tools/build_android.sh --arch=armv8 --toolchain=clang --with_cv=ON --with_extra=ON +# FP16 +./lite/tools/build_android.sh --arch=armv8 --toolchain=clang --with_cv=ON --with_extra=ON --with_arm82_fp16=ON +``` + +**注意**：编译Paddle-Lite获得预测库时，需要打开`--with_cv=ON --with_extra=ON`两个选项，`--arch`表示`arm`版本，这里指定为armv8，更多编译命令介绍请参考[链接](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_andriod.html#id2)。 + +直接下载预测库并解压后，可以得到`inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/`文件夹，通过编译Paddle-Lite得到的预测库位于`Paddle-Lite/build.lite.android.armv8.gcc/inference_lite_lib.android.armv8/`文件夹下。 +预测库的文件目录如下： + +``` +inference_lite_lib.android.armv8/ +|-- cxx C++ 预测库和头文件 +| |-- include C++ 头文件 +| | |-- paddle_api.h +| | |-- paddle_image_preprocess.h +| | |-- paddle_lite_factory_helper.h +| | |-- paddle_place.h +| | |-- paddle_use_kernels.h +| | |-- paddle_use_ops.h +| | `-- paddle_use_passes.h +| `-- lib C++预测库 +| |-- libpaddle_api_light_bundled.a C++静态库 +| `-- libpaddle_light_api_shared.so C++动态库 +|-- java Java预测库 +| |-- jar +| | `-- PaddlePredictor.jar +| |-- so +| | `-- libpaddle_lite_jni.so +| `-- src +|-- demo C++和Java示例代码 +| |-- cxx C++ 预测库demo +| `-- java Java 预测库demo +``` + +## 2 开始运行 + +### 2.1 模型准备 + + +#### 2.1.1 模型准备 + +```shell +# 进入lite_ppshitu目录 +cd $PaddleClas/deploy/lite_shitu +wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/lite/ppshitu_lite_models_v1.0.tar +tar -xf ppshitu_lite_models_v1.0.tar +rm -f ppshitu_lite_models_v1.0.tar +``` + +#### 2.1.2将yaml文件转换成json文件 + +```shell +# 如果测试单张图像 +python generate_json_config.py --det_model_path ppshitu_lite_models_v1.0/mainbody_PPLCNet_x2_5_640_quant_v1.0_lite.nb --rec_model_path ppshitu_lite_models_v1.0/general_PPLCNet_x2_5_quant_v1.0_lite.nb --rec_label_path ppshitu_lite_models_v1.0/label.txt --img_path images/demo.jpg +# or +# 如果测试多张图像 +python generate_json_config.py --det_model_path ppshitu_lite_models_v1.0/mainbody_PPLCNet_x2_5_640_quant_v1.0_lite.nb --rec_model_path ppshitu_lite_models_v1.0/general_PPLCNet_x2_5_quant_v1.0_lite.nb --rec_label_path ppshitu_lite_models_v1.0/label.txt --img_dir images + +# 执行完成后，会在lit_shitu下生成shitu_config.json配置文件 + +``` + +### 2.2 与手机联调 + +首先需要进行一些准备工作。 +1. 准备一台arm8的安卓手机，如果编译的预测库是armv7，则需要arm7的手机，并修改Makefile中`ARM_ABI=arm7`。 +2. 电脑上安装ADB工具，用于调试。 ADB安装方式如下： + + 2.1. MAC电脑安装ADB: + + ```shell + brew cask install android-platform-tools + ``` + 2.2. Linux安装ADB + ```shell + sudo apt update + sudo apt install -y wget adb + ``` + 2.3. Window安装ADB + + win上安装需要去谷歌的安卓平台下载ADB软件包进行安装：[链接](https://developer.android.com/studio) + +3. 手机连接电脑后，开启手机`USB调试`选项，选择`文件传输`模式，在电脑终端中输入： + +```shell +adb devices +``` +如果有device输出，则表示安装成功，如下所示： +``` +List of devices attached +744be294 device +``` + +4. 编译lite部署代码生成移动端可执行文件 + +```shell +cd $PaddleClas/deploy/lite_shitu +# ${lite prediction library path}下载的Paddle-Lite库路径 +inference_lite_path=${lite prediction library path}/inference_lite_lib.android.armv8.gcc.c++_static.with_extra.with_cv/ +mkdir $inference_lite_path/demo/cxx/ppshitu_lite + +cp -r * $inference_lite_path/demo/cxx/ppshitu_lite +cd $inference_lite_path/demo/cxx/ppshitu_lite + +# 执行编译，等待完成后得到可执行文件main +make ARM_ABI=arm8 +#如果是arm7，则执行 make ARM_ABI = arm7 (或者在Makefile中修改该项) +``` + +5. 准备优化后的模型、预测库文件、测试图像。 + +```shell +mkdir deploy +mv ppshitu_lite_models_v1.0 deploy/ +mv images deploy/ +mv shitu_config.json deploy/ +cp pp_shitu deploy/ + +# 将C++预测动态库so文件复制到deploy文件夹中 +cp ../../../cxx/lib/libpaddle_light_api_shared.so deploy/ +``` + +执行完成后，deploy文件夹下将有如下文件格式： + +```shell +deploy/ +|-- ppshitu_lite_models_v1.0/ +| |--mainbody_PPLCNet_x2_5_640_v1.0_lite.nb 优化后的主体检测模型文件 +| |--general_PPLCNet_x2_5_quant_v1.0_lite.nb 优化后的识别模型文件 +| |--label.txt 识别模型的label文件 +|-- images/ +| |--demo.jpg 图片文件 +| ... 图片文件 +|-- pp_shitu 生成的移动端执行文件 +|-- shitu_config.json 执行时参数配置文件 +|-- libpaddle_light_api_shared.so Paddle-Lite库文件 +``` + +**注意：** +* `shitu_config.json` 包含了目标检测的超参数，请按需进行修改 + +6. 启动调试，上述步骤完成后就可以使用ADB将文件夹 `deploy/` push到手机上运行，步骤如下： + +```shell +# 将上述deploy文件夹push到手机上 +adb push deploy /data/local/tmp/ + +adb shell +cd /data/local/tmp/deploy +export LD_LIBRARY_PATH=/data/local/tmp/deploy:$LD_LIBRARY_PATH + +# 修改权限为可执行 +chmod 777 pp_shitu +# 执行程序 +./pp_shitu shitu_config.json +``` + +如果对代码做了修改，则需要重新编译并push到手机上。 + +运行效果如下： + +![](../../docs/images/ppshitu_lite_demo.png) + +## FAQ +Q1：如果想更换模型怎么办，需要重新按照流程走一遍吗？ +A1：如果已经走通了上述步骤，更换模型只需要替换 `.nb` 模型文件即可，同时要注意修改下配置文件中的 `.nb` 文件路径以及类别映射文件（如有必要）。 + +Q2：换一个图测试怎么做？ +A2：替换 deploy 下的测试图像为你想要测试的图像，并重新生成json配置文件（或者直接修改图像路径），使用 ADB 再次 push 到手机上即可。 diff --git a/exprements/PaddleClas/deploy/lite_shitu/generate_json_config.py b/exprements/PaddleClas/deploy/lite_shitu/generate_json_config.py new file mode 100644 index 00000000..1525cdab --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/generate_json_config.py @@ -0,0 +1,138 @@ +import argparse +import json +import os + +import yaml + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument( + '--yaml_path', type=str, default='../configs/inference_drink.yaml') + parser.add_argument( + '--img_dir', + type=str, + default=None, + help='The dir path for inference images') + parser.add_argument( + '--img_path', + type=str, + default=None, + help='The dir path for inference images') + parser.add_argument( + '--det_model_path', + type=str, + default='./det.nb', + help="The model path for mainbody detection") + parser.add_argument( + '--rec_model_path', + type=str, + default='./rec.nb', + help="The rec model path") + parser.add_argument( + '--rec_label_path', + type=str, + default='./label.txt', + help='The rec model label') + parser.add_argument( + '--arch', + type=str, + default='PicoDet', + help='The model structure for detection model') + parser.add_argument( + '--fpn-stride', + type=list, + default=[8, 16, 32, 64], + help="The fpn strid for detection model") + parser.add_argument( + '--keep_top_k', + type=int, + default=100, + help='The params for nms(postprocess for detection)') + parser.add_argument( + '--nms-name', + type=str, + default='MultiClassNMS', + help='The nms name for postprocess of detection model') + parser.add_argument( + '--nms_threshold', + type=float, + default=0.5, + help='The nms nms_threshold for detection postprocess') + parser.add_argument( + '--nms_top_k', + type=int, + default=1000, + help='The nms_top_k in postprocess of detection model') + parser.add_argument( + '--score_threshold', + type=float, + default=0.3, + help='The score_threshold for postprocess of detection') + args = parser.parse_args() + return args + + +def main(): + args = parse_args() + config_yaml = yaml.safe_load(open(args.yaml_path)) + config_json = {} + config_json["Global"] = {} + config_json["Global"][ + "infer_imgs"] = args.img_path if args.img_path else config_yaml[ + "Global"]["infer_imgs"] + if args.img_dir is not None: + config_json["Global"]["infer_imgs_dir"] = args.img_dir + config_json["Global"]["infer_imgs"] = None + else: + config_json["Global"][ + "infer_imgs"] = args.img_path if args.img_path else config_yaml[ + "Global"]["infer_imgs"] + config_json["Global"]["batch_size"] = config_yaml["Global"]["batch_size"] + config_json["Global"]["cpu_num_threads"] = min( + config_yaml["Global"]["cpu_num_threads"], 4) + config_json["Global"]["image_shape"] = config_yaml["Global"]["image_shape"] + config_json["Global"]["det_model_path"] = args.det_model_path + config_json["Global"]["rec_model_path"] = args.rec_model_path + config_json["Global"]["rec_label_path"] = args.rec_label_path + config_json["Global"]["label_list"] = config_yaml["Global"]["labe_list"] + config_json["Global"]["rec_nms_thresold"] = config_yaml["Global"][ + "rec_nms_thresold"] + config_json["Global"]["max_det_results"] = config_yaml["Global"][ + "max_det_results"] + config_json["Global"]["det_fpn_stride"] = args.fpn_stride + config_json["Global"]["det_arch"] = args.arch + config_json["Global"]["return_k"] = config_yaml["IndexProcess"]["return_k"] + + # config_json["DetPreProcess"] = config_yaml["DetPreProcess"] + config_json["DetPreProcess"] = {} + config_json["DetPreProcess"]["transform_ops"] = [] + for x in config_yaml["DetPreProcess"]["transform_ops"]: + k = list(x.keys())[0] + y = x[k] + y['type'] = k + config_json["DetPreProcess"]["transform_ops"].append(y) + + config_json["DetPostProcess"] = { + "keep_top_k": args.keep_top_k, + "name": args.nms_name, + "nms_threshold": args.nms_threshold, + "nms_top_k": args.nms_top_k, + "score_threshold": args.score_threshold + } + # config_json["RecPreProcess"] = config_yaml["RecPreProcess"] + config_json["RecPreProcess"] = {} + config_json["RecPreProcess"]["transform_ops"] = [] + for x in config_yaml["RecPreProcess"]["transform_ops"]: + k = list(x.keys())[0] + y = x[k] + if y is not None: + y["type"] = k + config_json["RecPreProcess"]["transform_ops"].append(y) + + with open('shitu_config.json', 'w') as fd: + json.dump(config_json, fd, indent=4) + + +if __name__ == '__main__': + main() diff --git a/exprements/PaddleClas/deploy/lite_shitu/images/demo.jpg b/exprements/PaddleClas/deploy/lite_shitu/images/demo.jpg new file mode 100644 index 00000000..075dc31d Binary files /dev/null and b/exprements/PaddleClas/deploy/lite_shitu/images/demo.jpg differ diff --git a/exprements/PaddleClas/deploy/lite_shitu/include/config_parser.h b/exprements/PaddleClas/deploy/lite_shitu/include/config_parser.h new file mode 100644 index 00000000..dca0e5a6 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/include/config_parser.h @@ -0,0 +1,92 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once +#include +#include +#include +#include +#include + +#include "json/json.h" + +#ifdef _WIN32 +#define OS_PATH_SEP "\\" +#else +#define OS_PATH_SEP "/" +#endif + +namespace PPShiTu { + +void load_jsonf(std::string jsonfile, Json::Value& jsondata); + +// Inference model configuration parser +class ConfigPaser { + public: + ConfigPaser() {} + + ~ConfigPaser() {} + + bool load_config(const Json::Value& config) { + + // Get model arch : YOLO, SSD, RetinaNet, RCNN, Face + if (config["Global"].isMember("det_arch")) { + arch_ = config["Global"]["det_arch"].as(); + } else { + std::cerr << "Please set model arch," + << "support value : YOLO, SSD, RetinaNet, RCNN, Face." + << std::endl; + return false; + } + + // Get Preprocess for preprocessing + if (config.isMember("DetPreProcess")) { + preprocess_info_ = config["DetPreProcess"]["transform_ops"]; + } else { + std::cerr << "Please set Preprocess." << std::endl; + return false; + } + // Get label_list for visualization + if (config["Global"].isMember("label_list")) { + label_list_.clear(); + for (auto item : config["Global"]["label_list"]) { + label_list_.emplace_back(item.as()); + } + } else { + std::cerr << "Please set label_list." << std::endl; + return false; + } + + // Get NMS for postprocess + if (config.isMember("DetPostProcess")) { + nms_info_ = config["DetPostProcess"]; + } + // Get fpn_stride in PicoDet + if (config["Global"].isMember("det_fpn_stride")) { + fpn_stride_.clear(); + for (auto item : config["Global"]["det_fpn_stride"]) { + fpn_stride_.emplace_back(item.as()); + } + } + + return true; + } + std::string arch_; + Json::Value preprocess_info_; + Json::Value nms_info_; + std::vector label_list_; + std::vector fpn_stride_; +}; + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/include/object_detector.h b/exprements/PaddleClas/deploy/lite_shitu/include/object_detector.h new file mode 100644 index 00000000..779cc89d --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/include/object_detector.h @@ -0,0 +1,102 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include "json/json.h" + +#include "paddle_api.h" // NOLINT + +#include "include/config_parser.h" +#include "include/preprocess_op.h" +#include "include/utils.h" +#include "include/picodet_postprocess.h" + +using namespace paddle::lite_api; // NOLINT + +namespace PPShiTu { + +// Generate visualization colormap for each class +std::vector GenerateColorMap(int num_class); + +// Visualiztion Detection Result +cv::Mat VisualizeResult(const cv::Mat& img, + const std::vector& results, + const std::vector& lables, + const std::vector& colormap, + const bool is_rbox); + +class ObjectDetector { + public: + explicit ObjectDetector(const Json::Value& config, + const std::string& model_dir, + int cpu_threads = 1, + const int batch_size = 1) { + config_.load_config(config); + printf("config created\n"); + preprocessor_.Init(config_.preprocess_info_); + printf("before object detector\n"); + if(config["Global"]["det_model_path"].as().empty()){ + std::cout << "Please set [det_model_path] in config file" << std::endl; + exit(-1); + } + LoadModel(config["Global"]["det_model_path"].as(), cpu_threads); + printf("create object detector\n"); } + + // Load Paddle inference model + void LoadModel(std::string model_file, int num_theads); + + // Run predictor + void Predict(const std::vector& imgs, + const int warmup = 0, + const int repeats = 1, + std::vector* result = nullptr, + std::vector* bbox_num = nullptr, + std::vector* times = nullptr); + + // Get Model Label list + const std::vector& GetLabelList() const { + return config_.label_list_; + } + + private: + // Preprocess image and copy data to input buffer + void Preprocess(const cv::Mat& image_mat); + // Postprocess result + void Postprocess(const std::vector mats, + std::vector* result, + std::vector bbox_num, + bool is_rbox); + + std::shared_ptr predictor_; + Preprocessor preprocessor_; + ImageBlob inputs_; + std::vector output_data_; + std::vector out_bbox_num_data_; + float threshold_; + ConfigPaser config_; + +}; + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/include/picodet_postprocess.h b/exprements/PaddleClas/deploy/lite_shitu/include/picodet_postprocess.h new file mode 100644 index 00000000..758795b5 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/include/picodet_postprocess.h @@ -0,0 +1,38 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include +#include +#include +#include +#include +#include + +#include "include/utils.h" + +namespace PPShiTu { + +void PicoDetPostProcess(std::vector* results, + std::vector outs, + std::vector fpn_stride, + std::vector im_shape, + std::vector scale_factor, + float score_threshold = 0.3, + float nms_threshold = 0.5, + int num_class = 80, + int reg_max = 7); + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/include/preprocess_op.h b/exprements/PaddleClas/deploy/lite_shitu/include/preprocess_op.h new file mode 100644 index 00000000..f7050fa8 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/include/preprocess_op.h @@ -0,0 +1,188 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include "json/json.h" + +namespace PPShiTu { + +// Object for storing all preprocessed data +class ImageBlob { + public: + // image width and height + std::vector im_shape_; + // Buffer for image data after preprocessing + std::vector im_data_; + // in net data shape(after pad) + std::vector in_net_shape_; + // Evaluation image width and height + // std::vector eval_im_size_f_; + // Scale factor for image size to origin image size + std::vector scale_factor_; +}; + +// Abstraction of preprocessing opration class +class PreprocessOp { + public: + virtual void Init(const Json::Value& item) = 0; + virtual void Run(cv::Mat* im, ImageBlob* data) = 0; +}; + +class InitInfo : public PreprocessOp { + public: + virtual void Init(const Json::Value& item) {} + virtual void Run(cv::Mat* im, ImageBlob* data); +}; + +class NormalizeImage : public PreprocessOp { + public: + virtual void Init(const Json::Value& item) { + mean_.clear(); + scale_.clear(); + for (auto tmp : item["mean"]) { + mean_.emplace_back(tmp.as()); + } + for (auto tmp : item["std"]) { + scale_.emplace_back(tmp.as()); + } + is_scale_ = item["is_scale"].as(); + } + + virtual void Run(cv::Mat* im, ImageBlob* data); + + private: + // CHW or HWC + std::vector mean_; + std::vector scale_; + bool is_scale_; +}; + +class Permute : public PreprocessOp { + public: + virtual void Init(const Json::Value& item) {} + virtual void Run(cv::Mat* im, ImageBlob* data); +}; + +class Resize : public PreprocessOp { + public: + virtual void Init(const Json::Value& item) { + interp_ = item["interp"].as(); + // max_size_ = item["target_size"].as(); + keep_ratio_ = item["keep_ratio"].as(); + target_size_.clear(); + for (auto tmp : item["target_size"]) { + target_size_.emplace_back(tmp.as()); + } + } + + // Compute best resize scale for x-dimension, y-dimension + std::pair GenerateScale(const cv::Mat& im); + + virtual void Run(cv::Mat* im, ImageBlob* data); + + private: + int interp_; + bool keep_ratio_; + std::vector target_size_; + std::vector in_net_shape_; +}; + +// Models with FPN need input shape % stride == 0 +class PadStride : public PreprocessOp { + public: + virtual void Init(const Json::Value& item) { + stride_ = item["stride"].as(); + } + + virtual void Run(cv::Mat* im, ImageBlob* data); + + private: + int stride_; +}; + +class TopDownEvalAffine : public PreprocessOp { + public: + virtual void Init(const Json::Value& item) { + trainsize_.clear(); + for (auto tmp : item["trainsize"]) { + trainsize_.emplace_back(tmp.as()); + } + } + + virtual void Run(cv::Mat* im, ImageBlob* data); + + private: + int interp_ = 1; + std::vector trainsize_; +}; + +void CropImg(cv::Mat& img, + cv::Mat& crop_img, + std::vector& area, + std::vector& center, + std::vector& scale, + float expandratio = 0.15); + +class Preprocessor { + public: + void Init(const Json::Value& config_node) { + // initialize image info at first + ops_["InitInfo"] = std::make_shared(); + for (const auto& item : config_node) { + auto op_name = item["type"].as(); + + ops_[op_name] = CreateOp(op_name); + ops_[op_name]->Init(item); + } + } + + std::shared_ptr CreateOp(const std::string& name) { + if (name == "DetResize") { + return std::make_shared(); + } else if (name == "DetPermute") { + return std::make_shared(); + } else if (name == "DetNormalizeImage") { + return std::make_shared(); + } else if (name == "DetPadStride") { + // use PadStride instead of PadBatch + return std::make_shared(); + } else if (name == "TopDownEvalAffine") { + return std::make_shared(); + } + std::cerr << "can not find function of OP: " << name + << " and return: nullptr" << std::endl; + return nullptr; + } + + void Run(cv::Mat* im, ImageBlob* data); + + public: + static const std::vector RUN_ORDER; + + private: + std::unordered_map> ops_; +}; + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/include/recognition.h b/exprements/PaddleClas/deploy/lite_shitu/include/recognition.h new file mode 100644 index 00000000..0c45e946 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/include/recognition.h @@ -0,0 +1,115 @@ +// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once +#include "paddle_api.h" // NOLINT +#include "json/json.h" +#include +#include +#include +#include +#include +#include +#include +#include +#include + +using namespace paddle::lite_api; // NOLINT +using namespace std; + +namespace PPShiTu { + +struct RESULT { + std::string class_name; + int class_id; + float score; +}; + +class Recognition { + +public: + explicit Recognition(const Json::Value &config_file) { + MobileConfig config; + if (config_file["Global"]["rec_model_path"].as().empty()) { + std::cout << "Please set [rec_model_path] in config file" << std::endl; + exit(-1); + } + config.set_model_from_file( + config_file["Global"]["rec_model_path"].as()); + this->predictor = CreatePaddlePredictor(config); + + if (config_file["Global"]["rec_label_path"].as().empty()) { + std::cout << "Please set [rec_label_path] in config file" << std::endl; + exit(-1); + } + LoadLabel(config_file["Global"]["rec_label_path"].as()); + SetPreProcessParam(config_file["RecPreProcess"]["transform_ops"]); + if (!config_file["Global"].isMember("return_k")){ + this->topk = config_file["Global"]["return_k"].as(); + } + printf("rec model create!\n"); + } + + void LoadLabel(std::string path) { + std::ifstream file; + std::vector label_list; + file.open(path); + while (file) { + std::string line; + std::getline(file, line); + std::string::size_type pos = line.find(" "); + if (pos != std::string::npos) { + line = line.substr(pos); + } + this->label_list.push_back(line); + } + file.clear(); + file.close(); + } + + void SetPreProcessParam(const Json::Value &config_file) { + for (const auto &item : config_file) { + auto op_name = item["type"].as(); + if (op_name == "ResizeImage") { + this->size = item["size"].as(); + } else if (op_name == "NormalizeImage") { + this->mean.clear(); + this->std.clear(); + for (auto tmp : item["mean"]) { + this->mean.emplace_back(tmp.as()); + } + for (auto tmp : item["std"]) { + this->std.emplace_back(1 / tmp.as()); + } + this->scale = item["scale"].as(); + } + } + } + + std::vector RunRecModel(const cv::Mat &img, double &cost_time); + std::vector PostProcess(const float *output_data, int output_size, + cv::Mat &output_image); + cv::Mat ResizeImage(const cv::Mat &img); + void NeonMeanScale(const float *din, float *dout, int size); + +private: + std::shared_ptr predictor; + std::vector label_list; + std::vector mean = {0.485f, 0.456f, 0.406f}; + std::vector std = {1 / 0.229f, 1 / 0.224f, 1 / 0.225f}; + double scale = 0.00392157; + float size = 224; + int topk = 5; +}; +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/include/utils.h b/exprements/PaddleClas/deploy/lite_shitu/include/utils.h new file mode 100644 index 00000000..18a04cf3 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/include/utils.h @@ -0,0 +1,43 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include +#include +#include +#include +#include +#include +#include +#include + +namespace PPShiTu { + +// Object Detection Result +struct ObjectResult { + // Rectangle coordinates of detected object: left, right, top, down + std::vector rect; + // Class id of detected object + int class_id; + // Confidence of detected object + float confidence; + + // RecModel result + std::vector rec_result; +}; + +void nms(std::vector &input_boxes, float nms_threshold, bool rec_nms=false); + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/src/config_parser.cc b/exprements/PaddleClas/deploy/lite_shitu/src/config_parser.cc new file mode 100644 index 00000000..d98b2f90 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/src/config_parser.cc @@ -0,0 +1,32 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "include/config_parser.h" + +namespace PPShiTu { + +void load_jsonf(std::string jsonfile, Json::Value &jsondata) { + std::ifstream ifs; + ifs.open(jsonfile); + + Json::CharReaderBuilder builder; + builder["collectComments"] = true; + JSONCPP_STRING errs; + if (!parseFromStream(builder, ifs, &jsondata, &errs)) { + std::cout << errs << std::endl; + return; + } +} + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/src/main.cc b/exprements/PaddleClas/deploy/lite_shitu/src/main.cc new file mode 100644 index 00000000..690162a4 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/src/main.cc @@ -0,0 +1,225 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "include/config_parser.h" +#include "include/object_detector.h" +#include "include/preprocess_op.h" +#include "include/recognition.h" +#include "json/json.h" + +Json::Value RT_Config; + +static std::string DirName(const std::string &filepath) { + auto pos = filepath.rfind(OS_PATH_SEP); + if (pos == std::string::npos) { + return ""; + } + return filepath.substr(0, pos); +} + +static bool PathExists(const std::string &path) { + struct stat buffer; + return (stat(path.c_str(), &buffer) == 0); +} + +static void MkDir(const std::string &path) { + if (PathExists(path)) + return; + int ret = 0; + ret = mkdir(path.c_str(), 0755); + if (ret != 0) { + std::string path_error(path); + path_error += " mkdir failed!"; + throw std::runtime_error(path_error); + } +} + +static void MkDirs(const std::string &path) { + if (path.empty()) + return; + if (PathExists(path)) + return; + + MkDirs(DirName(path)); + MkDir(path); +} + +void DetPredictImage(const std::vector &batch_imgs, + std::vector &im_result, + const int batch_size_det, const int max_det_num, + const bool run_benchmark, PPShiTu::ObjectDetector *det) { + std::vector det_t = {0, 0, 0}; + int steps = ceil(float(batch_imgs.size()) / batch_size_det); + for (int idx = 0; idx < steps; idx++) { + int left_image_cnt = batch_imgs.size() - idx * batch_size_det; + if (left_image_cnt > batch_size_det) { + left_image_cnt = batch_size_det; + } + // Store all detected result + std::vector result; + std::vector bbox_num; + std::vector det_times; + + bool is_rbox = false; + if (run_benchmark) { + det->Predict(batch_imgs, 50, 50, &result, &bbox_num, &det_times); + } else { + det->Predict(batch_imgs, 0, 1, &result, &bbox_num, &det_times); + } + + int item_start_idx = 0; + for (int i = 0; i < left_image_cnt; i++) { + cv::Mat im = batch_imgs[i]; + // std::vector im_result; + int detect_num = 0; + for (int j = 0; j < min(bbox_num[i], max_det_num); j++) { + PPShiTu::ObjectResult item = result[item_start_idx + j]; + if (item.class_id == -1) { + continue; + } + detect_num += 1; + im_result.push_back(item); + } + item_start_idx = item_start_idx + bbox_num[i]; + } + + det_t[0] += det_times[0]; + det_t[1] += det_times[1]; + det_t[2] += det_times[2]; + } +} + +void PrintResult(const std::string &image_path, + std::vector &det_result) { + printf("%s:\n", image_path.c_str()); + for (int i = 0; i < det_result.size(); ++i) { + printf("\tresult%d: bbox[%d, %d, %d, %d], score: %f, label: %s\n", i, + det_result[i].rect[0], det_result[i].rect[1], det_result[i].rect[2], + det_result[i].rect[3], det_result[i].rec_result[0].score, + det_result[i].rec_result[0].class_name.c_str()); + } +} + +int main(int argc, char **argv) { + std::cout << "Usage: " << argv[0] + << " [config_path](option) [image_dir](option)\n"; + if (argc < 2) { + std::cout << "Usage: ./main det_runtime_config.json" << std::endl; + return -1; + } + std::string config_path = argv[1]; + std::string img_dir = ""; + + if (argc >= 3) { + img_dir = argv[2]; + } + // Parsing command-line + PPShiTu::load_jsonf(config_path, RT_Config); + if (RT_Config["Global"]["det_model_path"].as().empty()) { + std::cout << "Please set [det_model_path] in " << config_path << std::endl; + return -1; + } + + if (!RT_Config["Global"]["infer_imgs_dir"].as().empty() && + img_dir.empty()) { + img_dir = RT_Config["Global"]["infer_imgs_dir"].as(); + } + if (RT_Config["Global"]["infer_imgs"].as().empty() && + img_dir.empty()) { + std::cout << "Please set [infer_imgs] in " << config_path + << " Or use command: <" << argv[0] << " [shitu_config]" + << " [image_dir]>" << std::endl; + return -1; + } + // Load model and create a object detector + PPShiTu::ObjectDetector det( + RT_Config, RT_Config["Global"]["det_model_path"].as(), + RT_Config["Global"]["cpu_num_threads"].as(), + RT_Config["Global"]["batch_size"].as()); + // create rec model + PPShiTu::Recognition rec(RT_Config); + // Do inference on input image + + std::vector det_result; + std::vector batch_imgs; + double rec_time; + if (!RT_Config["Global"]["infer_imgs"].as().empty() || + !img_dir.empty()) { + std::vector all_img_paths; + std::vector cv_all_img_paths; + if (!RT_Config["Global"]["infer_imgs"].as().empty()) { + all_img_paths.push_back( + RT_Config["Global"]["infer_imgs"].as()); + if (RT_Config["Global"]["batch_size"].as() > 1) { + std::cout << "batch_size_det should be 1, when set `image_file`." + << std::endl; + return -1; + } + } else { + cv::glob(img_dir, + cv_all_img_paths); + for (const auto &img_path : cv_all_img_paths) { + all_img_paths.push_back(img_path); + } + } + for (int i = 0; i < all_img_paths.size(); ++i) { + std::string img_path = all_img_paths[i]; + cv::Mat srcimg = cv::imread(img_path, cv::IMREAD_COLOR); + if (!srcimg.data) { + std::cerr << "[ERROR] image read failed! image path: " << img_path + << "\n"; + exit(-1); + } + cv::cvtColor(srcimg, srcimg, cv::COLOR_BGR2RGB); + batch_imgs.push_back(srcimg); + DetPredictImage( + batch_imgs, det_result, RT_Config["Global"]["batch_size"].as(), + RT_Config["Global"]["max_det_results"].as(), false, &det); + + // add the whole image for recognition to improve recall + PPShiTu::ObjectResult result_whole_img = { + {0, 0, srcimg.cols, srcimg.rows}, 0, 1.0}; + det_result.push_back(result_whole_img); + + // get rec result + for (int j = 0; j < det_result.size(); ++j) { + int w = det_result[j].rect[2] - det_result[j].rect[0]; + int h = det_result[j].rect[3] - det_result[j].rect[1]; + cv::Rect rect(det_result[j].rect[0], det_result[j].rect[1], w, h); + cv::Mat crop_img = srcimg(rect); + std::vector result = + rec.RunRecModel(crop_img, rec_time); + det_result[j].rec_result.assign(result.begin(), result.end()); + } + // rec nms + PPShiTu::nms(det_result, + RT_Config["Global"]["rec_nms_thresold"].as(), true); + PrintResult(img_path, det_result); + batch_imgs.clear(); + det_result.clear(); + } + } + return 0; +} diff --git a/exprements/PaddleClas/deploy/lite_shitu/src/object_detector.cc b/exprements/PaddleClas/deploy/lite_shitu/src/object_detector.cc new file mode 100644 index 00000000..ffea31bb --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/src/object_detector.cc @@ -0,0 +1,329 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +#include +// for setprecision +#include +#include +#include "include/object_detector.h" + +namespace PPShiTu { + +// Load Model and create model predictor +void ObjectDetector::LoadModel(std::string model_file, int num_theads) { + MobileConfig config; + config.set_threads(num_theads); + config.set_model_from_file(model_file); + config.set_power_mode(LITE_POWER_HIGH); + + predictor_ = CreatePaddlePredictor(config); +} + +// Visualiztion MaskDetector results +cv::Mat VisualizeResult(const cv::Mat& img, + const std::vector& results, + const std::vector& lables, + const std::vector& colormap, + const bool is_rbox = false) { + cv::Mat vis_img = img.clone(); + for (int i = 0; i < results.size(); ++i) { + // Configure color and text size + std::ostringstream oss; + oss << std::setiosflags(std::ios::fixed) << std::setprecision(4); + oss << lables[results[i].class_id] << " "; + oss << results[i].confidence; + std::string text = oss.str(); + int c1 = colormap[3 * results[i].class_id + 0]; + int c2 = colormap[3 * results[i].class_id + 1]; + int c3 = colormap[3 * results[i].class_id + 2]; + cv::Scalar roi_color = cv::Scalar(c1, c2, c3); + int font_face = cv::FONT_HERSHEY_COMPLEX_SMALL; + double font_scale = 0.5f; + float thickness = 0.5; + cv::Size text_size = + cv::getTextSize(text, font_face, font_scale, thickness, nullptr); + cv::Point origin; + + if (is_rbox) { + // Draw object, text, and background + for (int k = 0; k < 4; k++) { + cv::Point pt1 = cv::Point(results[i].rect[(k * 2) % 8], + results[i].rect[(k * 2 + 1) % 8]); + cv::Point pt2 = cv::Point(results[i].rect[(k * 2 + 2) % 8], + results[i].rect[(k * 2 + 3) % 8]); + cv::line(vis_img, pt1, pt2, roi_color, 2); + } + } else { + int w = results[i].rect[2] - results[i].rect[0]; + int h = results[i].rect[3] - results[i].rect[1]; + cv::Rect roi = cv::Rect(results[i].rect[0], results[i].rect[1], w, h); + // Draw roi object, text, and background + cv::rectangle(vis_img, roi, roi_color, 2); + } + + origin.x = results[i].rect[0]; + origin.y = results[i].rect[1]; + + // Configure text background + cv::Rect text_back = cv::Rect(results[i].rect[0], + results[i].rect[1] - text_size.height, + text_size.width, + text_size.height); + // Draw text, and background + cv::rectangle(vis_img, text_back, roi_color, -1); + cv::putText(vis_img, + text, + origin, + font_face, + font_scale, + cv::Scalar(255, 255, 255), + thickness); + } + return vis_img; +} + +void ObjectDetector::Preprocess(const cv::Mat& ori_im) { + // Clone the image : keep the original mat for postprocess + cv::Mat im = ori_im.clone(); + // cv::cvtColor(im, im, cv::COLOR_BGR2RGB); + preprocessor_.Run(&im, &inputs_); +} + +void ObjectDetector::Postprocess(const std::vector mats, + std::vector* result, + std::vector bbox_num, + bool is_rbox = false) { + result->clear(); + int start_idx = 0; + for (int im_id = 0; im_id < mats.size(); im_id++) { + cv::Mat raw_mat = mats[im_id]; + int rh = 1; + int rw = 1; + if (config_.arch_ == "Face") { + rh = raw_mat.rows; + rw = raw_mat.cols; + } + for (int j = start_idx; j < start_idx + bbox_num[im_id]; j++) { + if (is_rbox) { + // Class id + int class_id = static_cast(round(output_data_[0 + j * 10])); + // Confidence score + float score = output_data_[1 + j * 10]; + int x1 = (output_data_[2 + j * 10] * rw); + int y1 = (output_data_[3 + j * 10] * rh); + int x2 = (output_data_[4 + j * 10] * rw); + int y2 = (output_data_[5 + j * 10] * rh); + int x3 = (output_data_[6 + j * 10] * rw); + int y3 = (output_data_[7 + j * 10] * rh); + int x4 = (output_data_[8 + j * 10] * rw); + int y4 = (output_data_[9 + j * 10] * rh); + + PPShiTu::ObjectResult result_item; + result_item.rect = {x1, y1, x2, y2, x3, y3, x4, y4}; + result_item.class_id = class_id; + result_item.confidence = score; + result->push_back(result_item); + } else { + // Class id + int class_id = static_cast(round(output_data_[0 + j * 6])); + // Confidence score + float score = output_data_[1 + j * 6]; + int xmin = (output_data_[2 + j * 6] * rw); + int ymin = (output_data_[3 + j * 6] * rh); + int xmax = (output_data_[4 + j * 6] * rw); + int ymax = (output_data_[5 + j * 6] * rh); + int wd = xmax - xmin; + int hd = ymax - ymin; + + PPShiTu::ObjectResult result_item; + result_item.rect = {xmin, ymin, xmax, ymax}; + result_item.class_id = class_id; + result_item.confidence = score; + result->push_back(result_item); + } + } + start_idx += bbox_num[im_id]; + } +} + +void ObjectDetector::Predict(const std::vector& imgs, + const int warmup, + const int repeats, + std::vector* result, + std::vector* bbox_num, + std::vector* times) { + auto preprocess_start = std::chrono::steady_clock::now(); + int batch_size = imgs.size(); + + // in_data_batch + std::vector in_data_all; + std::vector im_shape_all(batch_size * 2); + std::vector scale_factor_all(batch_size * 2); + // Preprocess image + for (int bs_idx = 0; bs_idx < batch_size; bs_idx++) { + cv::Mat im = imgs.at(bs_idx); + Preprocess(im); + im_shape_all[bs_idx * 2] = inputs_.im_shape_[0]; + im_shape_all[bs_idx * 2 + 1] = inputs_.im_shape_[1]; + + scale_factor_all[bs_idx * 2] = inputs_.scale_factor_[0]; + scale_factor_all[bs_idx * 2 + 1] = inputs_.scale_factor_[1]; + + // TODO: reduce cost time + in_data_all.insert( + in_data_all.end(), inputs_.im_data_.begin(), inputs_.im_data_.end()); + } + auto preprocess_end = std::chrono::steady_clock::now(); + std::vector output_data_list_; + // Prepare input tensor + + auto input_names = predictor_->GetInputNames(); + for (const auto& tensor_name : input_names) { + auto in_tensor = predictor_->GetInputByName(tensor_name); + if (tensor_name == "image") { + int rh = inputs_.in_net_shape_[0]; + int rw = inputs_.in_net_shape_[1]; + in_tensor->Resize({batch_size, 3, rh, rw}); + auto* inptr = in_tensor->mutable_data(); + std::copy_n(in_data_all.data(), in_data_all.size(), inptr); + } else if (tensor_name == "im_shape") { + in_tensor->Resize({batch_size, 2}); + auto* inptr = in_tensor->mutable_data(); + std::copy_n(im_shape_all.data(), im_shape_all.size(), inptr); + } else if (tensor_name == "scale_factor") { + in_tensor->Resize({batch_size, 2}); + auto* inptr = in_tensor->mutable_data(); + std::copy_n(scale_factor_all.data(), scale_factor_all.size(), inptr); + } + } + + // Run predictor + // warmup + for (int i = 0; i < warmup; i++) { + predictor_->Run(); + // Get output tensor + auto output_names = predictor_->GetOutputNames(); + if (config_.arch_ == "PicoDet") { + for (int j = 0; j < output_names.size(); j++) { + auto output_tensor = predictor_->GetTensor(output_names[j]); + const float* outptr = output_tensor->data(); + std::vector output_shape = output_tensor->shape(); + output_data_list_.push_back(outptr); + } + } else { + auto out_tensor = predictor_->GetTensor(output_names[0]); + auto out_bbox_num = predictor_->GetTensor(output_names[1]); + } + } + + bool is_rbox = false; + auto inference_start = std::chrono::steady_clock::now(); + for (int i = 0; i < repeats; i++) { + predictor_->Run(); + } + auto inference_end = std::chrono::steady_clock::now(); + auto postprocess_start = std::chrono::steady_clock::now(); + // Get output tensor + output_data_list_.clear(); + int num_class = 1; + int reg_max = 7; + auto output_names = predictor_->GetOutputNames(); + // TODO: Unified model output. + if (config_.arch_ == "PicoDet") { + for (int i = 0; i < output_names.size(); i++) { + auto output_tensor = predictor_->GetTensor(output_names[i]); + const float* outptr = output_tensor->data(); + std::vector output_shape = output_tensor->shape(); + if (i == 0) { + num_class = output_shape[2]; + } + if (i == config_.fpn_stride_.size()) { + reg_max = output_shape[2] / 4 - 1; + } + output_data_list_.push_back(outptr); + } + } else { + auto output_tensor = predictor_->GetTensor(output_names[0]); + auto output_shape = output_tensor->shape(); + auto out_bbox_num = predictor_->GetTensor(output_names[1]); + auto out_bbox_num_shape = out_bbox_num->shape(); + // Calculate output length + int output_size = 1; + for (int j = 0; j < output_shape.size(); ++j) { + output_size *= output_shape[j]; + } + is_rbox = output_shape[output_shape.size() - 1] % 10 == 0; + + if (output_size < 6) { + std::cerr << "[WARNING] No object detected." << std::endl; + } + output_data_.resize(output_size); + std::copy_n( + output_tensor->mutable_data(), output_size, output_data_.data()); + + int out_bbox_num_size = 1; + for (int j = 0; j < out_bbox_num_shape.size(); ++j) { + out_bbox_num_size *= out_bbox_num_shape[j]; + } + out_bbox_num_data_.resize(out_bbox_num_size); + std::copy_n(out_bbox_num->mutable_data(), + out_bbox_num_size, + out_bbox_num_data_.data()); + } + // Postprocessing result + + result->clear(); + if (config_.arch_ == "PicoDet") { + PPShiTu::PicoDetPostProcess( + result, output_data_list_, config_.fpn_stride_, + inputs_.im_shape_, inputs_.scale_factor_, + config_.nms_info_["score_threshold"].as(), + config_.nms_info_["nms_threshold"].as(), num_class, reg_max); + bbox_num->push_back(result->size()); + } else { + Postprocess(imgs, result, out_bbox_num_data_, is_rbox); + bbox_num->clear(); + for (int k = 0; k < out_bbox_num_data_.size(); k++) { + int tmp = out_bbox_num_data_[k]; + bbox_num->push_back(tmp); + } + } + auto postprocess_end = std::chrono::steady_clock::now(); + + std::chrono::duration preprocess_diff = + preprocess_end - preprocess_start; + times->push_back(double(preprocess_diff.count() * 1000)); + std::chrono::duration inference_diff = inference_end - inference_start; + times->push_back(double(inference_diff.count() / repeats * 1000)); + std::chrono::duration postprocess_diff = + postprocess_end - postprocess_start; + times->push_back(double(postprocess_diff.count() * 1000)); +} + +std::vector GenerateColorMap(int num_class) { + auto colormap = std::vector(3 * num_class, 0); + for (int i = 0; i < num_class; ++i) { + int j = 0; + int lab = i; + while (lab) { + colormap[i * 3] |= (((lab >> 0) & 1) << (7 - j)); + colormap[i * 3 + 1] |= (((lab >> 1) & 1) << (7 - j)); + colormap[i * 3 + 2] |= (((lab >> 2) & 1) << (7 - j)); + ++j; + lab >>= 3; + } + } + return colormap; +} + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/src/picodet_postprocess.cc b/exprements/PaddleClas/deploy/lite_shitu/src/picodet_postprocess.cc new file mode 100644 index 00000000..cde914c2 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/src/picodet_postprocess.cc @@ -0,0 +1,129 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +// The code is based on: +// https://github.com/RangiLyu/nanodet/blob/main/demo_mnn/nanodet_mnn.cpp + +#include "include/picodet_postprocess.h" +#include + +namespace PPShiTu { + +float fast_exp(float x) { + union { + uint32_t i; + float f; + } v{}; + v.i = (1 << 23) * (1.4426950409 * x + 126.93490512f); + return v.f; +} + +template +int activation_function_softmax(const _Tp *src, _Tp *dst, int length) { + const _Tp alpha = *std::max_element(src, src + length); + _Tp denominator{0}; + + for (int i = 0; i < length; ++i) { + dst[i] = fast_exp(src[i] - alpha); + denominator += dst[i]; + } + + for (int i = 0; i < length; ++i) { + dst[i] /= denominator; + } + + return 0; +} + +// PicoDet decode +PPShiTu::ObjectResult +disPred2Bbox(const float *&dfl_det, int label, float score, int x, int y, + int stride, std::vector im_shape, int reg_max) { + float ct_x = (x + 0.5) * stride; + float ct_y = (y + 0.5) * stride; + std::vector dis_pred; + dis_pred.resize(4); + for (int i = 0; i < 4; i++) { + float dis = 0; + float *dis_after_sm = new float[reg_max + 1]; + activation_function_softmax(dfl_det + i * (reg_max + 1), dis_after_sm, + reg_max + 1); + for (int j = 0; j < reg_max + 1; j++) { + dis += j * dis_after_sm[j]; + } + dis *= stride; + dis_pred[i] = dis; + delete[] dis_after_sm; + } + int xmin = (int)(std::max)(ct_x - dis_pred[0], .0f); + int ymin = (int)(std::max)(ct_y - dis_pred[1], .0f); + int xmax = (int)(std::min)(ct_x + dis_pred[2], (float)im_shape[0]); + int ymax = (int)(std::min)(ct_y + dis_pred[3], (float)im_shape[1]); + + PPShiTu::ObjectResult result_item; + result_item.rect = {xmin, ymin, xmax, ymax}; + result_item.class_id = label; + result_item.confidence = score; + + return result_item; +} + +void PicoDetPostProcess(std::vector *results, + std::vector outs, + std::vector fpn_stride, + std::vector im_shape, + std::vector scale_factor, float score_threshold, + float nms_threshold, int num_class, int reg_max) { + std::vector> bbox_results; + bbox_results.resize(num_class); + int in_h = im_shape[0], in_w = im_shape[1]; + for (int i = 0; i < fpn_stride.size(); ++i) { + int feature_h = ceil((float)in_h / fpn_stride[i]); + int feature_w = ceil((float)in_w / fpn_stride[i]); + for (int idx = 0; idx < feature_h * feature_w; idx++) { + const float *scores = outs[i] + (idx * num_class); + + int row = idx / feature_w; + int col = idx % feature_w; + float score = 0; + int cur_label = 0; + for (int label = 0; label < num_class; label++) { + if (scores[label] > score) { + score = scores[label]; + cur_label = label; + } + } + if (score > score_threshold) { + const float *bbox_pred = + outs[i + fpn_stride.size()] + (idx * 4 * (reg_max + 1)); + bbox_results[cur_label].push_back( + disPred2Bbox(bbox_pred, cur_label, score, col, row, fpn_stride[i], + im_shape, reg_max)); + } + } + } + for (int i = 0; i < (int)bbox_results.size(); i++) { + PPShiTu::nms(bbox_results[i], nms_threshold); + + for (auto box : bbox_results[i]) { + box.rect[0] = box.rect[0] / scale_factor[1]; + box.rect[2] = box.rect[2] / scale_factor[1]; + box.rect[1] = box.rect[1] / scale_factor[0]; + box.rect[3] = box.rect[3] / scale_factor[0]; + results->push_back(box); + } + } +} + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/src/preprocess_op.cc b/exprements/PaddleClas/deploy/lite_shitu/src/preprocess_op.cc new file mode 100644 index 00000000..9c74d6ee --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/src/preprocess_op.cc @@ -0,0 +1,185 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include +#include + +#include "include/preprocess_op.h" + +namespace PPShiTu { + +void InitInfo::Run(cv::Mat* im, ImageBlob* data) { + data->im_shape_ = {static_cast(im->rows), + static_cast(im->cols)}; + data->scale_factor_ = {1., 1.}; + data->in_net_shape_ = {static_cast(im->rows), + static_cast(im->cols)}; +} + +void NormalizeImage::Run(cv::Mat* im, ImageBlob* data) { + double e = 1.0; + if (is_scale_) { + e *= 1./255.0; + } + (*im).convertTo(*im, CV_32FC3, e); + for (int h = 0; h < im->rows; h++) { + for (int w = 0; w < im->cols; w++) { + im->at(h, w)[0] = + (im->at(h, w)[0] - mean_[0]) / scale_[0]; + im->at(h, w)[1] = + (im->at(h, w)[1] - mean_[1]) / scale_[1]; + im->at(h, w)[2] = + (im->at(h, w)[2] - mean_[2]) / scale_[2]; + } + } +} + +void Permute::Run(cv::Mat* im, ImageBlob* data) { + (*im).convertTo(*im, CV_32FC3); + int rh = im->rows; + int rw = im->cols; + int rc = im->channels(); + (data->im_data_).resize(rc * rh * rw); + float* base = (data->im_data_).data(); + for (int i = 0; i < rc; ++i) { + cv::extractChannel(*im, cv::Mat(rh, rw, CV_32FC1, base + i * rh * rw), i); + } +} + +void Resize::Run(cv::Mat* im, ImageBlob* data) { + auto resize_scale = GenerateScale(*im); + data->im_shape_ = {static_cast(im->cols * resize_scale.first), + static_cast(im->rows * resize_scale.second)}; + data->in_net_shape_ = {static_cast(im->cols * resize_scale.first), + static_cast(im->rows * resize_scale.second)}; + cv::resize( + *im, *im, cv::Size(), resize_scale.first, resize_scale.second, interp_); + data->im_shape_ = { + static_cast(im->rows), static_cast(im->cols), + }; + data->scale_factor_ = { + resize_scale.second, resize_scale.first, + }; +} + +std::pair Resize::GenerateScale(const cv::Mat& im) { + std::pair resize_scale; + int origin_w = im.cols; + int origin_h = im.rows; + + if (keep_ratio_) { + int im_size_max = std::max(origin_w, origin_h); + int im_size_min = std::min(origin_w, origin_h); + int target_size_max = + *std::max_element(target_size_.begin(), target_size_.end()); + int target_size_min = + *std::min_element(target_size_.begin(), target_size_.end()); + float scale_min = + static_cast(target_size_min) / static_cast(im_size_min); + float scale_max = + static_cast(target_size_max) / static_cast(im_size_max); + float scale_ratio = std::min(scale_min, scale_max); + resize_scale = {scale_ratio, scale_ratio}; + } else { + resize_scale.first = + static_cast(target_size_[1]) / static_cast(origin_w); + resize_scale.second = + static_cast(target_size_[0]) / static_cast(origin_h); + } + return resize_scale; +} + +void PadStride::Run(cv::Mat* im, ImageBlob* data) { + if (stride_ <= 0) { + return; + } + int rc = im->channels(); + int rh = im->rows; + int rw = im->cols; + int nh = (rh / stride_) * stride_ + (rh % stride_ != 0) * stride_; + int nw = (rw / stride_) * stride_ + (rw % stride_ != 0) * stride_; + cv::copyMakeBorder( + *im, *im, 0, nh - rh, 0, nw - rw, cv::BORDER_CONSTANT, cv::Scalar(0)); + data->in_net_shape_ = { + static_cast(im->rows), static_cast(im->cols), + }; +} + +void TopDownEvalAffine::Run(cv::Mat* im, ImageBlob* data) { + cv::resize(*im, *im, cv::Size(trainsize_[0], trainsize_[1]), 0, 0, interp_); + // todo: Simd::ResizeBilinear(); + data->in_net_shape_ = { + static_cast(trainsize_[1]), static_cast(trainsize_[0]), + }; +} + +// Preprocessor op running order +const std::vector Preprocessor::RUN_ORDER = {"InitInfo", + "DetTopDownEvalAffine", + "DetResize", + "DetNormalizeImage", + "DetPadStride", + "DetPermute"}; + +void Preprocessor::Run(cv::Mat* im, ImageBlob* data) { + for (const auto& name : RUN_ORDER) { + if (ops_.find(name) != ops_.end()) { + ops_[name]->Run(im, data); + } + } +} + +void CropImg(cv::Mat& img, + cv::Mat& crop_img, + std::vector& area, + std::vector& center, + std::vector& scale, + float expandratio) { + int crop_x1 = std::max(0, area[0]); + int crop_y1 = std::max(0, area[1]); + int crop_x2 = std::min(img.cols - 1, area[2]); + int crop_y2 = std::min(img.rows - 1, area[3]); + + int center_x = (crop_x1 + crop_x2) / 2.; + int center_y = (crop_y1 + crop_y2) / 2.; + int half_h = (crop_y2 - crop_y1) / 2.; + int half_w = (crop_x2 - crop_x1) / 2.; + + if (half_h * 3 > half_w * 4) { + half_w = static_cast(half_h * 0.75); + } else { + half_h = static_cast(half_w * 4 / 3); + } + + crop_x1 = + std::max(0, center_x - static_cast(half_w * (1 + expandratio))); + crop_y1 = + std::max(0, center_y - static_cast(half_h * (1 + expandratio))); + crop_x2 = std::min(img.cols - 1, + static_cast(center_x + half_w * (1 + expandratio))); + crop_y2 = std::min(img.rows - 1, + static_cast(center_y + half_h * (1 + expandratio))); + crop_img = + img(cv::Range(crop_y1, crop_y2 + 1), cv::Range(crop_x1, crop_x2 + 1)); + + center.clear(); + center.emplace_back((crop_x1 + crop_x2) / 2); + center.emplace_back((crop_y1 + crop_y2) / 2); + scale.clear(); + scale.emplace_back((crop_x2 - crop_x1)); + scale.emplace_back((crop_y2 - crop_y1)); +} + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/lite_shitu/src/recognition.cc b/exprements/PaddleClas/deploy/lite_shitu/src/recognition.cc new file mode 100644 index 00000000..0e711f38 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/src/recognition.cc @@ -0,0 +1,143 @@ +// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "include/recognition.h" + +namespace PPShiTu { +std::vector Recognition::RunRecModel(const cv::Mat &img, + double &cost_time) { + + // Read img + cv::Mat resize_image = ResizeImage(img); + + cv::Mat img_fp; + resize_image.convertTo(img_fp, CV_32FC3, scale); + + // Prepare input data from image + std::unique_ptr input_tensor(std::move(this->predictor->GetInput(0))); + input_tensor->Resize({1, 3, img_fp.rows, img_fp.cols}); + auto *data0 = input_tensor->mutable_data(); + + const float *dimg = reinterpret_cast(img_fp.data); + NeonMeanScale(dimg, data0, img_fp.rows * img_fp.cols); + + auto start = std::chrono::system_clock::now(); + // Run predictor + this->predictor->Run(); + + // Get output and post process + std::unique_ptr output_tensor( + std::move(this->predictor->GetOutput(1))); + auto *output_data = output_tensor->data(); + auto end = std::chrono::system_clock::now(); + auto duration = + std::chrono::duration_cast(end - start); + cost_time = double(duration.count()) * + std::chrono::microseconds::period::num / + std::chrono::microseconds::period::den; + + int output_size = 1; + for (auto dim : output_tensor->shape()) { + output_size *= dim; + } + + cv::Mat output_image; + auto results = PostProcess(output_data, output_size, output_image); + return results; +} + +void Recognition::NeonMeanScale(const float *din, float *dout, int size) { + + if (this->mean.size() != 3 || this->std.size() != 3) { + std::cerr << "[ERROR] mean or scale size must equal to 3\n"; + exit(1); + } + float32x4_t vmean0 = vdupq_n_f32(mean[0]); + float32x4_t vmean1 = vdupq_n_f32(mean[1]); + float32x4_t vmean2 = vdupq_n_f32(mean[2]); + float32x4_t vscale0 = vdupq_n_f32(std[0]); + float32x4_t vscale1 = vdupq_n_f32(std[1]); + float32x4_t vscale2 = vdupq_n_f32(std[2]); + + float *dout_c0 = dout; + float *dout_c1 = dout + size; + float *dout_c2 = dout + size * 2; + + int i = 0; + for (; i < size - 3; i += 4) { + float32x4x3_t vin3 = vld3q_f32(din); + float32x4_t vsub0 = vsubq_f32(vin3.val[0], vmean0); + float32x4_t vsub1 = vsubq_f32(vin3.val[1], vmean1); + float32x4_t vsub2 = vsubq_f32(vin3.val[2], vmean2); + float32x4_t vs0 = vmulq_f32(vsub0, vscale0); + float32x4_t vs1 = vmulq_f32(vsub1, vscale1); + float32x4_t vs2 = vmulq_f32(vsub2, vscale2); + vst1q_f32(dout_c0, vs0); + vst1q_f32(dout_c1, vs1); + vst1q_f32(dout_c2, vs2); + + din += 12; + dout_c0 += 4; + dout_c1 += 4; + dout_c2 += 4; + } + for (; i < size; i++) { + *(dout_c0++) = (*(din++) - this->mean[0]) * this->std[0]; + *(dout_c1++) = (*(din++) - this->mean[1]) * this->std[1]; + *(dout_c2++) = (*(din++) - this->mean[2]) * this->std[2]; + } +} + +cv::Mat Recognition::ResizeImage(const cv::Mat &img) { + cv::Mat resize_img; + cv::resize(img, resize_img, cv::Size(this->size, this->size)); + return resize_img; +} +std::vector Recognition::PostProcess(const float *output_data, + int output_size, + cv::Mat &output_image) { + + int max_indices[this->topk]; + double max_scores[this->topk]; + for (int i = 0; i < this->topk; i++) { + max_indices[i] = 0; + max_scores[i] = 0; + } + for (int i = 0; i < output_size; i++) { + float score = output_data[i]; + int index = i; + for (int j = 0; j < this->topk; j++) { + if (score > max_scores[j]) { + index += max_indices[j]; + max_indices[j] = index - max_indices[j]; + index -= max_indices[j]; + score += max_scores[j]; + max_scores[j] = score - max_scores[j]; + score -= max_scores[j]; + } + } + } + + std::vector results(this->topk); + for (int i = 0; i < results.size(); i++) { + results[i].class_name = "Unknown"; + if (max_indices[i] >= 0 && max_indices[i] < this->label_list.size()) { + results[i].class_name = this->label_list[max_indices[i]]; + } + results[i].score = max_scores[i]; + results[i].class_id = max_indices[i]; + } + return results; +} +} diff --git a/exprements/PaddleClas/deploy/lite_shitu/src/utils.cc b/exprements/PaddleClas/deploy/lite_shitu/src/utils.cc new file mode 100644 index 00000000..3bc46177 --- /dev/null +++ b/exprements/PaddleClas/deploy/lite_shitu/src/utils.cc @@ -0,0 +1,57 @@ +// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "include/utils.h" + +namespace PPShiTu { + +void nms(std::vector &input_boxes, float nms_threshold, + bool rec_nms) { + if (!rec_nms) { + std::sort(input_boxes.begin(), input_boxes.end(), + [](ObjectResult a, ObjectResult b) { + return a.confidence > b.confidence; + }); + } else { + std::sort(input_boxes.begin(), input_boxes.end(), + [](ObjectResult a, ObjectResult b) { + return a.rec_result[0].score > b.rec_result[0].score; + }); + } + std::vector vArea(input_boxes.size()); + for (int i = 0; i < int(input_boxes.size()); ++i) { + vArea[i] = (input_boxes.at(i).rect[2] - input_boxes.at(i).rect[0] + 1) * + (input_boxes.at(i).rect[3] - input_boxes.at(i).rect[1] + 1); + } + for (int i = 0; i < int(input_boxes.size()); ++i) { + for (int j = i + 1; j < int(input_boxes.size());) { + float xx1 = (std::max)(input_boxes[i].rect[0], input_boxes[j].rect[0]); + float yy1 = (std::max)(input_boxes[i].rect[1], input_boxes[j].rect[1]); + float xx2 = (std::min)(input_boxes[i].rect[2], input_boxes[j].rect[2]); + float yy2 = (std::min)(input_boxes[i].rect[3], input_boxes[j].rect[3]); + float w = (std::max)(float(0), xx2 - xx1 + 1); + float h = (std::max)(float(0), yy2 - yy1 + 1); + float inter = w * h; + float ovr = inter / (vArea[i] + vArea[j] - inter); + if (ovr >= nms_threshold) { + input_boxes.erase(input_boxes.begin() + j); + vArea.erase(vArea.begin() + j); + } else { + j++; + } + } + } +} + +} // namespace PPShiTu diff --git a/exprements/PaddleClas/deploy/paddle2onnx/readme.md b/exprements/PaddleClas/deploy/paddle2onnx/readme.md new file mode 100644 index 00000000..d1307ea8 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddle2onnx/readme.md @@ -0,0 +1,61 @@ +# paddle2onnx 模型转化与预测 + +本章节介绍 ResNet50_vd 模型如何转化为 ONNX 模型，并基于 ONNX 引擎预测。 + +## 1. 环境准备 + +需要准备 Paddle2ONNX 模型转化环境，和 ONNX 模型预测环境。 + +Paddle2ONNX 支持将 PaddlePaddle 模型格式转化到 ONNX 模型格式，算子目前稳定支持导出 ONNX Opset 9~11，部分Paddle算子支持更低的ONNX Opset转换。 +更多细节可参考 [Paddle2ONNX](https://github.com/PaddlePaddle/Paddle2ONNX/blob/develop/README_zh.md) + +- 安装 Paddle2ONNX +``` +python3.7 -m pip install paddle2onnx +``` + +- 安装 ONNX 运行时 +``` +python3.7 -m pip install onnxruntime +``` + +## 2. 模型转换 + +- ResNet50_vd inference模型下载 + +``` +cd deploy +mkdir models && cd models +wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet50_vd_infer.tar && tar xf ResNet50_vd_infer.tar +cd .. +``` + +- 模型转换 + +使用 Paddle2ONNX 将 Paddle 静态图模型转换为 ONNX 模型格式： +``` +paddle2onnx --model_dir=./models/ResNet50_vd_infer/ \ +--model_filename=inference.pdmodel \ +--params_filename=inference.pdiparams \ +--save_file=./models/ResNet50_vd_infer/inference.onnx \ +--opset_version=10 \ +--enable_onnx_checker=True +``` + +执行完毕后，ONNX 模型 `inference.onnx` 会被保存在 `./models/ResNet50_vd_infer/` 路径下 + +## 3. onnx 预测 + +执行如下命令： +``` +python3.7 python/predict_cls.py \ +-c configs/inference_cls.yaml \ +-o Global.use_onnx=True \ +-o Global.use_gpu=False \ +-o Global.inference_model_dir=./models/ResNet50_vd_infer +``` + +结果如下： +``` +ILSVRC2012_val_00000010.jpeg: class id(s): [153, 204, 229, 332, 155], score(s): [0.69, 0.10, 0.02, 0.01, 0.01], label_name(s): ['Maltese dog, Maltese terrier, Maltese', 'Lhasa, Lhasa apso', 'Old English sheepdog, bobtail', 'Angora, Angora rabbit', 'Shih-Tzu'] +``` diff --git a/exprements/PaddleClas/deploy/paddleserving/__init__.py b/exprements/PaddleClas/deploy/paddleserving/__init__.py new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/deploy/paddleserving/classification_web_service.py b/exprements/PaddleClas/deploy/paddleserving/classification_web_service.py new file mode 100644 index 00000000..75eb335a --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/classification_web_service.py @@ -0,0 +1,74 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import sys +from paddle_serving_app.reader import Sequential, URL2Image, Resize, CenterCrop, RGB2BGR, Transpose, Div, Normalize, Base64ToImage +try: + from paddle_serving_server_gpu.web_service import WebService, Op +except ImportError: + from paddle_serving_server.web_service import WebService, Op +import logging +import numpy as np +import base64, cv2 + + +class ImagenetOp(Op): + def init_op(self): + self.seq = Sequential([ + Resize(256), CenterCrop(224), RGB2BGR(), Transpose((2, 0, 1)), + Div(255), Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], + True) + ]) + self.label_dict = {} + label_idx = 0 + with open("imagenet.label") as fin: + for line in fin: + self.label_dict[label_idx] = line.strip() + label_idx += 1 + + def preprocess(self, input_dicts, data_id, log_id): + (_, input_dict), = input_dicts.items() + batch_size = len(input_dict.keys()) + imgs = [] + for key in input_dict.keys(): + data = base64.b64decode(input_dict[key].encode('utf8')) + data = np.fromstring(data, np.uint8) + im = cv2.imdecode(data, cv2.IMREAD_COLOR) + img = self.seq(im) + imgs.append(img[np.newaxis, :].copy()) + input_imgs = np.concatenate(imgs, axis=0) + return {"inputs": input_imgs}, False, None, "" + + def postprocess(self, input_dicts, fetch_dict, data_id, log_id): + score_list = fetch_dict["prediction"] + result = {"label": [], "prob": []} + for score in score_list: + score = score.tolist() + max_score = max(score) + result["label"].append(self.label_dict[score.index(max_score)] + .strip().replace(",", "")) + result["prob"].append(max_score) + result["label"] = str(result["label"]) + result["prob"] = str(result["prob"]) + return result, None, "" + + +class ImageService(WebService): + def get_pipeline_response(self, read_op): + image_op = ImagenetOp(name="imagenet", input_ops=[read_op]) + return image_op + + +uci_service = ImageService(name="imagenet") +uci_service.prepare_pipeline_config("config.yml") +uci_service.run_service() diff --git a/exprements/PaddleClas/deploy/paddleserving/config.yml b/exprements/PaddleClas/deploy/paddleserving/config.yml new file mode 100644 index 00000000..d9f464dd --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/config.yml @@ -0,0 +1,33 @@ +#worker_num, 最大并发数。当build_dag_each_worker=True时, 框架会创建worker_num个进程，每个进程内构建grpcSever和DAG +##当build_dag_each_worker=False时，框架会设置主线程grpc线程池的max_workers=worker_num +worker_num: 1 + +#http端口, rpc_port和http_port不允许同时为空。当rpc_port可用且http_port为空时，不自动生成http_port +http_port: 18080 +rpc_port: 9993 + +dag: + #op资源类型, True, 为线程模型；False，为进程模型 + is_thread_op: False +op: + imagenet: + #并发数，is_thread_op=True时，为线程并发；否则为进程并发 + concurrency: 1 + + #当op配置没有server_endpoints时，从local_service_conf读取本地服务配置 + local_service_conf: + + #uci模型路径 + model_config: ResNet50_vd_serving + + #计算硬件类型: 空缺时由devices决定(CPU/GPU)，0=cpu, 1=gpu, 2=tensorRT, 3=arm cpu, 4=kunlun xpu + device_type: 1 + + #计算硬件ID，当devices为""或不写时为CPU预测；当devices为"0", "0,1,2"时为GPU预测，表示使用的GPU卡 + devices: "0" # "0,1" + + #client类型，包括brpc, grpc和local_predictor.local_predictor不启动Serving服务，进程内预测 + client_type: local_predictor + + #Fetch结果列表，以client_config中fetch_var的alias_name为准 + fetch_list: ["prediction"] diff --git a/exprements/PaddleClas/deploy/paddleserving/cpu_utilization.py b/exprements/PaddleClas/deploy/paddleserving/cpu_utilization.py new file mode 100644 index 00000000..984c7237 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/cpu_utilization.py @@ -0,0 +1,4 @@ +import psutil +cpu_utilization=psutil.cpu_percent(1,False) +print('CPU_UTILIZATION:', cpu_utilization) + diff --git a/exprements/PaddleClas/deploy/paddleserving/daisy.jpg b/exprements/PaddleClas/deploy/paddleserving/daisy.jpg new file mode 100644 index 00000000..7edeca63 Binary files /dev/null and b/exprements/PaddleClas/deploy/paddleserving/daisy.jpg differ diff --git a/exprements/PaddleClas/deploy/paddleserving/imagenet.label b/exprements/PaddleClas/deploy/paddleserving/imagenet.label new file mode 100644 index 00000000..d7146735 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/imagenet.label @@ -0,0 +1,1000 @@ +tench, Tinca tinca, +goldfish, Carassius auratus, +great white shark, white shark, man-eater, man-eating shark, Carcharodon carcharias, +tiger shark, Galeocerdo cuvieri, +hammerhead, hammerhead shark, +electric ray, crampfish, numbfish, torpedo, +stingray, +cock, +hen, +ostrich, Struthio camelus, +brambling, Fringilla montifringilla, +goldfinch, Carduelis carduelis, +house finch, linnet, Carpodacus mexicanus, +junco, snowbird, +indigo bunting, indigo finch, indigo bird, Passerina cyanea, +robin, American robin, Turdus migratorius, +bulbul, +jay, +magpie, +chickadee, +water ouzel, dipper, +kite, +bald eagle, American eagle, Haliaeetus leucocephalus, +vulture, +great grey owl, great gray owl, Strix nebulosa, +European fire salamander, Salamandra salamandra, +common newt, Triturus vulgaris, +eft, +spotted salamander, Ambystoma maculatum, +axolotl, mud puppy, Ambystoma mexicanum, +bullfrog, Rana catesbeiana, +tree frog, tree-frog, +tailed frog, bell toad, ribbed toad, tailed toad, Ascaphus trui, +loggerhead, loggerhead turtle, Caretta caretta, +leatherback turtle, leatherback, leathery turtle, Dermochelys coriacea, +mud turtle, +terrapin, +box turtle, box tortoise, +banded gecko, +common iguana, iguana, Iguana iguana, +American chameleon, anole, Anolis carolinensis, +whiptail, whiptail lizard, +agama, +frilled lizard, Chlamydosaurus kingi, +alligator lizard, +Gila monster, Heloderma suspectum, +green lizard, Lacerta viridis, +African chameleon, Chamaeleo chamaeleon, +Komodo dragon, Komodo lizard, dragon lizard, giant lizard, Varanus komodoensis, +African crocodile, Nile crocodile, Crocodylus niloticus, +American alligator, Alligator mississipiensis, +triceratops, +thunder snake, worm snake, Carphophis amoenus, +ringneck snake, ring-necked snake, ring snake, +hognose snake, puff adder, sand viper, +green snake, grass snake, +king snake, kingsnake, +garter snake, grass snake, +water snake, +vine snake, +night snake, Hypsiglena torquata, +boa constrictor, Constrictor constrictor, +rock python, rock snake, Python sebae, +Indian cobra, Naja naja, +green mamba, +sea snake, +horned viper, cerastes, sand viper, horned asp, Cerastes cornutus, +diamondback, diamondback rattlesnake, Crotalus adamanteus, +sidewinder, horned rattlesnake, Crotalus cerastes, +trilobite, +harvestman, daddy longlegs, Phalangium opilio, +scorpion, +black and gold garden spider, Argiope aurantia, +barn spider, Araneus cavaticus, +garden spider, Aranea diademata, +black widow, Latrodectus mactans, +tarantula, +wolf spider, hunting spider, +tick, +centipede, +black grouse, +ptarmigan, +ruffed grouse, partridge, Bonasa umbellus, +prairie chicken, prairie grouse, prairie fowl, +peacock, +quail, +partridge, +African grey, African gray, Psittacus erithacus, +macaw, +sulphur-crested cockatoo, Kakatoe galerita, Cacatua galerita, +lorikeet, +coucal, +bee eater, +hornbill, +hummingbird, +jacamar, +toucan, +drake, +red-breasted merganser, Mergus serrator, +goose, +black swan, Cygnus atratus, +tusker, +echidna, spiny anteater, anteater, +platypus, duckbill, duckbilled platypus, duck-billed platypus, Ornithorhynchus anatinus, +wallaby, brush kangaroo, +koala, koala bear, kangaroo bear, native bear, Phascolarctos cinereus, +wombat, +jellyfish, +sea anemone, anemone, +brain coral, +flatworm, platyhelminth, +nematode, nematode worm, roundworm, +conch, +snail, +slug, +sea slug, nudibranch, +chiton, coat-of-mail shell, sea cradle, polyplacophore, +chambered nautilus, pearly nautilus, nautilus, +Dungeness crab, Cancer magister, +rock crab, Cancer irroratus, +fiddler crab, +king crab, Alaska crab, Alaskan king crab, Alaska king crab, Paralithodes camtschatica, +American lobster, Northern lobster, Maine lobster, Homarus americanus, +spiny lobster, langouste, rock lobster, crawfish, crayfish, sea crawfish, +crayfish, crawfish, crawdad, crawdaddy, +hermit crab, +isopod, +white stork, Ciconia ciconia, +black stork, Ciconia nigra, +spoonbill, +flamingo, +little blue heron, Egretta caerulea, +American egret, great white heron, Egretta albus, +bittern, +crane, +limpkin, Aramus pictus, +European gallinule, Porphyrio porphyrio, +American coot, marsh hen, mud hen, water hen, Fulica americana, +bustard, +ruddy turnstone, Arenaria interpres, +red-backed sandpiper, dunlin, Erolia alpina, +redshank, Tringa totanus, +dowitcher, +oystercatcher, oyster catcher, +pelican, +king penguin, Aptenodytes patagonica, +albatross, mollymawk, +grey whale, gray whale, devilfish, Eschrichtius gibbosus, Eschrichtius robustus, +killer whale, killer, orca, grampus, sea wolf, Orcinus orca, +dugong, Dugong dugon, +sea lion, +Chihuahua, +Japanese spaniel, +Maltese dog, Maltese terrier, Maltese, +Pekinese, Pekingese, Peke, +Shih-Tzu, +Blenheim spaniel, +papillon, +toy terrier, +Rhodesian ridgeback, +Afghan hound, Afghan, +basset, basset hound, +beagle, +bloodhound, sleuthhound, +bluetick, +black-and-tan coonhound, +Walker hound, Walker foxhound, +English foxhound, +redbone, +borzoi, Russian wolfhound, +Irish wolfhound, +Italian greyhound, +whippet, +Ibizan hound, Ibizan Podenco, +Norwegian elkhound, elkhound, +otterhound, otter hound, +Saluki, gazelle hound, +Scottish deerhound, deerhound, +Weimaraner, +Staffordshire bullterrier, Staffordshire bull terrier, +American Staffordshire terrier, Staffordshire terrier, American pit bull terrier, pit bull terrier, +Bedlington terrier, +Border terrier, +Kerry blue terrier, +Irish terrier, +Norfolk terrier, +Norwich terrier, +Yorkshire terrier, +wire-haired fox terrier, +Lakeland terrier, +Sealyham terrier, Sealyham, +Airedale, Airedale terrier, +cairn, cairn terrier, +Australian terrier, +Dandie Dinmont, Dandie Dinmont terrier, +Boston bull, Boston terrier, +miniature schnauzer, +giant schnauzer, +standard schnauzer, +Scotch terrier, Scottish terrier, Scottie, +Tibetan terrier, chrysanthemum dog, +silky terrier, Sydney silky, +soft-coated wheaten terrier, +West Highland white terrier, +Lhasa, Lhasa apso, +flat-coated retriever, +curly-coated retriever, +golden retriever, +Labrador retriever, +Chesapeake Bay retriever, +German short-haired pointer, +vizsla, Hungarian pointer, +English setter, +Irish setter, red setter, +Gordon setter, +Brittany spaniel, +clumber, clumber spaniel, +English springer, English springer spaniel, +Welsh springer spaniel, +cocker spaniel, English cocker spaniel, cocker, +Sussex spaniel, +Irish water spaniel, +kuvasz, +schipperke, +groenendael, +malinois, +briard, +kelpie, +komondor, +Old English sheepdog, bobtail, +Shetland sheepdog, Shetland sheep dog, Shetland, +collie, +Border collie, +Bouvier des Flandres, Bouviers des Flandres, +Rottweiler, +German shepherd, German shepherd dog, German police dog, alsatian, +Doberman, Doberman pinscher, +miniature pinscher, +Greater Swiss Mountain dog, +Bernese mountain dog, +Appenzeller, +EntleBucher, +boxer, +bull mastiff, +Tibetan mastiff, +French bulldog, +Great Dane, +Saint Bernard, St Bernard, +Eskimo dog, husky, +malamute, malemute, Alaskan malamute, +Siberian husky, +dalmatian, coach dog, carriage dog, +affenpinscher, monkey pinscher, monkey dog, +basenji, +pug, pug-dog, +Leonberg, +Newfoundland, Newfoundland dog, +Great Pyrenees, +Samoyed, Samoyede, +Pomeranian, +chow, chow chow, +keeshond, +Brabancon griffon, +Pembroke, Pembroke Welsh corgi, +Cardigan, Cardigan Welsh corgi, +toy poodle, +miniature poodle, +standard poodle, +Mexican hairless, +timber wolf, grey wolf, gray wolf, Canis lupus, +white wolf, Arctic wolf, Canis lupus tundrarum, +red wolf, maned wolf, Canis rufus, Canis niger, +coyote, prairie wolf, brush wolf, Canis latrans, +dingo, warrigal, warragal, Canis dingo, +dhole, Cuon alpinus, +African hunting dog, hyena dog, Cape hunting dog, Lycaon pictus, +hyena, hyaena, +red fox, Vulpes vulpes, +kit fox, Vulpes macrotis, +Arctic fox, white fox, Alopex lagopus, +grey fox, gray fox, Urocyon cinereoargenteus, +tabby, tabby cat, +tiger cat, +Persian cat, +Siamese cat, Siamese, +Egyptian cat, +cougar, puma, catamount, mountain lion, painter, panther, Felis concolor, +lynx, catamount, +leopard, Panthera pardus, +snow leopard, ounce, Panthera uncia, +jaguar, panther, Panthera onca, Felis onca, +lion, king of beasts, Panthera leo, +tiger, Panthera tigris, +cheetah, chetah, Acinonyx jubatus, +brown bear, bruin, Ursus arctos, +American black bear, black bear, Ursus americanus, Euarctos americanus, +ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus, +sloth bear, Melursus ursinus, Ursus ursinus, +mongoose, +meerkat, mierkat, +tiger beetle, +ladybug, ladybeetle, lady beetle, ladybird, ladybird beetle, +ground beetle, carabid beetle, +long-horned beetle, longicorn, longicorn beetle, +leaf beetle, chrysomelid, +dung beetle, +rhinoceros beetle, +weevil, +fly, +bee, +ant, emmet, pismire, +grasshopper, hopper, +cricket, +walking stick, walkingstick, stick insect, +cockroach, roach, +mantis, mantid, +cicada, cicala, +leafhopper, +lacewing, lacewing fly, +"dragonfly, darning needle, devils darning needle, sewing needle, snake feeder, snake doctor, mosquito hawk, skeeter hawk", +damselfly, +admiral, +ringlet, ringlet butterfly, +monarch, monarch butterfly, milkweed butterfly, Danaus plexippus, +cabbage butterfly, +sulphur butterfly, sulfur butterfly, +lycaenid, lycaenid butterfly, +starfish, sea star, +sea urchin, +sea cucumber, holothurian, +wood rabbit, cottontail, cottontail rabbit, +hare, +Angora, Angora rabbit, +hamster, +porcupine, hedgehog, +fox squirrel, eastern fox squirrel, Sciurus niger, +marmot, +beaver, +guinea pig, Cavia cobaya, +sorrel, +zebra, +hog, pig, grunter, squealer, Sus scrofa, +wild boar, boar, Sus scrofa, +warthog, +hippopotamus, hippo, river horse, Hippopotamus amphibius, +ox, +water buffalo, water ox, Asiatic buffalo, Bubalus bubalis, +bison, +ram, tup, +bighorn, bighorn sheep, cimarron, Rocky Mountain bighorn, Rocky Mountain sheep, Ovis canadensis, +ibex, Capra ibex, +hartebeest, +impala, Aepyceros melampus, +gazelle, +Arabian camel, dromedary, Camelus dromedarius, +llama, +weasel, +mink, +polecat, fitch, foulmart, foumart, Mustela putorius, +black-footed ferret, ferret, Mustela nigripes, +otter, +skunk, polecat, wood pussy, +badger, +armadillo, +three-toed sloth, ai, Bradypus tridactylus, +orangutan, orang, orangutang, Pongo pygmaeus, +gorilla, Gorilla gorilla, +chimpanzee, chimp, Pan troglodytes, +gibbon, Hylobates lar, +siamang, Hylobates syndactylus, Symphalangus syndactylus, +guenon, guenon monkey, +patas, hussar monkey, Erythrocebus patas, +baboon, +macaque, +langur, +colobus, colobus monkey, +proboscis monkey, Nasalis larvatus, +marmoset, +capuchin, ringtail, Cebus capucinus, +howler monkey, howler, +titi, titi monkey, +spider monkey, Ateles geoffroyi, +squirrel monkey, Saimiri sciureus, +Madagascar cat, ring-tailed lemur, Lemur catta, +indri, indris, Indri indri, Indri brevicaudatus, +Indian elephant, Elephas maximus, +African elephant, Loxodonta africana, +lesser panda, red panda, panda, bear cat, cat bear, Ailurus fulgens, +giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca, +barracouta, snoek, +eel, +coho, cohoe, coho salmon, blue jack, silver salmon, Oncorhynchus kisutch, +rock beauty, Holocanthus tricolor, +anemone fish, +sturgeon, +gar, garfish, garpike, billfish, Lepisosteus osseus, +lionfish, +puffer, pufferfish, blowfish, globefish, +abacus, +abaya, +"academic gown, academic robe, judges robe", +accordion, piano accordion, squeeze box, +acoustic guitar, +aircraft carrier, carrier, flattop, attack aircraft carrier, +airliner, +airship, dirigible, +altar, +ambulance, +amphibian, amphibious vehicle, +analog clock, +apiary, bee house, +apron, +ashcan, trash can, garbage can, wastebin, ash bin, ash-bin, ashbin, dustbin, trash barrel, trash bin, +assault rifle, assault gun, +backpack, back pack, knapsack, packsack, rucksack, haversack, +bakery, bakeshop, bakehouse, +balance beam, beam, +balloon, +ballpoint, ballpoint pen, ballpen, Biro, +Band Aid, +banjo, +bannister, banister, balustrade, balusters, handrail, +barbell, +barber chair, +barbershop, +barn, +barometer, +barrel, cask, +barrow, garden cart, lawn cart, wheelbarrow, +baseball, +basketball, +bassinet, +bassoon, +bathing cap, swimming cap, +bath towel, +bathtub, bathing tub, bath, tub, +beach wagon, station wagon, wagon, estate car, beach waggon, station waggon, waggon, +beacon, lighthouse, beacon light, pharos, +beaker, +bearskin, busby, shako, +beer bottle, +beer glass, +bell cote, bell cot, +bib, +bicycle-built-for-two, tandem bicycle, tandem, +bikini, two-piece, +binder, ring-binder, +binoculars, field glasses, opera glasses, +birdhouse, +boathouse, +bobsled, bobsleigh, bob, +bolo tie, bolo, bola tie, bola, +bonnet, poke bonnet, +bookcase, +bookshop, bookstore, bookstall, +bottlecap, +bow, +bow tie, bow-tie, bowtie, +brass, memorial tablet, plaque, +brassiere, bra, bandeau, +breakwater, groin, groyne, mole, bulwark, seawall, jetty, +breastplate, aegis, egis, +broom, +bucket, pail, +buckle, +bulletproof vest, +bullet train, bullet, +butcher shop, meat market, +cab, hack, taxi, taxicab, +caldron, cauldron, +candle, taper, wax light, +cannon, +canoe, +can opener, tin opener, +cardigan, +car mirror, +carousel, carrousel, merry-go-round, roundabout, whirligig, +"carpenters kit, tool kit", +carton, +car wheel, +cash machine, cash dispenser, automated teller machine, automatic teller machine, automated teller, automatic teller, ATM, +cassette, +cassette player, +castle, +catamaran, +CD player, +cello, violoncello, +cellular telephone, cellular phone, cellphone, cell, mobile phone, +chain, +chainlink fence, +chain mail, ring mail, mail, chain armor, chain armour, ring armor, ring armour, +chain saw, chainsaw, +chest, +chiffonier, commode, +chime, bell, gong, +china cabinet, china closet, +Christmas stocking, +church, church building, +cinema, movie theater, movie theatre, movie house, picture palace, +cleaver, meat cleaver, chopper, +cliff dwelling, +cloak, +clog, geta, patten, sabot, +cocktail shaker, +coffee mug, +coffeepot, +coil, spiral, volute, whorl, helix, +combination lock, +computer keyboard, keypad, +confectionery, confectionary, candy store, +container ship, containership, container vessel, +convertible, +corkscrew, bottle screw, +cornet, horn, trumpet, trump, +cowboy boot, +cowboy hat, ten-gallon hat, +cradle, +crane, +crash helmet, +crate, +crib, cot, +Crock Pot, +croquet ball, +crutch, +cuirass, +dam, dike, dyke, +desk, +desktop computer, +dial telephone, dial phone, +diaper, nappy, napkin, +digital clock, +digital watch, +dining table, board, +dishrag, dishcloth, +dishwasher, dish washer, dishwashing machine, +disk brake, disc brake, +dock, dockage, docking facility, +dogsled, dog sled, dog sleigh, +dome, +doormat, welcome mat, +drilling platform, offshore rig, +drum, membranophone, tympan, +drumstick, +dumbbell, +Dutch oven, +electric fan, blower, +electric guitar, +electric locomotive, +entertainment center, +envelope, +espresso maker, +face powder, +feather boa, boa, +file, file cabinet, filing cabinet, +fireboat, +fire engine, fire truck, +fire screen, fireguard, +flagpole, flagstaff, +flute, transverse flute, +folding chair, +football helmet, +forklift, +fountain, +fountain pen, +four-poster, +freight car, +French horn, horn, +frying pan, frypan, skillet, +fur coat, +garbage truck, dustcart, +gasmask, respirator, gas helmet, +gas pump, gasoline pump, petrol pump, island dispenser, +goblet, +go-kart, +golf ball, +golfcart, golf cart, +gondola, +gong, tam-tam, +gown, +grand piano, grand, +greenhouse, nursery, glasshouse, +grille, radiator grille, +grocery store, grocery, food market, market, +guillotine, +hair slide, +hair spray, +half track, +hammer, +hamper, +hand blower, blow dryer, blow drier, hair dryer, hair drier, +hand-held computer, hand-held microcomputer, +handkerchief, hankie, hanky, hankey, +hard disc, hard disk, fixed disk, +harmonica, mouth organ, harp, mouth harp, +harp, +harvester, reaper, +hatchet, +holster, +home theater, home theatre, +honeycomb, +hook, claw, +hoopskirt, crinoline, +horizontal bar, high bar, +horse cart, horse-cart, +hourglass, +iPod, +iron, smoothing iron, +"jack-o-lantern", +jean, blue jean, denim, +jeep, landrover, +jersey, T-shirt, tee shirt, +jigsaw puzzle, +jinrikisha, ricksha, rickshaw, +joystick, +kimono, +knee pad, +knot, +lab coat, laboratory coat, +ladle, +lampshade, lamp shade, +laptop, laptop computer, +lawn mower, mower, +lens cap, lens cover, +letter opener, paper knife, paperknife, +library, +lifeboat, +lighter, light, igniter, ignitor, +limousine, limo, +liner, ocean liner, +lipstick, lip rouge, +Loafer, +lotion, +loudspeaker, speaker, speaker unit, loudspeaker system, speaker system, +"loupe, jewelers loupe", +lumbermill, sawmill, +magnetic compass, +mailbag, postbag, +mailbox, letter box, +maillot, +maillot, tank suit, +manhole cover, +maraca, +marimba, xylophone, +mask, +matchstick, +maypole, +maze, labyrinth, +measuring cup, +medicine chest, medicine cabinet, +megalith, megalithic structure, +microphone, mike, +microwave, microwave oven, +military uniform, +milk can, +minibus, +miniskirt, mini, +minivan, +missile, +mitten, +mixing bowl, +mobile home, manufactured home, +Model T, +modem, +monastery, +monitor, +moped, +mortar, +mortarboard, +mosque, +mosquito net, +motor scooter, scooter, +mountain bike, all-terrain bike, off-roader, +mountain tent, +mouse, computer mouse, +mousetrap, +moving van, +muzzle, +nail, +neck brace, +necklace, +nipple, +notebook, notebook computer, +obelisk, +oboe, hautboy, hautbois, +ocarina, sweet potato, +odometer, hodometer, mileometer, milometer, +oil filter, +organ, pipe organ, +oscilloscope, scope, cathode-ray oscilloscope, CRO, +overskirt, +oxcart, +oxygen mask, +packet, +paddle, boat paddle, +paddlewheel, paddle wheel, +padlock, +paintbrush, +"pajama, pyjama, pjs, jammies", +palace, +panpipe, pandean pipe, syrinx, +paper towel, +parachute, chute, +parallel bars, bars, +park bench, +parking meter, +passenger car, coach, carriage, +patio, terrace, +pay-phone, pay-station, +pedestal, plinth, footstall, +pencil box, pencil case, +pencil sharpener, +perfume, essence, +Petri dish, +photocopier, +pick, plectrum, plectron, +pickelhaube, +picket fence, paling, +pickup, pickup truck, +pier, +piggy bank, penny bank, +pill bottle, +pillow, +ping-pong ball, +pinwheel, +pirate, pirate ship, +pitcher, ewer, +"plane, carpenters plane, woodworking plane", +planetarium, +plastic bag, +plate rack, +plow, plough, +"plunger, plumbers helper", +Polaroid camera, Polaroid Land camera, +pole, +police van, police wagon, paddy wagon, patrol wagon, wagon, black Maria, +poncho, +pool table, billiard table, snooker table, +pop bottle, soda bottle, +pot, flowerpot, +"potters wheel", +power drill, +prayer rug, prayer mat, +printer, +prison, prison house, +projectile, missile, +projector, +puck, hockey puck, +punching bag, punch bag, punching ball, punchball, +purse, +quill, quill pen, +quilt, comforter, comfort, puff, +racer, race car, racing car, +racket, racquet, +radiator, +radio, wireless, +radio telescope, radio reflector, +rain barrel, +recreational vehicle, RV, R.V., +reel, +reflex camera, +refrigerator, icebox, +remote control, remote, +restaurant, eating house, eating place, eatery, +revolver, six-gun, six-shooter, +rifle, +rocking chair, rocker, +rotisserie, +rubber eraser, rubber, pencil eraser, +rugby ball, +rule, ruler, +running shoe, +safe, +safety pin, +saltshaker, salt shaker, +sandal, +sarong, +sax, saxophone, +scabbard, +scale, weighing machine, +school bus, +schooner, +scoreboard, +screen, CRT screen, +screw, +screwdriver, +seat belt, seatbelt, +sewing machine, +shield, buckler, +shoe shop, shoe-shop, shoe store, +shoji, +shopping basket, +shopping cart, +shovel, +shower cap, +shower curtain, +ski, +ski mask, +sleeping bag, +slide rule, slipstick, +sliding door, +slot, one-armed bandit, +snorkel, +snowmobile, +snowplow, snowplough, +soap dispenser, +soccer ball, +sock, +solar dish, solar collector, solar furnace, +sombrero, +soup bowl, +space bar, +space heater, +space shuttle, +spatula, +speedboat, +"spider web, spiders web", +spindle, +sports car, sport car, +spotlight, spot, +stage, +steam locomotive, +steel arch bridge, +steel drum, +stethoscope, +stole, +stone wall, +stopwatch, stop watch, +stove, +strainer, +streetcar, tram, tramcar, trolley, trolley car, +stretcher, +studio couch, day bed, +stupa, tope, +submarine, pigboat, sub, U-boat, +suit, suit of clothes, +sundial, +sunglass, +sunglasses, dark glasses, shades, +sunscreen, sunblock, sun blocker, +suspension bridge, +swab, swob, mop, +sweatshirt, +swimming trunks, bathing trunks, +swing, +switch, electric switch, electrical switch, +syringe, +table lamp, +tank, army tank, armored combat vehicle, armoured combat vehicle, +tape player, +teapot, +teddy, teddy bear, +television, television system, +tennis ball, +thatch, thatched roof, +theater curtain, theatre curtain, +thimble, +thresher, thrasher, threshing machine, +throne, +tile roof, +toaster, +tobacco shop, tobacconist shop, tobacconist, +toilet seat, +torch, +totem pole, +tow truck, tow car, wrecker, +toyshop, +tractor, +trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi, +tray, +trench coat, +tricycle, trike, velocipede, +trimaran, +tripod, +triumphal arch, +trolleybus, trolley coach, trackless trolley, +trombone, +tub, vat, +turnstile, +typewriter keyboard, +umbrella, +unicycle, monocycle, +upright, upright piano, +vacuum, vacuum cleaner, +vase, +vault, +velvet, +vending machine, +vestment, +viaduct, +violin, fiddle, +volleyball, +waffle iron, +wall clock, +wallet, billfold, notecase, pocketbook, +wardrobe, closet, press, +warplane, military plane, +washbasin, handbasin, washbowl, lavabo, wash-hand basin, +washer, automatic washer, washing machine, +water bottle, +water jug, +water tower, +whiskey jug, +whistle, +wig, +window screen, +window shade, +Windsor tie, +wine bottle, +wing, +wok, +wooden spoon, +wool, woolen, woollen, +worm fence, snake fence, snake-rail fence, Virginia fence, +wreck, +yawl, +yurt, +web site, website, internet site, site, +comic book, +crossword puzzle, crossword, +street sign, +traffic light, traffic signal, stoplight, +book jacket, dust cover, dust jacket, dust wrapper, +menu, +plate, +guacamole, +consomme, +hot pot, hotpot, +trifle, +ice cream, icecream, +ice lolly, lolly, lollipop, popsicle, +French loaf, +bagel, beigel, +pretzel, +cheeseburger, +hotdog, hot dog, red hot, +mashed potato, +head cabbage, +broccoli, +cauliflower, +zucchini, courgette, +spaghetti squash, +acorn squash, +butternut squash, +cucumber, cuke, +artichoke, globe artichoke, +bell pepper, +cardoon, +mushroom, +Granny Smith, +strawberry, +orange, +lemon, +fig, +pineapple, ananas, +banana, +jackfruit, jak, jack, +custard apple, +pomegranate, +hay, +carbonara, +chocolate sauce, chocolate syrup, +dough, +meat loaf, meatloaf, +pizza, pizza pie, +potpie, +burrito, +red wine, +espresso, +cup, +eggnog, +alp, +bubble, +cliff, drop, drop-off, +coral reef, +geyser, +lakeside, lakeshore, +promontory, headland, head, foreland, +sandbar, sand bar, +seashore, coast, seacoast, sea-coast, +valley, vale, +volcano, +ballplayer, baseball player, +groom, bridegroom, +scuba diver, +rapeseed, +daisy, +"yellow ladys slipper, yellow lady-slipper, Cypripedium calceolus, Cypripedium parviflorum", +corn, +acorn, +hip, rose hip, rosehip, +buckeye, horse chestnut, conker, +coral fungus, +agaric, +gyromitra, +stinkhorn, carrion fungus, +earthstar, +hen-of-the-woods, hen of the woods, Polyporus frondosus, Grifola frondosa, +bolete, +ear, spike, capitulum, +toilet tissue, toilet paper, bathroom tissue diff --git a/exprements/PaddleClas/deploy/paddleserving/imgs/results.png b/exprements/PaddleClas/deploy/paddleserving/imgs/results.png new file mode 100644 index 00000000..ad44083c Binary files /dev/null and b/exprements/PaddleClas/deploy/paddleserving/imgs/results.png differ diff --git a/exprements/PaddleClas/deploy/paddleserving/imgs/results_shitu.png b/exprements/PaddleClas/deploy/paddleserving/imgs/results_shitu.png new file mode 100644 index 00000000..5c099499 Binary files /dev/null and b/exprements/PaddleClas/deploy/paddleserving/imgs/results_shitu.png differ diff --git a/exprements/PaddleClas/deploy/paddleserving/imgs/start_server.png b/exprements/PaddleClas/deploy/paddleserving/imgs/start_server.png new file mode 100644 index 00000000..8294e19b Binary files /dev/null and b/exprements/PaddleClas/deploy/paddleserving/imgs/start_server.png differ diff --git a/exprements/PaddleClas/deploy/paddleserving/imgs/start_server_shitu.png b/exprements/PaddleClas/deploy/paddleserving/imgs/start_server_shitu.png new file mode 100644 index 00000000..f2709ec0 Binary files /dev/null and b/exprements/PaddleClas/deploy/paddleserving/imgs/start_server_shitu.png differ diff --git a/exprements/PaddleClas/deploy/paddleserving/pipeline_http_client.py b/exprements/PaddleClas/deploy/paddleserving/pipeline_http_client.py new file mode 100644 index 00000000..290d647c --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/pipeline_http_client.py @@ -0,0 +1,20 @@ +import requests +import json +import base64 +import os + + +def cv2_to_base64(image): + return base64.b64encode(image).decode('utf8') + + +if __name__ == "__main__": + url = "http://127.0.0.1:18080/imagenet/prediction" + with open(os.path.join(".", "daisy.jpg"), 'rb') as file: + image_data1 = file.read() + image = cv2_to_base64(image_data1) + + data = {"key": ["image"], "value": [image]} + for i in range(1): + r = requests.post(url=url, data=json.dumps(data)) + print(r.json()) diff --git a/exprements/PaddleClas/deploy/paddleserving/pipeline_rpc_client.py b/exprements/PaddleClas/deploy/paddleserving/pipeline_rpc_client.py new file mode 100644 index 00000000..75bcae0d --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/pipeline_rpc_client.py @@ -0,0 +1,33 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +try: + from paddle_serving_server_gpu.pipeline import PipelineClient +except ImportError: + from paddle_serving_server.pipeline import PipelineClient +import base64 + +client = PipelineClient() +client.connect(['127.0.0.1:9993']) + +def cv2_to_base64(image): + return base64.b64encode(image).decode('utf8') + +if __name__ == "__main__": + with open("daisy.jpg", 'rb') as file: + image_data = file.read() + image = cv2_to_base64(image_data) + + for i in range(1): + ret = client.predict(feed_dict={"image": image}, fetch=["label", "prob"]) + print(ret) diff --git a/exprements/PaddleClas/deploy/paddleserving/readme.md b/exprements/PaddleClas/deploy/paddleserving/readme.md new file mode 100644 index 00000000..a2fdec2d --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/readme.md @@ -0,0 +1 @@ +../../docs/zh_CN/inference_deployment/paddle_serving_deploy.md \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/__init__.py b/exprements/PaddleClas/deploy/paddleserving/recognition/__init__.py new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/config.yml b/exprements/PaddleClas/deploy/paddleserving/recognition/config.yml new file mode 100644 index 00000000..6ecc32e2 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/recognition/config.yml @@ -0,0 +1,43 @@ +#worker_num, 最大并发数。当build_dag_each_worker=True时, 框架会创建worker_num个进程，每个进程内构建grpcSever和DAG +##当build_dag_each_worker=False时，框架会设置主线程grpc线程池的max_workers=worker_num +worker_num: 1 + +#http端口, rpc_port和http_port不允许同时为空。当rpc_port可用且http_port为空时，不自动生成http_port +http_port: 18081 +rpc_port: 9994 + +dag: + #op资源类型, True, 为线程模型；False，为进程模型 + is_thread_op: False +op: + rec: + #并发数，is_thread_op=True时，为线程并发；否则为进程并发 + concurrency: 1 + + #当op配置没有server_endpoints时，从local_service_conf读取本地服务配置 + local_service_conf: + + #uci模型路径 + model_config: ../../models/general_PPLCNet_x2_5_lite_v1.0_serving + + #计算硬件类型: 空缺时由devices决定(CPU/GPU)，0=cpu, 1=gpu, 2=tensorRT, 3=arm cpu, 4=kunlun xpu + device_type: 1 + + #计算硬件ID，当devices为""或不写时为CPU预测；当devices为"0", "0,1,2"时为GPU预测，表示使用的GPU卡 + devices: "0" # "0,1" + + #client类型，包括brpc, grpc和local_predictor.local_predictor不启动Serving服务，进程内预测 + client_type: local_predictor + + #Fetch结果列表，以client_config中fetch_var的alias_name为准 + fetch_list: ["features"] + + det: + concurrency: 1 + local_service_conf: + client_type: local_predictor + device_type: 1 + devices: '0' + fetch_list: + - save_infer_model/scale_0.tmp_1 + model_config: ../../models/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_serving/ diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/daoxiangcunjinzhubing_6.jpg b/exprements/PaddleClas/deploy/paddleserving/recognition/daoxiangcunjinzhubing_6.jpg new file mode 100644 index 00000000..fc64a953 Binary files /dev/null and b/exprements/PaddleClas/deploy/paddleserving/recognition/daoxiangcunjinzhubing_6.jpg differ diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/label_list.txt b/exprements/PaddleClas/deploy/paddleserving/recognition/label_list.txt new file mode 100644 index 00000000..35e26a62 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/recognition/label_list.txt @@ -0,0 +1,2 @@ +foreground +background \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/pipeline_http_client.py b/exprements/PaddleClas/deploy/paddleserving/recognition/pipeline_http_client.py new file mode 100644 index 00000000..efc0f3af --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/recognition/pipeline_http_client.py @@ -0,0 +1,21 @@ +import requests +import json +import base64 +import os + +imgpath = "../../drink_dataset_v1.0/test_images/001.jpeg" + +def cv2_to_base64(image): + return base64.b64encode(image).decode('utf8') + +if __name__ == "__main__": + url = "http://127.0.0.1:18081/recognition/prediction" + + with open(os.path.join(".", imgpath), 'rb') as file: + image_data1 = file.read() + image = cv2_to_base64(image_data1) + data = {"key": ["image"], "value": [image]} + + for i in range(1): + r = requests.post(url=url, data=json.dumps(data)) + print(r.json()) diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/pipeline_rpc_client.py b/exprements/PaddleClas/deploy/paddleserving/recognition/pipeline_rpc_client.py new file mode 100644 index 00000000..50a1e42c --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/recognition/pipeline_rpc_client.py @@ -0,0 +1,34 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +try: + from paddle_serving_server_gpu.pipeline import PipelineClient +except ImportError: + from paddle_serving_server.pipeline import PipelineClient +import base64 + +client = PipelineClient() +client.connect(['127.0.0.1:9994']) +imgpath = "../../drink_dataset_v1.0/test_images/001.jpeg" + +def cv2_to_base64(image): + return base64.b64encode(image).decode('utf8') + +if __name__ == "__main__": + with open(imgpath, 'rb') as file: + image_data = file.read() + image = cv2_to_base64(image_data) + + for i in range(1): + ret = client.predict(feed_dict={"image": image}, fetch=["result"]) + print(ret) diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/recognition_web_service.py b/exprements/PaddleClas/deploy/paddleserving/recognition/recognition_web_service.py new file mode 100644 index 00000000..4a3478b6 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/recognition/recognition_web_service.py @@ -0,0 +1,209 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from paddle_serving_server.web_service import WebService, Op +import logging +import numpy as np +import sys +import cv2 +from paddle_serving_app.reader import * +import base64 +import os +import faiss +import pickle +import json + + +class DetOp(Op): + def init_op(self): + self.img_preprocess = Sequential([ + BGR2RGB(), Div(255.0), + Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], False), + Resize((640, 640)), Transpose((2, 0, 1)) + ]) + + self.img_postprocess = RCNNPostprocess("label_list.txt", "output") + self.threshold = 0.2 + self.max_det_results = 5 + + def generate_scale(self, im): + """ + Args: + im (np.ndarray): image (np.ndarray) + Returns: + im_scale_x: the resize ratio of X + im_scale_y: the resize ratio of Y + """ + target_size = [640, 640] + origin_shape = im.shape[:2] + resize_h, resize_w = target_size + im_scale_y = resize_h / float(origin_shape[0]) + im_scale_x = resize_w / float(origin_shape[1]) + return im_scale_y, im_scale_x + + def preprocess(self, input_dicts, data_id, log_id): + (_, input_dict), = input_dicts.items() + imgs = [] + raw_imgs = [] + for key in input_dict.keys(): + data = base64.b64decode(input_dict[key].encode('utf8')) + raw_imgs.append(data) + data = np.fromstring(data, np.uint8) + raw_im = cv2.imdecode(data, cv2.IMREAD_COLOR) + + im_scale_y, im_scale_x = self.generate_scale(raw_im) + im = self.img_preprocess(raw_im) + + im_shape = np.array(im.shape[1:]).reshape(-1) + scale_factor = np.array([im_scale_y, im_scale_x]).reshape(-1) + imgs.append({ + "image": im[np.newaxis, :], + "im_shape": im_shape[np.newaxis, :], + "scale_factor": scale_factor[np.newaxis, :], + }) + self.raw_img = raw_imgs + + feed_dict = { + "image": np.concatenate( + [x["image"] for x in imgs], axis=0), + "im_shape": np.concatenate( + [x["im_shape"] for x in imgs], axis=0), + "scale_factor": np.concatenate( + [x["scale_factor"] for x in imgs], axis=0) + } + return feed_dict, False, None, "" + + def postprocess(self, input_dicts, fetch_dict, data_id, log_id): + boxes = self.img_postprocess(fetch_dict, visualize=False) + boxes.sort(key=lambda x: x["score"], reverse=True) + boxes = filter(lambda x: x["score"] >= self.threshold, + boxes[:self.max_det_results]) + boxes = list(boxes) + for i in range(len(boxes)): + boxes[i]["bbox"][2] += boxes[i]["bbox"][0] - 1 + boxes[i]["bbox"][3] += boxes[i]["bbox"][1] - 1 + result = json.dumps(boxes) + res_dict = {"bbox_result": result, "image": self.raw_img} + return res_dict, None, "" + + +class RecOp(Op): + def init_op(self): + self.seq = Sequential([ + BGR2RGB(), Resize((224, 224)), Div(255), + Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], + False), Transpose((2, 0, 1)) + ]) + + index_dir = "../../drink_dataset_v1.0/index" + assert os.path.exists(os.path.join( + index_dir, "vector.index")), "vector.index not found ..." + assert os.path.exists(os.path.join( + index_dir, "id_map.pkl")), "id_map.pkl not found ... " + + self.searcher = faiss.read_index( + os.path.join(index_dir, "vector.index")) + + with open(os.path.join(index_dir, "id_map.pkl"), "rb") as fd: + self.id_map = pickle.load(fd) + + self.rec_nms_thresold = 0.05 + self.rec_score_thres = 0.5 + self.feature_normalize = True + self.return_k = 1 + + def preprocess(self, input_dicts, data_id, log_id): + (_, input_dict), = input_dicts.items() + raw_img = input_dict["image"][0] + data = np.frombuffer(raw_img, np.uint8) + origin_img = cv2.imdecode(data, cv2.IMREAD_COLOR) + dt_boxes = input_dict["bbox_result"] + boxes = json.loads(dt_boxes) + boxes.append({ + "category_id": 0, + "score": 1.0, + "bbox": [0, 0, origin_img.shape[1], origin_img.shape[0]] + }) + self.det_boxes = boxes + + #construct batch images for rec + imgs = [] + for box in boxes: + box = [int(x) for x in box["bbox"]] + im = origin_img[box[1]:box[3], box[0]:box[2]].copy() + img = self.seq(im) + imgs.append(img[np.newaxis, :].copy()) + + input_imgs = np.concatenate(imgs, axis=0) + return {"x": input_imgs}, False, None, "" + + def nms_to_rec_results(self, results, thresh=0.1): + filtered_results = [] + x1 = np.array([r["bbox"][0] for r in results]).astype("float32") + y1 = np.array([r["bbox"][1] for r in results]).astype("float32") + x2 = np.array([r["bbox"][2] for r in results]).astype("float32") + y2 = np.array([r["bbox"][3] for r in results]).astype("float32") + scores = np.array([r["rec_scores"] for r in results]) + + areas = (x2 - x1 + 1) * (y2 - y1 + 1) + order = scores.argsort()[::-1] + while order.size > 0: + i = order[0] + xx1 = np.maximum(x1[i], x1[order[1:]]) + yy1 = np.maximum(y1[i], y1[order[1:]]) + xx2 = np.minimum(x2[i], x2[order[1:]]) + yy2 = np.minimum(y2[i], y2[order[1:]]) + + w = np.maximum(0.0, xx2 - xx1 + 1) + h = np.maximum(0.0, yy2 - yy1 + 1) + inter = w * h + ovr = inter / (areas[i] + areas[order[1:]] - inter) + inds = np.where(ovr <= thresh)[0] + order = order[inds + 1] + filtered_results.append(results[i]) + return filtered_results + + def postprocess(self, input_dicts, fetch_dict, data_id, log_id): + batch_features = fetch_dict["features"] + + if self.feature_normalize: + feas_norm = np.sqrt( + np.sum(np.square(batch_features), axis=1, keepdims=True)) + batch_features = np.divide(batch_features, feas_norm) + + scores, docs = self.searcher.search(batch_features, self.return_k) + + results = [] + for i in range(scores.shape[0]): + pred = {} + if scores[i][0] >= self.rec_score_thres: + pred["bbox"] = [int(x) for x in self.det_boxes[i]["bbox"]] + pred["rec_docs"] = self.id_map[docs[i][0]].split()[1] + pred["rec_scores"] = scores[i][0] + results.append(pred) + + #do nms + results = self.nms_to_rec_results(results, self.rec_nms_thresold) + return {"result": str(results)}, None, "" + + +class RecognitionService(WebService): + def get_pipeline_response(self, read_op): + det_op = DetOp(name="det", input_ops=[read_op]) + rec_op = RecOp(name="rec", input_ops=[det_op]) + return rec_op + + +product_recog_service = RecognitionService(name="recognition") +product_recog_service.prepare_pipeline_config("config.yml") +product_recog_service.run_service() diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/run_cpp_serving.sh b/exprements/PaddleClas/deploy/paddleserving/recognition/run_cpp_serving.sh new file mode 100644 index 00000000..affca99c --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/recognition/run_cpp_serving.sh @@ -0,0 +1,7 @@ +nohup python3 -m paddle_serving_server.serve \ +--model ../../models/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_serving \ + --port 9293 >>log_mainbody_detection.txt 1&>2 & + +nohup python3 -m paddle_serving_server.serve \ +--model ../../models/general_PPLCNet_x2_5_lite_v1.0_serving \ +--port 9294 >>log_feature_extraction.txt 1&>2 & diff --git a/exprements/PaddleClas/deploy/paddleserving/recognition/test_cpp_serving_client.py b/exprements/PaddleClas/deploy/paddleserving/recognition/test_cpp_serving_client.py new file mode 100644 index 00000000..a2bf1ae3 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/recognition/test_cpp_serving_client.py @@ -0,0 +1,202 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import sys +import numpy as np + +from paddle_serving_client import Client +from paddle_serving_app.reader import * +import cv2 +import faiss +import os +import pickle + + +class MainbodyDetect(): + """ + pp-shitu mainbody detect. + include preprocess, process, postprocess + return detect results + Attention: Postprocess include num limit and box filter; no nms + """ + + def __init__(self): + self.preprocess = DetectionSequential([ + DetectionFile2Image(), DetectionNormalize( + [0.485, 0.456, 0.406], [0.229, 0.224, 0.225], True), + DetectionResize( + (640, 640), False, interpolation=2), DetectionTranspose( + (2, 0, 1)) + ]) + + self.client = Client() + self.client.load_client_config( + "../../models/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_client/serving_client_conf.prototxt" + ) + self.client.connect(['127.0.0.1:9293']) + + self.max_det_result = 5 + self.conf_threshold = 0.2 + + def predict(self, imgpath): + im, im_info = self.preprocess(imgpath) + im_shape = np.array(im.shape[1:]).reshape(-1) + scale_factor = np.array(list(im_info['scale_factor'])).reshape(-1) + + fetch_map = self.client.predict( + feed={ + "image": im, + "im_shape": im_shape, + "scale_factor": scale_factor, + }, + fetch=["save_infer_model/scale_0.tmp_1"], + batch=False) + return self.postprocess(fetch_map, imgpath) + + def postprocess(self, fetch_map, imgpath): + #1. get top max_det_result + det_results = fetch_map["save_infer_model/scale_0.tmp_1"] + if len(det_results) > self.max_det_result: + boxes_reserved = fetch_map[ + "save_infer_model/scale_0.tmp_1"][:self.max_det_result] + else: + boxes_reserved = det_results + + #2. do conf threshold + boxes_list = [] + for i in range(boxes_reserved.shape[0]): + if (boxes_reserved[i, 1]) > self.conf_threshold: + boxes_list.append(boxes_reserved[i, :]) + + #3. add origin image box + origin_img = cv2.imread(imgpath) + boxes_list.append( + np.array([0, 1.0, 0, 0, origin_img.shape[1], origin_img.shape[0]])) + return np.array(boxes_list) + + +class ObjectRecognition(): + """ + pp-shitu object recognion for all objects detected by MainbodyDetect. + include preprocess, process, postprocess + preprocess include preprocess for each image and batching. + Batch process + postprocess include retrieval and nms + """ + + def __init__(self): + self.client = Client() + self.client.load_client_config( + "../../models/general_PPLCNet_x2_5_lite_v1.0_client/serving_client_conf.prototxt" + ) + self.client.connect(["127.0.0.1:9294"]) + + self.seq = Sequential([ + BGR2RGB(), Resize((224, 224)), Div(255), + Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], + False), Transpose((2, 0, 1)) + ]) + + self.searcher, self.id_map = self.init_index() + + self.rec_nms_thresold = 0.05 + self.rec_score_thres = 0.5 + self.feature_normalize = True + self.return_k = 1 + + def init_index(self): + index_dir = "../../drink_dataset_v1.0/index" + assert os.path.exists(os.path.join( + index_dir, "vector.index")), "vector.index not found ..." + assert os.path.exists(os.path.join( + index_dir, "id_map.pkl")), "id_map.pkl not found ... " + + searcher = faiss.read_index(os.path.join(index_dir, "vector.index")) + + with open(os.path.join(index_dir, "id_map.pkl"), "rb") as fd: + id_map = pickle.load(fd) + return searcher, id_map + + def predict(self, det_boxes, imgpath): + #1. preprocess + batch_imgs = [] + origin_img = cv2.imread(imgpath) + for i in range(det_boxes.shape[0]): + box = det_boxes[i] + x1, y1, x2, y2 = [int(x) for x in box[2:]] + cropped_img = origin_img[y1:y2, x1:x2, :].copy() + tmp = self.seq(cropped_img) + batch_imgs.append(tmp) + batch_imgs = np.array(batch_imgs) + + #2. process + fetch_map = self.client.predict( + feed={"x": batch_imgs}, fetch=["features"], batch=True) + batch_features = fetch_map["features"] + + #3. postprocess + if self.feature_normalize: + feas_norm = np.sqrt( + np.sum(np.square(batch_features), axis=1, keepdims=True)) + batch_features = np.divide(batch_features, feas_norm) + scores, docs = self.searcher.search(batch_features, self.return_k) + + results = [] + for i in range(scores.shape[0]): + pred = {} + if scores[i][0] >= self.rec_score_thres: + pred["bbox"] = [int(x) for x in det_boxes[i, 2:]] + pred["rec_docs"] = self.id_map[docs[i][0]].split()[1] + pred["rec_scores"] = scores[i][0] + results.append(pred) + return self.nms_to_rec_results(results) + + def nms_to_rec_results(self, results): + filtered_results = [] + x1 = np.array([r["bbox"][0] for r in results]).astype("float32") + y1 = np.array([r["bbox"][1] for r in results]).astype("float32") + x2 = np.array([r["bbox"][2] for r in results]).astype("float32") + y2 = np.array([r["bbox"][3] for r in results]).astype("float32") + scores = np.array([r["rec_scores"] for r in results]) + + areas = (x2 - x1 + 1) * (y2 - y1 + 1) + order = scores.argsort()[::-1] + while order.size > 0: + i = order[0] + xx1 = np.maximum(x1[i], x1[order[1:]]) + yy1 = np.maximum(y1[i], y1[order[1:]]) + xx2 = np.minimum(x2[i], x2[order[1:]]) + yy2 = np.minimum(y2[i], y2[order[1:]]) + + w = np.maximum(0.0, xx2 - xx1 + 1) + h = np.maximum(0.0, yy2 - yy1 + 1) + inter = w * h + ovr = inter / (areas[i] + areas[order[1:]] - inter) + inds = np.where(ovr <= self.rec_nms_thresold)[0] + order = order[inds + 1] + filtered_results.append(results[i]) + return filtered_results + + +if __name__ == "__main__": + det = MainbodyDetect() + rec = ObjectRecognition() + + #1. get det_results + imgpath = "../../drink_dataset_v1.0/test_images/001.jpeg" + det_results = det.predict(imgpath) + + #2. get rec_results + rec_results = rec.predict(det_results, imgpath) + print(rec_results) diff --git a/exprements/PaddleClas/deploy/paddleserving/run_cpp_serving.sh b/exprements/PaddleClas/deploy/paddleserving/run_cpp_serving.sh new file mode 100644 index 00000000..05794b7d --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/run_cpp_serving.sh @@ -0,0 +1,2 @@ +#run cls server: +nohup python3 -m paddle_serving_server.serve --model ResNet50_vd_serving --port 9292 & diff --git a/exprements/PaddleClas/deploy/paddleserving/test_cpp_serving_client.py b/exprements/PaddleClas/deploy/paddleserving/test_cpp_serving_client.py new file mode 100644 index 00000000..50794b36 --- /dev/null +++ b/exprements/PaddleClas/deploy/paddleserving/test_cpp_serving_client.py @@ -0,0 +1,52 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import sys +from paddle_serving_client import Client + +#app +from paddle_serving_app.reader import Sequential, URL2Image, Resize +from paddle_serving_app.reader import CenterCrop, RGB2BGR, Transpose, Div, Normalize +import time + +client = Client() +client.load_client_config("./ResNet50_vd_serving/serving_server_conf.prototxt") +client.connect(["127.0.0.1:9292"]) + +label_dict = {} +label_idx = 0 +with open("imagenet.label") as fin: + for line in fin: + label_dict[label_idx] = line.strip() + label_idx += 1 + +#preprocess +seq = Sequential([ + URL2Image(), Resize(256), CenterCrop(224), RGB2BGR(), Transpose((2, 0, 1)), + Div(255), Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], True) +]) + +start = time.time() +image_file = "https://paddle-serving.bj.bcebos.com/imagenet-example/daisy.jpg" +for i in range(1): + img = seq(image_file) + fetch_map = client.predict( + feed={"inputs": img}, fetch=["prediction"], batch=False) + + prob = max(fetch_map["prediction"][0]) + label = label_dict[fetch_map["prediction"][0].tolist().index(prob)].strip( + ).replace(",", "") + print("prediction: {}, probability: {}".format(label, prob)) +end = time.time() +print(end - start) diff --git a/exprements/PaddleClas/deploy/python/__init__.py b/exprements/PaddleClas/deploy/python/__init__.py new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/deploy/python/build_gallery.py b/exprements/PaddleClas/deploy/python/build_gallery.py new file mode 100644 index 00000000..7b69a04d --- /dev/null +++ b/exprements/PaddleClas/deploy/python/build_gallery.py @@ -0,0 +1,214 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import os +import sys + +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +import cv2 +import faiss +import numpy as np +from tqdm import tqdm +import pickle + +from python.predict_rec import RecPredictor + +from utils import logger +from utils import config + + +def split_datafile(data_file, image_root, delimiter="\t"): + ''' + data_file: image path and info, which can be splitted by spacer + image_root: image path root + delimiter: delimiter + ''' + gallery_images = [] + gallery_docs = [] + with open(data_file, 'r', encoding='utf-8') as f: + lines = f.readlines() + for _, ori_line in enumerate(lines): + line = ori_line.strip().split(delimiter) + text_num = len(line) + assert text_num >= 2, f"line({ori_line}) must be splitted into at least 2 parts, but got {text_num}" + image_file = os.path.join(image_root, line[0]) + + gallery_images.append(image_file) + gallery_docs.append(ori_line.strip()) + + return gallery_images, gallery_docs + + +class GalleryBuilder(object): + def __init__(self, config): + + self.config = config + self.rec_predictor = RecPredictor(config) + assert 'IndexProcess' in config.keys(), "Index config not found ... " + self.build(config['IndexProcess']) + + def build(self, config): + ''' + build index from scratch + ''' + operation_method = config.get("index_operation", "new").lower() + + gallery_images, gallery_docs = split_datafile( + config['data_file'], config['image_root'], config['delimiter']) + + # when remove data in index, do not need extract fatures + if operation_method != "remove": + gallery_features = self._extract_features(gallery_images, config) + assert operation_method in [ + "new", "remove", "append" + ], "Only append, remove and new operation are supported" + + # vector.index: faiss index file + # id_map.pkl: use this file to map id to image_doc + if operation_method in ["remove", "append"]: + # if remove or append, vector.index and id_map.pkl must exist + assert os.path.join( + config["index_dir"], "vector.index" + ), "The vector.index dose not exist in {} when 'index_operation' is not None".format( + config["index_dir"]) + assert os.path.join( + config["index_dir"], "id_map.pkl" + ), "The id_map.pkl dose not exist in {} when 'index_operation' is not None".format( + config["index_dir"]) + index = faiss.read_index( + os.path.join(config["index_dir"], "vector.index")) + with open(os.path.join(config["index_dir"], "id_map.pkl"), + 'rb') as fd: + ids = pickle.load(fd) + assert index.ntotal == len(ids.keys( + )), "data number in index is not equal in in id_map" + else: + if not os.path.exists(config["index_dir"]): + os.makedirs(config["index_dir"], exist_ok=True) + index_method = config.get("index_method", "HNSW32") + + # if IVF method, cal ivf number automaticlly + if index_method == "IVF": + index_method = index_method + str( + min(int(len(gallery_images) // 8), 65536)) + ",Flat" + + # for binary index, add B at head of index_method + if config["dist_type"] == "hamming": + index_method = "B" + index_method + + #dist_type + dist_type = faiss.METRIC_INNER_PRODUCT if config[ + "dist_type"] == "IP" else faiss.METRIC_L2 + + #build index + if config["dist_type"] == "hamming": + index = faiss.index_binary_factory(config["embedding_size"], + index_method) + else: + index = faiss.index_factory(config["embedding_size"], + index_method, dist_type) + index = faiss.IndexIDMap2(index) + ids = {} + + if config["index_method"] == "HNSW32": + logger.warning( + "The HNSW32 method dose not support 'remove' operation") + + if operation_method != "remove": + # calculate id for new data + start_id = max(ids.keys()) + 1 if ids else 0 + ids_now = ( + np.arange(0, len(gallery_images)) + start_id).astype(np.int64) + + # only train when new index file + if operation_method == "new": + if config["dist_type"] == "hamming": + index.add(gallery_features) + else: + index.train(gallery_features) + + if not config["dist_type"] == "hamming": + index.add_with_ids(gallery_features, ids_now) + + for i, d in zip(list(ids_now), gallery_docs): + ids[i] = d + else: + if config["index_method"] == "HNSW32": + raise RuntimeError( + "The index_method: HNSW32 dose not support 'remove' operation" + ) + # remove ids in id_map, remove index data in faiss index + remove_ids = list( + filter(lambda k: ids.get(k) in gallery_docs, ids.keys())) + remove_ids = np.asarray(remove_ids) + index.remove_ids(remove_ids) + for k in remove_ids: + del ids[k] + + # store faiss index file and id_map file + if config["dist_type"] == "hamming": + faiss.write_index_binary( + index, os.path.join(config["index_dir"], "vector.index")) + else: + faiss.write_index( + index, os.path.join(config["index_dir"], "vector.index")) + + with open(os.path.join(config["index_dir"], "id_map.pkl"), 'wb') as fd: + pickle.dump(ids, fd) + + def _extract_features(self, gallery_images, config): + # extract gallery features + if config["dist_type"] == "hamming": + gallery_features = np.zeros( + [len(gallery_images), config['embedding_size'] // 8], + dtype=np.uint8) + else: + gallery_features = np.zeros( + [len(gallery_images), config['embedding_size']], + dtype=np.float32) + + #construct batch imgs and do inference + batch_size = config.get("batch_size", 32) + batch_img = [] + for i, image_file in enumerate(tqdm(gallery_images)): + img = cv2.imread(image_file) + if img is None: + logger.error("img empty, please check {}".format(image_file)) + exit() + img = img[:, :, ::-1] + batch_img.append(img) + + if (i + 1) % batch_size == 0: + rec_feat = self.rec_predictor.predict(batch_img) + gallery_features[i - batch_size + 1:i + 1, :] = rec_feat + batch_img = [] + + if len(batch_img) > 0: + rec_feat = self.rec_predictor.predict(batch_img) + gallery_features[-len(batch_img):, :] = rec_feat + batch_img = [] + + return gallery_features + + +def main(config): + GalleryBuilder(config) + return + + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config(args.config, overrides=args.override, show=True) + main(config) diff --git a/exprements/PaddleClas/deploy/python/det_preprocess.py b/exprements/PaddleClas/deploy/python/det_preprocess.py new file mode 100644 index 00000000..5f5760ec --- /dev/null +++ b/exprements/PaddleClas/deploy/python/det_preprocess.py @@ -0,0 +1,205 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import cv2 +import numpy as np + + +def decode_image(im_file, im_info): + """read rgb image + Args: + im_file (str|np.ndarray): input can be image path or np.ndarray + im_info (dict): info of image + Returns: + im (np.ndarray): processed image (np.ndarray) + im_info (dict): info of processed image + """ + if isinstance(im_file, str): + with open(im_file, 'rb') as f: + im_read = f.read() + data = np.frombuffer(im_read, dtype='uint8') + im = cv2.imdecode(data, 1) # BGR mode, but need RGB mode + im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB) + else: + im = im_file + im_info['im_shape'] = np.array(im.shape[:2], dtype=np.float32) + im_info['scale_factor'] = np.array([1., 1.], dtype=np.float32) + return im, im_info + + +class DetResize(object): + """resize image by target_size and max_size + Args: + target_size (int): the target size of image + keep_ratio (bool): whether keep_ratio or not, default true + interp (int): method of resize + """ + + def __init__( + self, + target_size, + keep_ratio=True, + interp=cv2.INTER_LINEAR, ): + if isinstance(target_size, int): + target_size = [target_size, target_size] + self.target_size = target_size + self.keep_ratio = keep_ratio + self.interp = interp + + def __call__(self, im, im_info): + """ + Args: + im (np.ndarray): image (np.ndarray) + im_info (dict): info of image + Returns: + im (np.ndarray): processed image (np.ndarray) + im_info (dict): info of processed image + """ + assert len(self.target_size) == 2 + assert self.target_size[0] > 0 and self.target_size[1] > 0 + im_channel = im.shape[2] + im_scale_y, im_scale_x = self.generate_scale(im) + # set image_shape + im_info['input_shape'][1] = int(im_scale_y * im.shape[0]) + im_info['input_shape'][2] = int(im_scale_x * im.shape[1]) + im = cv2.resize( + im, + None, + None, + fx=im_scale_x, + fy=im_scale_y, + interpolation=self.interp) + im_info['im_shape'] = np.array(im.shape[:2]).astype('float32') + im_info['scale_factor'] = np.array( + [im_scale_y, im_scale_x]).astype('float32') + return im, im_info + + def generate_scale(self, im): + """ + Args: + im (np.ndarray): image (np.ndarray) + Returns: + im_scale_x: the resize ratio of X + im_scale_y: the resize ratio of Y + """ + origin_shape = im.shape[:2] + im_c = im.shape[2] + if self.keep_ratio: + im_size_min = np.min(origin_shape) + im_size_max = np.max(origin_shape) + target_size_min = np.min(self.target_size) + target_size_max = np.max(self.target_size) + im_scale = float(target_size_min) / float(im_size_min) + if np.round(im_scale * im_size_max) > target_size_max: + im_scale = float(target_size_max) / float(im_size_max) + im_scale_x = im_scale + im_scale_y = im_scale + else: + resize_h, resize_w = self.target_size + im_scale_y = resize_h / float(origin_shape[0]) + im_scale_x = resize_w / float(origin_shape[1]) + return im_scale_y, im_scale_x + + +class DetNormalizeImage(object): + """normalize image + Args: + mean (list): im - mean + std (list): im / std + is_scale (bool): whether need im / 255 + is_channel_first (bool): if True: image shape is CHW, else: HWC + """ + + def __init__(self, mean, std, is_scale=True): + self.mean = mean + self.std = std + self.is_scale = is_scale + + def __call__(self, im, im_info): + """ + Args: + im (np.ndarray): image (np.ndarray) + im_info (dict): info of image + Returns: + im (np.ndarray): processed image (np.ndarray) + im_info (dict): info of processed image + """ + im = im.astype(np.float32, copy=False) + mean = np.array(self.mean)[np.newaxis, np.newaxis, :] + std = np.array(self.std)[np.newaxis, np.newaxis, :] + + if self.is_scale: + im = im / 255.0 + + im -= mean + im /= std + return im, im_info + + +class DetPermute(object): + """permute image + Args: + to_bgr (bool): whether convert RGB to BGR + channel_first (bool): whether convert HWC to CHW + """ + + def __init__(self, ): + super().__init__() + + def __call__(self, im, im_info): + """ + Args: + im (np.ndarray): image (np.ndarray) + im_info (dict): info of image + Returns: + im (np.ndarray): processed image (np.ndarray) + im_info (dict): info of processed image + """ + im = im.transpose((2, 0, 1)).copy() + return im, im_info + + +class DetPadStride(object): + """ padding image for model with FPN , instead PadBatch(pad_to_stride, pad_gt) in original config + Args: + stride (bool): model with FPN need image shape % stride == 0 + """ + + def __init__(self, stride=0): + self.coarsest_stride = stride + + def __call__(self, im, im_info): + """ + Args: + im (np.ndarray): image (np.ndarray) + im_info (dict): info of image + Returns: + im (np.ndarray): processed image (np.ndarray) + im_info (dict): info of processed image + """ + coarsest_stride = self.coarsest_stride + if coarsest_stride <= 0: + return im, im_info + im_c, im_h, im_w = im.shape + pad_h = int(np.ceil(float(im_h) / coarsest_stride) * coarsest_stride) + pad_w = int(np.ceil(float(im_w) / coarsest_stride) * coarsest_stride) + padding_im = np.zeros((im_c, pad_h, pad_w), dtype=np.float32) + padding_im[:, :im_h, :im_w] = im + return padding_im, im_info + + +def det_preprocess(im, im_info, preprocess_ops): + for operator in preprocess_ops: + im, im_info = operator(im, im_info) + return im, im_info diff --git a/exprements/PaddleClas/deploy/python/postprocess.py b/exprements/PaddleClas/deploy/python/postprocess.py new file mode 100644 index 00000000..d26cbaa9 --- /dev/null +++ b/exprements/PaddleClas/deploy/python/postprocess.py @@ -0,0 +1,161 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import copy +import shutil +from functools import partial +import importlib +import numpy as np +import paddle +import paddle.nn.functional as F + + +def build_postprocess(config): + if config is None: + return None + + mod = importlib.import_module(__name__) + config = copy.deepcopy(config) + + main_indicator = config.pop( + "main_indicator") if "main_indicator" in config else None + main_indicator = main_indicator if main_indicator else "" + + func_list = [] + for func in config: + func_list.append(getattr(mod, func)(**config[func])) + return PostProcesser(func_list, main_indicator) + + +class PostProcesser(object): + def __init__(self, func_list, main_indicator="Topk"): + self.func_list = func_list + self.main_indicator = main_indicator + + def __call__(self, x, image_file=None): + rtn = None + for func in self.func_list: + tmp = func(x, image_file) + if type(func).__name__ in self.main_indicator: + rtn = tmp + return rtn + + +class Topk(object): + def __init__(self, topk=1, class_id_map_file=None): + assert isinstance(topk, (int, )) + self.class_id_map = self.parse_class_id_map(class_id_map_file) + self.topk = topk + + def parse_class_id_map(self, class_id_map_file): + if class_id_map_file is None: + return None + + if not os.path.exists(class_id_map_file): + print( + "Warning: If want to use your own label_dict, please input legal path!\nOtherwise label_names will be empty!" + ) + return None + + try: + class_id_map = {} + with open(class_id_map_file, "r") as fin: + lines = fin.readlines() + for line in lines: + partition = line.split("\n")[0].partition(" ") + class_id_map[int(partition[0])] = str(partition[-1]) + except Exception as ex: + print(ex) + class_id_map = None + return class_id_map + + def __call__(self, x, file_names=None, multilabel=False): + if file_names is not None: + assert x.shape[0] == len(file_names) + y = [] + for idx, probs in enumerate(x): + index = probs.argsort(axis=0)[-self.topk:][::-1].astype( + "int32") if not multilabel else np.where( + probs >= 0.5)[0].astype("int32") + clas_id_list = [] + score_list = [] + label_name_list = [] + for i in index: + clas_id_list.append(i.item()) + score_list.append(probs[i].item()) + if self.class_id_map is not None: + label_name_list.append(self.class_id_map[i.item()]) + result = { + "class_ids": clas_id_list, + "scores": np.around( + score_list, decimals=5).tolist(), + } + if file_names is not None: + result["file_name"] = file_names[idx] + if label_name_list is not None: + result["label_names"] = label_name_list + y.append(result) + return y + + +class MultiLabelTopk(Topk): + def __init__(self, topk=1, class_id_map_file=None): + super().__init__() + + def __call__(self, x, file_names=None): + return super().__call__(x, file_names, multilabel=True) + + +class SavePreLabel(object): + def __init__(self, save_dir): + if save_dir is None: + raise Exception( + "Please specify save_dir if SavePreLabel specified.") + self.save_dir = partial(os.path.join, save_dir) + + def __call__(self, x, file_names=None): + if file_names is None: + return + assert x.shape[0] == len(file_names) + for idx, probs in enumerate(x): + index = probs.argsort(axis=0)[-1].astype("int32") + self.save(index, file_names[idx]) + + def save(self, id, image_file): + output_dir = self.save_dir(str(id)) + os.makedirs(output_dir, exist_ok=True) + shutil.copy(image_file, output_dir) + + +class Binarize(object): + def __init__(self, method="round"): + self.method = method + self.unit = np.array([[128, 64, 32, 16, 8, 4, 2, 1]]).T + + def __call__(self, x, file_names=None): + if self.method == "round": + x = np.round(x + 1).astype("uint8") - 1 + + if self.method == "sign": + x = ((np.sign(x) + 1) / 2).astype("uint8") + + embedding_size = x.shape[1] + assert embedding_size % 8 == 0, "The Binary index only support vectors with sizes multiple of 8" + + byte = np.zeros([x.shape[0], embedding_size // 8], dtype=np.uint8) + for i in range(embedding_size // 8): + byte[:, i:i + 1] = np.dot(x[:, i * 8:(i + 1) * 8], self.unit) + + return byte diff --git a/exprements/PaddleClas/deploy/python/predict_cls.py b/exprements/PaddleClas/deploy/python/predict_cls.py new file mode 100644 index 00000000..574caa3e --- /dev/null +++ b/exprements/PaddleClas/deploy/python/predict_cls.py @@ -0,0 +1,153 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import os +import sys + +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +import cv2 +import numpy as np + +from utils import logger +from utils import config +from utils.predictor import Predictor +from utils.get_image_list import get_image_list +from python.preprocess import create_operators +from python.postprocess import build_postprocess + + +class ClsPredictor(Predictor): + def __init__(self, config): + super().__init__(config["Global"]) + + self.preprocess_ops = [] + self.postprocess = None + if "PreProcess" in config: + if "transform_ops" in config["PreProcess"]: + self.preprocess_ops = create_operators(config["PreProcess"][ + "transform_ops"]) + if "PostProcess" in config: + self.postprocess = build_postprocess(config["PostProcess"]) + + # for whole_chain project to test each repo of paddle + self.benchmark = config["Global"].get("benchmark", False) + if self.benchmark: + import auto_log + import os + pid = os.getpid() + size = config["PreProcess"]["transform_ops"][1]["CropImage"][ + "size"] + self.auto_logger = auto_log.AutoLogger( + model_name=config["Global"].get("model_name", "cls"), + model_precision='fp16' + if config["Global"]["use_fp16"] else 'fp32', + batch_size=config["Global"].get("batch_size", 1), + data_shape=[3, size, size], + save_path=config["Global"].get("save_log_path", + "./auto_log.log"), + inference_config=self.config, + pids=pid, + process_name=None, + gpu_ids=None, + time_keys=[ + 'preprocess_time', 'inference_time', 'postprocess_time' + ], + warmup=2) + + def predict(self, images): + use_onnx = self.args.get("use_onnx", False) + if not use_onnx: + input_names = self.predictor.get_input_names() + input_tensor = self.predictor.get_input_handle(input_names[0]) + + output_names = self.predictor.get_output_names() + output_tensor = self.predictor.get_output_handle(output_names[0]) + else: + input_names = self.predictor.get_inputs()[0].name + output_names = self.predictor.get_outputs()[0].name + + if self.benchmark: + self.auto_logger.times.start() + if not isinstance(images, (list, )): + images = [images] + for idx in range(len(images)): + for ops in self.preprocess_ops: + images[idx] = ops(images[idx]) + image = np.array(images) + if self.benchmark: + self.auto_logger.times.stamp() + + if not use_onnx: + input_tensor.copy_from_cpu(image) + self.predictor.run() + batch_output = output_tensor.copy_to_cpu() + else: + batch_output = self.predictor.run( + output_names=[output_names], + input_feed={input_names: image})[0] + + if self.benchmark: + self.auto_logger.times.stamp() + if self.postprocess is not None: + batch_output = self.postprocess(batch_output) + if self.benchmark: + self.auto_logger.times.end(stamp=True) + return batch_output + + +def main(config): + cls_predictor = ClsPredictor(config) + image_list = get_image_list(config["Global"]["infer_imgs"]) + + batch_imgs = [] + batch_names = [] + cnt = 0 + for idx, img_path in enumerate(image_list): + img = cv2.imread(img_path) + if img is None: + logger.warning( + "Image file failed to read and has been skipped. The path: {}". + format(img_path)) + else: + img = img[:, :, ::-1] + batch_imgs.append(img) + img_name = os.path.basename(img_path) + batch_names.append(img_name) + cnt += 1 + + if cnt % config["Global"]["batch_size"] == 0 or (idx + 1 + ) == len(image_list): + if len(batch_imgs) == 0: + continue + batch_results = cls_predictor.predict(batch_imgs) + for number, result_dict in enumerate(batch_results): + filename = batch_names[number] + clas_ids = result_dict["class_ids"] + scores_str = "[{}]".format(", ".join("{:.2f}".format( + r) for r in result_dict["scores"])) + label_names = result_dict["label_names"] + print("{}:\tclass id(s): {}, score(s): {}, label_name(s): {}". + format(filename, clas_ids, scores_str, label_names)) + batch_imgs = [] + batch_names = [] + if cls_predictor.benchmark: + cls_predictor.auto_logger.report() + return + + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config(args.config, overrides=args.override, show=True) + main(config) diff --git a/exprements/PaddleClas/deploy/python/predict_det.py b/exprements/PaddleClas/deploy/python/predict_det.py new file mode 100644 index 00000000..e4e0a24a --- /dev/null +++ b/exprements/PaddleClas/deploy/python/predict_det.py @@ -0,0 +1,157 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import os +import sys + +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +from utils import logger +from utils import config +from utils.predictor import Predictor +from utils.get_image_list import get_image_list +from det_preprocess import det_preprocess +from preprocess import create_operators + +import os +import argparse +import time +import yaml +import ast +from functools import reduce +import cv2 +import numpy as np +import paddle + + +class DetPredictor(Predictor): + def __init__(self, config): + super().__init__(config["Global"], + config["Global"]["det_inference_model_dir"]) + + self.preprocess_ops = create_operators(config["DetPreProcess"][ + "transform_ops"]) + self.config = config + + def preprocess(self, img): + im_info = { + 'scale_factor': np.array( + [1., 1.], dtype=np.float32), + 'im_shape': np.array( + img.shape[:2], dtype=np.float32), + 'input_shape': self.config["Global"]["image_shape"], + "scale_factor": np.array( + [1., 1.], dtype=np.float32) + } + im, im_info = det_preprocess(img, im_info, self.preprocess_ops) + inputs = self.create_inputs(im, im_info) + return inputs + + def create_inputs(self, im, im_info): + """generate input for different model type + Args: + im (np.ndarray): image (np.ndarray) + im_info (dict): info of image + model_arch (str): model type + Returns: + inputs (dict): input of model + """ + inputs = {} + inputs['image'] = np.array((im, )).astype('float32') + inputs['im_shape'] = np.array( + (im_info['im_shape'], )).astype('float32') + inputs['scale_factor'] = np.array( + (im_info['scale_factor'], )).astype('float32') + + return inputs + + def parse_det_results(self, pred, threshold, label_list): + max_det_results = self.config["Global"]["max_det_results"] + keep_indexes = pred[:, 1].argsort()[::-1][:max_det_results] + results = [] + for idx in keep_indexes: + single_res = pred[idx] + class_id = int(single_res[0]) + score = single_res[1] + bbox = single_res[2:] + if score < threshold: + continue + label_name = label_list[class_id] + results.append({ + "class_id": class_id, + "score": score, + "bbox": bbox, + "label_name": label_name, + }) + return results + + def predict(self, image, threshold=0.5, run_benchmark=False): + ''' + Args: + image (str/np.ndarray): path of image/ np.ndarray read by cv2 + threshold (float): threshold of predicted box' score + Returns: + results (dict): include 'boxes': np.ndarray: shape:[N,6], N: number of box, + matix element:[class, score, x_min, y_min, x_max, y_max] + MaskRCNN's results include 'masks': np.ndarray: + shape: [N, im_h, im_w] + ''' + inputs = self.preprocess(image) + np_boxes = None + input_names = self.predictor.get_input_names() + + for i in range(len(input_names)): + input_tensor = self.predictor.get_input_handle(input_names[i]) + input_tensor.copy_from_cpu(inputs[input_names[i]]) + + t1 = time.time() + self.predictor.run() + output_names = self.predictor.get_output_names() + boxes_tensor = self.predictor.get_output_handle(output_names[0]) + np_boxes = boxes_tensor.copy_to_cpu() + t2 = time.time() + + print("Inference: {} ms per batch image".format((t2 - t1) * 1000.0)) + + # do not perform postprocess in benchmark mode + results = [] + if reduce(lambda x, y: x * y, np_boxes.shape) < 6: + print('[WARNNING] No object detected.') + results = np.array([]) + else: + results = np_boxes + + results = self.parse_det_results(results, + self.config["Global"]["threshold"], + self.config["Global"]["labe_list"]) + return results + + +def main(config): + det_predictor = DetPredictor(config) + image_list = get_image_list(config["Global"]["infer_imgs"]) + + assert config["Global"]["batch_size"] == 1 + for idx, image_file in enumerate(image_list): + img = cv2.imread(image_file)[:, :, ::-1] + output = det_predictor.predict(img) + print(output) + + return + + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config(args.config, overrides=args.override, show=True) + main(config) diff --git a/exprements/PaddleClas/deploy/python/predict_rec.py b/exprements/PaddleClas/deploy/python/predict_rec.py new file mode 100644 index 00000000..e2433581 --- /dev/null +++ b/exprements/PaddleClas/deploy/python/predict_rec.py @@ -0,0 +1,148 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import os +import sys + +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +import cv2 +import numpy as np + +from utils import logger +from utils import config +from utils.predictor import Predictor +from utils.get_image_list import get_image_list +from preprocess import create_operators +from postprocess import build_postprocess + + +class RecPredictor(Predictor): + def __init__(self, config): + super().__init__(config["Global"], + config["Global"]["rec_inference_model_dir"]) + self.preprocess_ops = create_operators(config["RecPreProcess"][ + "transform_ops"]) + self.postprocess = build_postprocess(config["RecPostProcess"]) + self.benchmark = config["Global"].get("benchmark", False) + + if self.benchmark: + import auto_log + pid = os.getpid() + self.auto_logger = auto_log.AutoLogger( + model_name=config["Global"].get("model_name", "rec"), + model_precision='fp16' + if config["Global"]["use_fp16"] else 'fp32', + batch_size=config["Global"].get("batch_size", 1), + data_shape=[3, 224, 224], + save_path=config["Global"].get("save_log_path", + "./auto_log.log"), + inference_config=self.config, + pids=pid, + process_name=None, + gpu_ids=None, + time_keys=[ + 'preprocess_time', 'inference_time', 'postprocess_time' + ], + warmup=2) + + def predict(self, images, feature_normalize=True): + use_onnx = self.args.get("use_onnx", False) + if not use_onnx: + input_names = self.predictor.get_input_names() + input_tensor = self.predictor.get_input_handle(input_names[0]) + + output_names = self.predictor.get_output_names() + output_tensor = self.predictor.get_output_handle(output_names[0]) + else: + input_names = self.predictor.get_inputs()[0].name + output_names = self.predictor.get_outputs()[0].name + + if self.benchmark: + self.auto_logger.times.start() + if not isinstance(images, (list, )): + images = [images] + for idx in range(len(images)): + for ops in self.preprocess_ops: + images[idx] = ops(images[idx]) + image = np.array(images) + if self.benchmark: + self.auto_logger.times.stamp() + + if not use_onnx: + input_tensor.copy_from_cpu(image) + self.predictor.run() + batch_output = output_tensor.copy_to_cpu() + else: + batch_output = self.predictor.run( + output_names=[output_names], + input_feed={input_names: image})[0] + + if self.benchmark: + self.auto_logger.times.stamp() + + if feature_normalize: + feas_norm = np.sqrt( + np.sum(np.square(batch_output), axis=1, keepdims=True)) + batch_output = np.divide(batch_output, feas_norm) + + if self.postprocess is not None: + batch_output = self.postprocess(batch_output) + + if self.benchmark: + self.auto_logger.times.end(stamp=True) + return batch_output + + +def main(config): + rec_predictor = RecPredictor(config) + image_list = get_image_list(config["Global"]["infer_imgs"]) + + batch_imgs = [] + batch_names = [] + cnt = 0 + for idx, img_path in enumerate(image_list): + img = cv2.imread(img_path) + if img is None: + logger.warning( + "Image file failed to read and has been skipped. The path: {}". + format(img_path)) + else: + img = img[:, :, ::-1] + batch_imgs.append(img) + img_name = os.path.basename(img_path) + batch_names.append(img_name) + cnt += 1 + + if cnt % config["Global"]["batch_size"] == 0 or (idx + 1 + ) == len(image_list): + if len(batch_imgs) == 0: + continue + + batch_results = rec_predictor.predict(batch_imgs) + for number, result_dict in enumerate(batch_results): + filename = batch_names[number] + print("{}:\t{}".format(filename, result_dict)) + batch_imgs = [] + batch_names = [] + if rec_predictor.benchmark: + rec_predictor.auto_logger.report() + + return + + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config(args.config, overrides=args.override, show=True) + main(config) diff --git a/exprements/PaddleClas/deploy/python/predict_system.py b/exprements/PaddleClas/deploy/python/predict_system.py new file mode 100644 index 00000000..23987553 --- /dev/null +++ b/exprements/PaddleClas/deploy/python/predict_system.py @@ -0,0 +1,151 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import os +import sys + +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +import copy +import cv2 +import numpy as np +import faiss +import pickle + +from python.predict_rec import RecPredictor +from python.predict_det import DetPredictor + +from utils import logger +from utils import config +from utils.get_image_list import get_image_list +from utils.draw_bbox import draw_bbox_results + + +class SystemPredictor(object): + def __init__(self, config): + + self.config = config + self.rec_predictor = RecPredictor(config) + self.det_predictor = DetPredictor(config) + + assert 'IndexProcess' in config.keys(), "Index config not found ... " + self.return_k = self.config['IndexProcess']['return_k'] + + index_dir = self.config["IndexProcess"]["index_dir"] + assert os.path.exists(os.path.join( + index_dir, "vector.index")), "vector.index not found ..." + assert os.path.exists(os.path.join( + index_dir, "id_map.pkl")), "id_map.pkl not found ... " + + if config['IndexProcess'].get("dist_type") == "hamming": + self.Searcher = faiss.read_index_binary( + os.path.join(index_dir, "vector.index")) + else: + self.Searcher = faiss.read_index( + os.path.join(index_dir, "vector.index")) + + with open(os.path.join(index_dir, "id_map.pkl"), "rb") as fd: + self.id_map = pickle.load(fd) + + def append_self(self, results, shape): + results.append({ + "class_id": 0, + "score": 1.0, + "bbox": + np.array([0, 0, shape[1], shape[0]]), # xmin, ymin, xmax, ymax + "label_name": "foreground", + }) + return results + + def nms_to_rec_results(self, results, thresh=0.1): + filtered_results = [] + x1 = np.array([r["bbox"][0] for r in results]).astype("float32") + y1 = np.array([r["bbox"][1] for r in results]).astype("float32") + x2 = np.array([r["bbox"][2] for r in results]).astype("float32") + y2 = np.array([r["bbox"][3] for r in results]).astype("float32") + scores = np.array([r["rec_scores"] for r in results]) + + areas = (x2 - x1 + 1) * (y2 - y1 + 1) + order = scores.argsort()[::-1] + while order.size > 0: + i = order[0] + xx1 = np.maximum(x1[i], x1[order[1:]]) + yy1 = np.maximum(y1[i], y1[order[1:]]) + xx2 = np.minimum(x2[i], x2[order[1:]]) + yy2 = np.minimum(y2[i], y2[order[1:]]) + + w = np.maximum(0.0, xx2 - xx1 + 1) + h = np.maximum(0.0, yy2 - yy1 + 1) + inter = w * h + ovr = inter / (areas[i] + areas[order[1:]] - inter) + inds = np.where(ovr <= thresh)[0] + order = order[inds + 1] + filtered_results.append(results[i]) + + return filtered_results + + def predict(self, img): + output = [] + # st1: get all detection results + results = self.det_predictor.predict(img) + + # st2: add the whole image for recognition to improve recall + results = self.append_self(results, img.shape) + + # st3: recognition process, use score_thres to ensure accuracy + for result in results: + preds = {} + xmin, ymin, xmax, ymax = result["bbox"].astype("int") + crop_img = img[ymin:ymax, xmin:xmax, :].copy() + rec_results = self.rec_predictor.predict(crop_img) + preds["bbox"] = [xmin, ymin, xmax, ymax] + scores, docs = self.Searcher.search(rec_results, self.return_k) + + # just top-1 result will be returned for the final + if self.config["IndexProcess"]["dist_type"] == "hamming": + if scores[0][0] <= self.config["IndexProcess"][ + "hamming_radius"]: + preds["rec_docs"] = self.id_map[docs[0][0]].split()[1] + preds["rec_scores"] = scores[0][0] + output.append(preds) + else: + if scores[0][0] >= self.config["IndexProcess"]["score_thres"]: + preds["rec_docs"] = self.id_map[docs[0][0]].split()[1] + preds["rec_scores"] = scores[0][0] + output.append(preds) + + # st5: nms to the final results to avoid fetching duplicate results + output = self.nms_to_rec_results( + output, self.config["Global"]["rec_nms_thresold"]) + + return output + + +def main(config): + system_predictor = SystemPredictor(config) + image_list = get_image_list(config["Global"]["infer_imgs"]) + + assert config["Global"]["batch_size"] == 1 + for idx, image_file in enumerate(image_list): + img = cv2.imread(image_file)[:, :, ::-1] + output = system_predictor.predict(img) + draw_bbox_results(img, output, image_file) + print(output) + return + + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config(args.config, overrides=args.override, show=True) + main(config) diff --git a/exprements/PaddleClas/deploy/python/preprocess.py b/exprements/PaddleClas/deploy/python/preprocess.py new file mode 100644 index 00000000..1da32ad6 --- /dev/null +++ b/exprements/PaddleClas/deploy/python/preprocess.py @@ -0,0 +1,337 @@ +""" +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +""" + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +from __future__ import unicode_literals + +from functools import partial +import six +import math +import random +import cv2 +import numpy as np +import importlib +from PIL import Image + +from python.det_preprocess import DetNormalizeImage, DetPadStride, DetPermute, DetResize + + +def create_operators(params): + """ + create operators based on the config + + Args: + params(list): a dict list, used to create some operators + """ + assert isinstance(params, list), ('operator config should be a list') + mod = importlib.import_module(__name__) + ops = [] + for operator in params: + assert isinstance(operator, + dict) and len(operator) == 1, "yaml format error" + op_name = list(operator)[0] + param = {} if operator[op_name] is None else operator[op_name] + op = getattr(mod, op_name)(**param) + ops.append(op) + + return ops + + +class UnifiedResize(object): + def __init__(self, interpolation=None, backend="cv2"): + _cv2_interp_from_str = { + 'nearest': cv2.INTER_NEAREST, + 'bilinear': cv2.INTER_LINEAR, + 'area': cv2.INTER_AREA, + 'bicubic': cv2.INTER_CUBIC, + 'lanczos': cv2.INTER_LANCZOS4 + } + _pil_interp_from_str = { + 'nearest': Image.NEAREST, + 'bilinear': Image.BILINEAR, + 'bicubic': Image.BICUBIC, + 'box': Image.BOX, + 'lanczos': Image.LANCZOS, + 'hamming': Image.HAMMING + } + + def _pil_resize(src, size, resample): + pil_img = Image.fromarray(src) + pil_img = pil_img.resize(size, resample) + return np.asarray(pil_img) + + if backend.lower() == "cv2": + if isinstance(interpolation, str): + interpolation = _cv2_interp_from_str[interpolation.lower()] + # compatible with opencv < version 4.4.0 + elif interpolation is None: + interpolation = cv2.INTER_LINEAR + self.resize_func = partial(cv2.resize, interpolation=interpolation) + elif backend.lower() == "pil": + if isinstance(interpolation, str): + interpolation = _pil_interp_from_str[interpolation.lower()] + self.resize_func = partial(_pil_resize, resample=interpolation) + else: + logger.warning( + f"The backend of Resize only support \"cv2\" or \"PIL\". \"f{backend}\" is unavailable. Use \"cv2\" instead." + ) + self.resize_func = cv2.resize + + def __call__(self, src, size): + return self.resize_func(src, size) + + +class OperatorParamError(ValueError): + """ OperatorParamError + """ + pass + + +class DecodeImage(object): + """ decode image """ + + def __init__(self, to_rgb=True, to_np=False, channel_first=False): + self.to_rgb = to_rgb + self.to_np = to_np # to numpy + self.channel_first = channel_first # only enabled when to_np is True + + def __call__(self, img): + if six.PY2: + assert type(img) is str and len( + img) > 0, "invalid input 'img' in DecodeImage" + else: + assert type(img) is bytes and len( + img) > 0, "invalid input 'img' in DecodeImage" + data = np.frombuffer(img, dtype='uint8') + img = cv2.imdecode(data, 1) + if self.to_rgb: + assert img.shape[2] == 3, 'invalid shape of image[%s]' % ( + img.shape) + img = img[:, :, ::-1] + + if self.channel_first: + img = img.transpose((2, 0, 1)) + + return img + + +class ResizeImage(object): + """ resize image """ + + def __init__(self, + size=None, + resize_short=None, + interpolation=None, + backend="cv2"): + if resize_short is not None and resize_short > 0: + self.resize_short = resize_short + self.w = None + self.h = None + elif size is not None: + self.resize_short = None + self.w = size if type(size) is int else size[0] + self.h = size if type(size) is int else size[1] + else: + raise OperatorParamError("invalid params for ReisizeImage for '\ + 'both 'size' and 'resize_short' are None") + + self._resize_func = UnifiedResize( + interpolation=interpolation, backend=backend) + + def __call__(self, img): + img_h, img_w = img.shape[:2] + if self.resize_short is not None: + percent = float(self.resize_short) / min(img_w, img_h) + w = int(round(img_w * percent)) + h = int(round(img_h * percent)) + else: + w = self.w + h = self.h + return self._resize_func(img, (w, h)) + + +class CropImage(object): + """ crop image """ + + def __init__(self, size): + if type(size) is int: + self.size = (size, size) + else: + self.size = size # (h, w) + + def __call__(self, img): + w, h = self.size + img_h, img_w = img.shape[:2] + + if img_h < h or img_w < w: + raise Exception( + f"The size({h}, {w}) of CropImage must be greater than size({img_h}, {img_w}) of image. Please check image original size and size of ResizeImage if used." + ) + + w_start = (img_w - w) // 2 + h_start = (img_h - h) // 2 + + w_end = w_start + w + h_end = h_start + h + return img[h_start:h_end, w_start:w_end, :] + + +class RandCropImage(object): + """ random crop image """ + + def __init__(self, + size, + scale=None, + ratio=None, + interpolation=None, + backend="cv2"): + if type(size) is int: + self.size = (size, size) # (h, w) + else: + self.size = size + + self.scale = [0.08, 1.0] if scale is None else scale + self.ratio = [3. / 4., 4. / 3.] if ratio is None else ratio + + self._resize_func = UnifiedResize( + interpolation=interpolation, backend=backend) + + def __call__(self, img): + size = self.size + scale = self.scale + ratio = self.ratio + + aspect_ratio = math.sqrt(random.uniform(*ratio)) + w = 1. * aspect_ratio + h = 1. / aspect_ratio + + img_h, img_w = img.shape[:2] + + bound = min((float(img_w) / img_h) / (w**2), + (float(img_h) / img_w) / (h**2)) + scale_max = min(scale[1], bound) + scale_min = min(scale[0], bound) + + target_area = img_w * img_h * random.uniform(scale_min, scale_max) + target_size = math.sqrt(target_area) + w = int(target_size * w) + h = int(target_size * h) + + i = random.randint(0, img_w - w) + j = random.randint(0, img_h - h) + + img = img[j:j + h, i:i + w, :] + + return self._resize_func(img, size) + + +class RandFlipImage(object): + """ random flip image + flip_code: + 1: Flipped Horizontally + 0: Flipped Vertically + -1: Flipped Horizontally & Vertically + """ + + def __init__(self, flip_code=1): + assert flip_code in [-1, 0, 1 + ], "flip_code should be a value in [-1, 0, 1]" + self.flip_code = flip_code + + def __call__(self, img): + if random.randint(0, 1) == 1: + return cv2.flip(img, self.flip_code) + else: + return img + + +class AutoAugment(object): + def __init__(self): + self.policy = ImageNetPolicy() + + def __call__(self, img): + from PIL import Image + img = np.ascontiguousarray(img) + img = Image.fromarray(img) + img = self.policy(img) + img = np.asarray(img) + + +class NormalizeImage(object): + """ normalize image such as substract mean, divide std + """ + + def __init__(self, + scale=None, + mean=None, + std=None, + order='chw', + output_fp16=False, + channel_num=3): + if isinstance(scale, str): + scale = eval(scale) + assert channel_num in [ + 3, 4 + ], "channel number of input image should be set to 3 or 4." + self.channel_num = channel_num + self.output_dtype = 'float16' if output_fp16 else 'float32' + self.scale = np.float32(scale if scale is not None else 1.0 / 255.0) + self.order = order + mean = mean if mean is not None else [0.485, 0.456, 0.406] + std = std if std is not None else [0.229, 0.224, 0.225] + + shape = (3, 1, 1) if self.order == 'chw' else (1, 1, 3) + self.mean = np.array(mean).reshape(shape).astype('float32') + self.std = np.array(std).reshape(shape).astype('float32') + + def __call__(self, img): + from PIL import Image + if isinstance(img, Image.Image): + img = np.array(img) + + assert isinstance(img, + np.ndarray), "invalid input 'img' in NormalizeImage" + + img = (img.astype('float32') * self.scale - self.mean) / self.std + + if self.channel_num == 4: + img_h = img.shape[1] if self.order == 'chw' else img.shape[0] + img_w = img.shape[2] if self.order == 'chw' else img.shape[1] + pad_zeros = np.zeros( + (1, img_h, img_w)) if self.order == 'chw' else np.zeros( + (img_h, img_w, 1)) + img = (np.concatenate( + (img, pad_zeros), axis=0) + if self.order == 'chw' else np.concatenate( + (img, pad_zeros), axis=2)) + return img.astype(self.output_dtype) + + +class ToCHWImage(object): + """ convert hwc image to chw image + """ + + def __init__(self): + pass + + def __call__(self, img): + from PIL import Image + if isinstance(img, Image.Image): + img = np.array(img) + + return img.transpose((2, 0, 1)) diff --git a/exprements/PaddleClas/deploy/shell/predict.sh b/exprements/PaddleClas/deploy/shell/predict.sh new file mode 100644 index 00000000..f0f59f4a --- /dev/null +++ b/exprements/PaddleClas/deploy/shell/predict.sh @@ -0,0 +1,18 @@ +# classification +python3.7 python/predict_cls.py -c configs/inference_cls.yaml + +# multilabel_classification +#python3.7 python/predict_cls.py -c configs/inference_multilabel_cls.yaml + +# feature extractor +# python3.7 python/predict_rec.py -c configs/inference_rec.yaml + +# detection +# python3.7 python/predict_det.py -c configs/inference_rec.yaml + + +# build system +#python3.7 python/build_gallery.py -c configs/build_logo.yaml + +# inference system +# python3.7 python/predict_system.py -c configs/inference_logo.yaml diff --git a/exprements/PaddleClas/deploy/slim/README.md b/exprements/PaddleClas/deploy/slim/README.md new file mode 100644 index 00000000..eed8aa3b --- /dev/null +++ b/exprements/PaddleClas/deploy/slim/README.md @@ -0,0 +1,144 @@ + +## Slim功能介绍 +复杂的模型有利于提高模型的性能，但也导致模型中存在一定冗余。此部分提供精简模型的功能，包括两部分：模型量化（量化训练、离线量化）、模型剪枝。 + +其中模型量化将全精度缩减到定点数减少这种冗余，达到减少模型计算复杂度，提高模型推理性能的目的。 +模型量化可以在基本不损失模型的精度的情况下，将FP32精度的模型参数转换为Int8精度，减小模型参数大小并加速计算，使用量化后的模型在移动端等部署时更具备速度优势。 + +模型剪枝将CNN中不重要的卷积核裁剪掉，减少模型参数量，从而降低模型计算复杂度。 + +本教程将介绍如何使用飞桨模型压缩库PaddleSlim做PaddleClas模型的压缩。 +[PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim) 集成了模型剪枝、量化（包括量化训练和离线量化）、蒸馏和神经网络搜索等多种业界常用且领先的模型压缩功能，如果您感兴趣，可以关注并了解。 + +在开始本教程之前，建议先了解[PaddleClas模型的训练方法](../../docs/zh_CN/tutorials/getting_started.md)以及[PaddleSlim](https://paddleslim.readthedocs.io/zh_CN/latest/index.html) + + +## 快速开始 +当训练出一个模型后，如果希望进一步的压缩模型大小并加速预测，可使用量化或者剪枝的方法压缩模型。 + +模型压缩主要包括五个步骤： +1. 安装 PaddleSlim +2. 准备训练好的模型 +3. 模型压缩 +4. 导出量化推理模型 +5. 量化模型预测部署 + +### 1. 安装PaddleSlim + +* 可以通过pip install的方式进行安装。 + +```bash +pip install paddleslim -i https://pypi.tuna.tsinghua.edu.cn/simple +``` + +* 如果获取PaddleSlim的最新特性，可以从源码安装。 + +```bash +git clone https://github.com/PaddlePaddle/PaddleSlim.git +cd Paddleslim +python3.7 setup.py install +``` + +### 2. 准备训练好的模型 + +PaddleClas提供了一系列训练好的[模型](../../docs/zh_CN/models/models_intro.md)，如果待量化的模型不在列表中，需要按照[常规训练](../../docs/zh_CN/tutorials/getting_started.md)方法得到训练好的模型。 + +### 3. 模型压缩 + +进入PaddleClas根目录 + +```bash +cd PaddleClas +``` + +`slim`训练相关代码已经集成到`ppcls/engine/`下，离线量化代码位于`deploy/slim/quant_post_static.py`。 + +#### 3.1 模型量化 + +量化训练包括离线量化训练和在线量化训练，在线量化训练效果更好，需加载预训练模型，在定义好量化策略后即可对模型进行量化。 + +##### 3.1.1 在线量化训练 + +训练指令如下： + +* CPU/单卡GPU + +以CPU为例，若使用GPU，则将命令中改成`cpu`改成`gpu` + +```bash +python3.7 tools/train.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml -o Global.device=cpu +``` + +其中`yaml`文件解析详见[参考文档](../../docs/zh_CN/tutorials/config_description.md)。为了保证精度，`yaml`文件中已经使用`pretrained model`. + + +* 单机多卡/多机多卡启动 + +```bash +export CUDA_VISIBLE_DEVICES=0,1,2,3 +python3.7 -m paddle.distributed.launch \ + --gpus="0,1,2,3" \ + tools/train.py \ + -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +``` + +##### 3.1.2 离线量化 + +**注意**：目前离线量化，必须使用已经训练好的模型，导出的`inference model`进行量化。一般模型导出`inference model`可参考[教程](../../docs/zh_CN/inference.md). + +一般来说，离线量化损失模型精度较多。 + +生成`inference model`后，离线量化运行方式如下 + +```bash +python3.7 deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./deploy/models/class_ResNet50_vd_ImageNet_infer +``` + +`Global.save_inference_dir`是`inference model`存放的目录。 + +执行成功后，在`Global.save_inference_dir`的目录下，生成`quant_post_static_model`文件夹，其中存储生成的离线量化模型，其可以直接进行预测部署，无需再重新导出模型。 + +#### 3.2 模型剪枝 + +训练指令如下： + +- CPU/单卡GPU + +以CPU为例，若使用GPU，则将命令中改成`cpu`改成`gpu` + +```bash +python3.7 tools/train.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml -o Global.device=cpu +``` + +- 单机单卡/单机多卡/多机多卡启动 + +```bash +export CUDA_VISIBLE_DEVICES=0,1,2,3 +python3.7 -m paddle.distributed.launch \ + --gpus="0,1,2,3" \ + tools/train.py \ + -c ppcls/configs/slim/ResNet50_vd_prune.yaml +``` + +### 4. 导出模型 + +在得到在线量化训练、模型剪枝保存的模型后，可以将其导出为inference model，用于预测部署，以模型剪枝为例： + +```bash +python3.7 tools/export.py \ + -c ppcls/configs/slim/ResNet50_vd_prune.yaml \ + -o Global.pretrained_model=./output/ResNet50_vd/best_model \ + -o Global.save_inference_dir=./inference +``` + + +### 5. 模型部署 + +上述步骤导出的模型可以通过PaddleLite的opt模型转换工具完成模型转换。 +模型部署的可参考 [移动端模型部署](../lite/readme.md) + + +## 训练超参数建议 + +* 量化训练时，建议加载常规训练得到的预训练模型，加速量化训练收敛。 +* 量化训练时，建议初始学习率修改为常规训练的`1/20~1/10`，同时将训练epoch数修改为常规训练的`1/5~1/2`，学习率策略方面，加上Warmup，其他配置信息不建议修改。 diff --git a/exprements/PaddleClas/deploy/slim/README_en.md b/exprements/PaddleClas/deploy/slim/README_en.md new file mode 100644 index 00000000..d7a978f4 --- /dev/null +++ b/exprements/PaddleClas/deploy/slim/README_en.md @@ -0,0 +1,144 @@ + +## Introduction to Slim + +Generally, a more complex model would achive better performance in the task, but it also leads to some redundancy in the model. This part provides the function of compressing the model, including two parts: model quantization (offline quantization training and online quantization training) and model pruning. +Quantization is a technique that reduces this redundancy by reducing the full precision data to a fixed number, so as to reduce model calculation complexity and improve model inference performance. + +Model pruning cuts off the unimportant convolution kernel in CNN to reduce the amount of model parameters, so as to reduce the computational complexity of the model. + +It is recommended that you could understand following pages before reading this example： +- [The training strategy of PaddleClas models](../../docs/en/tutorials/getting_started_en.md) +- [PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim) + +## Quick Start + After training a model, if you want to further compress the model size and speed up the prediction, you can use quantization or pruning to compress the model according to the following steps. + +1. Install PaddleSlim +2. Prepare trained model +3. Model compression +4. Export inference model +5. Deploy quantization inference model + + +### 1. Install PaddleSlim + +* Install by pip. + +```bash +pip install paddleslim -i https://pypi.tuna.tsinghua.edu.cn/simple +``` + +* Install from source code to get the lastest features. + +```bash +git clone https://github.com/PaddlePaddle/PaddleSlim.git +cd Paddleslim +python setup.py install +``` + + +### 2. Download Pretrain Model +PaddleClas provides a series of trained [models](../../docs/en/models/models_intro_en.md). +If the model to be quantified is not in the list, you need to follow the [Regular Training](../../docs/en/tutorials/getting_started_en.md) method to get the trained model. + +### 3. Model Compression + +Go to the root directory of PaddleClas + +```bash +cd PaddleClase +``` + +The training related codes have been integrated into `ppcls/engine/`. The offline quantization code is located in `deploy/slim/quant_post_static.py` + +#### 3.1 Model Quantization + +Quantization training includes offline quantization and online quantization training. + +##### 3.1.1 Online quantization training + +Online quantization training is more effective. It is necessary to load the pre-trained model. +After the quantization strategy is defined, the model can be quantified. + +The training command is as follow: + +* CPU/Single GPU + +If using GPU, change the `cpu` to `gpu` in the following command. + +```bash +python3.7 tools/train.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml -o Global.device=cpu +``` + +The description of `yaml` file can be found in this [doc](../../docs/en/tutorials/config_en.md). To get better accuracy, the `pretrained model`is used in `yaml`. + + +* Distributed training + +```bash +export CUDA_VISIBLE_DEVICES=0,1,2,3 +python3.7 -m paddle.distributed.launch \ + --gpus="0,1,2,3" \ + tools/train.py \ + -m train \ + -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +``` + +##### 3.1.2 Offline quantization + +**Attention**: At present, offline quantization must use `inference model` as input, which can be exported by trained model. The process of exporting `inference model` for trained model can refer to this [doc](../../docs/en/inference.md). + +Generally speaking, the offline quantization gets more loss of accuracy than online qutization training. + +After getting `inference model`, we can run following command to get offline quantization model. + +``` +python3.7 deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./deploy/models/class_ResNet50_vd_ImageNet_infer +``` + +`Global.save_inference_dir` is the directory storing the `inference model`. + +If run successfully, the directory `quant_post_static_model` is generated in `Global.save_inference_dir`, which stores the offline quantization model that can be used for deploy directly. + +#### 3.2 Model Pruning + +- CPU/Single GPU + +If using GPU, change the `cpu` to `gpu` in the following command. + +```bash +python3.7 tools/train.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml -o Global.device=cpu +``` + +- Distributed training + +```bash +export CUDA_VISIBLE_DEVICES=0,1,2,3 +python3.7 -m paddle.distributed.launch \ + --gpus="0,1,2,3" \ + tools/train.py \ + -c ppcls/configs/slim/ResNet50_vd_prune.yaml +``` + + + +### 4. Export inference model + +After getting the compressed model, we can export it as inference model for predictive deployment. Using pruned model as example: + +```bash +python3.7 tools/export.py \ + -c ppcls/configs/slim/ResNet50_vd_prune.yaml \ + -o Global.pretrained_model=./output/ResNet50_vd/best_model + -o Global.save_inference_dir=./inference +``` + +### 5. Deploy +The derived model can be converted through the `opt tool` of PaddleLite. + +For compresed model deployment, please refer to [Mobile terminal model deployment](../lite/readme_en.md) + +## Notes: + +* In quantitative training, it is suggested to load the pre-trained model obtained from conventional training to accelerate the convergence of quantitative training. +* In quantitative training, it is suggested that the initial learning rate should be changed to `1 / 20 ~ 1 / 10` of the conventional training, and the training epoch number should be changed to `1 / 5 ~ 1 / 2` of the conventional training. In terms of learning rate strategy, it's better to train with warmup, other configuration information is not recommended to be changed. diff --git a/exprements/PaddleClas/deploy/slim/quant_post_static.py b/exprements/PaddleClas/deploy/slim/quant_post_static.py new file mode 100644 index 00000000..5c846979 --- /dev/null +++ b/exprements/PaddleClas/deploy/slim/quant_post_static.py @@ -0,0 +1,74 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import os +import sys + +import numpy as np +import paddle +import paddleslim +from paddle.jit import to_static +from paddleslim.analysis import dygraph_flops as flops + +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../../'))) +from paddleslim.dygraph.quant import QAT + +from ppcls.data import build_dataloader +from ppcls.utils import config as conf +from ppcls.utils.logger import init_logger + + +def main(): + args = conf.parse_args() + config = conf.get_config(args.config, overrides=args.override, show=False) + + assert os.path.exists( + os.path.join(config["Global"]["save_inference_dir"], + 'inference.pdmodel')) and os.path.exists( + os.path.join(config["Global"]["save_inference_dir"], + 'inference.pdiparams')) + config["DataLoader"]["Eval"]["sampler"]["batch_size"] = 1 + config["DataLoader"]["Eval"]["loader"]["num_workers"] = 0 + init_logger() + device = paddle.set_device("cpu") + train_dataloader = build_dataloader(config["DataLoader"], "Eval", device, + False) + + def sample_generator(loader): + def __reader__(): + for indx, data in enumerate(loader): + images = np.array(data[0]) + yield images + + return __reader__ + + paddle.enable_static() + place = paddle.CPUPlace() + exe = paddle.static.Executor(place) + paddleslim.quant.quant_post_static( + executor=exe, + model_dir=config["Global"]["save_inference_dir"], + model_filename='inference.pdmodel', + params_filename='inference.pdiparams', + quantize_model_path=os.path.join( + config["Global"]["save_inference_dir"], "quant_post_static_model"), + sample_generator=sample_generator(train_dataloader), + batch_nums=10) + + +if __name__ == "__main__": + main() diff --git a/exprements/PaddleClas/deploy/utils/__init__.py b/exprements/PaddleClas/deploy/utils/__init__.py new file mode 100644 index 00000000..baf14a9e --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/__init__.py @@ -0,0 +1,5 @@ +from . import logger +from . import config +from . import get_image_list +from . import predictor +from . import encode_decode \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/utils/config.py b/exprements/PaddleClas/deploy/utils/config.py new file mode 100644 index 00000000..7e7ffb79 --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/config.py @@ -0,0 +1,197 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import copy +import argparse +import yaml + +from utils import logger + +__all__ = ['get_config'] + + +class AttrDict(dict): + def __getattr__(self, key): + return self[key] + + def __setattr__(self, key, value): + if key in self.__dict__: + self.__dict__[key] = value + else: + self[key] = value + + def __deepcopy__(self, content): + return copy.deepcopy(dict(self)) + + +def create_attr_dict(yaml_config): + from ast import literal_eval + for key, value in yaml_config.items(): + if type(value) is dict: + yaml_config[key] = value = AttrDict(value) + if isinstance(value, str): + try: + value = literal_eval(value) + except BaseException: + pass + if isinstance(value, AttrDict): + create_attr_dict(yaml_config[key]) + else: + yaml_config[key] = value + + +def parse_config(cfg_file): + """Load a config file into AttrDict""" + with open(cfg_file, 'r') as fopen: + yaml_config = AttrDict(yaml.load(fopen, Loader=yaml.SafeLoader)) + create_attr_dict(yaml_config) + return yaml_config + + +def print_dict(d, delimiter=0): + """ + Recursively visualize a dict and + indenting acrrording by the relationship of keys. + """ + placeholder = "-" * 60 + for k, v in sorted(d.items()): + if isinstance(v, dict): + logger.info("{}{} : ".format(delimiter * " ", + logger.coloring(k, "HEADER"))) + print_dict(v, delimiter + 4) + elif isinstance(v, list) and len(v) >= 1 and isinstance(v[0], dict): + logger.info("{}{} : ".format(delimiter * " ", + logger.coloring(str(k), "HEADER"))) + for value in v: + print_dict(value, delimiter + 4) + else: + logger.info("{}{} : {}".format(delimiter * " ", + logger.coloring(k, "HEADER"), + logger.coloring(v, "OKGREEN"))) + if k.isupper(): + logger.info(placeholder) + + +def print_config(config): + """ + visualize configs + Arguments: + config: configs + """ + logger.advertise() + print_dict(config) + + +def override(dl, ks, v): + """ + Recursively replace dict of list + Args: + dl(dict or list): dict or list to be replaced + ks(list): list of keys + v(str): value to be replaced + """ + + def str2num(v): + try: + return eval(v) + except Exception: + return v + + assert isinstance(dl, (list, dict)), ("{} should be a list or a dict") + assert len(ks) > 0, ('lenght of keys should larger than 0') + if isinstance(dl, list): + k = str2num(ks[0]) + if len(ks) == 1: + assert k < len(dl), ('index({}) out of range({})'.format(k, dl)) + dl[k] = str2num(v) + else: + override(dl[k], ks[1:], v) + else: + if len(ks) == 1: + # assert ks[0] in dl, ('{} is not exist in {}'.format(ks[0], dl)) + if not ks[0] in dl: + logger.warning('A new filed ({}) detected!'.format(ks[0])) + dl[ks[0]] = str2num(v) + else: + override(dl[ks[0]], ks[1:], v) + + +def override_config(config, options=None): + """ + Recursively override the config + Args: + config(dict): dict to be replaced + options(list): list of pairs(key0.key1.idx.key2=value) + such as: [ + 'topk=2', + 'VALID.transforms.1.ResizeImage.resize_short=300' + ] + Returns: + config(dict): replaced config + """ + if options is not None: + for opt in options: + assert isinstance(opt, str), ( + "option({}) should be a str".format(opt)) + assert "=" in opt, ( + "option({}) should contain a =" + "to distinguish between key and value".format(opt)) + pair = opt.split('=') + assert len(pair) == 2, ("there can be only a = in the option") + key, value = pair + keys = key.split('.') + override(config, keys, value) + return config + + +def get_config(fname, overrides=None, show=True): + """ + Read config from file + """ + assert os.path.exists(fname), ( + 'config file({}) is not exist'.format(fname)) + config = parse_config(fname) + override_config(config, overrides) + if show: + print_config(config) + # check_config(config) + return config + + +def parser(): + parser = argparse.ArgumentParser("generic-image-rec train script") + parser.add_argument( + '-c', + '--config', + type=str, + default='configs/config.yaml', + help='config file path') + parser.add_argument( + '-o', + '--override', + action='append', + default=[], + help='config options to be overridden') + parser.add_argument( + '-v', + '--verbose', + action='store_true', + help='wheather print the config info') + return parser + + +def parse_args(): + args = parser().parse_args() + return args diff --git a/exprements/PaddleClas/deploy/utils/draw_bbox.py b/exprements/PaddleClas/deploy/utils/draw_bbox.py new file mode 100644 index 00000000..9c0bcf61 --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/draw_bbox.py @@ -0,0 +1,61 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import numpy as np +import cv2 +from PIL import Image, ImageDraw, ImageFont + + +def draw_bbox_results(image, + results, + input_path, + font_path="./utils/simfang.ttf", + save_dir=None): + if isinstance(image, np.ndarray): + image = Image.fromarray(image) + draw = ImageDraw.Draw(image) + font_size = 18 + font = ImageFont.truetype(font_path, font_size, encoding="utf-8") + + color = (0, 102, 255) + + for result in results: + # empty results + if result["rec_docs"] is None: + continue + + xmin, ymin, xmax, ymax = result["bbox"] + text = "{}, {:.2f}".format(result["rec_docs"], result["rec_scores"]) + th = font_size + tw = font.getsize(text)[0] + # tw = int(len(result["rec_docs"]) * font_size) + 60 + start_y = max(0, ymin - th) + + draw.rectangle( + [(xmin + 1, start_y), (xmin + tw + 1, start_y + th)], fill=color) + + draw.text((xmin + 1, start_y), text, fill=(255, 255, 255), font=font) + + draw.rectangle( + [(xmin, ymin), (xmax, ymax)], outline=(255, 0, 0), width=2) + + image_name = os.path.basename(input_path) + if save_dir is None: + save_dir = "output" + os.makedirs(save_dir, exist_ok=True) + output_path = os.path.join(save_dir, image_name) + + image.save(output_path, quality=95) + return np.array(image) diff --git a/exprements/PaddleClas/deploy/utils/encode_decode.py b/exprements/PaddleClas/deploy/utils/encode_decode.py new file mode 100644 index 00000000..d76a529f --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/encode_decode.py @@ -0,0 +1,31 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import base64 + +import numpy as np + + +def np_to_b64(images): + img_str = base64.b64encode(images).decode('utf8') + return img_str, images.shape + + +def b64_to_np(b64str, revert_params): + shape = revert_params["shape"] + dtype = revert_params["dtype"] + dtype = getattr(np, dtype) if isinstance(str, type(dtype)) else dtype + data = base64.b64decode(b64str.encode('utf8')) + data = np.fromstring(data, dtype).reshape(shape) + return data \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/utils/get_image_list.py b/exprements/PaddleClas/deploy/utils/get_image_list.py new file mode 100644 index 00000000..6f10935a --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/get_image_list.py @@ -0,0 +1,49 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import argparse +import base64 +import numpy as np + + +def get_image_list(img_file): + imgs_lists = [] + if img_file is None or not os.path.exists(img_file): + raise Exception("not found any img file in {}".format(img_file)) + + img_end = ['jpg', 'png', 'jpeg', 'JPEG', 'JPG', 'bmp'] + if os.path.isfile(img_file) and img_file.split('.')[-1] in img_end: + imgs_lists.append(img_file) + elif os.path.isdir(img_file): + for single_file in os.listdir(img_file): + if single_file.split('.')[-1] in img_end: + imgs_lists.append(os.path.join(img_file, single_file)) + if len(imgs_lists) == 0: + raise Exception("not found any img file in {}".format(img_file)) + imgs_lists = sorted(imgs_lists) + return imgs_lists + + +def get_image_list_from_label_file(image_path, label_file_path): + imgs_lists = [] + gt_labels = [] + with open(label_file_path, "r") as fin: + lines = fin.readlines() + for line in lines: + image_name, label = line.strip("\n").split() + label = int(label) + imgs_lists.append(os.path.join(image_path, image_name)) + gt_labels.append(int(label)) + return imgs_lists, gt_labels diff --git a/exprements/PaddleClas/deploy/utils/imagenet1k_label_list.txt b/exprements/PaddleClas/deploy/utils/imagenet1k_label_list.txt new file mode 100644 index 00000000..376e1802 --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/imagenet1k_label_list.txt @@ -0,0 +1,1000 @@ +0 tench, Tinca tinca +1 goldfish, Carassius auratus +2 great white shark, white shark, man-eater, man-eating shark, Carcharodon carcharias +3 tiger shark, Galeocerdo cuvieri +4 hammerhead, hammerhead shark +5 electric ray, crampfish, numbfish, torpedo +6 stingray +7 cock +8 hen +9 ostrich, Struthio camelus +10 brambling, Fringilla montifringilla +11 goldfinch, Carduelis carduelis +12 house finch, linnet, Carpodacus mexicanus +13 junco, snowbird +14 indigo bunting, indigo finch, indigo bird, Passerina cyanea +15 robin, American robin, Turdus migratorius +16 bulbul +17 jay +18 magpie +19 chickadee +20 water ouzel, dipper +21 kite +22 bald eagle, American eagle, Haliaeetus leucocephalus +23 vulture +24 great grey owl, great gray owl, Strix nebulosa +25 European fire salamander, Salamandra salamandra +26 common newt, Triturus vulgaris +27 eft +28 spotted salamander, Ambystoma maculatum +29 axolotl, mud puppy, Ambystoma mexicanum +30 bullfrog, Rana catesbeiana +31 tree frog, tree-frog +32 tailed frog, bell toad, ribbed toad, tailed toad, Ascaphus trui +33 loggerhead, loggerhead turtle, Caretta caretta +34 leatherback turtle, leatherback, leathery turtle, Dermochelys coriacea +35 mud turtle +36 terrapin +37 box turtle, box tortoise +38 banded gecko +39 common iguana, iguana, Iguana iguana +40 American chameleon, anole, Anolis carolinensis +41 whiptail, whiptail lizard +42 agama +43 frilled lizard, Chlamydosaurus kingi +44 alligator lizard +45 Gila monster, Heloderma suspectum +46 green lizard, Lacerta viridis +47 African chameleon, Chamaeleo chamaeleon +48 Komodo dragon, Komodo lizard, dragon lizard, giant lizard, Varanus komodoensis +49 African crocodile, Nile crocodile, Crocodylus niloticus +50 American alligator, Alligator mississipiensis +51 triceratops +52 thunder snake, worm snake, Carphophis amoenus +53 ringneck snake, ring-necked snake, ring snake +54 hognose snake, puff adder, sand viper +55 green snake, grass snake +56 king snake, kingsnake +57 garter snake, grass snake +58 water snake +59 vine snake +60 night snake, Hypsiglena torquata +61 boa constrictor, Constrictor constrictor +62 rock python, rock snake, Python sebae +63 Indian cobra, Naja naja +64 green mamba +65 sea snake +66 horned viper, cerastes, sand viper, horned asp, Cerastes cornutus +67 diamondback, diamondback rattlesnake, Crotalus adamanteus +68 sidewinder, horned rattlesnake, Crotalus cerastes +69 trilobite +70 harvestman, daddy longlegs, Phalangium opilio +71 scorpion +72 black and gold garden spider, Argiope aurantia +73 barn spider, Araneus cavaticus +74 garden spider, Aranea diademata +75 black widow, Latrodectus mactans +76 tarantula +77 wolf spider, hunting spider +78 tick +79 centipede +80 black grouse +81 ptarmigan +82 ruffed grouse, partridge, Bonasa umbellus +83 prairie chicken, prairie grouse, prairie fowl +84 peacock +85 quail +86 partridge +87 African grey, African gray, Psittacus erithacus +88 macaw +89 sulphur-crested cockatoo, Kakatoe galerita, Cacatua galerita +90 lorikeet +91 coucal +92 bee eater +93 hornbill +94 hummingbird +95 jacamar +96 toucan +97 drake +98 red-breasted merganser, Mergus serrator +99 goose +100 black swan, Cygnus atratus +101 tusker +102 echidna, spiny anteater, anteater +103 platypus, duckbill, duckbilled platypus, duck-billed platypus, Ornithorhynchus anatinus +104 wallaby, brush kangaroo +105 koala, koala bear, kangaroo bear, native bear, Phascolarctos cinereus +106 wombat +107 jellyfish +108 sea anemone, anemone +109 brain coral +110 flatworm, platyhelminth +111 nematode, nematode worm, roundworm +112 conch +113 snail +114 slug +115 sea slug, nudibranch +116 chiton, coat-of-mail shell, sea cradle, polyplacophore +117 chambered nautilus, pearly nautilus, nautilus +118 Dungeness crab, Cancer magister +119 rock crab, Cancer irroratus +120 fiddler crab +121 king crab, Alaska crab, Alaskan king crab, Alaska king crab, Paralithodes camtschatica +122 American lobster, Northern lobster, Maine lobster, Homarus americanus +123 spiny lobster, langouste, rock lobster, crawfish, crayfish, sea crawfish +124 crayfish, crawfish, crawdad, crawdaddy +125 hermit crab +126 isopod +127 white stork, Ciconia ciconia +128 black stork, Ciconia nigra +129 spoonbill +130 flamingo +131 little blue heron, Egretta caerulea +132 American egret, great white heron, Egretta albus +133 bittern +134 crane +135 limpkin, Aramus pictus +136 European gallinule, Porphyrio porphyrio +137 American coot, marsh hen, mud hen, water hen, Fulica americana +138 bustard +139 ruddy turnstone, Arenaria interpres +140 red-backed sandpiper, dunlin, Erolia alpina +141 redshank, Tringa totanus +142 dowitcher +143 oystercatcher, oyster catcher +144 pelican +145 king penguin, Aptenodytes patagonica +146 albatross, mollymawk +147 grey whale, gray whale, devilfish, Eschrichtius gibbosus, Eschrichtius robustus +148 killer whale, killer, orca, grampus, sea wolf, Orcinus orca +149 dugong, Dugong dugon +150 sea lion +151 Chihuahua +152 Japanese spaniel +153 Maltese dog, Maltese terrier, Maltese +154 Pekinese, Pekingese, Peke +155 Shih-Tzu +156 Blenheim spaniel +157 papillon +158 toy terrier +159 Rhodesian ridgeback +160 Afghan hound, Afghan +161 basset, basset hound +162 beagle +163 bloodhound, sleuthhound +164 bluetick +165 black-and-tan coonhound +166 Walker hound, Walker foxhound +167 English foxhound +168 redbone +169 borzoi, Russian wolfhound +170 Irish wolfhound +171 Italian greyhound +172 whippet +173 Ibizan hound, Ibizan Podenco +174 Norwegian elkhound, elkhound +175 otterhound, otter hound +176 Saluki, gazelle hound +177 Scottish deerhound, deerhound +178 Weimaraner +179 Staffordshire bullterrier, Staffordshire bull terrier +180 American Staffordshire terrier, Staffordshire terrier, American pit bull terrier, pit bull terrier +181 Bedlington terrier +182 Border terrier +183 Kerry blue terrier +184 Irish terrier +185 Norfolk terrier +186 Norwich terrier +187 Yorkshire terrier +188 wire-haired fox terrier +189 Lakeland terrier +190 Sealyham terrier, Sealyham +191 Airedale, Airedale terrier +192 cairn, cairn terrier +193 Australian terrier +194 Dandie Dinmont, Dandie Dinmont terrier +195 Boston bull, Boston terrier +196 miniature schnauzer +197 giant schnauzer +198 standard schnauzer +199 Scotch terrier, Scottish terrier, Scottie +200 Tibetan terrier, chrysanthemum dog +201 silky terrier, Sydney silky +202 soft-coated wheaten terrier +203 West Highland white terrier +204 Lhasa, Lhasa apso +205 flat-coated retriever +206 curly-coated retriever +207 golden retriever +208 Labrador retriever +209 Chesapeake Bay retriever +210 German short-haired pointer +211 vizsla, Hungarian pointer +212 English setter +213 Irish setter, red setter +214 Gordon setter +215 Brittany spaniel +216 clumber, clumber spaniel +217 English springer, English springer spaniel +218 Welsh springer spaniel +219 cocker spaniel, English cocker spaniel, cocker +220 Sussex spaniel +221 Irish water spaniel +222 kuvasz +223 schipperke +224 groenendael +225 malinois +226 briard +227 kelpie +228 komondor +229 Old English sheepdog, bobtail +230 Shetland sheepdog, Shetland sheep dog, Shetland +231 collie +232 Border collie +233 Bouvier des Flandres, Bouviers des Flandres +234 Rottweiler +235 German shepherd, German shepherd dog, German police dog, alsatian +236 Doberman, Doberman pinscher +237 miniature pinscher +238 Greater Swiss Mountain dog +239 Bernese mountain dog +240 Appenzeller +241 EntleBucher +242 boxer +243 bull mastiff +244 Tibetan mastiff +245 French bulldog +246 Great Dane +247 Saint Bernard, St Bernard +248 Eskimo dog, husky +249 malamute, malemute, Alaskan malamute +250 Siberian husky +251 dalmatian, coach dog, carriage dog +252 affenpinscher, monkey pinscher, monkey dog +253 basenji +254 pug, pug-dog +255 Leonberg +256 Newfoundland, Newfoundland dog +257 Great Pyrenees +258 Samoyed, Samoyede +259 Pomeranian +260 chow, chow chow +261 keeshond +262 Brabancon griffon +263 Pembroke, Pembroke Welsh corgi +264 Cardigan, Cardigan Welsh corgi +265 toy poodle +266 miniature poodle +267 standard poodle +268 Mexican hairless +269 timber wolf, grey wolf, gray wolf, Canis lupus +270 white wolf, Arctic wolf, Canis lupus tundrarum +271 red wolf, maned wolf, Canis rufus, Canis niger +272 coyote, prairie wolf, brush wolf, Canis latrans +273 dingo, warrigal, warragal, Canis dingo +274 dhole, Cuon alpinus +275 African hunting dog, hyena dog, Cape hunting dog, Lycaon pictus +276 hyena, hyaena +277 red fox, Vulpes vulpes +278 kit fox, Vulpes macrotis +279 Arctic fox, white fox, Alopex lagopus +280 grey fox, gray fox, Urocyon cinereoargenteus +281 tabby, tabby cat +282 tiger cat +283 Persian cat +284 Siamese cat, Siamese +285 Egyptian cat +286 cougar, puma, catamount, mountain lion, painter, panther, Felis concolor +287 lynx, catamount +288 leopard, Panthera pardus +289 snow leopard, ounce, Panthera uncia +290 jaguar, panther, Panthera onca, Felis onca +291 lion, king of beasts, Panthera leo +292 tiger, Panthera tigris +293 cheetah, chetah, Acinonyx jubatus +294 brown bear, bruin, Ursus arctos +295 American black bear, black bear, Ursus americanus, Euarctos americanus +296 ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus +297 sloth bear, Melursus ursinus, Ursus ursinus +298 mongoose +299 meerkat, mierkat +300 tiger beetle +301 ladybug, ladybeetle, lady beetle, ladybird, ladybird beetle +302 ground beetle, carabid beetle +303 long-horned beetle, longicorn, longicorn beetle +304 leaf beetle, chrysomelid +305 dung beetle +306 rhinoceros beetle +307 weevil +308 fly +309 bee +310 ant, emmet, pismire +311 grasshopper, hopper +312 cricket +313 walking stick, walkingstick, stick insect +314 cockroach, roach +315 mantis, mantid +316 cicada, cicala +317 leafhopper +318 lacewing, lacewing fly +319 dragonfly, darning needle, devil's darning needle, sewing needle, snake feeder, snake doctor, mosquito hawk, skeeter hawk +320 damselfly +321 admiral +322 ringlet, ringlet butterfly +323 monarch, monarch butterfly, milkweed butterfly, Danaus plexippus +324 cabbage butterfly +325 sulphur butterfly, sulfur butterfly +326 lycaenid, lycaenid butterfly +327 starfish, sea star +328 sea urchin +329 sea cucumber, holothurian +330 wood rabbit, cottontail, cottontail rabbit +331 hare +332 Angora, Angora rabbit +333 hamster +334 porcupine, hedgehog +335 fox squirrel, eastern fox squirrel, Sciurus niger +336 marmot +337 beaver +338 guinea pig, Cavia cobaya +339 sorrel +340 zebra +341 hog, pig, grunter, squealer, Sus scrofa +342 wild boar, boar, Sus scrofa +343 warthog +344 hippopotamus, hippo, river horse, Hippopotamus amphibius +345 ox +346 water buffalo, water ox, Asiatic buffalo, Bubalus bubalis +347 bison +348 ram, tup +349 bighorn, bighorn sheep, cimarron, Rocky Mountain bighorn, Rocky Mountain sheep, Ovis canadensis +350 ibex, Capra ibex +351 hartebeest +352 impala, Aepyceros melampus +353 gazelle +354 Arabian camel, dromedary, Camelus dromedarius +355 llama +356 weasel +357 mink +358 polecat, fitch, foulmart, foumart, Mustela putorius +359 black-footed ferret, ferret, Mustela nigripes +360 otter +361 skunk, polecat, wood pussy +362 badger +363 armadillo +364 three-toed sloth, ai, Bradypus tridactylus +365 orangutan, orang, orangutang, Pongo pygmaeus +366 gorilla, Gorilla gorilla +367 chimpanzee, chimp, Pan troglodytes +368 gibbon, Hylobates lar +369 siamang, Hylobates syndactylus, Symphalangus syndactylus +370 guenon, guenon monkey +371 patas, hussar monkey, Erythrocebus patas +372 baboon +373 macaque +374 langur +375 colobus, colobus monkey +376 proboscis monkey, Nasalis larvatus +377 marmoset +378 capuchin, ringtail, Cebus capucinus +379 howler monkey, howler +380 titi, titi monkey +381 spider monkey, Ateles geoffroyi +382 squirrel monkey, Saimiri sciureus +383 Madagascar cat, ring-tailed lemur, Lemur catta +384 indri, indris, Indri indri, Indri brevicaudatus +385 Indian elephant, Elephas maximus +386 African elephant, Loxodonta africana +387 lesser panda, red panda, panda, bear cat, cat bear, Ailurus fulgens +388 giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca +389 barracouta, snoek +390 eel +391 coho, cohoe, coho salmon, blue jack, silver salmon, Oncorhynchus kisutch +392 rock beauty, Holocanthus tricolor +393 anemone fish +394 sturgeon +395 gar, garfish, garpike, billfish, Lepisosteus osseus +396 lionfish +397 puffer, pufferfish, blowfish, globefish +398 abacus +399 abaya +400 academic gown, academic robe, judge's robe +401 accordion, piano accordion, squeeze box +402 acoustic guitar +403 aircraft carrier, carrier, flattop, attack aircraft carrier +404 airliner +405 airship, dirigible +406 altar +407 ambulance +408 amphibian, amphibious vehicle +409 analog clock +410 apiary, bee house +411 apron +412 ashcan, trash can, garbage can, wastebin, ash bin, ash-bin, ashbin, dustbin, trash barrel, trash bin +413 assault rifle, assault gun +414 backpack, back pack, knapsack, packsack, rucksack, haversack +415 bakery, bakeshop, bakehouse +416 balance beam, beam +417 balloon +418 ballpoint, ballpoint pen, ballpen, Biro +419 Band Aid +420 banjo +421 bannister, banister, balustrade, balusters, handrail +422 barbell +423 barber chair +424 barbershop +425 barn +426 barometer +427 barrel, cask +428 barrow, garden cart, lawn cart, wheelbarrow +429 baseball +430 basketball +431 bassinet +432 bassoon +433 bathing cap, swimming cap +434 bath towel +435 bathtub, bathing tub, bath, tub +436 beach wagon, station wagon, wagon, estate car, beach waggon, station waggon, waggon +437 beacon, lighthouse, beacon light, pharos +438 beaker +439 bearskin, busby, shako +440 beer bottle +441 beer glass +442 bell cote, bell cot +443 bib +444 bicycle-built-for-two, tandem bicycle, tandem +445 bikini, two-piece +446 binder, ring-binder +447 binoculars, field glasses, opera glasses +448 birdhouse +449 boathouse +450 bobsled, bobsleigh, bob +451 bolo tie, bolo, bola tie, bola +452 bonnet, poke bonnet +453 bookcase +454 bookshop, bookstore, bookstall +455 bottlecap +456 bow +457 bow tie, bow-tie, bowtie +458 brass, memorial tablet, plaque +459 brassiere, bra, bandeau +460 breakwater, groin, groyne, mole, bulwark, seawall, jetty +461 breastplate, aegis, egis +462 broom +463 bucket, pail +464 buckle +465 bulletproof vest +466 bullet train, bullet +467 butcher shop, meat market +468 cab, hack, taxi, taxicab +469 caldron, cauldron +470 candle, taper, wax light +471 cannon +472 canoe +473 can opener, tin opener +474 cardigan +475 car mirror +476 carousel, carrousel, merry-go-round, roundabout, whirligig +477 carpenter's kit, tool kit +478 carton +479 car wheel +480 cash machine, cash dispenser, automated teller machine, automatic teller machine, automated teller, automatic teller, ATM +481 cassette +482 cassette player +483 castle +484 catamaran +485 CD player +486 cello, violoncello +487 cellular telephone, cellular phone, cellphone, cell, mobile phone +488 chain +489 chainlink fence +490 chain mail, ring mail, mail, chain armor, chain armour, ring armor, ring armour +491 chain saw, chainsaw +492 chest +493 chiffonier, commode +494 chime, bell, gong +495 china cabinet, china closet +496 Christmas stocking +497 church, church building +498 cinema, movie theater, movie theatre, movie house, picture palace +499 cleaver, meat cleaver, chopper +500 cliff dwelling +501 cloak +502 clog, geta, patten, sabot +503 cocktail shaker +504 coffee mug +505 coffeepot +506 coil, spiral, volute, whorl, helix +507 combination lock +508 computer keyboard, keypad +509 confectionery, confectionary, candy store +510 container ship, containership, container vessel +511 convertible +512 corkscrew, bottle screw +513 cornet, horn, trumpet, trump +514 cowboy boot +515 cowboy hat, ten-gallon hat +516 cradle +517 crane +518 crash helmet +519 crate +520 crib, cot +521 Crock Pot +522 croquet ball +523 crutch +524 cuirass +525 dam, dike, dyke +526 desk +527 desktop computer +528 dial telephone, dial phone +529 diaper, nappy, napkin +530 digital clock +531 digital watch +532 dining table, board +533 dishrag, dishcloth +534 dishwasher, dish washer, dishwashing machine +535 disk brake, disc brake +536 dock, dockage, docking facility +537 dogsled, dog sled, dog sleigh +538 dome +539 doormat, welcome mat +540 drilling platform, offshore rig +541 drum, membranophone, tympan +542 drumstick +543 dumbbell +544 Dutch oven +545 electric fan, blower +546 electric guitar +547 electric locomotive +548 entertainment center +549 envelope +550 espresso maker +551 face powder +552 feather boa, boa +553 file, file cabinet, filing cabinet +554 fireboat +555 fire engine, fire truck +556 fire screen, fireguard +557 flagpole, flagstaff +558 flute, transverse flute +559 folding chair +560 football helmet +561 forklift +562 fountain +563 fountain pen +564 four-poster +565 freight car +566 French horn, horn +567 frying pan, frypan, skillet +568 fur coat +569 garbage truck, dustcart +570 gasmask, respirator, gas helmet +571 gas pump, gasoline pump, petrol pump, island dispenser +572 goblet +573 go-kart +574 golf ball +575 golfcart, golf cart +576 gondola +577 gong, tam-tam +578 gown +579 grand piano, grand +580 greenhouse, nursery, glasshouse +581 grille, radiator grille +582 grocery store, grocery, food market, market +583 guillotine +584 hair slide +585 hair spray +586 half track +587 hammer +588 hamper +589 hand blower, blow dryer, blow drier, hair dryer, hair drier +590 hand-held computer, hand-held microcomputer +591 handkerchief, hankie, hanky, hankey +592 hard disc, hard disk, fixed disk +593 harmonica, mouth organ, harp, mouth harp +594 harp +595 harvester, reaper +596 hatchet +597 holster +598 home theater, home theatre +599 honeycomb +600 hook, claw +601 hoopskirt, crinoline +602 horizontal bar, high bar +603 horse cart, horse-cart +604 hourglass +605 iPod +606 iron, smoothing iron +607 jack-o'-lantern +608 jean, blue jean, denim +609 jeep, landrover +610 jersey, T-shirt, tee shirt +611 jigsaw puzzle +612 jinrikisha, ricksha, rickshaw +613 joystick +614 kimono +615 knee pad +616 knot +617 lab coat, laboratory coat +618 ladle +619 lampshade, lamp shade +620 laptop, laptop computer +621 lawn mower, mower +622 lens cap, lens cover +623 letter opener, paper knife, paperknife +624 library +625 lifeboat +626 lighter, light, igniter, ignitor +627 limousine, limo +628 liner, ocean liner +629 lipstick, lip rouge +630 Loafer +631 lotion +632 loudspeaker, speaker, speaker unit, loudspeaker system, speaker system +633 loupe, jeweler's loupe +634 lumbermill, sawmill +635 magnetic compass +636 mailbag, postbag +637 mailbox, letter box +638 maillot +639 maillot, tank suit +640 manhole cover +641 maraca +642 marimba, xylophone +643 mask +644 matchstick +645 maypole +646 maze, labyrinth +647 measuring cup +648 medicine chest, medicine cabinet +649 megalith, megalithic structure +650 microphone, mike +651 microwave, microwave oven +652 military uniform +653 milk can +654 minibus +655 miniskirt, mini +656 minivan +657 missile +658 mitten +659 mixing bowl +660 mobile home, manufactured home +661 Model T +662 modem +663 monastery +664 monitor +665 moped +666 mortar +667 mortarboard +668 mosque +669 mosquito net +670 motor scooter, scooter +671 mountain bike, all-terrain bike, off-roader +672 mountain tent +673 mouse, computer mouse +674 mousetrap +675 moving van +676 muzzle +677 nail +678 neck brace +679 necklace +680 nipple +681 notebook, notebook computer +682 obelisk +683 oboe, hautboy, hautbois +684 ocarina, sweet potato +685 odometer, hodometer, mileometer, milometer +686 oil filter +687 organ, pipe organ +688 oscilloscope, scope, cathode-ray oscilloscope, CRO +689 overskirt +690 oxcart +691 oxygen mask +692 packet +693 paddle, boat paddle +694 paddlewheel, paddle wheel +695 padlock +696 paintbrush +697 pajama, pyjama, pj's, jammies +698 palace +699 panpipe, pandean pipe, syrinx +700 paper towel +701 parachute, chute +702 parallel bars, bars +703 park bench +704 parking meter +705 passenger car, coach, carriage +706 patio, terrace +707 pay-phone, pay-station +708 pedestal, plinth, footstall +709 pencil box, pencil case +710 pencil sharpener +711 perfume, essence +712 Petri dish +713 photocopier +714 pick, plectrum, plectron +715 pickelhaube +716 picket fence, paling +717 pickup, pickup truck +718 pier +719 piggy bank, penny bank +720 pill bottle +721 pillow +722 ping-pong ball +723 pinwheel +724 pirate, pirate ship +725 pitcher, ewer +726 plane, carpenter's plane, woodworking plane +727 planetarium +728 plastic bag +729 plate rack +730 plow, plough +731 plunger, plumber's helper +732 Polaroid camera, Polaroid Land camera +733 pole +734 police van, police wagon, paddy wagon, patrol wagon, wagon, black Maria +735 poncho +736 pool table, billiard table, snooker table +737 pop bottle, soda bottle +738 pot, flowerpot +739 potter's wheel +740 power drill +741 prayer rug, prayer mat +742 printer +743 prison, prison house +744 projectile, missile +745 projector +746 puck, hockey puck +747 punching bag, punch bag, punching ball, punchball +748 purse +749 quill, quill pen +750 quilt, comforter, comfort, puff +751 racer, race car, racing car +752 racket, racquet +753 radiator +754 radio, wireless +755 radio telescope, radio reflector +756 rain barrel +757 recreational vehicle, RV, R.V. +758 reel +759 reflex camera +760 refrigerator, icebox +761 remote control, remote +762 restaurant, eating house, eating place, eatery +763 revolver, six-gun, six-shooter +764 rifle +765 rocking chair, rocker +766 rotisserie +767 rubber eraser, rubber, pencil eraser +768 rugby ball +769 rule, ruler +770 running shoe +771 safe +772 safety pin +773 saltshaker, salt shaker +774 sandal +775 sarong +776 sax, saxophone +777 scabbard +778 scale, weighing machine +779 school bus +780 schooner +781 scoreboard +782 screen, CRT screen +783 screw +784 screwdriver +785 seat belt, seatbelt +786 sewing machine +787 shield, buckler +788 shoe shop, shoe-shop, shoe store +789 shoji +790 shopping basket +791 shopping cart +792 shovel +793 shower cap +794 shower curtain +795 ski +796 ski mask +797 sleeping bag +798 slide rule, slipstick +799 sliding door +800 slot, one-armed bandit +801 snorkel +802 snowmobile +803 snowplow, snowplough +804 soap dispenser +805 soccer ball +806 sock +807 solar dish, solar collector, solar furnace +808 sombrero +809 soup bowl +810 space bar +811 space heater +812 space shuttle +813 spatula +814 speedboat +815 spider web, spider's web +816 spindle +817 sports car, sport car +818 spotlight, spot +819 stage +820 steam locomotive +821 steel arch bridge +822 steel drum +823 stethoscope +824 stole +825 stone wall +826 stopwatch, stop watch +827 stove +828 strainer +829 streetcar, tram, tramcar, trolley, trolley car +830 stretcher +831 studio couch, day bed +832 stupa, tope +833 submarine, pigboat, sub, U-boat +834 suit, suit of clothes +835 sundial +836 sunglass +837 sunglasses, dark glasses, shades +838 sunscreen, sunblock, sun blocker +839 suspension bridge +840 swab, swob, mop +841 sweatshirt +842 swimming trunks, bathing trunks +843 swing +844 switch, electric switch, electrical switch +845 syringe +846 table lamp +847 tank, army tank, armored combat vehicle, armoured combat vehicle +848 tape player +849 teapot +850 teddy, teddy bear +851 television, television system +852 tennis ball +853 thatch, thatched roof +854 theater curtain, theatre curtain +855 thimble +856 thresher, thrasher, threshing machine +857 throne +858 tile roof +859 toaster +860 tobacco shop, tobacconist shop, tobacconist +861 toilet seat +862 torch +863 totem pole +864 tow truck, tow car, wrecker +865 toyshop +866 tractor +867 trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +868 tray +869 trench coat +870 tricycle, trike, velocipede +871 trimaran +872 tripod +873 triumphal arch +874 trolleybus, trolley coach, trackless trolley +875 trombone +876 tub, vat +877 turnstile +878 typewriter keyboard +879 umbrella +880 unicycle, monocycle +881 upright, upright piano +882 vacuum, vacuum cleaner +883 vase +884 vault +885 velvet +886 vending machine +887 vestment +888 viaduct +889 violin, fiddle +890 volleyball +891 waffle iron +892 wall clock +893 wallet, billfold, notecase, pocketbook +894 wardrobe, closet, press +895 warplane, military plane +896 washbasin, handbasin, washbowl, lavabo, wash-hand basin +897 washer, automatic washer, washing machine +898 water bottle +899 water jug +900 water tower +901 whiskey jug +902 whistle +903 wig +904 window screen +905 window shade +906 Windsor tie +907 wine bottle +908 wing +909 wok +910 wooden spoon +911 wool, woolen, woollen +912 worm fence, snake fence, snake-rail fence, Virginia fence +913 wreck +914 yawl +915 yurt +916 web site, website, internet site, site +917 comic book +918 crossword puzzle, crossword +919 street sign +920 traffic light, traffic signal, stoplight +921 book jacket, dust cover, dust jacket, dust wrapper +922 menu +923 plate +924 guacamole +925 consomme +926 hot pot, hotpot +927 trifle +928 ice cream, icecream +929 ice lolly, lolly, lollipop, popsicle +930 French loaf +931 bagel, beigel +932 pretzel +933 cheeseburger +934 hotdog, hot dog, red hot +935 mashed potato +936 head cabbage +937 broccoli +938 cauliflower +939 zucchini, courgette +940 spaghetti squash +941 acorn squash +942 butternut squash +943 cucumber, cuke +944 artichoke, globe artichoke +945 bell pepper +946 cardoon +947 mushroom +948 Granny Smith +949 strawberry +950 orange +951 lemon +952 fig +953 pineapple, ananas +954 banana +955 jackfruit, jak, jack +956 custard apple +957 pomegranate +958 hay +959 carbonara +960 chocolate sauce, chocolate syrup +961 dough +962 meat loaf, meatloaf +963 pizza, pizza pie +964 potpie +965 burrito +966 red wine +967 espresso +968 cup +969 eggnog +970 alp +971 bubble +972 cliff, drop, drop-off +973 coral reef +974 geyser +975 lakeside, lakeshore +976 promontory, headland, head, foreland +977 sandbar, sand bar +978 seashore, coast, seacoast, sea-coast +979 valley, vale +980 volcano +981 ballplayer, baseball player +982 groom, bridegroom +983 scuba diver +984 rapeseed +985 daisy +986 yellow lady's slipper, yellow lady-slipper, Cypripedium calceolus, Cypripedium parviflorum +987 corn +988 acorn +989 hip, rose hip, rosehip +990 buckeye, horse chestnut, conker +991 coral fungus +992 agaric +993 gyromitra +994 stinkhorn, carrion fungus +995 earthstar +996 hen-of-the-woods, hen of the woods, Polyporus frondosus, Grifola frondosa +997 bolete +998 ear, spike, capitulum +999 toilet tissue, toilet paper, bathroom tissue diff --git a/exprements/PaddleClas/deploy/utils/logger.py b/exprements/PaddleClas/deploy/utils/logger.py new file mode 100644 index 00000000..ece85262 --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/logger.py @@ -0,0 +1,120 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +import os +import datetime + +logging.basicConfig( + level=logging.INFO, + format="%(asctime)s %(levelname)s: %(message)s", + datefmt="%Y-%m-%d %H:%M:%S") + + +def time_zone(sec, fmt): + real_time = datetime.datetime.now() + return real_time.timetuple() + + +logging.Formatter.converter = time_zone +_logger = logging.getLogger(__name__) + +Color = { + 'RED': '\033[31m', + 'HEADER': '\033[35m', # deep purple + 'PURPLE': '\033[95m', # purple + 'OKBLUE': '\033[94m', + 'OKGREEN': '\033[92m', + 'WARNING': '\033[93m', + 'FAIL': '\033[91m', + 'ENDC': '\033[0m' +} + + +def coloring(message, color="OKGREEN"): + assert color in Color.keys() + if os.environ.get('PADDLECLAS_COLORING', False): + return Color[color] + str(message) + Color["ENDC"] + else: + return message + + +def anti_fleet(log): + """ + logs will print multi-times when calling Fleet API. + Only display single log and ignore the others. + """ + + def wrapper(fmt, *args): + if int(os.getenv("PADDLE_TRAINER_ID", 0)) == 0: + log(fmt, *args) + + return wrapper + + +@anti_fleet +def info(fmt, *args): + _logger.info(fmt, *args) + + +@anti_fleet +def warning(fmt, *args): + _logger.warning(coloring(fmt, "RED"), *args) + + +@anti_fleet +def error(fmt, *args): + _logger.error(coloring(fmt, "FAIL"), *args) + + +def scaler(name, value, step, writer): + """ + This function will draw a scalar curve generated by the visualdl. + Usage: Install visualdl: pip3 install visualdl==2.0.0b4 + and then: + visualdl --logdir ./scalar --host 0.0.0.0 --port 8830 + to preview loss corve in real time. + """ + writer.add_scalar(tag=name, step=step, value=value) + + +def advertise(): + """ + Show the advertising message like the following: + + =========================================================== + == PaddleClas is powered by PaddlePaddle ! == + =========================================================== + == == + == For more info please go to the following website. == + == == + == https://github.com/PaddlePaddle/PaddleClas == + =========================================================== + + """ + copyright = "PaddleClas is powered by PaddlePaddle !" + ad = "For more info please go to the following website." + website = "https://github.com/PaddlePaddle/PaddleClas" + AD_LEN = 6 + len(max([copyright, ad, website], key=len)) + + info( + coloring("\n{0}\n{1}\n{2}\n{3}\n{4}\n{5}\n{6}\n{7}\n".format( + "=" * (AD_LEN + 4), + "=={}==".format(copyright.center(AD_LEN)), + "=" * (AD_LEN + 4), + "=={}==".format(' ' * AD_LEN), + "=={}==".format(ad.center(AD_LEN)), + "=={}==".format(' ' * AD_LEN), + "=={}==".format(website.center(AD_LEN)), + "=" * (AD_LEN + 4), ), "RED")) diff --git a/exprements/PaddleClas/deploy/utils/predictor.py b/exprements/PaddleClas/deploy/utils/predictor.py new file mode 100644 index 00000000..d44ae03d --- /dev/null +++ b/exprements/PaddleClas/deploy/utils/predictor.py @@ -0,0 +1,92 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import os +import argparse +import base64 +import shutil +import cv2 +import numpy as np + +from paddle.inference import Config +from paddle.inference import create_predictor + + +class Predictor(object): + def __init__(self, args, inference_model_dir=None): + # HALF precission predict only work when using tensorrt + if args.use_fp16 is True: + assert args.use_tensorrt is True + self.args = args + if self.args.get("use_onnx", False): + self.predictor, self.config = self.create_onnx_predictor( + args, inference_model_dir) + else: + self.predictor, self.config = self.create_paddle_predictor( + args, inference_model_dir) + + def predict(self, image): + raise NotImplementedError + + def create_paddle_predictor(self, args, inference_model_dir=None): + if inference_model_dir is None: + inference_model_dir = args.inference_model_dir + params_file = os.path.join(inference_model_dir, "inference.pdiparams") + model_file = os.path.join(inference_model_dir, "inference.pdmodel") + config = Config(model_file, params_file) + + if args.use_gpu: + config.enable_use_gpu(args.gpu_mem, 0) + else: + config.disable_gpu() + if args.enable_mkldnn: + # cache 10 different shapes for mkldnn to avoid memory leak + config.set_mkldnn_cache_capacity(10) + config.enable_mkldnn() + config.set_cpu_math_library_num_threads(args.cpu_num_threads) + + if args.enable_profile: + config.enable_profile() + config.disable_glog_info() + config.switch_ir_optim(args.ir_optim) # default true + if args.use_tensorrt: + config.enable_tensorrt_engine( + precision_mode=Config.Precision.Half + if args.use_fp16 else Config.Precision.Float32, + max_batch_size=args.batch_size, + workspace_size=1 << 30, + min_subgraph_size=30) + + config.enable_memory_optim() + # use zero copy + config.switch_use_feed_fetch_ops(False) + predictor = create_predictor(config) + + return predictor, config + + def create_onnx_predictor(self, args, inference_model_dir=None): + import onnxruntime as ort + if inference_model_dir is None: + inference_model_dir = args.inference_model_dir + model_file = os.path.join(inference_model_dir, "inference.onnx") + config = ort.SessionOptions() + if args.use_gpu: + raise ValueError( + "onnx inference now only supports cpu! please specify use_gpu false." + ) + else: + config.intra_op_num_threads = args.cpu_num_threads + if args.ir_optim: + config.graph_optimization_level = ort.GraphOptimizationLevel.ORT_ENABLE_ALL + predictor = ort.InferenceSession(model_file, sess_options=config) + return predictor, config diff --git a/exprements/PaddleClas/deploy/utils/simfang.ttf b/exprements/PaddleClas/deploy/utils/simfang.ttf new file mode 100644 index 00000000..2b59eae4 Binary files /dev/null and b/exprements/PaddleClas/deploy/utils/simfang.ttf differ diff --git a/exprements/PaddleClas/deploy/vector_search/Makefile b/exprements/PaddleClas/deploy/vector_search/Makefile new file mode 100644 index 00000000..ec2df598 --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/Makefile @@ -0,0 +1,15 @@ +CXX=g++ + +ifeq ($(OS),Windows_NT) + postfix=dll +else + postfix=so +endif + +all : index + +index : src/config.h src/graph.h src/data.h interface.cc + ${CXX} -shared -fPIC interface.cc -o index.${postfix} -std=c++11 -Ofast -march=native -g -flto -funroll-loops -DOMP -fopenmp + +clean : + rm index.${postfix} \ No newline at end of file diff --git a/exprements/PaddleClas/deploy/vector_search/README.md b/exprements/PaddleClas/deploy/vector_search/README.md new file mode 100644 index 00000000..afa1dc28 --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/README.md @@ -0,0 +1,93 @@ +# 向量检索 + +**注意**：由于系统适配性问题，在新版本中，此检索算法将被废弃。新版本中将使用[faiss](https://github.com/facebookresearch/faiss)，整体检索的过程保持不变，但建立索引及检索时的yaml文件有所修改。 +## 1. 简介 + +一些垂域识别任务（如车辆、商品等）需要识别的类别数较大，往往采用基于检索的方式，通过查询向量与底库向量进行快速的最近邻搜索，获得匹配的预测类别。向量检索模块提供基础的近似最近邻搜索算法，基于百度自研的Möbius算法，一种基于图的近似最近邻搜索算法，用于最大内积搜索 (MIPS)。 该模块提供python接口，支持numpy和 tensor类型向量,支持L2和Inner Product距离计算。 + +Mobius 算法细节详见论文 （[Möbius Transformation for Fast Inner Product Search on Graph](http://research.baidu.com/Public/uploads/5e189d36b5cf6.PDF), [Code](https://github.com/sunbelbd/mobius)） + + + +## 2. 安装 + +### 2.1 直接使用提供的库文件 + +该文件夹下有已经编译好的`index.so`(gcc8.2.0下编译，用于Linux)以及`index.dll`(gcc10.3.0下编译，用于Windows)，可以跳过2.2与2.3节，直接使用。 + +如果因为gcc版本过低或者环境不兼容的问题，导致库文件无法使用，则需要在不同的平台下手动编译库文件。 + +**注意：** +请确保您的 C++ 编译器支持 C++11 标准。 + + +### 2.2 Linux上编译生成库文件 + +运行下面的命令，安装gcc与g++。 + +```shell +sudo apt-get update +sudo apt-get upgrade -y +sudo apt-get install build-essential gcc g++ +``` + +可以通过命令`gcc -v`查看gcc版本。 + +进入该文件夹，直接运行`make`即可，如果希望重新生成`index.so`文件，可以首先使用`make clean`清除已经生成的缓存，再使用`make`生成更新之后的库文件。 + +### 2.3 Windows上编译生成库文件 + +Windows上首先需要安装gcc编译工具，推荐使用[TDM-GCC](https://jmeubank.github.io/tdm-gcc/articles/2020-03/9.2.0-release)，进入官网之后，可以选择合适的版本进行下载。推荐下载[tdm64-gcc-10.3.0-2.exe](https://github.com/jmeubank/tdm-gcc/releases/download/v10.3.0-tdm64-2/tdm64-gcc-10.3.0-2.exe)。 + +下载完成之后，按照默认的安装步骤进行安装即可。这里有3点需要注意： +1. 向量检索模块依赖于openmp，因此在安装到`choose components`步骤的时候，需要勾选上`openmp`的安装选项，否则之后编译的时候会报错`libgomp.spec: No such file or directory`，[参考链接](https://github.com/dmlc/xgboost/issues/1027) +2. 安装过程中会提示是否需要添加到系统的环境变量中，这里建议勾选上，否则之后使用的时候还需要手动添加系统环境变量。 +3. Linux上的编译命令为`make`，Windows上为`mingw32-make`，这里需要区分一下。 + + +安装完成后，可以打开一个命令行终端，通过命令`gcc -v`查看gcc版本。 + +在该文件夹(deploy/vector_search)下，运行命令`mingw32-make`，即可生成`index.dll`库文件。如果希望重新生成`index.dll`文件，可以首先使用`mingw32-make clean`清除已经生成的缓存，再使用`mingw32-make`生成更新之后的库文件。 + +### 2.4 MacOS上编译生成库文件 + +运行下面的命令，安装gcc与g++: + +```shell +brew install gcc +``` +#### 注意： +1. 若提示 `Error: Running Homebrew as root is extremely dangerous and no longer supported...`, 参考该[链接](https://jingyan.baidu.com/article/e52e3615057a2840c60c519c.html)处理 +2. 若提示 `Error: Failure while executing; `tar --extract --no-same-owner --file...`， 参考该[链接](https://blog.csdn.net/Dawn510/article/details/117787358)处理 + +在安装之后编译后的可执行程序会被复制到/usr/local/bin下面，查看这个文件夹下的gcc： +``` +ls /usr/local/bin/gcc* +``` +可以看到本地gcc对应的版本号为gcc-11，编译命令如下: (如果本地gcc版本为gcc-9, 则相应命令修改为`CXX=g++-9 make`) +``` +CXX=g++-11 make +``` + +## 3. 快速使用 + + import numpy as np + from interface import Graph_Index + + # 随机产生样本 + index_vectors = np.random.rand(100000,128).astype(np.float32) + query_vector = np.random.rand(128).astype(np.float32) + index_docs = ["ID_"+str(i) for i in range(100000)] + + # 初始化索引结构 + indexer = Graph_Index(dist_type="IP") #支持"IP"和"L2" + indexer.build(gallery_vectors=index_vectors, gallery_docs=index_docs, pq_size=100, index_path='test') + + # 查询 + scores, docs = indexer.search(query=query_vector, return_k=10, search_budget=100) + print(scores) + print(docs) + + # 保存与加载 + indexer.dump(index_path="test") + indexer.load(index_path="test") diff --git a/exprements/PaddleClas/deploy/vector_search/README_en.md b/exprements/PaddleClas/deploy/vector_search/README_en.md new file mode 100644 index 00000000..aecadfd9 --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/README_en.md @@ -0,0 +1,97 @@ +# Vector search + +**Attention**: Due to the system adaptability problem, this retrieval algorithm will be abandoned in the new version. [faiss](https://github.com/facebookresearch/faiss) will be used in the new version. The use process of the overall retrieval system base will remain unchanged, but the yaml files for build indexes and retrieval will be modified. + +## 1. Introduction + +Some vertical domain recognition tasks (e.g., vehicles, commodities, etc.) require a large number of recognized categories, and often use a retrieval-based approach to obtain matching predicted categories by performing a fast nearest neighbor search with query vectors and underlying library vectors. The vector search module provides the basic approximate nearest neighbor search algorithm based on Baidu's self-developed Möbius algorithm, a graph-based approximate nearest neighbor search algorithm for maximum inner product search (MIPS). This module provides python interface, supports numpy and tensor type vectors, and supports L2 and Inner Product distance calculation. + +Details of the Mobius algorithm can be found in the paper.（[Möbius Transformation for Fast Inner Product Search on Graph](http://research.baidu.com/Public/uploads/5e189d36b5cf6.PDF), [Code](https://github.com/sunbelbd/mobius)） + +## 2. Installation + +### 2.1 Use the provided library files directly + +This folder contains the compiled `index.so` (compiled under gcc8.2.0 for Linux) and `index.dll` (compiled under gcc10.3.0 for Windows), which can be used directly, skipping sections 2.2 and 2.3. + +If the library files are not available due to a low gcc version or an incompatible environment, you need to manually compile the library files under a different platform. + +**Note：** Make sure that C++ compiler supports the C++11 standard. + +### 2.2 Compile and generate library files on Linux + +Run the following command to install gcc and g++. + +``` +sudo apt-get update +sudo apt-get upgrade -y +sudo apt-get install build-essential gcc g++ +``` + +Check the gcc version by the command `gcc -v`. + +`make` can be operated directly. If you wish to regenerate the `index.so`, you can first use `make clean` to clear the cache, and then use `make` to generate the updated library file. + +### 2.3 Compile and generate library files on Windows + +You need to install gcc compiler tool first, we recommend using [TDM-GCC](https://jmeubank.github.io/tdm-gcc/articles/2020-03/9.2.0-release), you can choose the right version on the official website. We recommend downloading [tdm64-gcc-10.3.0-2.exe](https://github.com/jmeubank/tdm-gcc/releases/download/v10.3.0-tdm64-2/tdm64-gcc-10.3.0-2.exe). + +After the downloading, follow the default installation steps to install. There are 3 points to note here: + +1. The vector search module depends on openmp, so you need to check the `openmp` installation option when going on to `choose components` step, otherwise it will report an error `libgomp.spec: No such file or directory`, [reference link](https://github.com/dmlc/xgboost/issues/1027) +2. When being asked whether to add to the system environment variables, it is recommended to check here, otherwise you need to add the system environment variables manually later. +3. The compile command is `make` on Linux and `mingw32-make` on Windows, so you need to distinguish here. + +After installation, you can open a command line terminal and check the gcc version with the command `gcc -v`. + +Run the command `mingw32-make` to generate the `index.dll` library file under the folder (deploy/vector_search). If you want to regenerate the `index.dll` file, you can first use `mingw32-make clean` to clear the cache, and then use `mingw32-make` to generate the updated library file. + +### 2.4 Compile and generate library files on MacOS + +Run the following command to install gcc and g++: + +``` +brew install gcc +``` + +#### Caution： + +1. If prompted with `Error: Running Homebrew as root is extremely dangerous and no longer supported... `, refer to this [link](https://jingyan.baidu.com/article/e52e3615057a2840c60c519c.html) +2. If prompted with `Error: Failure while executing; tar --extract --no-same-owner --file... `, refer to this [link](https://blog.csdn.net/Dawn510/article/details/117787358). + +After installation the compiled executable is copied under /usr/local/bin, look at the gcc in this folder: + +``` +ls /usr/local/bin/gcc* +``` + +The local gcc version is gcc-11, and the compile command is as follows: (If the local gcc version is gcc-9, the corresponding command should be `CXX=g++-9 make`) + +``` +CXX=g++-11 make +``` + +## 3. Quick use + +``` +import numpy as np +from interface import Graph_Index + +# Random sample generation +index_vectors = np.random.rand(100000,128).astype(np.float32) +query_vector = np.random.rand(128).astype(np.float32) +index_docs = ["ID_"+str(i) for i in range(100000)] + +# Initialize index structure +indexer = Graph_Index(dist_type="IP") #support "IP" and "L2" +indexer.build(gallery_vectors=index_vectors, gallery_docs=index_docs, pq_size=100, index_path='test') + +# Query +scores, docs = indexer.search(query=query_vector, return_k=10, search_budget=100) +print(scores) +print(docs) + +# Save and load +indexer.dump(index_path="test") +indexer.load(index_path="test") +``` diff --git a/exprements/PaddleClas/deploy/vector_search/__init__.py b/exprements/PaddleClas/deploy/vector_search/__init__.py new file mode 100644 index 00000000..4ee26dfb --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/__init__.py @@ -0,0 +1 @@ +from .interface import Graph_Index diff --git a/exprements/PaddleClas/deploy/vector_search/index.dll b/exprements/PaddleClas/deploy/vector_search/index.dll new file mode 100644 index 00000000..5d547535 Binary files /dev/null and b/exprements/PaddleClas/deploy/vector_search/index.dll differ diff --git a/exprements/PaddleClas/deploy/vector_search/index.so b/exprements/PaddleClas/deploy/vector_search/index.so new file mode 100644 index 00000000..d58d4f74 Binary files /dev/null and b/exprements/PaddleClas/deploy/vector_search/index.so differ diff --git a/exprements/PaddleClas/deploy/vector_search/interface.cc b/exprements/PaddleClas/deploy/vector_search/interface.cc new file mode 100644 index 00000000..40136686 --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/interface.cc @@ -0,0 +1,266 @@ +//MIT License +// +//Copyright (c) 2021 Mobius Authors + +//Permission is hereby granted, free of charge, to any person obtaining a copy +//of this software and associated documentation files (the "Software"), to deal +//in the Software without restriction, including without limitation the rights +//to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +//copies of the Software, and to permit persons to whom the Software is +//furnished to do so, subject to the following conditions: + +//The above copyright notice and this permission notice shall be included in all +//copies or substantial portions of the Software. + +//THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +//IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +//FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +//AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +//LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +//OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +//SOFTWARE. + +//from https://github.com/sunbelbd/mobius/blob/e2d166547d61d791da8f06747a63b9cd38f02c71/main.cc + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include"src/data.h" +#include"src/graph.h" + +struct IndexContext{ + void* graph; + void* data; +}; + + +int topk = 0; +int display_topk = 1; +int build_idx_offset = 0; +int query_idx_offset = 0; + +void flush_add_buffer( + std::vector>>>& add_buffer, + GraphWrapper* graph){ + #pragma omp parallel for + for(int i = 0;i < add_buffer.size();++i){ + auto& idx = add_buffer[i].first; + auto& point = add_buffer[i].second; + graph->add_vertex_lock(idx,point); + } + add_buffer.clear(); +} + + +extern "C"{ +// for mobius IP index +void build_mobius_index(float* dense_mat,int row,int dim, int pq_size, double mobius_pow , const char* prefix){ + std::unique_ptr data; + std::unique_ptr data_original; + std::unique_ptr graph; + int topk = 0; + int display_topk = 1; + int build_idx_offset = 0; + int query_idx_offset = 0; + + ++row; + data = std::unique_ptr(new Data(row,dim)); + graph = std::unique_ptr(new FixedDegreeGraph<3>(data.get())); + graph->set_construct_pq_size(pq_size); + + std::vector>>> add_buffer; + + ((FixedDegreeGraph<3>*)graph.get())->get_data()->mobius_pow = mobius_pow; + data_original = std::unique_ptr(new Data(row,dim)); + + std::vector> dummy_mobius_point; + for(int i = 0;i < dim;++i) + dummy_mobius_point.push_back(std::make_pair(i,0)); + + //idx += build_idx_offset; + + for(int i = 0;i < row - 1;++i){ + + std::vector> point; + point.reserve(dim); + for(int j = 0;j < dim;++j) + point.push_back(std::make_pair(j,dense_mat[i * dim + j])); + + data_original->add(i,point); + data->add_mobius(i,point); + if(i < 1000){ + graph->add_vertex(i,point); + }else{ + add_buffer.push_back(std::make_pair(i,point)); + } + if(add_buffer.size() >= 1000000) + flush_add_buffer(add_buffer,graph.get()); + } + flush_add_buffer(add_buffer,graph.get()); + graph->add_vertex(row - 1,dummy_mobius_point); + data.swap(data_original); + + std::string str = std::string(prefix); + data->dump(str + ".data"); + graph->dump(str + ".graph"); + +} + +void load_mobius_index_prefix(int row,int dim,IndexContext* index_context,const char* prefix){ + std::string str = std::string(prefix); + + ++row; + Data* data = new Data(row,dim); + GraphWrapper* graph = new FixedDegreeGraph<1>(data); + + //idx += build_idx_offset; + data->load(str + ".data"); + graph->load(str + ".graph"); + + ((FixedDegreeGraph<1>*)graph)->search_start_point = row - 1; + ((FixedDegreeGraph<1>*)graph)->ignore_startpoint = true; + + index_context->graph = graph; + index_context->data = data; +} + +void save_mobius_index_prefix(IndexContext* index_context,const char* prefix){ + std::string str = std::string(prefix); + Data* data = (Data*)(index_context->data); + GraphWrapper* graph = (GraphWrapper*)(index_context->graph); + + data->dump(str + ".data"); + graph->dump(str + ".graph"); +} + +void search_mobius_index(float* dense_vec,int dim,int search_budget,int return_k, IndexContext* index_context,idx_t* ret_id,double* ret_score){ + int topk = 0; + int display_topk = 1; + int build_idx_offset = 0; + int query_idx_offset = 0; + + Data* data = reinterpret_cast(index_context->data); + GraphWrapper* graph = reinterpret_cast(index_context->graph); + + + //auto flag = (data==NULL); + //std::cout<> point; + point.reserve(dim); + for(int j = 0;j < dim;++j) + point.push_back(std::make_pair(j,dense_vec[j])); + std::vector topN; + std::vector score; + graph->search_top_k_with_score(point,search_budget,topN,score); + for(int i = 0;i < topN.size() && i < return_k;++i){ + ret_id[i] = topN[i]; + ret_score[i] = score[i]; + } +} + + +// For L2 index +void build_l2_index(float* dense_mat,int row,int dim, int pq_size, const char* prefix){ + std::unique_ptr data; + std::unique_ptr graph; + int topk = 0; + int display_topk = 1; + int build_idx_offset = 0; + int query_idx_offset = 0; + + data = std::unique_ptr(new Data(row,dim)); + graph = std::unique_ptr(new FixedDegreeGraph<3>(data.get())); + graph->set_construct_pq_size(pq_size); + + std::vector>>> add_buffer; + + for(int i = 0;i < row;++i){ + std::vector> point; + point.reserve(dim); + for(int j = 0;j < dim;++j) + point.push_back(std::make_pair(j,dense_mat[i * dim + j])); + data->add(i,point); + if(i < 1000){ + graph->add_vertex(i,point); + }else{ + add_buffer.push_back(std::make_pair(i,point)); + } + if(add_buffer.size() >= 1000000) + flush_add_buffer(add_buffer,graph.get()); + } + flush_add_buffer(add_buffer,graph.get()); + + std::string str = std::string(prefix); + data->dump(str + ".data"); + graph->dump(str + ".graph"); + +} + +void load_l2_index_prefix(int row,int dim,IndexContext* index_context,const char* prefix){ + std::string str = std::string(prefix); + + Data* data = new Data(row,dim); + GraphWrapper* graph = new FixedDegreeGraph<3>(data); + + //idx += build_idx_offset; + + data->load(str + ".data"); + graph->load(str + ".graph"); + + index_context->graph = graph; + index_context->data = data; +} + +void save_l2_index_prefix(IndexContext* index_context,const char* prefix){ + std::string str = std::string(prefix); + Data* data = (Data*)(index_context->data); + GraphWrapper* graph = (GraphWrapper*)(index_context->graph); + + data->dump(str + ".data"); + graph->dump(str + ".graph"); +} + + + +void search_l2_index(float* dense_vec,int dim,int search_budget,int return_k, IndexContext* index_context,idx_t* ret_id,double* ret_score){ + int topk = 0; + int display_topk = 1; + int build_idx_offset = 0; + int query_idx_offset = 0; + + Data* data = reinterpret_cast(index_context->data); + GraphWrapper* graph = reinterpret_cast(index_context->graph); + + std::vector> point; + point.reserve(dim); + for(int j = 0;j < dim;++j) + point.push_back(std::make_pair(j,dense_vec[j])); + std::vector topN; + std::vector score; + graph->search_top_k_with_score(point,search_budget,topN,score); + for(int i = 0;i < topN.size() && i < return_k;++i){ +// printf("%d: (%zu, %f)\n",i,topN[i],score[i]); + ret_id[i] = topN[i]; + ret_score[i] = score[i]; + } +} + + +void release_context(IndexContext* index_context){ + delete (Data*)(index_context->data); + delete (GraphWrapper*)(index_context->graph); +} + +} // extern "C" + diff --git a/exprements/PaddleClas/deploy/vector_search/interface.py b/exprements/PaddleClas/deploy/vector_search/interface.py new file mode 100644 index 00000000..8dcd86f5 --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/interface.py @@ -0,0 +1,272 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import ctypes +import paddle +import numpy.ctypeslib as ctl +import numpy as np +import os +import sys +import json +import platform + +from ctypes import * +from numpy.ctypeslib import ndpointer + +__dir__ = os.path.dirname(os.path.abspath(__file__)) +winmode = None +if platform.system() == "Windows": + lib_filename = "index.dll" + if sys.version_info.minor >= 8: + winmode = 0x8 +else: + lib_filename = "index.so" +so_path = os.path.join(__dir__, lib_filename) +try: + if winmode is not None: + lib = ctypes.CDLL(so_path, winmode=winmode) + else: + lib = ctypes.CDLL(so_path) +except Exception as ex: + readme_path = os.path.join(__dir__, "README.md") + print( + f"Error happened when load lib {so_path} with msg {ex},\nplease refer to {readme_path} to rebuild your library." + ) + exit(-1) + + +class IndexContext(Structure): + _fields_ = [("graph", c_void_p), ("data", c_void_p)] + + +# for mobius IP index +build_mobius_index = lib.build_mobius_index +build_mobius_index.restype = None +build_mobius_index.argtypes = [ + ctl.ndpointer( + np.float32, flags='aligned, c_contiguous'), ctypes.c_int, ctypes.c_int, + ctypes.c_int, ctypes.c_double, ctypes.c_char_p +] + +search_mobius_index = lib.search_mobius_index +search_mobius_index.restype = None +search_mobius_index.argtypes = [ + ctl.ndpointer( + np.float32, flags='aligned, c_contiguous'), ctypes.c_int, ctypes.c_int, + ctypes.c_int, POINTER(IndexContext), ctl.ndpointer( + np.uint64, flags='aligned, c_contiguous'), ctl.ndpointer( + np.float64, flags='aligned, c_contiguous') +] + +load_mobius_index_prefix = lib.load_mobius_index_prefix +load_mobius_index_prefix.restype = None +load_mobius_index_prefix.argtypes = [ + ctypes.c_int, ctypes.c_int, POINTER(IndexContext), ctypes.c_char_p +] + +save_mobius_index_prefix = lib.save_mobius_index_prefix +save_mobius_index_prefix.restype = None +save_mobius_index_prefix.argtypes = [POINTER(IndexContext), ctypes.c_char_p] + +# for L2 index +build_l2_index = lib.build_l2_index +build_l2_index.restype = None +build_l2_index.argtypes = [ + ctl.ndpointer( + np.float32, flags='aligned, c_contiguous'), ctypes.c_int, ctypes.c_int, + ctypes.c_int, ctypes.c_char_p +] + +search_l2_index = lib.search_l2_index +search_l2_index.restype = None +search_l2_index.argtypes = [ + ctl.ndpointer( + np.float32, flags='aligned, c_contiguous'), ctypes.c_int, ctypes.c_int, + ctypes.c_int, POINTER(IndexContext), ctl.ndpointer( + np.uint64, flags='aligned, c_contiguous'), ctl.ndpointer( + np.float64, flags='aligned, c_contiguous') +] + +load_l2_index_prefix = lib.load_l2_index_prefix +load_l2_index_prefix.restype = None +load_l2_index_prefix.argtypes = [ + ctypes.c_int, ctypes.c_int, POINTER(IndexContext), ctypes.c_char_p +] + +save_l2_index_prefix = lib.save_l2_index_prefix +save_l2_index_prefix.restype = None +save_l2_index_prefix.argtypes = [POINTER(IndexContext), ctypes.c_char_p] + +release_context = lib.release_context +release_context.restype = None +release_context.argtypes = [POINTER(IndexContext)] + + +class Graph_Index(object): + """ + graph index + """ + + def __init__(self, dist_type="IP"): + self.dim = 0 + self.total_num = 0 + self.dist_type = dist_type + self.mobius_pow = 2.0 + self.index_context = IndexContext(0, 0) + self.gallery_doc_dict = {} + self.with_attr = False + assert dist_type in ["IP", "L2"], "Only support IP and L2 distance ..." + + def build(self, + gallery_vectors, + gallery_docs=[], + pq_size=100, + index_path='graph_index/', + append_index=False): + """ + build index + """ + if paddle.is_tensor(gallery_vectors): + gallery_vectors = gallery_vectors.numpy() + assert gallery_vectors.ndim == 2, "Input vector must be 2D ..." + + self.total_num = gallery_vectors.shape[0] + self.dim = gallery_vectors.shape[1] + + assert (len(gallery_docs) == self.total_num + if len(gallery_docs) > 0 else True) + + print("training index -> num: {}, dim: {}, dist_type: {}".format( + self.total_num, self.dim, self.dist_type)) + + if not os.path.exists(index_path): + os.makedirs(index_path) + + if self.dist_type == "IP": + build_mobius_index( + gallery_vectors, self.total_num, self.dim, pq_size, + self.mobius_pow, + create_string_buffer((index_path + "/index").encode('utf-8'))) + load_mobius_index_prefix( + self.total_num, self.dim, + ctypes.byref(self.index_context), + create_string_buffer((index_path + "/index").encode('utf-8'))) + else: + build_l2_index( + gallery_vectors, self.total_num, self.dim, pq_size, + create_string_buffer((index_path + "/index").encode('utf-8'))) + load_l2_index_prefix( + self.total_num, self.dim, + ctypes.byref(self.index_context), + create_string_buffer((index_path + "/index").encode('utf-8'))) + + self.gallery_doc_dict = {} + if len(gallery_docs) > 0: + self.with_attr = True + for i in range(gallery_vectors.shape[0]): + self.gallery_doc_dict[str(i)] = gallery_docs[i] + + self.gallery_doc_dict["total_num"] = self.total_num + self.gallery_doc_dict["dim"] = self.dim + self.gallery_doc_dict["dist_type"] = self.dist_type + self.gallery_doc_dict["with_attr"] = self.with_attr + + output_path = os.path.join(index_path, "info.json") + if append_index is True and os.path.exists(output_path): + with open(output_path, "r") as fin: + lines = fin.readlines()[0] + ori_gallery_doc_dict = json.loads(lines) + assert ori_gallery_doc_dict["dist_type"] == self.gallery_doc_dict[ + "dist_type"] + assert ori_gallery_doc_dict["dim"] == self.gallery_doc_dict["dim"] + assert ori_gallery_doc_dict["with_attr"] == self.gallery_doc_dict[ + "with_attr"] + offset = ori_gallery_doc_dict["total_num"] + for i in range(0, self.gallery_doc_dict["total_num"]): + ori_gallery_doc_dict[str(i + offset)] = self.gallery_doc_dict[ + str(i)] + + ori_gallery_doc_dict["total_num"] += self.gallery_doc_dict[ + "total_num"] + self.gallery_doc_dict = ori_gallery_doc_dict + with open(output_path, "w") as f: + json.dump(self.gallery_doc_dict, f) + + print("finished creating index ...") + + def search(self, query, return_k=10, search_budget=100): + """ + search + """ + ret_id = np.zeros(return_k, dtype=np.uint64) + ret_score = np.zeros(return_k, dtype=np.float64) + + if paddle.is_tensor(query): + query = query.numpy() + if self.dist_type == "IP": + search_mobius_index(query, self.dim, search_budget, return_k, + ctypes.byref(self.index_context), ret_id, + ret_score) + else: + search_l2_index(query, self.dim, search_budget, return_k, + ctypes.byref(self.index_context), ret_id, + ret_score) + + ret_id = ret_id.tolist() + ret_doc = [] + if self.with_attr: + for i in range(return_k): + ret_doc.append(self.gallery_doc_dict[str(ret_id[i])]) + return ret_score, ret_doc + else: + return ret_score, ret_id + + def dump(self, index_path): + + if not os.path.exists(index_path): + os.makedirs(index_path) + + if self.dist_type == "IP": + save_mobius_index_prefix( + ctypes.byref(self.index_context), + create_string_buffer((index_path + "/index").encode('utf-8'))) + else: + save_l2_index_prefix( + ctypes.byref(self.index_context), + create_string_buffer((index_path + "/index").encode('utf-8'))) + + with open(index_path + "/info.json", "w") as f: + json.dump(self.gallery_doc_dict, f) + + def load(self, index_path): + self.gallery_doc_dict = {} + + with open(index_path + "/info.json", "r") as f: + self.gallery_doc_dict = json.load(f) + + self.total_num = self.gallery_doc_dict["total_num"] + self.dim = self.gallery_doc_dict["dim"] + self.dist_type = self.gallery_doc_dict["dist_type"] + self.with_attr = self.gallery_doc_dict["with_attr"] + + if self.dist_type == "IP": + load_mobius_index_prefix( + self.total_num, self.dim, + ctypes.byref(self.index_context), + create_string_buffer((index_path + "/index").encode('utf-8'))) + else: + load_l2_index_prefix( + self.total_num, self.dim, + ctypes.byref(self.index_context), + create_string_buffer((index_path + "/index").encode('utf-8'))) diff --git a/exprements/PaddleClas/deploy/vector_search/src/config.h b/exprements/PaddleClas/deploy/vector_search/src/config.h new file mode 100644 index 00000000..352d6843 --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/src/config.h @@ -0,0 +1,43 @@ +//MIT License +// +//Copyright (c) 2021 Mobius Authors + +//Permission is hereby granted, free of charge, to any person obtaining a copy +//of this software and associated documentation files (the "Software"), to deal +//in the Software without restriction, including without limitation the rights +//to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +//copies of the Software, and to permit persons to whom the Software is +//furnished to do so, subject to the following conditions: + +//The above copyright notice and this permission notice shall be included in all +//copies or substantial portions of the Software. + +//THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +//IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +//FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +//AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +//LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +//OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +//SOFTWARE. + + +//from https://github.com/sunbelbd/mobius/blob/e2d166547d61d791da8f06747a63b9cd38f02c71/config.h + +#pragma once + + +typedef float value_t; +//typedef double dist_t; +typedef float dist_t; +typedef size_t idx_t; +typedef int UINT; + + +#define ACC_BATCH_SIZE 4096 +#define FIXED_DEGREE 31 +#define FIXED_DEGREE_SHIFT 5 + + +//for construction +#define SEARCH_DEGREE 15 +#define CONSTRUCT_SEARCH_BUDGET 150 diff --git a/exprements/PaddleClas/deploy/vector_search/src/data.h b/exprements/PaddleClas/deploy/vector_search/src/data.h new file mode 100644 index 00000000..ea65e70a --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/src/data.h @@ -0,0 +1,365 @@ +//MIT License +// +//Copyright (c) 2021 Mobius Authors + +//Permission is hereby granted, free of charge, to any person obtaining a copy +//of this software and associated documentation files (the "Software"), to deal +//in the Software without restriction, including without limitation the rights +//to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +//copies of the Software, and to permit persons to whom the Software is +//furnished to do so, subject to the following conditions: + +//The above copyright notice and this permission notice shall be included in all +//copies or substantial portions of the Software. + +//THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +//IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +//FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +//AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +//LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +//OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +//SOFTWARE. + +//from https://github.com/sunbelbd/mobius/blob/e2d166547d61d791da8f06747a63b9cd38f02c71/data.h + +#pragma once + +#include +#include +#include + +#include"config.h" + +#define ZERO_EPS 1e-10 + +#define _SCALE_WORLD_DENSE_DATA + +#ifdef _SCALE_WORLD_DENSE_DATA +//dense data +class Data{ +private: + std::unique_ptr data; + size_t num; + size_t curr_num = 0; + int dim; + +public: + value_t mobius_pow = 2; + value_t max_ip_norm = 1; + value_t max_ip_norm2 = 1; + + Data(size_t num, int dim) : num(num),dim(dim){ + data = std::unique_ptr(new value_t[num * dim]); + memset(data.get(),0,sizeof(value_t) * num * dim); + } + + value_t* get(idx_t idx) const{ + return data.get() + idx * dim; + } + + template + dist_t ipwrap_l2_query_distance(idx_t a,T& v) const{ + auto pa = get(a); + dist_t ret = 0; + dist_t normu = 0; + for(int i = 0;i < dim;++i){ + auto diff = (*(pa + i) / max_ip_norm) - v[i]; + ret += diff * diff; + normu += (*(pa + i)) * (*(pa + i)); + } + ret += 1 - normu / max_ip_norm2; + return ret; + } + + template + dist_t ipwrap_l2_build_distance(idx_t a,T& v) const{ + auto pa = get(a); + dist_t ret = 0; + dist_t normu = 0; + dist_t normv = 0; + for(int i = 0;i < dim;++i){ + auto diff = *(pa + i) - v[i]; + ret += diff * diff; + normu += (*(pa + i)) * (*(pa + i)); + normv += v[i] * v[i]; + } + dist_t wrap_termu = sqrt(1 - normu / max_ip_norm2); + dist_t wrap_termv = sqrt(1 - normv / max_ip_norm2); + dist_t diff_wrap = wrap_termu - wrap_termv; + ret = ret / max_ip_norm2 + diff_wrap * diff_wrap; + return ret; + } + + template + dist_t l2_distance(idx_t a,T& v) const{ + auto pa = get(a); + dist_t ret = 0; + for(int i = 0;i < dim;++i){ + auto diff = *(pa + i) - v[i]; + ret += diff * diff; + } + return ret; + } + + template + dist_t negative_inner_prod_distance(idx_t a,T& v) const{ + auto pa = get(a); + dist_t ret = 0; + for(int i = 0;i < dim;++i){ + ret -= (*(pa + i)) * v[i]; + } + return ret; + } + + template + dist_t negative_cosine_distance(idx_t a,T& v) const{ + auto pa = get(a); + dist_t ret = 0; + value_t lena = 0,lenv = 0; + for(int i = 0;i < dim;++i){ + ret += (*(pa + i)) * v[i]; + lena += (*(pa + i)) * (*(pa + i)); + lenv += v[i] * v[i]; + } + int sign = ret < 0 ? 1 : -1; +// return sign * (ret * ret / lena);// / lenv); + return sign * (ret * ret / lena / lenv); + } + + template + dist_t mobius_l2_distance(idx_t a,T& v) const{ + auto pa = get(a); + dist_t ret = 0; + value_t lena = 0,lenv = 0; + for(int i = 0;i < dim;++i){ + lena += (*(pa + i)) * (*(pa + i)); + lenv += v[i] * v[i]; + } + value_t modifier_a = pow(lena,0.5 * mobius_pow); + value_t modifier_v = pow(lenv,0.5 * mobius_pow); + if(fabs(modifier_a) < ZERO_EPS) + modifier_a = 1; + if(fabs(modifier_v) < ZERO_EPS) + modifier_v = 1; + for(int i = 0;i < dim;++i){ + value_t tmp = (*(pa + i)) / modifier_a - v[i] / modifier_v; + ret += tmp * tmp; + } + return ret; + } + + template + dist_t real_nn(T& v) const{ + dist_t minn = 1e100; + for(size_t i = 0;i < curr_num;++i){ + auto res = l2_distance(i,v); + if(res < minn){ + minn = res; + } + } + return minn; + } + + std::vector organize_point_mobius(const std::vector>& v){ + std::vector ret(dim,0); + value_t lena = 0; + for(const auto& p : v){ +// ret[p.first] = p.second; + lena += p.second * p.second; + } + value_t modifier_a = pow(lena,0.5 * mobius_pow); + if(fabs(modifier_a) < ZERO_EPS) + modifier_a = 1; + for(const auto& p : v){ + ret[p.first] = p.second / modifier_a; + } + return std::move(ret); + } + + std::vector organize_point(const std::vector>& v){ + std::vector ret(dim,0); + for(const auto& p : v){ + if(p.first >= dim) + printf("error %d %d\n",p.first,dim); + ret[p.first] = p.second; + } + return std::move(ret); + } + + value_t vec_sum2(const std::vector>& v){ + value_t ret = 0; + for(const auto& p : v){ + if(p.first >= dim) + printf("error %d %d\n",p.first,dim); + ret += p.second * p.second; + } + return std::move(ret); + } + + + void add(idx_t idx, std::vector>& value){ + //printf("adding %zu\n",idx); + //for(auto p : value) + // printf("%zu %d %f\n",idx,p.first,p.second); + curr_num = std::max(curr_num,idx); + auto p = get(idx); + for(const auto& v : value) + *(p + v.first) = v.second; + } + + void add_mobius(idx_t idx, std::vector>& value){ + //printf("adding %zu\n",idx); + //for(auto p : value) + // printf("%zu %d %f\n",idx,p.first,p.second); + curr_num = std::max(curr_num,idx); + auto p = get(idx); + value_t lena = 0; + for(const auto& v : value){ + *(p + v.first) = v.second; + lena += v.second * v.second; + } + value_t modifier_a = pow(lena,0.5 * mobius_pow); + if(fabs(modifier_a) < ZERO_EPS) + modifier_a = 1; + for(const auto& v : value){ + *(p + v.first) = v.second / modifier_a; + } + } + + inline size_t max_vertices(){ + return num; + } + + inline size_t curr_vertices(){ + return curr_num; + } + + void print(){ + for(int i = 0;i < num && i < 10;++i) + printf("%f ",*(data.get() + i)); + printf("\n"); + } + + int get_dim(){ + return dim; + } + + void dump(std::string path = "bfsg.data"){ + FILE* fp = fopen(path.c_str(),"wb"); + fwrite(data.get(),sizeof(value_t) * num * dim,1,fp); + fclose(fp); + } + + void load(std::string path = "bfsg.data"){ + curr_num = num; + FILE* fp = fopen(path.c_str(),"rb"); + auto cnt = fread(data.get(),sizeof(value_t) * num * dim,1,fp); + fclose(fp); + } + +}; +template<> +dist_t Data::ipwrap_l2_build_distance(idx_t a,idx_t& b) const{ + auto pa = get(a); + auto pb = get(b); + dist_t ret = 0; + dist_t normu = 0; + dist_t normv = 0; + for(int i = 0;i < dim;++i){ + auto diff = *(pa + i) - *(pb + i); + ret += diff * diff; + normu += (*(pa + i)) * (*(pa + i)); + normv += (*(pb + i)) * (*(pb + i)); + } + dist_t wrap_termu = sqrt(1 - normu / max_ip_norm2); + dist_t wrap_termv = sqrt(1 - normv / max_ip_norm2); + dist_t diff_wrap = wrap_termu - wrap_termv; + ret = ret / max_ip_norm2 + diff_wrap * diff_wrap; + return ret; +} +template<> +dist_t Data::ipwrap_l2_query_distance(idx_t a,idx_t& b) const{ + auto pa = get(a); + auto pb = get(b); + dist_t ret = 0; + dist_t normu = 0; + for(int i = 0;i < dim;++i){ + auto diff = (*(pa + i) / max_ip_norm) - *(pb + i); + ret += diff * diff; + normu += (*(pa + i)) * (*(pa + i)); + } + ret += 1 - normu / max_ip_norm2; + return ret; +} +template<> +dist_t Data::l2_distance(idx_t a,idx_t& b) const{ + auto pa = get(a), + pb = get(b); + dist_t ret = 0; + for(int i = 0;i < dim;++i){ + auto diff = *(pa + i) - *(pb + i); + ret += diff * diff; + } + return ret; +} + +template<> +dist_t Data::negative_inner_prod_distance(idx_t a,idx_t& b) const{ + auto pa = get(a), + pb = get(b); + dist_t ret = 0; + for(int i = 0;i < dim;++i){ + ret -= (*(pa + i)) * (*(pb + i)); + } + return ret; +} + +template<> +dist_t Data::negative_cosine_distance(idx_t a,idx_t& b) const{ + auto pa = get(a), + pb = get(b); + dist_t ret = 0; + value_t lena = 0,lenv = 0; + for(int i = 0;i < dim;++i){ + ret += (*(pa + i)) * (*(pb + i)); + lena += (*(pa + i)) * (*(pa + i)); + lenv += (*(pb + i)) * (*(pb + i)); + } + int sign = ret < 0 ? 1 : -1; +// return sign * (ret * ret / lena); + return sign * (ret * ret / lena / lenv); +} + +template<> +dist_t Data::mobius_l2_distance(idx_t a,idx_t& b) const{ + auto pa = get(a), + pb = get(b); + dist_t ret = 0; + value_t lena = 0,lenv = 0; + for(int i = 0;i < dim;++i){ + lena += (*(pa + i)) * (*(pa + i)); + lenv += (*(pb + i)) * (*(pb + i)); + } + value_t modifier_a = pow(lena,0.5 * mobius_pow); + value_t modifier_v = pow(lenv,0.5 * mobius_pow); + if(fabs(modifier_a) < ZERO_EPS) + modifier_a = 1; + if(fabs(modifier_v) < ZERO_EPS) + modifier_v = 1; + for(int i = 0;i < dim;++i){ + value_t tmp = (*(pa + i)) / modifier_a - (*(pb + i)) / modifier_v; + ret += tmp * tmp; + } + return ret; +} + +#else +//sparse data +class Data{ +public: + //TODO + +}; +#endif + + diff --git a/exprements/PaddleClas/deploy/vector_search/src/graph.h b/exprements/PaddleClas/deploy/vector_search/src/graph.h new file mode 100644 index 00000000..d54353ba --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/src/graph.h @@ -0,0 +1,636 @@ +//MIT License +// +//Copyright (c) 2021 Mobius Authors + +//Permission is hereby granted, free of charge, to any person obtaining a copy +//of this software and associated documentation files (the "Software"), to deal +//in the Software without restriction, including without limitation the rights +//to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +//copies of the Software, and to permit persons to whom the Software is +//furnished to do so, subject to the following conditions: + +//The above copyright notice and this permission notice shall be included in all +//copies or substantial portions of the Software. + +//THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +//IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +//FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +//AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +//LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +//OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +//SOFTWARE. + +//from https://github.com/sunbelbd/mobius/blob/e2d166547d61d791da8f06747a63b9cd38f02c71/graph.h + + +#pragma once +#include +#include +#include +#include +#include +#include +#include +#include + +#include"config.h" +#include"data.h" + +#ifdef OMP +#include +#endif + +typedef unsigned int vl_type; + +class VisitedList { +public: + vl_type curV; + vl_type *mass; + unsigned int numelements; + + VisitedList(int numelements1) { + curV = 1; + numelements = numelements1; + mass = new vl_type[numelements]; + memset(mass, 0, sizeof(vl_type) * numelements); + } + + void reset() { + ++curV; + if (curV == 0) { + curV = 1; + memset(mass, 0, sizeof(vl_type) * numelements); + } + }; + + ~VisitedList() { delete mass; } +}; + +struct GraphMeasures{ + int distance_cnt = 0; +}; + +class GraphWrapper{ +public: + virtual void add_vertex(idx_t vertex_id,std::vector>& point) = 0; + virtual void add_vertex_lock(idx_t vertex_id,std::vector>& point) = 0; + virtual void search_top_k(const std::vector>& query,int k,std::vector& result) = 0; + virtual void search_top_k_with_score(const std::vector>& query,int k,std::vector& result,std::vector& score){} + + virtual void dump(std::string path = "bfsg.graph") = 0; + virtual void load(std::string path = "bfsg.graph") = 0; + virtual ~GraphWrapper(){} + virtual void set_construct_pq_size(int size){}; + GraphMeasures measures; +}; + +template +class FixedDegreeGraph : public GraphWrapper{ +private: + const int degree = SEARCH_DEGREE; + const int flexible_degree = FIXED_DEGREE; + const int vertex_offset_shift = FIXED_DEGREE_SHIFT; + std::vector edges; + std::vector edge_dist; + Data* data; + std::mt19937_64 rand_gen = std::mt19937_64(1234567);//std::random_device{}()); + std::vector edge_mutex;//do not push back on this vector, it will destroy the mutex + + bool debug = false; + VisitedList* p_visited = NULL; + #ifdef OMP + std::vector visited_pool; + #endif + int construct_pq_size = CONSTRUCT_SEARCH_BUDGET; + + + void rank_and_switch_ordered(idx_t v_id,idx_t u_id){ + //We assume the neighbors of v_ids in edges[offset] are sorted + //by the distance to v_id ascendingly when it is full + //NOTICE: before it is full, it is unsorted + auto curr_dist = pair_distance(v_id,u_id); + auto offset = ((size_t)v_id) << vertex_offset_shift; + int degree = edges[offset]; + std::vector neighbor; + neighbor.reserve(degree + 1); + for(int i = 0;i < degree;++i) + neighbor.push_back(edges[offset + i + 1]); + neighbor.push_back(u_id); + neighbor = edge_selection_filter_neighbor(neighbor,v_id,flexible_degree); + edges[offset] = neighbor.size(); + for(int i = 0;i < neighbor.size();++i) + edges[offset + i + 1] = neighbor[i]; + return; + //We assert edges[offset] > 0 here + if(curr_dist >= edge_dist[offset + edges[offset]]){ + return; + } + edges[offset + edges[offset]] = u_id; + edge_dist[offset + edges[offset]] = curr_dist; + for(size_t i = offset + edges[offset] - 1;i > offset;--i){ + if(edge_dist[i] > edge_dist[i + 1]){ + std::swap(edges[i],edges[i + 1]); + std::swap(edge_dist[i],edge_dist[i + 1]); + }else{ + break; + } + } + } + + void rank_and_switch(idx_t v_id,idx_t u_id){ + rank_and_switch_ordered(v_id,u_id); + //TODO: + //Implement an unordered version to compare with + } + + template + dist_t distance(idx_t a,T& b){ + if(dist_type == 0) + return data->l2_distance(a,b); + else if(dist_type == 1) + return data->negative_inner_prod_distance(a,b); + else if(dist_type == 2) + return data->negative_cosine_distance(a,b); + else if(dist_type == 3) + return data->l2_distance(a,b); + else if(dist_type == 4) + return data->ipwrap_l2_build_distance(a,b); + else if(dist_type == 5) + return data->ipwrap_l2_query_distance(a,b); + else{ + // should not happen + fprintf(stderr,"unsupported dist_type %d\n",dist_type); + return 0; + } + } + + void compute_distance_naive(size_t offset,std::vector& dists){ + dists.resize(edges[offset]); + auto degree = edges[offset]; + for(int i = 0;i < degree;++i){ + dists[i] = distance(offset >> vertex_offset_shift,edges[offset + i + 1]); + } + } + + void compute_distance(size_t offset,std::vector& dists){ + compute_distance_naive(offset,dists); + } + + template + dist_t pair_distance_naive(idx_t a,T& b){ + ++measures.distance_cnt; + return distance(a,b); + } + + template + dist_t pair_distance(idx_t a,T& b){ + return pair_distance_naive(a,b); + } + + + void qsort(size_t l,size_t r){ + auto mid = (l + r) >> 1; + int i = l,j = r; + auto k = edge_dist[mid]; + do{ + while(edge_dist[i] < k) ++i; + while(k < edge_dist[j]) --j; + if(i <= j){ + std::swap(edge_dist[i],edge_dist[j]); + std::swap(edges[i],edges[j]); + ++i; + --j; + } + }while(i <= j); + if(i < r)qsort(i,r); + if(l < j)qsort(l,j); + } + + void rank_edges(size_t offset){ + std::vector dists; + compute_distance(offset,dists); + for(int i = 0;i < dists.size();++i) + edge_dist[offset + i + 1] = dists[i]; + qsort(offset + 1,offset + dists.size()); + //TODO: + //use a heap in the edge_dist + } + + void add_edge_lock(idx_t v_id,idx_t u_id){ + edge_mutex[v_id].lock(); + auto offset = ((size_t)v_id) << vertex_offset_shift; + if(edges[offset] < flexible_degree){ + ++edges[offset]; + edges[offset + edges[offset]] = u_id; + }else{ + rank_and_switch(v_id,u_id); + } + edge_mutex[v_id].unlock(); + } + + void add_edge(idx_t v_id,idx_t u_id){ + auto offset = ((size_t)v_id) << vertex_offset_shift; + if(edges[offset] < flexible_degree){ + ++edges[offset]; + edges[offset + edges[offset]] = u_id; + }else{ + rank_and_switch(v_id,u_id); + } + } + +public: + long long total_explore_cnt = 0; + int total_explore_times = 0; + + size_t search_start_point = 0; + bool ignore_startpoint = false; + + FixedDegreeGraph(Data* data) : data(data){ + auto num_vertices = data->max_vertices(); + edges = std::vector(((size_t)num_vertices) << vertex_offset_shift); + edge_dist = std::vector(((size_t)num_vertices) << vertex_offset_shift); + edge_mutex = std::vector(num_vertices); + p_visited = new VisitedList(num_vertices + 5); + #ifdef OMP + int n_threads = 1; + #pragma omp parallel + #pragma omp master + { + n_threads = omp_get_num_threads(); + } + visited_pool.resize(n_threads); + for(int i = 0;i < n_threads;++i) + visited_pool[i] = new VisitedList(num_vertices + 5); + #endif + } + + void set_construct_pq_size(int size){ + construct_pq_size = size; + } + + std::vector edge_selection_filter_neighbor(std::vector& neighbor,idx_t vertex_id,int desired_size){ + std::vector filtered_neighbor; + std::vector dists(neighbor.size()); + for(int i = 0;i < dists.size();++i) + dists[i] = pair_distance(vertex_id,neighbor[i]); + std::vector idx(neighbor.size()); + for(int i = 0;i < idx.size();++i) + idx[i] = i; + std::sort(idx.begin(),idx.end(),[&](int a,int b){return dists[a] < dists[b];}); + for(int i = 0;i < idx.size();++i){ + dist_t cur_dist = dists[idx[i]]; + bool pass = true; + for(auto neighbor_id : filtered_neighbor){ + if(cur_dist > pair_distance(neighbor_id,neighbor[idx[i]])){ + pass = false; + break; + } + } + if(pass){ + filtered_neighbor.push_back(neighbor[idx[i]]); + if(filtered_neighbor.size() >= desired_size) + break; + }else{ + } + } + return std::move(filtered_neighbor); + } + + void add_vertex_lock(idx_t vertex_id,std::vector>& point){ + std::vector neighbor; + search_top_k_lock(point,construct_pq_size,neighbor); + auto offset = ((size_t)vertex_id) << vertex_offset_shift; + int num_neighbors = degree < neighbor.size() ? degree : neighbor.size(); + edge_mutex[vertex_id].lock(); + // TODO: + // it is possible to save this space --- edges[offset] + // by set the last number in the range as + // a large number - current degree + if(neighbor.size() >= degree) + neighbor = edge_selection_filter_neighbor(neighbor,vertex_id,degree); + edges[offset] = neighbor.size(); + + for(int i = 0;i < neighbor.size() && i < degree;++i){ + edges[offset + i + 1] = neighbor[i]; + } + edge_mutex[vertex_id].unlock(); + for(int i = 0;i < neighbor.size() && i < degree;++i){ + add_edge_lock(neighbor[i],vertex_id); + } + } + void add_vertex(idx_t vertex_id,std::vector>& point){ + std::vector neighbor; + search_top_k(point,construct_pq_size,neighbor); + auto offset = ((size_t)vertex_id) << vertex_offset_shift; + int num_neighbors = degree < neighbor.size() ? degree : neighbor.size(); + // TODO: + // it is possible to save this space --- edges[offset] + // by set the last number in the range as + // a large number - current degree + if(neighbor.size() >= degree){ + neighbor = edge_selection_filter_neighbor(neighbor,vertex_id,degree); + } + edges[offset] = neighbor.size(); + + for(int i = 0;i < neighbor.size() && i < degree;++i){ + edges[offset + i + 1] = neighbor[i]; + } + for(int i = 0;i < neighbor.size() && i < degree;++i){ + add_edge(neighbor[i],vertex_id); + } + } + + void astar_multi_start_search_lock(const std::vector>& query,int k,std::vector& result){ + std::priority_queue,std::vector>,std::greater>> q; + const int num_start_point = 1; + + auto converted_query = dist_type == 3 ? data->organize_point_mobius(query) : data->organize_point(query); + #ifdef OMP + int tid = omp_get_thread_num(); + auto& p_visited = visited_pool[tid]; + #endif + + p_visited->reset(); + auto tag = p_visited->curV; + for(int i = 0;i < num_start_point && i < data->curr_vertices();++i){ + auto start = search_start_point;//rand_gen() % data->curr_vertices(); + if(p_visited->mass[start] == tag) + continue; + p_visited->mass[start] = tag; + q.push(std::make_pair(pair_distance_naive(start,converted_query),start)); + } + std::priority_queue> topk; + const int max_step = 1000000; + bool found_min_node = false; + dist_t min_dist = 1e100; + int explore_cnt = 0; + for(int iter = 0;iter < max_step && !q.empty();++iter){ + auto now = q.top(); + if(topk.size() == k && topk.top().first < now.first){ + break; + } + ++explore_cnt; + min_dist = std::min(min_dist,now.first); + q.pop(); + if(ignore_startpoint == false || iter != 0) + topk.push(now); + if(topk.size() > k) + topk.pop(); + edge_mutex[now.second].lock(); + auto offset = ((size_t)now.second) << vertex_offset_shift; + auto degree = edges[offset]; + + for(int i = 0;i < degree;++i){ + auto start = edges[offset + i + 1]; + if(p_visited->mass[start] == tag) + continue; + p_visited->mass[start] = tag; + auto dist = pair_distance_naive(start,converted_query); + if(topk.empty() || dist < topk.top().first || topk.size() < k) + q.push(std::make_pair(dist,start)); + } + edge_mutex[now.second].unlock(); + } + total_explore_cnt += explore_cnt; + ++total_explore_times; + result.resize(topk.size()); + int i = result.size() - 1; + while(!topk.empty()){ + result[i] = (topk.top().second); + topk.pop(); + --i; + } + } + + void astar_no_heap_search(const std::vector>& query,std::vector& result){ + const int num_start_point = 1; + std::pair q_top = std::make_pair(10000000000,0); + auto converted_query = dist_type == 3 ? data->organize_point_mobius(query) : data->organize_point(query); + p_visited->reset(); + auto tag = p_visited->curV; + for(int i = 0;i < num_start_point && i < data->curr_vertices();++i){ + auto start = search_start_point;//rand_gen() % data->curr_vertices(); + p_visited->mass[start] = tag; + if(ignore_startpoint == false){ + q_top = (std::make_pair(pair_distance_naive(start,converted_query),start)); + }else{ + auto offset = ((size_t)start) << vertex_offset_shift; + auto degree = edges[offset]; + + for(int i = 1;i <= degree;++i){ + p_visited->mass[edges[offset + i]] = tag; + auto dis = pair_distance_naive(edges[offset + i],converted_query); + if(dis < q_top.first) + q_top = (std::make_pair(dis,start)); + } + } + } + const int max_step = 1000000; + bool found_min_node = false; + dist_t min_dist = 1e100; + int explore_cnt = 0; + for(int iter = 0;iter < max_step;++iter){ + ++explore_cnt; + auto offset = ((size_t)q_top.second) << vertex_offset_shift; + auto degree = edges[offset]; + + bool changed = false; + for(int i = 0;i < degree;++i){ + auto start = edges[offset + i + 1]; + if(p_visited->mass[start] == tag) + continue; + p_visited->mass[start] = tag; + auto dist = pair_distance_naive(start,converted_query); + if(dist < q_top.first){ + q_top = (std::make_pair(dist,start)); + changed = true; + } + } + if(changed == false) + break; + } + total_explore_cnt += explore_cnt; + ++total_explore_times; + result.resize(1); + result[0] = q_top.second; + } + + void astar_multi_start_search_with_score(const std::vector>& query,int k,std::vector& result,std::vector& score){ + std::priority_queue,std::vector>,std::greater>> q; + const int num_start_point = 1; + + auto converted_query = dist_type == 3 ? data->organize_point_mobius(query) : data->organize_point(query); + p_visited->reset(); + auto tag = p_visited->curV; + for(int i = 0;i < num_start_point && i < data->curr_vertices();++i){ + auto start = search_start_point;//rand_gen() % data->curr_vertices(); + if(p_visited->mass[start] == tag) + continue; + p_visited->mass[start] = tag; + q.push(std::make_pair(pair_distance_naive(start,converted_query),start)); + } + std::priority_queue> topk; + const int max_step = 1000000; + bool found_min_node = false; + dist_t min_dist = 1e100; + int explore_cnt = 0; + for(int iter = 0;iter < max_step && !q.empty();++iter){ + auto now = q.top(); + if(topk.size() == k && topk.top().first < now.first){ + break; + } + ++explore_cnt; + min_dist = std::min(min_dist,now.first); + q.pop(); + if(ignore_startpoint == false || iter != 0) + topk.push(now); + if(topk.size() > k) + topk.pop(); + auto offset = ((size_t)now.second) << vertex_offset_shift; + auto degree = edges[offset]; + + for(int i = 0;i < degree;++i){ + auto start = edges[offset + i + 1]; + if(p_visited->mass[start] == tag) + continue; + p_visited->mass[start] = tag; + auto dist = pair_distance_naive(start,converted_query); + if(topk.empty() || dist < topk.top().first || topk.size() < k) + q.push(std::make_pair(dist,start)); + } + } + total_explore_cnt += explore_cnt; + ++total_explore_times; + result.resize(topk.size()); + score.resize(topk.size()); + int i = result.size() - 1; + while(!topk.empty()){ + result[i] = (topk.top().second); + score[i] = -(topk.top().first); + topk.pop(); + --i; + } + } + + void astar_multi_start_search(const std::vector>& query,int k,std::vector& result){ + std::priority_queue,std::vector>,std::greater>> q; + const int num_start_point = 1; + + auto converted_query = dist_type == 3 ? data->organize_point_mobius(query) : data->organize_point(query); + p_visited->reset(); + auto tag = p_visited->curV; + for(int i = 0;i < num_start_point && i < data->curr_vertices();++i){ + auto start = search_start_point;//rand_gen() % data->curr_vertices(); + if(p_visited->mass[start] == tag) + continue; + p_visited->mass[start] = tag; + q.push(std::make_pair(pair_distance_naive(start,converted_query),start)); + } + std::priority_queue> topk; + const int max_step = 1000000; + bool found_min_node = false; + dist_t min_dist = 1e100; + int explore_cnt = 0; + for(int iter = 0;iter < max_step && !q.empty();++iter){ + auto now = q.top(); + if(topk.size() == k && topk.top().first < now.first){ + break; + } + ++explore_cnt; + min_dist = std::min(min_dist,now.first); + q.pop(); + if(ignore_startpoint == false || iter != 0) + topk.push(now); + if(topk.size() > k) + topk.pop(); + auto offset = ((size_t)now.second) << vertex_offset_shift; + auto degree = edges[offset]; + + for(int i = 0;i < degree;++i){ + auto start = edges[offset + i + 1]; + if(p_visited->mass[start] == tag) + continue; + p_visited->mass[start] = tag; + auto dist = pair_distance_naive(start,converted_query); + if(topk.empty() || dist < topk.top().first || topk.size() < k) + q.push(std::make_pair(dist,start)); + } + } + total_explore_cnt += explore_cnt; + ++total_explore_times; + result.resize(topk.size()); + int i = result.size() - 1; + while(!topk.empty()){ + result[i] = (topk.top().second); + topk.pop(); + --i; + } + } + + void search_top_k(const std::vector>& query,int k,std::vector& result){ + if(k == 1) + astar_no_heap_search(query,result); + else + astar_multi_start_search(query,k,result); + } + + void search_top_k_with_score(const std::vector>& query,int k,std::vector& result,std::vector& score){ + astar_multi_start_search_with_score(query,k,result,score); + } + + void search_top_k_lock(const std::vector>& query,int k,std::vector& result){ + astar_multi_start_search_lock(query,k,result); + } + + void print_stat(){ + auto n = data->max_vertices(); + size_t sum = 0; + std::vector histogram(2 * degree + 1,0); + for(size_t i = 0;i < n;++i){ + sum += edges[i << vertex_offset_shift]; + int tmp = edges[i << vertex_offset_shift]; + if(tmp > 2 * degree + 1) + fprintf(stderr,"[ERROR] node %zu has %d degree\n",i,tmp); + ++histogram[edges[i << vertex_offset_shift]]; + if(tmp != degree) + fprintf(stderr,"[INFO] %zu has degree %d\n",i,tmp); + } + fprintf(stderr,"[INFO] #vertices %zu, avg degree %f\n",n,sum * 1.0 / n); + std::unordered_set visited; + fprintf(stderr,"[INFO] degree histogram:\n"); + for(int i = 0;i <= 2 * degree + 1;++i) + fprintf(stderr,"[INFO] %d:\t%zu\n",i,histogram[i]); + + } + + void print_edges(int x){ + for(size_t i = 0;i < x;++i){ + size_t offset = i << vertex_offset_shift; + int degree = edges[offset]; + fprintf(stderr,"%d (%d): ",i,degree); + for(int j = 1;j <= degree;++j) + fprintf(stderr,"(%zu,%f) ",edges[offset + j],edge_dist[offset + j]); + fprintf(stderr,"\n"); + } + } + + void dump(std::string path = "bfsg.graph"){ + FILE* fp = fopen(path.c_str(),"wb"); + size_t num_vertices = data->max_vertices(); + fwrite(&edges[0],sizeof(edges[0]) * (num_vertices << vertex_offset_shift),1,fp); + fclose(fp); + } + + void load(std::string path = "bfsg.graph"){ + FILE* fp = fopen(path.c_str(),"rb"); + size_t num_vertices = data->max_vertices(); + auto cnt = fread(&edges[0],sizeof(edges[0]) * (num_vertices << vertex_offset_shift),1,fp); + fclose(fp); + } + + Data* get_data(){ + return data; + } + +}; + diff --git a/exprements/PaddleClas/deploy/vector_search/test.py b/exprements/PaddleClas/deploy/vector_search/test.py new file mode 100644 index 00000000..06951b47 --- /dev/null +++ b/exprements/PaddleClas/deploy/vector_search/test.py @@ -0,0 +1,34 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import numpy as np +from interface import Graph_Index + +# 随机产生样本 +index_vectors = np.random.rand(100000,128).astype(np.float32) +query_vector = np.random.rand(128).astype(np.float32) +index_docs = ["ID_"+str(i) for i in range(100000)] + +# 初始化索引结构 +indexer = Graph_Index(dist_type="IP") #支持"IP"和"L2" +indexer.build(gallery_vectors=index_vectors, gallery_docs=index_docs, pq_size=100, index_path='test') + +# 查询 +scores, docs = indexer.search(query=query_vector, return_k=10, search_budget=100) +print(scores) +print(docs) + +# 保存与加载 +indexer.dump(index_path="test") +indexer.load(index_path="test") diff --git a/exprements/PaddleClas/hubconf.py b/exprements/PaddleClas/hubconf.py new file mode 100644 index 00000000..b7f76745 --- /dev/null +++ b/exprements/PaddleClas/hubconf.py @@ -0,0 +1,788 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +dependencies = ['paddle'] + +import paddle +import os +import sys + + +class _SysPathG(object): + """ + _SysPathG used to add/clean path for sys.path. Making sure minimal pkgs dependents by skiping parent dirs. + + __enter__ + add path into sys.path + __exit__ + clean user's sys.path to avoid unexpect behaviors + """ + + def __init__(self, path): + self.path = path + + def __enter__(self, ): + sys.path.insert(0, self.path) + + def __exit__(self, type, value, traceback): + _p = sys.path.pop(0) + assert _p == self.path, 'Make sure sys.path cleaning {} correctly.'.format( + self.path) + + +with _SysPathG(os.path.dirname(os.path.abspath(__file__)), ): + import ppcls + import ppcls.arch.backbone as backbone + + def ppclas_init(): + if ppcls.utils.logger._logger is None: + ppcls.utils.logger.init_logger() + + ppclas_init() + + def _load_pretrained_parameters(model, name): + url = 'https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/{}_pretrained.pdparams'.format( + name) + path = paddle.utils.download.get_weights_path_from_url(url) + model.set_state_dict(paddle.load(path)) + return model + + def alexnet(pretrained=False, **kwargs): + """ + AlexNet + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `AlexNet` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.AlexNet(**kwargs) + + return model + + def vgg11(pretrained=False, **kwargs): + """ + VGG11 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` + Returns: + model: nn.Layer. Specific `VGG11` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.VGG11(**kwargs) + + return model + + def vgg13(pretrained=False, **kwargs): + """ + VGG13 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` + Returns: + model: nn.Layer. Specific `VGG13` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.VGG13(**kwargs) + + return model + + def vgg16(pretrained=False, **kwargs): + """ + VGG16 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` + Returns: + model: nn.Layer. Specific `VGG16` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.VGG16(**kwargs) + + return model + + def vgg19(pretrained=False, **kwargs): + """ + VGG19 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` + Returns: + model: nn.Layer. Specific `VGG19` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.VGG19(**kwargs) + + return model + + def resnet18(pretrained=False, **kwargs): + """ + ResNet18 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + input_image_channel: int=3. The number of input image channels + data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') + Returns: + model: nn.Layer. Specific `ResNet18` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNet18(**kwargs) + + return model + + def resnet34(pretrained=False, **kwargs): + """ + ResNet34 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + input_image_channel: int=3. The number of input image channels + data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') + Returns: + model: nn.Layer. Specific `ResNet34` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNet34(**kwargs) + + return model + + def resnet50(pretrained=False, **kwargs): + """ + ResNet50 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + input_image_channel: int=3. The number of input image channels + data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') + Returns: + model: nn.Layer. Specific `ResNet50` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNet50(**kwargs) + + return model + + def resnet101(pretrained=False, **kwargs): + """ + ResNet101 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + input_image_channel: int=3. The number of input image channels + data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') + Returns: + model: nn.Layer. Specific `ResNet101` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNet101(**kwargs) + + return model + + def resnet152(pretrained=False, **kwargs): + """ + ResNet152 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + input_image_channel: int=3. The number of input image channels + data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') + Returns: + model: nn.Layer. Specific `ResNet152` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNet152(**kwargs) + + return model + + def squeezenet1_0(pretrained=False, **kwargs): + """ + SqueezeNet1_0 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `SqueezeNet1_0` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.SqueezeNet1_0(**kwargs) + + return model + + def squeezenet1_1(pretrained=False, **kwargs): + """ + SqueezeNet1_1 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `SqueezeNet1_1` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.SqueezeNet1_1(**kwargs) + + return model + + def densenet121(pretrained=False, **kwargs): + """ + DenseNet121 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + dropout: float=0. Probability of setting units to zero. + bn_size: int=4. The number of channals per group + Returns: + model: nn.Layer. Specific `DenseNet121` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.DenseNet121(**kwargs) + + return model + + def densenet161(pretrained=False, **kwargs): + """ + DenseNet161 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + dropout: float=0. Probability of setting units to zero. + bn_size: int=4. The number of channals per group + Returns: + model: nn.Layer. Specific `DenseNet161` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.DenseNet161(**kwargs) + + return model + + def densenet169(pretrained=False, **kwargs): + """ + DenseNet169 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + dropout: float=0. Probability of setting units to zero. + bn_size: int=4. The number of channals per group + Returns: + model: nn.Layer. Specific `DenseNet169` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.DenseNet169(**kwargs) + + return model + + def densenet201(pretrained=False, **kwargs): + """ + DenseNet201 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + dropout: float=0. Probability of setting units to zero. + bn_size: int=4. The number of channals per group + Returns: + model: nn.Layer. Specific `DenseNet201` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.DenseNet201(**kwargs) + + return model + + def densenet264(pretrained=False, **kwargs): + """ + DenseNet264 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + dropout: float=0. Probability of setting units to zero. + bn_size: int=4. The number of channals per group + Returns: + model: nn.Layer. Specific `DenseNet264` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.DenseNet264(**kwargs) + + return model + + def inceptionv3(pretrained=False, **kwargs): + """ + InceptionV3 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `InceptionV3` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.InceptionV3(**kwargs) + + return model + + def inceptionv4(pretrained=False, **kwargs): + """ + InceptionV4 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `InceptionV4` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.InceptionV4(**kwargs) + + return model + + def googlenet(pretrained=False, **kwargs): + """ + GoogLeNet + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `GoogLeNet` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.GoogLeNet(**kwargs) + + return model + + def shufflenetv2_x0_25(pretrained=False, **kwargs): + """ + ShuffleNetV2_x0_25 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ShuffleNetV2_x0_25` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ShuffleNetV2_x0_25(**kwargs) + + return model + + def mobilenetv1(pretrained=False, **kwargs): + """ + MobileNetV1 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV1` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV1(**kwargs) + + return model + + def mobilenetv1_x0_25(pretrained=False, **kwargs): + """ + MobileNetV1_x0_25 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV1_x0_25` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV1_x0_25(**kwargs) + + return model + + def mobilenetv1_x0_5(pretrained=False, **kwargs): + """ + MobileNetV1_x0_5 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV1_x0_5` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV1_x0_5(**kwargs) + + return model + + def mobilenetv1_x0_75(pretrained=False, **kwargs): + """ + MobileNetV1_x0_75 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV1_x0_75` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV1_x0_75(**kwargs) + + return model + + def mobilenetv2_x0_25(pretrained=False, **kwargs): + """ + MobileNetV2_x0_25 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV2_x0_25` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV2_x0_25(**kwargs) + + return model + + def mobilenetv2_x0_5(pretrained=False, **kwargs): + """ + MobileNetV2_x0_5 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV2_x0_5` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV2_x0_5(**kwargs) + + return model + + def mobilenetv2_x0_75(pretrained=False, **kwargs): + """ + MobileNetV2_x0_75 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV2_x0_75` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV2_x0_75(**kwargs) + + return model + + def mobilenetv2_x1_5(pretrained=False, **kwargs): + """ + MobileNetV2_x1_5 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV2_x1_5` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV2_x1_5(**kwargs) + + return model + + def mobilenetv2_x2_0(pretrained=False, **kwargs): + """ + MobileNetV2_x2_0 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV2_x2_0` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV2_x2_0(**kwargs) + + return model + + def mobilenetv3_large_x0_35(pretrained=False, **kwargs): + """ + MobileNetV3_large_x0_35 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x0_35` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_large_x0_35(**kwargs) + + return model + + def mobilenetv3_large_x0_5(pretrained=False, **kwargs): + """ + MobileNetV3_large_x0_5 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x0_5` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_large_x0_5(**kwargs) + + return model + + def mobilenetv3_large_x0_75(pretrained=False, **kwargs): + """ + MobileNetV3_large_x0_75 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x0_75` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_large_x0_75(**kwargs) + + return model + + def mobilenetv3_large_x1_0(pretrained=False, **kwargs): + """ + MobileNetV3_large_x1_0 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x1_0` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_large_x1_0(**kwargs) + + return model + + def mobilenetv3_large_x1_25(pretrained=False, **kwargs): + """ + MobileNetV3_large_x1_25 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x1_25` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_large_x1_25(**kwargs) + + return model + + def mobilenetv3_small_x0_35(pretrained=False, **kwargs): + """ + MobileNetV3_small_x0_35 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x0_35` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_small_x0_35(**kwargs) + + return model + + def mobilenetv3_small_x0_5(pretrained=False, **kwargs): + """ + MobileNetV3_small_x0_5 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x0_5` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_small_x0_5(**kwargs) + + return model + + def mobilenetv3_small_x0_75(pretrained=False, **kwargs): + """ + MobileNetV3_small_x0_75 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x0_75` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_small_x0_75(**kwargs) + + return model + + def mobilenetv3_small_x1_0(pretrained=False, **kwargs): + """ + MobileNetV3_small_x1_0 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x1_0` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_small_x1_0(**kwargs) + + return model + + def mobilenetv3_small_x1_25(pretrained=False, **kwargs): + """ + MobileNetV3_small_x1_25 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x1_25` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.MobileNetV3_small_x1_25(**kwargs) + + return model + + def resnext101_32x4d(pretrained=False, **kwargs): + """ + ResNeXt101_32x4d + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ResNeXt101_32x4d` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNeXt101_32x4d(**kwargs) + + return model + + def resnext101_64x4d(pretrained=False, **kwargs): + """ + ResNeXt101_64x4d + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ResNeXt101_64x4d` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNeXt101_64x4d(**kwargs) + + return model + + def resnext152_32x4d(pretrained=False, **kwargs): + """ + ResNeXt152_32x4d + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ResNeXt152_32x4d` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNeXt152_32x4d(**kwargs) + + return model + + def resnext152_64x4d(pretrained=False, **kwargs): + """ + ResNeXt152_64x4d + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ResNeXt152_64x4d` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNeXt152_64x4d(**kwargs) + + return model + + def resnext50_32x4d(pretrained=False, **kwargs): + """ + ResNeXt50_32x4d + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ResNeXt50_32x4d` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNeXt50_32x4d(**kwargs) + + return model + + def resnext50_64x4d(pretrained=False, **kwargs): + """ + ResNeXt50_64x4d + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ResNeXt50_64x4d` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.ResNeXt50_64x4d(**kwargs) + + return model + + def darknet53(pretrained=False, **kwargs): + """ + DarkNet53 + Args: + pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. + kwargs: + class_dim: int=1000. Output dim of last fc layer. + Returns: + model: nn.Layer. Specific `ResNeXt50_64x4d` model depends on args. + """ + kwargs.update({'pretrained': pretrained}) + model = backbone.DarkNet53(**kwargs) + + return model diff --git a/exprements/PaddleClas/paddleclas.py b/exprements/PaddleClas/paddleclas.py new file mode 100644 index 00000000..bfad1931 --- /dev/null +++ b/exprements/PaddleClas/paddleclas.py @@ -0,0 +1,572 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import sys +__dir__ = os.path.dirname(__file__) +sys.path.append(os.path.join(__dir__, "")) +sys.path.append(os.path.join(__dir__, "deploy")) + +from typing import Union, Generator +import argparse +import shutil +import textwrap +import tarfile +import requests +import warnings +from functools import partial +from difflib import SequenceMatcher + +import cv2 +import numpy as np +from tqdm import tqdm +from prettytable import PrettyTable + +from deploy.python.predict_cls import ClsPredictor +from deploy.utils.get_image_list import get_image_list +from deploy.utils import config + +from ppcls.arch.backbone import * +from ppcls.utils.logger import init_logger + +# for building model with loading pretrained weights from backbone +init_logger() + +__all__ = ["PaddleClas"] + +BASE_DIR = os.path.expanduser("~/.paddleclas/") +BASE_INFERENCE_MODEL_DIR = os.path.join(BASE_DIR, "inference_model") +BASE_IMAGES_DIR = os.path.join(BASE_DIR, "images") +BASE_DOWNLOAD_URL = "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/{}_infer.tar" +MODEL_SERIES = { + "AlexNet": ["AlexNet"], + "DarkNet": ["DarkNet53"], + "DeiT": [ + "DeiT_base_distilled_patch16_224", "DeiT_base_distilled_patch16_384", + "DeiT_base_patch16_224", "DeiT_base_patch16_384", + "DeiT_small_distilled_patch16_224", "DeiT_small_patch16_224", + "DeiT_tiny_distilled_patch16_224", "DeiT_tiny_patch16_224" + ], + "DenseNet": [ + "DenseNet121", "DenseNet161", "DenseNet169", "DenseNet201", + "DenseNet264" + ], + "DLA": [ + "DLA46_c", "DLA60x_c", "DLA34", "DLA60", "DLA60x", "DLA102", "DLA102x", + "DLA102x2", "DLA169" + ], + "DPN": ["DPN68", "DPN92", "DPN98", "DPN107", "DPN131"], + "EfficientNet": [ + "EfficientNetB0", "EfficientNetB0_small", "EfficientNetB1", + "EfficientNetB2", "EfficientNetB3", "EfficientNetB4", "EfficientNetB5", + "EfficientNetB6", "EfficientNetB7" + ], + "ESNet": ["ESNet_x0_25", "ESNet_x0_5", "ESNet_x0_75", "ESNet_x1_0"], + "GhostNet": + ["GhostNet_x0_5", "GhostNet_x1_0", "GhostNet_x1_3", "GhostNet_x1_3_ssld"], + "HarDNet": ["HarDNet39_ds", "HarDNet68_ds", "HarDNet68", "HarDNet85"], + "HRNet": [ + "HRNet_W18_C", "HRNet_W30_C", "HRNet_W32_C", "HRNet_W40_C", + "HRNet_W44_C", "HRNet_W48_C", "HRNet_W64_C", "HRNet_W18_C_ssld", + "HRNet_W48_C_ssld" + ], + "Inception": ["GoogLeNet", "InceptionV3", "InceptionV4"], + "MixNet": ["MixNet_S", "MixNet_M", "MixNet_L"], + "MobileNetV1": [ + "MobileNetV1_x0_25", "MobileNetV1_x0_5", "MobileNetV1_x0_75", + "MobileNetV1", "MobileNetV1_ssld" + ], + "MobileNetV2": [ + "MobileNetV2_x0_25", "MobileNetV2_x0_5", "MobileNetV2_x0_75", + "MobileNetV2", "MobileNetV2_x1_5", "MobileNetV2_x2_0", + "MobileNetV2_ssld" + ], + "MobileNetV3": [ + "MobileNetV3_small_x0_35", "MobileNetV3_small_x0_5", + "MobileNetV3_small_x0_75", "MobileNetV3_small_x1_0", + "MobileNetV3_small_x1_25", "MobileNetV3_large_x0_35", + "MobileNetV3_large_x0_5", "MobileNetV3_large_x0_75", + "MobileNetV3_large_x1_0", "MobileNetV3_large_x1_25", + "MobileNetV3_small_x1_0_ssld", "MobileNetV3_large_x1_0_ssld" + ], + "PPLCNet": [ + "PPLCNet_x0_25", "PPLCNet_x0_35", "PPLCNet_x0_5", "PPLCNet_x0_75", + "PPLCNet_x1_0", "PPLCNet_x1_5", "PPLCNet_x2_0", "PPLCNet_x2_5" + ], + "RedNet": ["RedNet26", "RedNet38", "RedNet50", "RedNet101", "RedNet152"], + "RegNet": ["RegNetX_4GF"], + "Res2Net": [ + "Res2Net50_14w_8s", "Res2Net50_26w_4s", "Res2Net50_vd_26w_4s", + "Res2Net200_vd_26w_4s", "Res2Net101_vd_26w_4s", + "Res2Net50_vd_26w_4s_ssld", "Res2Net101_vd_26w_4s_ssld", + "Res2Net200_vd_26w_4s_ssld" + ], + "ResNeSt": ["ResNeSt50", "ResNeSt50_fast_1s1x64d"], + "ResNet": [ + "ResNet18", "ResNet18_vd", "ResNet34", "ResNet34_vd", "ResNet50", + "ResNet50_vc", "ResNet50_vd", "ResNet50_vd_v2", "ResNet101", + "ResNet101_vd", "ResNet152", "ResNet152_vd", "ResNet200_vd", + "ResNet34_vd_ssld", "ResNet50_vd_ssld", "ResNet50_vd_ssld_v2", + "ResNet101_vd_ssld", "Fix_ResNet50_vd_ssld_v2", "ResNet50_ACNet_deploy" + ], + "ResNeXt": [ + "ResNeXt50_32x4d", "ResNeXt50_vd_32x4d", "ResNeXt50_64x4d", + "ResNeXt50_vd_64x4d", "ResNeXt101_32x4d", "ResNeXt101_vd_32x4d", + "ResNeXt101_32x8d_wsl", "ResNeXt101_32x16d_wsl", + "ResNeXt101_32x32d_wsl", "ResNeXt101_32x48d_wsl", + "Fix_ResNeXt101_32x48d_wsl", "ResNeXt101_64x4d", "ResNeXt101_vd_64x4d", + "ResNeXt152_32x4d", "ResNeXt152_vd_32x4d", "ResNeXt152_64x4d", + "ResNeXt152_vd_64x4d" + ], + "ReXNet": + ["ReXNet_1_0", "ReXNet_1_3", "ReXNet_1_5", "ReXNet_2_0", "ReXNet_3_0"], + "SENet": [ + "SENet154_vd", "SE_HRNet_W64_C_ssld", "SE_ResNet18_vd", + "SE_ResNet34_vd", "SE_ResNet50_vd", "SE_ResNeXt50_32x4d", + "SE_ResNeXt50_vd_32x4d", "SE_ResNeXt101_32x4d" + ], + "ShuffleNetV2": [ + "ShuffleNetV2_swish", "ShuffleNetV2_x0_25", "ShuffleNetV2_x0_33", + "ShuffleNetV2_x0_5", "ShuffleNetV2_x1_0", "ShuffleNetV2_x1_5", + "ShuffleNetV2_x2_0" + ], + "SqueezeNet": ["SqueezeNet1_0", "SqueezeNet1_1"], + "SwinTransformer": [ + "SwinTransformer_large_patch4_window7_224_22kto1k", + "SwinTransformer_large_patch4_window12_384_22kto1k", + "SwinTransformer_base_patch4_window7_224_22kto1k", + "SwinTransformer_base_patch4_window12_384_22kto1k", + "SwinTransformer_base_patch4_window12_384", + "SwinTransformer_base_patch4_window7_224", + "SwinTransformer_small_patch4_window7_224", + "SwinTransformer_tiny_patch4_window7_224" + ], + "Twins": [ + "pcpvt_small", "pcpvt_base", "pcpvt_large", "alt_gvt_small", + "alt_gvt_base", "alt_gvt_large" + ], + "VGG": ["VGG11", "VGG13", "VGG16", "VGG19"], + "VisionTransformer": [ + "ViT_base_patch16_224", "ViT_base_patch16_384", "ViT_base_patch32_384", + "ViT_large_patch16_224", "ViT_large_patch16_384", + "ViT_large_patch32_384", "ViT_small_patch16_224" + ], + "Xception": [ + "Xception41", "Xception41_deeplab", "Xception65", "Xception65_deeplab", + "Xception71" + ] +} + + +class ImageTypeError(Exception): + """ImageTypeError. + """ + + def __init__(self, message=""): + super().__init__(message) + + +class InputModelError(Exception): + """InputModelError. + """ + + def __init__(self, message=""): + super().__init__(message) + + +def init_config(model_name, + inference_model_dir, + use_gpu=True, + batch_size=1, + topk=5, + **kwargs): + imagenet1k_map_path = os.path.join( + os.path.abspath(__dir__), "ppcls/utils/imagenet1k_label_list.txt") + cfg = { + "Global": { + "infer_imgs": kwargs["infer_imgs"] + if "infer_imgs" in kwargs else False, + "model_name": model_name, + "inference_model_dir": inference_model_dir, + "batch_size": batch_size, + "use_gpu": use_gpu, + "enable_mkldnn": kwargs["enable_mkldnn"] + if "enable_mkldnn" in kwargs else False, + "cpu_num_threads": kwargs["cpu_num_threads"] + if "cpu_num_threads" in kwargs else 1, + "enable_benchmark": False, + "use_fp16": kwargs["use_fp16"] if "use_fp16" in kwargs else False, + "ir_optim": True, + "use_tensorrt": kwargs["use_tensorrt"] + if "use_tensorrt" in kwargs else False, + "gpu_mem": kwargs["gpu_mem"] if "gpu_mem" in kwargs else 8000, + "enable_profile": False + }, + "PreProcess": { + "transform_ops": [{ + "ResizeImage": { + "resize_short": kwargs["resize_short"] + if "resize_short" in kwargs else 256 + } + }, { + "CropImage": { + "size": kwargs["crop_size"] + if "crop_size" in kwargs else 224 + } + }, { + "NormalizeImage": { + "scale": 0.00392157, + "mean": [0.485, 0.456, 0.406], + "std": [0.229, 0.224, 0.225], + "order": '' + } + }, { + "ToCHWImage": None + }] + }, + "PostProcess": { + "main_indicator": "Topk", + "Topk": { + "topk": topk, + "class_id_map_file": imagenet1k_map_path + } + } + } + if "save_dir" in kwargs: + if kwargs["save_dir"] is not None: + cfg["PostProcess"]["SavePreLabel"] = { + "save_dir": kwargs["save_dir"] + } + if "class_id_map_file" in kwargs: + if kwargs["class_id_map_file"] is not None: + cfg["PostProcess"]["Topk"]["class_id_map_file"] = kwargs[ + "class_id_map_file"] + + cfg = config.AttrDict(cfg) + config.create_attr_dict(cfg) + return cfg + + +def args_cfg(): + def str2bool(v): + return v.lower() in ("true", "t", "1") + + parser = argparse.ArgumentParser() + parser.add_argument( + "--infer_imgs", + type=str, + required=True, + help="The image(s) to be predicted.") + parser.add_argument( + "--model_name", type=str, help="The model name to be used.") + parser.add_argument( + "--inference_model_dir", + type=str, + help="The directory of model files. Valid when model_name not specifed." + ) + parser.add_argument( + "--use_gpu", type=str, default=True, help="Whether use GPU.") + parser.add_argument("--gpu_mem", type=int, default=8000, help="") + parser.add_argument( + "--enable_mkldnn", + type=str2bool, + default=False, + help="Whether use MKLDNN. Valid when use_gpu is False") + parser.add_argument("--cpu_num_threads", type=int, default=1, help="") + parser.add_argument( + "--use_tensorrt", type=str2bool, default=False, help="") + parser.add_argument("--use_fp16", type=str2bool, default=False, help="") + parser.add_argument( + "--batch_size", type=int, default=1, help="Batch size. Default by 1.") + parser.add_argument( + "--topk", + type=int, + default=5, + help="Return topk score(s) and corresponding results. Default by 5.") + parser.add_argument( + "--class_id_map_file", + type=str, + help="The path of file that map class_id and label.") + parser.add_argument( + "--save_dir", + type=str, + help="The directory to save prediction results as pre-label.") + parser.add_argument( + "--resize_short", + type=int, + default=256, + help="Resize according to short size.") + parser.add_argument( + "--crop_size", type=int, default=224, help="Centor crop size.") + + args = parser.parse_args() + return vars(args) + + +def print_info(): + """Print list of supported models in formatted. + """ + table = PrettyTable(["Series", "Name"]) + try: + sz = os.get_terminal_size() + width = sz.columns - 30 if sz.columns > 50 else 10 + except OSError: + width = 100 + for series in MODEL_SERIES: + names = textwrap.fill(" ".join(MODEL_SERIES[series]), width=width) + table.add_row([series, names]) + width = len(str(table).split("\n")[0]) + print("{}".format("-" * width)) + print("Models supported by PaddleClas".center(width)) + print(table) + print("Powered by PaddlePaddle!".rjust(width)) + print("{}".format("-" * width)) + + +def get_model_names(): + """Get the model names list. + """ + model_names = [] + for series in MODEL_SERIES: + model_names += (MODEL_SERIES[series]) + return model_names + + +def similar_architectures(name="", names=[], thresh=0.1, topk=10): + """Find the most similar topk model names. + """ + scores = [] + for idx, n in enumerate(names): + if n.startswith("__"): + continue + score = SequenceMatcher(None, n.lower(), name.lower()).quick_ratio() + if score > thresh: + scores.append((idx, score)) + scores.sort(key=lambda x: x[1], reverse=True) + similar_names = [names[s[0]] for s in scores[:min(topk, len(scores))]] + return similar_names + + +def download_with_progressbar(url, save_path): + """Download from url with progressbar. + """ + if os.path.isfile(save_path): + os.remove(save_path) + response = requests.get(url, stream=True) + total_size_in_bytes = int(response.headers.get("content-length", 0)) + block_size = 1024 # 1 Kibibyte + progress_bar = tqdm(total=total_size_in_bytes, unit="iB", unit_scale=True) + with open(save_path, "wb") as file: + for data in response.iter_content(block_size): + progress_bar.update(len(data)) + file.write(data) + progress_bar.close() + if total_size_in_bytes == 0 or progress_bar.n != total_size_in_bytes or not os.path.isfile( + save_path): + raise Exception( + f"Something went wrong while downloading file from {url}") + + +def check_model_file(model_name): + """Check the model files exist and download and untar when no exist. + """ + storage_directory = partial(os.path.join, BASE_INFERENCE_MODEL_DIR, + model_name) + url = BASE_DOWNLOAD_URL.format(model_name) + + tar_file_name_list = [ + "inference.pdiparams", "inference.pdiparams.info", "inference.pdmodel" + ] + model_file_path = storage_directory("inference.pdmodel") + params_file_path = storage_directory("inference.pdiparams") + if not os.path.exists(model_file_path) or not os.path.exists( + params_file_path): + tmp_path = storage_directory(url.split("/")[-1]) + print(f"download {url} to {tmp_path}") + os.makedirs(storage_directory(), exist_ok=True) + download_with_progressbar(url, tmp_path) + with tarfile.open(tmp_path, "r") as tarObj: + for member in tarObj.getmembers(): + filename = None + for tar_file_name in tar_file_name_list: + if tar_file_name in member.name: + filename = tar_file_name + if filename is None: + continue + file = tarObj.extractfile(member) + with open(storage_directory(filename), "wb") as f: + f.write(file.read()) + os.remove(tmp_path) + if not os.path.exists(model_file_path) or not os.path.exists( + params_file_path): + raise Exception( + f"Something went wrong while praparing the model[{model_name}] files!" + ) + + return storage_directory() + + +class PaddleClas(object): + """PaddleClas. + """ + + print_info() + + def __init__(self, + model_name: str=None, + inference_model_dir: str=None, + use_gpu: bool=True, + batch_size: int=1, + topk: int=5, + **kwargs): + """Init PaddleClas with config. + + Args: + model_name (str, optional): The model name supported by PaddleClas. If specified, override config. Defaults to None. + inference_model_dir (str, optional): The directory that contained model file and params file to be used. If specified, override config. Defaults to None. + use_gpu (bool, optional): Whether use GPU. If specified, override config. Defaults to True. + batch_size (int, optional): The batch size to pridict. If specified, override config. Defaults to 1. + topk (int, optional): Return the top k prediction results with the highest score. Defaults to 5. + """ + super().__init__() + self._config = init_config(model_name, inference_model_dir, use_gpu, + batch_size, topk, **kwargs) + self._check_input_model() + self.cls_predictor = ClsPredictor(self._config) + + def get_config(self): + """Get the config. + """ + return self._config + + def _check_input_model(self): + """Check input model name or model files. + """ + candidate_model_names = get_model_names() + input_model_name = self._config.Global.get("model_name", None) + inference_model_dir = self._config.Global.get("inference_model_dir", + None) + if input_model_name is not None: + similar_names = similar_architectures(input_model_name, + candidate_model_names) + similar_names_str = ", ".join(similar_names) + if input_model_name not in candidate_model_names: + err = f"{input_model_name} is not provided by PaddleClas. \nMaybe you want: [{similar_names_str}]. \nIf you want to use your own model, please specify inference_model_dir!" + raise InputModelError(err) + self._config.Global.inference_model_dir = check_model_file( + input_model_name) + return + elif inference_model_dir is not None: + model_file_path = os.path.join(inference_model_dir, + "inference.pdmodel") + params_file_path = os.path.join(inference_model_dir, + "inference.pdiparams") + if not os.path.isfile(model_file_path) or not os.path.isfile( + params_file_path): + err = f"There is no model file or params file in this directory: {inference_model_dir}" + raise InputModelError(err) + return + else: + err = f"Please specify the model name supported by PaddleClas or directory contained model files(inference.pdmodel, inference.pdiparams)." + raise InputModelError(err) + return + + def predict(self, input_data: Union[str, np.array], + print_pred: bool=False) -> Generator[list, None, None]: + """Predict input_data. + + Args: + input_data (Union[str, np.array]): + When the type is str, it is the path of image, or the directory containing images, or the URL of image from Internet. + When the type is np.array, it is the image data whose channel order is RGB. + print_pred (bool, optional): Whether print the prediction result. Defaults to False. + + Raises: + ImageTypeError: Illegal input_data. + + Yields: + Generator[list, None, None]: + The prediction result(s) of input_data by batch_size. For every one image, + prediction result(s) is zipped as a dict, that includs topk "class_ids", "scores" and "label_names". + The format of batch prediction result(s) is as follow: [{"class_ids": [...], "scores": [...], "label_names": [...]}, ...] + """ + + if isinstance(input_data, np.ndarray): + yield self.cls_predictor.predict(input_data) + elif isinstance(input_data, str): + if input_data.startswith("http") or input_data.startswith("https"): + image_storage_dir = partial(os.path.join, BASE_IMAGES_DIR) + if not os.path.exists(image_storage_dir()): + os.makedirs(image_storage_dir()) + image_save_path = image_storage_dir("tmp.jpg") + download_with_progressbar(input_data, image_save_path) + input_data = image_save_path + warnings.warn( + f"Image to be predicted from Internet: {input_data}, has been saved to: {image_save_path}" + ) + image_list = get_image_list(input_data) + + batch_size = self._config.Global.get("batch_size", 1) + topk = self._config.PostProcess.Topk.get('topk', 1) + + img_list = [] + img_path_list = [] + cnt = 0 + for idx, img_path in enumerate(image_list): + img = cv2.imread(img_path) + if img is None: + warnings.warn( + f"Image file failed to read and has been skipped. The path: {img_path}" + ) + continue + img = img[:, :, ::-1] + img_list.append(img) + img_path_list.append(img_path) + cnt += 1 + + if cnt % batch_size == 0 or (idx + 1) == len(image_list): + preds = self.cls_predictor.predict(img_list) + + if print_pred and preds: + for idx, pred in enumerate(preds): + pred_str = ", ".join( + [f"{k}: {pred[k]}" for k in pred]) + print( + f"filename: {img_path_list[idx]}, top-{topk}, {pred_str}" + ) + + img_list = [] + img_path_list = [] + yield preds + else: + err = "Please input legal image! The type of image supported by PaddleClas are: NumPy.ndarray and string of local path or Ineternet URL" + raise ImageTypeError(err) + return + + +# for CLI +def main(): + """Function API used for commad line. + """ + cfg = args_cfg() + clas_engine = PaddleClas(**cfg) + res = clas_engine.predict(cfg["infer_imgs"], print_pred=True) + for _ in res: + pass + print("Predict complete!") + return + + +if __name__ == "__main__": + main() diff --git a/exprements/PaddleClas/ppcls/__init__.py b/exprements/PaddleClas/ppcls/__init__.py new file mode 100644 index 00000000..d6cdb6f8 --- /dev/null +++ b/exprements/PaddleClas/ppcls/__init__.py @@ -0,0 +1,20 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from . import optimizer + +from .arch import * +from .optimizer import * +from .data import * +from .utils import * diff --git a/exprements/PaddleClas/ppcls/arch/__init__.py b/exprements/PaddleClas/ppcls/arch/__init__.py new file mode 100644 index 00000000..2d5e29db --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/__init__.py @@ -0,0 +1,134 @@ +#copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +#Unless required by applicable law or agreed to in writing, software +#distributed under the License is distributed on an "AS IS" BASIS, +#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +#See the License for the specific language governing permissions and +#limitations under the License. + +import copy +import importlib + +import paddle.nn as nn +from paddle.jit import to_static +from paddle.static import InputSpec + +from . import backbone, gears +from .backbone import * +from .gears import build_gear +from .utils import * +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils import logger +from ppcls.utils.save_load import load_dygraph_pretrain +from ppcls.arch.slim import prune_model, quantize_model + +__all__ = ["build_model", "RecModel", "DistillationModel"] + + +def build_model(config): + arch_config = copy.deepcopy(config["Arch"]) + model_type = arch_config.pop("name") + mod = importlib.import_module(__name__) + arch = getattr(mod, model_type)(**arch_config) + if isinstance(arch, TheseusLayer): + prune_model(config, arch) + quantize_model(config, arch) + return arch + + +def apply_to_static(config, model): + support_to_static = config['Global'].get('to_static', False) + + if support_to_static: + specs = None + if 'image_shape' in config['Global']: + specs = [InputSpec([None] + config['Global']['image_shape'])] + model = to_static(model, input_spec=specs) + logger.info("Successfully to apply @to_static with specs: {}".format( + specs)) + return model + + +class RecModel(TheseusLayer): + def __init__(self, **config): + super().__init__() + backbone_config = config["Backbone"] + backbone_name = backbone_config.pop("name") + self.backbone = eval(backbone_name)(**backbone_config) + if "BackboneStopLayer" in config: + backbone_stop_layer = config["BackboneStopLayer"]["name"] + self.backbone.stop_after(backbone_stop_layer) + + if "Neck" in config: + self.neck = build_gear(config["Neck"]) + else: + self.neck = None + + if "Head" in config: + self.head = build_gear(config["Head"]) + else: + self.head = None + + def forward(self, x, label=None): + out = dict() + x = self.backbone(x) + out["backbone"] = x + if self.neck is not None: + x = self.neck(x) + out["neck"] = x + out["features"] = x + if self.head is not None: + y = self.head(x, label) + out["logits"] = y + return out + + +class DistillationModel(nn.Layer): + def __init__(self, + models=None, + pretrained_list=None, + freeze_params_list=None, + **kargs): + super().__init__() + assert isinstance(models, list) + self.model_list = [] + self.model_name_list = [] + if pretrained_list is not None: + assert len(pretrained_list) == len(models) + + if freeze_params_list is None: + freeze_params_list = [False] * len(models) + assert len(freeze_params_list) == len(models) + for idx, model_config in enumerate(models): + assert len(model_config) == 1 + key = list(model_config.keys())[0] + model_config = model_config[key] + model_name = model_config.pop("name") + model = eval(model_name)(**model_config) + + if freeze_params_list[idx]: + for param in model.parameters(): + param.trainable = False + self.model_list.append(self.add_sublayer(key, model)) + self.model_name_list.append(key) + + if pretrained_list is not None: + for idx, pretrained in enumerate(pretrained_list): + if pretrained is not None: + load_dygraph_pretrain( + self.model_name_list[idx], path=pretrained) + + def forward(self, x, label=None): + result_dict = dict() + for idx, model_name in enumerate(self.model_name_list): + if label is None: + result_dict[model_name] = self.model_list[idx](x) + else: + result_dict[model_name] = self.model_list[idx](x, label) + return result_dict diff --git a/exprements/PaddleClas/ppcls/arch/backbone/__init__.py b/exprements/PaddleClas/ppcls/arch/backbone/__init__.py new file mode 100644 index 00000000..dadac59a --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/__init__.py @@ -0,0 +1,82 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import sys +import inspect + +from ppcls.arch.backbone.legendary_models.mobilenet_v1 import MobileNetV1_x0_25, MobileNetV1_x0_5, MobileNetV1_x0_75, MobileNetV1 +from ppcls.arch.backbone.legendary_models.mobilenet_v3 import MobileNetV3_small_x0_35, MobileNetV3_small_x0_5, MobileNetV3_small_x0_75, MobileNetV3_small_x1_0, MobileNetV3_small_x1_25, MobileNetV3_large_x0_35, MobileNetV3_large_x0_5, MobileNetV3_large_x0_75, MobileNetV3_large_x1_0, MobileNetV3_large_x1_25 +from ppcls.arch.backbone.legendary_models.resnet import ResNet18, ResNet18_vd, ResNet34, ResNet34_vd, ResNet50, ResNet50_vd, ResNet101, ResNet101_vd, ResNet152, ResNet152_vd, ResNet200_vd +from ppcls.arch.backbone.legendary_models.vgg import VGG11, VGG13, VGG16, VGG19 +from ppcls.arch.backbone.legendary_models.inception_v3 import InceptionV3 +from ppcls.arch.backbone.legendary_models.hrnet import HRNet_W18_C, HRNet_W30_C, HRNet_W32_C, HRNet_W40_C, HRNet_W44_C, HRNet_W48_C, HRNet_W60_C, HRNet_W64_C, SE_HRNet_W64_C +from ppcls.arch.backbone.legendary_models.pp_lcnet import PPLCNet_x0_25, PPLCNet_x0_35, PPLCNet_x0_5, PPLCNet_x0_75, PPLCNet_x1_0, PPLCNet_x1_5, PPLCNet_x2_0, PPLCNet_x2_5 +from ppcls.arch.backbone.legendary_models.esnet import ESNet_x0_25, ESNet_x0_5, ESNet_x0_75, ESNet_x1_0 + +from ppcls.arch.backbone.model_zoo.resnet_vc import ResNet50_vc +from ppcls.arch.backbone.model_zoo.resnext import ResNeXt50_32x4d, ResNeXt50_64x4d, ResNeXt101_32x4d, ResNeXt101_64x4d, ResNeXt152_32x4d, ResNeXt152_64x4d +from ppcls.arch.backbone.model_zoo.resnext_vd import ResNeXt50_vd_32x4d, ResNeXt50_vd_64x4d, ResNeXt101_vd_32x4d, ResNeXt101_vd_64x4d, ResNeXt152_vd_32x4d, ResNeXt152_vd_64x4d +from ppcls.arch.backbone.model_zoo.res2net import Res2Net50_26w_4s, Res2Net50_14w_8s +from ppcls.arch.backbone.model_zoo.res2net_vd import Res2Net50_vd_26w_4s, Res2Net101_vd_26w_4s, Res2Net200_vd_26w_4s +from ppcls.arch.backbone.model_zoo.se_resnet_vd import SE_ResNet18_vd, SE_ResNet34_vd, SE_ResNet50_vd +from ppcls.arch.backbone.model_zoo.se_resnext_vd import SE_ResNeXt50_vd_32x4d, SE_ResNeXt50_vd_32x4d, SENet154_vd +from ppcls.arch.backbone.model_zoo.se_resnext import SE_ResNeXt50_32x4d, SE_ResNeXt101_32x4d, SE_ResNeXt152_64x4d +from ppcls.arch.backbone.model_zoo.dpn import DPN68, DPN92, DPN98, DPN107, DPN131 +from ppcls.arch.backbone.model_zoo.densenet import DenseNet121, DenseNet161, DenseNet169, DenseNet201, DenseNet264 +from ppcls.arch.backbone.model_zoo.efficientnet import EfficientNetB0, EfficientNetB1, EfficientNetB2, EfficientNetB3, EfficientNetB4, EfficientNetB5, EfficientNetB6, EfficientNetB7, EfficientNetB0_small +from ppcls.arch.backbone.model_zoo.resnest import ResNeSt50_fast_1s1x64d, ResNeSt50, ResNeSt101 +from ppcls.arch.backbone.model_zoo.googlenet import GoogLeNet +from ppcls.arch.backbone.model_zoo.mobilenet_v2 import MobileNetV2_x0_25, MobileNetV2_x0_5, MobileNetV2_x0_75, MobileNetV2, MobileNetV2_x1_5, MobileNetV2_x2_0 +from ppcls.arch.backbone.model_zoo.shufflenet_v2 import ShuffleNetV2_x0_25, ShuffleNetV2_x0_33, ShuffleNetV2_x0_5, ShuffleNetV2_x1_0, ShuffleNetV2_x1_5, ShuffleNetV2_x2_0, ShuffleNetV2_swish +from ppcls.arch.backbone.model_zoo.ghostnet import GhostNet_x0_5, GhostNet_x1_0, GhostNet_x1_3 +from ppcls.arch.backbone.model_zoo.alexnet import AlexNet +from ppcls.arch.backbone.model_zoo.inception_v4 import InceptionV4 +from ppcls.arch.backbone.model_zoo.xception import Xception41, Xception65, Xception71 +from ppcls.arch.backbone.model_zoo.xception_deeplab import Xception41_deeplab, Xception65_deeplab +from ppcls.arch.backbone.model_zoo.resnext101_wsl import ResNeXt101_32x8d_wsl, ResNeXt101_32x16d_wsl, ResNeXt101_32x32d_wsl, ResNeXt101_32x48d_wsl +from ppcls.arch.backbone.model_zoo.squeezenet import SqueezeNet1_0, SqueezeNet1_1 +from ppcls.arch.backbone.model_zoo.darknet import DarkNet53 +from ppcls.arch.backbone.model_zoo.regnet import RegNetX_200MF, RegNetX_4GF, RegNetX_32GF, RegNetY_200MF, RegNetY_4GF, RegNetY_32GF +from ppcls.arch.backbone.model_zoo.vision_transformer import ViT_small_patch16_224, ViT_base_patch16_224, ViT_base_patch16_384, ViT_base_patch32_384, ViT_large_patch16_224, ViT_large_patch16_384, ViT_large_patch32_384 +from ppcls.arch.backbone.model_zoo.distilled_vision_transformer import DeiT_tiny_patch16_224, DeiT_small_patch16_224, DeiT_base_patch16_224, DeiT_tiny_distilled_patch16_224, DeiT_small_distilled_patch16_224, DeiT_base_distilled_patch16_224, DeiT_base_patch16_384, DeiT_base_distilled_patch16_384 +from ppcls.arch.backbone.model_zoo.swin_transformer import SwinTransformer_tiny_patch4_window7_224, SwinTransformer_small_patch4_window7_224, SwinTransformer_base_patch4_window7_224, SwinTransformer_base_patch4_window12_384, SwinTransformer_large_patch4_window7_224, SwinTransformer_large_patch4_window12_384 +from ppcls.arch.backbone.model_zoo.mixnet import MixNet_S, MixNet_M, MixNet_L +from ppcls.arch.backbone.model_zoo.rexnet import ReXNet_1_0, ReXNet_1_3, ReXNet_1_5, ReXNet_2_0, ReXNet_3_0 +from ppcls.arch.backbone.model_zoo.gvt import pcpvt_small, pcpvt_base, pcpvt_large, alt_gvt_small, alt_gvt_base, alt_gvt_large +from ppcls.arch.backbone.model_zoo.levit import LeViT_128S, LeViT_128, LeViT_192, LeViT_256, LeViT_384 +from ppcls.arch.backbone.model_zoo.dla import DLA34, DLA46_c, DLA46x_c, DLA60, DLA60x, DLA60x_c, DLA102, DLA102x, DLA102x2, DLA169 +from ppcls.arch.backbone.model_zoo.rednet import RedNet26, RedNet38, RedNet50, RedNet101, RedNet152 +from ppcls.arch.backbone.model_zoo.tnt import TNT_small +from ppcls.arch.backbone.model_zoo.hardnet import HarDNet68, HarDNet85, HarDNet39_ds, HarDNet68_ds +from ppcls.arch.backbone.model_zoo.cspnet import CSPDarkNet53 +from ppcls.arch.backbone.model_zoo.pvt_v2 import PVT_V2_B0, PVT_V2_B1, PVT_V2_B2_Linear, PVT_V2_B2, PVT_V2_B3, PVT_V2_B4, PVT_V2_B5 +from ppcls.arch.backbone.variant_models.resnet_variant import ResNet50_last_stage_stride1 +from ppcls.arch.backbone.variant_models.vgg_variant import VGG19Sigmoid +from ppcls.arch.backbone.variant_models.pp_lcnet_variant import PPLCNet_x2_5_Tanh + + +# help whl get all the models' api (class type) and components' api (func type) +def get_apis(): + current_func = sys._getframe().f_code.co_name + current_module = sys.modules[__name__] + api = [] + for _, obj in inspect.getmembers(current_module, + inspect.isclass) + inspect.getmembers( + current_module, inspect.isfunction): + api.append(obj.__name__) + api.remove(current_func) + return api + + +__all__ = get_apis() diff --git a/exprements/PaddleClas/ppcls/arch/backbone/base/__init__.py b/exprements/PaddleClas/ppcls/arch/backbone/base/__init__.py new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/ppcls/arch/backbone/base/theseus_layer.py b/exprements/PaddleClas/ppcls/arch/backbone/base/theseus_layer.py new file mode 100644 index 00000000..908d9444 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/base/theseus_layer.py @@ -0,0 +1,301 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Tuple, List, Dict, Union, Callable, Any + +from paddle import nn +from ppcls.utils import logger + + +class Identity(nn.Layer): + def __init__(self): + super(Identity, self).__init__() + + def forward(self, inputs): + return inputs + + +class TheseusLayer(nn.Layer): + def __init__(self, *args, **kwargs): + super(TheseusLayer, self).__init__() + self.res_dict = {} + self.res_name = self.full_name() + self.pruner = None + self.quanter = None + + def _return_dict_hook(self, layer, input, output): + res_dict = {"output": output} + # 'list' is needed to avoid error raised by popping self.res_dict + for res_key in list(self.res_dict): + # clear the res_dict because the forward process may change according to input + res_dict[res_key] = self.res_dict.pop(res_key) + return res_dict + + def init_res(self, + stages_pattern, + return_patterns=None, + return_stages=None): + if return_patterns and return_stages: + msg = f"The 'return_patterns' would be ignored when 'return_stages' is set." + logger.warning(msg) + return_stages = None + + if return_stages is True: + return_patterns = stages_pattern + # return_stages is int or bool + if type(return_stages) is int: + return_stages = [return_stages] + if isinstance(return_stages, list): + if max(return_stages) > len(stages_pattern) or min( + return_stages) < 0: + msg = f"The 'return_stages' set error. Illegal value(s) have been ignored. The stages' pattern list is {stages_pattern}." + logger.warning(msg) + return_stages = [ + val for val in return_stages + if val >= 0 and val < len(stages_pattern) + ] + return_patterns = [stages_pattern[i] for i in return_stages] + + if return_patterns: + self.update_res(return_patterns) + + def replace_sub(self, *args, **kwargs) -> None: + msg = "The function 'replace_sub()' is deprecated, please use 'upgrade_sublayer()' instead." + logger.error(DeprecationWarning(msg)) + raise DeprecationWarning(msg) + + def upgrade_sublayer(self, + layer_name_pattern: Union[str, List[str]], + handle_func: Callable[[nn.Layer, str], nn.Layer] + ) -> Dict[str, nn.Layer]: + """use 'handle_func' to modify the sub-layer(s) specified by 'layer_name_pattern'. + + Args: + layer_name_pattern (Union[str, List[str]]): The name of layer to be modified by 'handle_func'. + handle_func (Callable[[nn.Layer, str], nn.Layer]): The function to modify target layer specified by 'layer_name_pattern'. The formal params are the layer(nn.Layer) and pattern(str) that is (a member of) layer_name_pattern (when layer_name_pattern is List type). And the return is the layer processed. + + Returns: + Dict[str, nn.Layer]: The key is the pattern and corresponding value is the result returned by 'handle_func()'. + + Examples: + + from paddle import nn + import paddleclas + + def rep_func(layer: nn.Layer, pattern: str): + new_layer = nn.Conv2D( + in_channels=layer._in_channels, + out_channels=layer._out_channels, + kernel_size=5, + padding=2 + ) + return new_layer + + net = paddleclas.MobileNetV1() + res = net.replace_sub(layer_name_pattern=["blocks[11].depthwise_conv.conv", "blocks[12].depthwise_conv.conv"], handle_func=rep_func) + print(res) + # {'blocks[11].depthwise_conv.conv': the corresponding new_layer, 'blocks[12].depthwise_conv.conv': the corresponding new_layer} + """ + + if not isinstance(layer_name_pattern, list): + layer_name_pattern = [layer_name_pattern] + + hit_layer_pattern_list = [] + for pattern in layer_name_pattern: + # parse pattern to find target layer and its parent + layer_list = parse_pattern_str(pattern=pattern, parent_layer=self) + if not layer_list: + continue + sub_layer_parent = layer_list[-2]["layer"] if len( + layer_list) > 1 else self + + sub_layer = layer_list[-1]["layer"] + sub_layer_name = layer_list[-1]["name"] + sub_layer_index = layer_list[-1]["index"] + + new_sub_layer = handle_func(sub_layer, pattern) + + if sub_layer_index: + getattr(sub_layer_parent, + sub_layer_name)[sub_layer_index] = new_sub_layer + else: + setattr(sub_layer_parent, sub_layer_name, new_sub_layer) + + hit_layer_pattern_list.append(pattern) + return hit_layer_pattern_list + + def stop_after(self, stop_layer_name: str) -> bool: + """stop forward and backward after 'stop_layer_name'. + + Args: + stop_layer_name (str): The name of layer that stop forward and backward after this layer. + + Returns: + bool: 'True' if successful, 'False' otherwise. + """ + + layer_list = parse_pattern_str(stop_layer_name, self) + if not layer_list: + return False + + parent_layer = self + for layer_dict in layer_list: + name, index = layer_dict["name"], layer_dict["index"] + if not set_identity(parent_layer, name, index): + msg = f"Failed to set the layers that after stop_layer_name('{stop_layer_name}') to IdentityLayer. The error layer's name is '{name}'." + logger.warning(msg) + return False + parent_layer = layer_dict["layer"] + + return True + + def update_res( + self, + return_patterns: Union[str, List[str]]) -> Dict[str, nn.Layer]: + """update the result(s) to be returned. + + Args: + return_patterns (Union[str, List[str]]): The name of layer to return output. + + Returns: + Dict[str, nn.Layer]: The pattern(str) and corresponding layer(nn.Layer) that have been set successfully. + """ + + # clear res_dict that could have been set + self.res_dict = {} + + class Handler(object): + def __init__(self, res_dict): + # res_dict is a reference + self.res_dict = res_dict + + def __call__(self, layer, pattern): + layer.res_dict = self.res_dict + layer.res_name = pattern + if hasattr(layer, "hook_remove_helper"): + layer.hook_remove_helper.remove() + layer.hook_remove_helper = layer.register_forward_post_hook( + save_sub_res_hook) + return layer + + handle_func = Handler(self.res_dict) + + hit_layer_pattern_list = self.upgrade_sublayer( + return_patterns, handle_func=handle_func) + + if hasattr(self, "hook_remove_helper"): + self.hook_remove_helper.remove() + self.hook_remove_helper = self.register_forward_post_hook( + self._return_dict_hook) + + return hit_layer_pattern_list + + +def save_sub_res_hook(layer, input, output): + layer.res_dict[layer.res_name] = output + + +def set_identity(parent_layer: nn.Layer, + layer_name: str, + layer_index: str=None) -> bool: + """set the layer specified by layer_name and layer_index to Indentity. + + Args: + parent_layer (nn.Layer): The parent layer of target layer specified by layer_name and layer_index. + layer_name (str): The name of target layer to be set to Indentity. + layer_index (str, optional): The index of target layer to be set to Indentity in parent_layer. Defaults to None. + + Returns: + bool: True if successfully, False otherwise. + """ + + stop_after = False + for sub_layer_name in parent_layer._sub_layers: + if stop_after: + parent_layer._sub_layers[sub_layer_name] = Identity() + continue + if sub_layer_name == layer_name: + stop_after = True + + if layer_index and stop_after: + stop_after = False + for sub_layer_index in parent_layer._sub_layers[ + layer_name]._sub_layers: + if stop_after: + parent_layer._sub_layers[layer_name][ + sub_layer_index] = Identity() + continue + if layer_index == sub_layer_index: + stop_after = True + + return stop_after + + +def parse_pattern_str(pattern: str, parent_layer: nn.Layer) -> Union[ + None, List[Dict[str, Union[nn.Layer, str, None]]]]: + """parse the string type pattern. + + Args: + pattern (str): The pattern to discribe layer. + parent_layer (nn.Layer): The root layer relative to the pattern. + + Returns: + Union[None, List[Dict[str, Union[nn.Layer, str, None]]]]: None if failed. If successfully, the members are layers parsed in order: + [ + {"layer": first layer, "name": first layer's name parsed, "index": first layer's index parsed if exist}, + {"layer": second layer, "name": second layer's name parsed, "index": second layer's index parsed if exist}, + ... + ] + """ + + pattern_list = pattern.split(".") + if not pattern_list: + msg = f"The pattern('{pattern}') is illegal. Please check and retry." + logger.warning(msg) + return None + + layer_list = [] + while len(pattern_list) > 0: + if '[' in pattern_list[0]: + target_layer_name = pattern_list[0].split('[')[0] + target_layer_index = pattern_list[0].split('[')[1].split(']')[0] + else: + target_layer_name = pattern_list[0] + target_layer_index = None + + target_layer = getattr(parent_layer, target_layer_name, None) + + if target_layer is None: + msg = f"Not found layer named('{target_layer_name}') specifed in pattern('{pattern}')." + logger.warning(msg) + return None + + if target_layer_index and target_layer: + if int(target_layer_index) < 0 or int(target_layer_index) >= len( + target_layer): + msg = f"Not found layer by index('{target_layer_index}') specifed in pattern('{pattern}'). The index should < {len(target_layer)} and > 0." + logger.warning(msg) + return None + + target_layer = target_layer[target_layer_index] + + layer_list.append({ + "layer": target_layer, + "name": target_layer_name, + "index": target_layer_index + }) + + pattern_list = pattern_list[1:] + parent_layer = target_layer + return layer_list diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/__init__.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/__init__.py new file mode 100644 index 00000000..1f837dac --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/__init__.py @@ -0,0 +1,6 @@ +from .resnet import ResNet18, ResNet34, ResNet50, ResNet101, ResNet152, ResNet18_vd, ResNet34_vd, ResNet50_vd, ResNet101_vd, ResNet152_vd +from .hrnet import HRNet_W18_C, HRNet_W30_C, HRNet_W32_C, HRNet_W40_C, HRNet_W44_C, HRNet_W48_C, HRNet_W64_C +from .mobilenet_v1 import MobileNetV1_x0_25, MobileNetV1_x0_5, MobileNetV1_x0_75, MobileNetV1 +from .mobilenet_v3 import MobileNetV3_small_x0_35, MobileNetV3_small_x0_5, MobileNetV3_small_x0_75, MobileNetV3_small_x1_0, MobileNetV3_small_x1_25, MobileNetV3_large_x0_35, MobileNetV3_large_x0_5, MobileNetV3_large_x0_75, MobileNetV3_large_x1_0, MobileNetV3_large_x1_25 +from .inception_v3 import InceptionV3 +from .vgg import VGG11, VGG13, VGG16, VGG19 diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/esnet.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/esnet.py new file mode 100644 index 00000000..e05e0ceb --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/esnet.py @@ -0,0 +1,369 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function +import math +import paddle +from paddle import ParamAttr, reshape, transpose, concat, split +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D +from paddle.nn.initializer import KaimingNormal +from paddle.regularizer import L2Decay + +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ESNet_x0_25": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_25_pretrained.pdparams", + "ESNet_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_5_pretrained.pdparams", + "ESNet_x0_75": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_75_pretrained.pdparams", + "ESNet_x1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x1_0_pretrained.pdparams", +} + +MODEL_STAGES_PATTERN = {"ESNet": ["blocks[2]", "blocks[9]", "blocks[12]"]} + +__all__ = list(MODEL_URLS.keys()) + + +def channel_shuffle(x, groups): + batch_size, num_channels, height, width = x.shape[0:4] + channels_per_group = num_channels // groups + x = reshape( + x=x, shape=[batch_size, groups, channels_per_group, height, width]) + x = transpose(x=x, perm=[0, 2, 1, 3, 4]) + x = reshape(x=x, shape=[batch_size, num_channels, height, width]) + return x + + +def make_divisible(v, divisor=8, min_value=None): + if min_value is None: + min_value = divisor + new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) + if new_v < 0.9 * v: + new_v += divisor + return new_v + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + in_channels, + out_channels, + kernel_size, + stride=1, + groups=1, + if_act=True): + super().__init__() + self.conv = Conv2D( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=(kernel_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(initializer=KaimingNormal()), + bias_attr=False) + + self.bn = BatchNorm( + out_channels, + param_attr=ParamAttr(regularizer=L2Decay(0.0)), + bias_attr=ParamAttr(regularizer=L2Decay(0.0))) + self.if_act = if_act + self.hardswish = nn.Hardswish() + + def forward(self, x): + x = self.conv(x) + x = self.bn(x) + if self.if_act: + x = self.hardswish(x) + return x + + +class SEModule(TheseusLayer): + def __init__(self, channel, reduction=4): + super().__init__() + self.avg_pool = AdaptiveAvgPool2D(1) + self.conv1 = Conv2D( + in_channels=channel, + out_channels=channel // reduction, + kernel_size=1, + stride=1, + padding=0) + self.relu = nn.ReLU() + self.conv2 = Conv2D( + in_channels=channel // reduction, + out_channels=channel, + kernel_size=1, + stride=1, + padding=0) + self.hardsigmoid = nn.Hardsigmoid() + + def forward(self, x): + identity = x + x = self.avg_pool(x) + x = self.conv1(x) + x = self.relu(x) + x = self.conv2(x) + x = self.hardsigmoid(x) + x = paddle.multiply(x=identity, y=x) + return x + + +class ESBlock1(TheseusLayer): + def __init__(self, in_channels, out_channels): + super().__init__() + self.pw_1_1 = ConvBNLayer( + in_channels=in_channels // 2, + out_channels=out_channels // 2, + kernel_size=1, + stride=1) + self.dw_1 = ConvBNLayer( + in_channels=out_channels // 2, + out_channels=out_channels // 2, + kernel_size=3, + stride=1, + groups=out_channels // 2, + if_act=False) + self.se = SEModule(out_channels) + + self.pw_1_2 = ConvBNLayer( + in_channels=out_channels, + out_channels=out_channels // 2, + kernel_size=1, + stride=1) + + def forward(self, x): + x1, x2 = split( + x, num_or_sections=[x.shape[1] // 2, x.shape[1] // 2], axis=1) + x2 = self.pw_1_1(x2) + x3 = self.dw_1(x2) + x3 = concat([x2, x3], axis=1) + x3 = self.se(x3) + x3 = self.pw_1_2(x3) + x = concat([x1, x3], axis=1) + return channel_shuffle(x, 2) + + +class ESBlock2(TheseusLayer): + def __init__(self, in_channels, out_channels): + super().__init__() + + # branch1 + self.dw_1 = ConvBNLayer( + in_channels=in_channels, + out_channels=in_channels, + kernel_size=3, + stride=2, + groups=in_channels, + if_act=False) + self.pw_1 = ConvBNLayer( + in_channels=in_channels, + out_channels=out_channels // 2, + kernel_size=1, + stride=1) + # branch2 + self.pw_2_1 = ConvBNLayer( + in_channels=in_channels, + out_channels=out_channels // 2, + kernel_size=1) + self.dw_2 = ConvBNLayer( + in_channels=out_channels // 2, + out_channels=out_channels // 2, + kernel_size=3, + stride=2, + groups=out_channels // 2, + if_act=False) + self.se = SEModule(out_channels // 2) + self.pw_2_2 = ConvBNLayer( + in_channels=out_channels // 2, + out_channels=out_channels // 2, + kernel_size=1) + self.concat_dw = ConvBNLayer( + in_channels=out_channels, + out_channels=out_channels, + kernel_size=3, + groups=out_channels) + self.concat_pw = ConvBNLayer( + in_channels=out_channels, out_channels=out_channels, kernel_size=1) + + def forward(self, x): + x1 = self.dw_1(x) + x1 = self.pw_1(x1) + x2 = self.pw_2_1(x) + x2 = self.dw_2(x2) + x2 = self.se(x2) + x2 = self.pw_2_2(x2) + x = concat([x1, x2], axis=1) + x = self.concat_dw(x) + x = self.concat_pw(x) + return x + + +class ESNet(TheseusLayer): + def __init__(self, + stages_pattern, + class_num=1000, + scale=1.0, + dropout_prob=0.2, + class_expand=1280, + return_patterns=None, + return_stages=None): + super().__init__() + self.scale = scale + self.class_num = class_num + self.class_expand = class_expand + stage_repeats = [3, 7, 3] + stage_out_channels = [ + -1, 24, make_divisible(116 * scale), make_divisible(232 * scale), + make_divisible(464 * scale), 1024 + ] + + self.conv1 = ConvBNLayer( + in_channels=3, + out_channels=stage_out_channels[1], + kernel_size=3, + stride=2) + self.max_pool = MaxPool2D(kernel_size=3, stride=2, padding=1) + + block_list = [] + for stage_id, num_repeat in enumerate(stage_repeats): + for i in range(num_repeat): + if i == 0: + block = ESBlock2( + in_channels=stage_out_channels[stage_id + 1], + out_channels=stage_out_channels[stage_id + 2]) + else: + block = ESBlock1( + in_channels=stage_out_channels[stage_id + 2], + out_channels=stage_out_channels[stage_id + 2]) + block_list.append(block) + self.blocks = nn.Sequential(*block_list) + + self.conv2 = ConvBNLayer( + in_channels=stage_out_channels[-2], + out_channels=stage_out_channels[-1], + kernel_size=1) + + self.avg_pool = AdaptiveAvgPool2D(1) + + self.last_conv = Conv2D( + in_channels=stage_out_channels[-1], + out_channels=self.class_expand, + kernel_size=1, + stride=1, + padding=0, + bias_attr=False) + self.hardswish = nn.Hardswish() + self.dropout = Dropout(p=dropout_prob, mode="downscale_in_infer") + self.flatten = nn.Flatten(start_axis=1, stop_axis=-1) + self.fc = Linear(self.class_expand, self.class_num) + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, x): + x = self.conv1(x) + x = self.max_pool(x) + x = self.blocks(x) + x = self.conv2(x) + x = self.avg_pool(x) + x = self.last_conv(x) + x = self.hardswish(x) + x = self.dropout(x) + x = self.flatten(x) + x = self.fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ESNet_x0_25(pretrained=False, use_ssld=False, **kwargs): + """ + ESNet_x0_25 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ESNet_x0_25` model depends on args. + """ + model = ESNet( + scale=0.25, stages_pattern=MODEL_STAGES_PATTERN["ESNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ESNet_x0_25"], use_ssld) + return model + + +def ESNet_x0_5(pretrained=False, use_ssld=False, **kwargs): + """ + ESNet_x0_5 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ESNet_x0_5` model depends on args. + """ + model = ESNet( + scale=0.5, stages_pattern=MODEL_STAGES_PATTERN["ESNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ESNet_x0_5"], use_ssld) + return model + + +def ESNet_x0_75(pretrained=False, use_ssld=False, **kwargs): + """ + ESNet_x0_75 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ESNet_x0_75` model depends on args. + """ + model = ESNet( + scale=0.75, stages_pattern=MODEL_STAGES_PATTERN["ESNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ESNet_x0_75"], use_ssld) + return model + + +def ESNet_x1_0(pretrained=False, use_ssld=False, **kwargs): + """ + ESNet_x1_0 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ESNet_x1_0` model depends on args. + """ + model = ESNet( + scale=1.0, stages_pattern=MODEL_STAGES_PATTERN["ESNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ESNet_x1_0"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/hrnet.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/hrnet.py new file mode 100644 index 00000000..c3f77590 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/hrnet.py @@ -0,0 +1,794 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import math +import paddle +from paddle import nn +from paddle import ParamAttr +from paddle.nn.functional import upsample +from paddle.nn.initializer import Uniform + +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer, Identity +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "HRNet_W18_C": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W18_C_pretrained.pdparams", + "HRNet_W30_C": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W30_C_pretrained.pdparams", + "HRNet_W32_C": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W32_C_pretrained.pdparams", + "HRNet_W40_C": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W40_C_pretrained.pdparams", + "HRNet_W44_C": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W44_C_pretrained.pdparams", + "HRNet_W48_C": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W48_C_pretrained.pdparams", + "HRNet_W64_C": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W64_C_pretrained.pdparams" +} + +MODEL_STAGES_PATTERN = {"HRNet": ["st4"]} + +__all__ = list(MODEL_URLS.keys()) + + +def _create_act(act): + if act == "hardswish": + return nn.Hardswish() + elif act == "relu": + return nn.ReLU() + elif act is None: + return Identity() + else: + raise RuntimeError( + "The activation function is not supported: {}".format(act)) + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + act="relu"): + super().__init__() + + self.conv = nn.Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + bias_attr=False) + self.bn = nn.BatchNorm(num_filters, act=None) + self.act = _create_act(act) + + def forward(self, x): + x = self.conv(x) + x = self.bn(x) + x = self.act(x) + return x + + +class BottleneckBlock(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + has_se, + stride=1, + downsample=False): + super().__init__() + + self.has_se = has_se + self.downsample = downsample + + self.conv1 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act="relu") + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + stride=stride, + act="relu") + self.conv3 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 4, + filter_size=1, + act=None) + + if self.downsample: + self.conv_down = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 4, + filter_size=1, + act=None) + + if self.has_se: + self.se = SELayer( + num_channels=num_filters * 4, + num_filters=num_filters * 4, + reduction_ratio=16) + self.relu = nn.ReLU() + + def forward(self, x, res_dict=None): + residual = x + x = self.conv1(x) + x = self.conv2(x) + x = self.conv3(x) + if self.downsample: + residual = self.conv_down(residual) + if self.has_se: + x = self.se(x) + x = paddle.add(x=residual, y=x) + x = self.relu(x) + return x + + +class BasicBlock(nn.Layer): + def __init__(self, num_channels, num_filters, has_se=False): + super().__init__() + + self.has_se = has_se + + self.conv1 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=3, + stride=1, + act="relu") + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + stride=1, + act=None) + + if self.has_se: + self.se = SELayer( + num_channels=num_filters, + num_filters=num_filters, + reduction_ratio=16) + self.relu = nn.ReLU() + + def forward(self, x): + residual = x + x = self.conv1(x) + x = self.conv2(x) + + if self.has_se: + x = self.se(x) + + x = paddle.add(x=residual, y=x) + x = self.relu(x) + return x + + +class SELayer(TheseusLayer): + def __init__(self, num_channels, num_filters, reduction_ratio): + super().__init__() + + self.avg_pool = nn.AdaptiveAvgPool2D(1) + + self._num_channels = num_channels + + med_ch = int(num_channels / reduction_ratio) + stdv = 1.0 / math.sqrt(num_channels * 1.0) + self.fc_squeeze = nn.Linear( + num_channels, + med_ch, + weight_attr=ParamAttr(initializer=Uniform(-stdv, stdv))) + self.relu = nn.ReLU() + stdv = 1.0 / math.sqrt(med_ch * 1.0) + self.fc_excitation = nn.Linear( + med_ch, + num_filters, + weight_attr=ParamAttr(initializer=Uniform(-stdv, stdv))) + self.sigmoid = nn.Sigmoid() + + def forward(self, x, res_dict=None): + residual = x + x = self.avg_pool(x) + x = paddle.squeeze(x, axis=[2, 3]) + x = self.fc_squeeze(x) + x = self.relu(x) + x = self.fc_excitation(x) + x = self.sigmoid(x) + x = paddle.unsqueeze(x, axis=[2, 3]) + x = residual * x + return x + + +class Stage(TheseusLayer): + def __init__(self, num_modules, num_filters, has_se=False): + super().__init__() + + self._num_modules = num_modules + + self.stage_func_list = nn.LayerList() + for i in range(num_modules): + self.stage_func_list.append( + HighResolutionModule( + num_filters=num_filters, has_se=has_se)) + + def forward(self, x, res_dict=None): + x = x + for idx in range(self._num_modules): + x = self.stage_func_list[idx](x) + return x + + +class HighResolutionModule(TheseusLayer): + def __init__(self, num_filters, has_se=False): + super().__init__() + + self.basic_block_list = nn.LayerList() + + for i in range(len(num_filters)): + self.basic_block_list.append( + nn.Sequential(* [ + BasicBlock( + num_channels=num_filters[i], + num_filters=num_filters[i], + has_se=has_se) for j in range(4) + ])) + + self.fuse_func = FuseLayers( + in_channels=num_filters, out_channels=num_filters) + + def forward(self, x, res_dict=None): + out = [] + for idx, xi in enumerate(x): + basic_block_list = self.basic_block_list[idx] + for basic_block_func in basic_block_list: + xi = basic_block_func(xi) + out.append(xi) + out = self.fuse_func(out) + return out + + +class FuseLayers(TheseusLayer): + def __init__(self, in_channels, out_channels): + super().__init__() + + self._actual_ch = len(in_channels) + self._in_channels = in_channels + + self.residual_func_list = nn.LayerList() + self.relu = nn.ReLU() + for i in range(len(in_channels)): + for j in range(len(in_channels)): + if j > i: + self.residual_func_list.append( + ConvBNLayer( + num_channels=in_channels[j], + num_filters=out_channels[i], + filter_size=1, + stride=1, + act=None)) + elif j < i: + pre_num_filters = in_channels[j] + for k in range(i - j): + if k == i - j - 1: + self.residual_func_list.append( + ConvBNLayer( + num_channels=pre_num_filters, + num_filters=out_channels[i], + filter_size=3, + stride=2, + act=None)) + pre_num_filters = out_channels[i] + else: + self.residual_func_list.append( + ConvBNLayer( + num_channels=pre_num_filters, + num_filters=out_channels[j], + filter_size=3, + stride=2, + act="relu")) + pre_num_filters = out_channels[j] + + def forward(self, x, res_dict=None): + out = [] + residual_func_idx = 0 + for i in range(len(self._in_channels)): + residual = x[i] + for j in range(len(self._in_channels)): + if j > i: + xj = self.residual_func_list[residual_func_idx](x[j]) + residual_func_idx += 1 + + xj = upsample(xj, scale_factor=2**(j - i), mode="nearest") + residual = paddle.add(x=residual, y=xj) + elif j < i: + xj = x[j] + for k in range(i - j): + xj = self.residual_func_list[residual_func_idx](xj) + residual_func_idx += 1 + + residual = paddle.add(x=residual, y=xj) + + residual = self.relu(residual) + out.append(residual) + + return out + + +class LastClsOut(TheseusLayer): + def __init__(self, + num_channel_list, + has_se, + num_filters_list=[32, 64, 128, 256]): + super().__init__() + + self.func_list = nn.LayerList() + for idx in range(len(num_channel_list)): + self.func_list.append( + BottleneckBlock( + num_channels=num_channel_list[idx], + num_filters=num_filters_list[idx], + has_se=has_se, + downsample=True)) + + def forward(self, x, res_dict=None): + out = [] + for idx, xi in enumerate(x): + xi = self.func_list[idx](xi) + out.append(xi) + return out + + +class HRNet(TheseusLayer): + """ + HRNet + Args: + width: int=18. Base channel number of HRNet. + has_se: bool=False. If 'True', add se module to HRNet. + class_num: int=1000. Output num of last fc layer. + Returns: + model: nn.Layer. Specific HRNet model depends on args. + """ + + def __init__(self, + stages_pattern, + width=18, + has_se=False, + class_num=1000, + return_patterns=None, + return_stages=None): + super().__init__() + + self.width = width + self.has_se = has_se + self._class_num = class_num + + channels_2 = [self.width, self.width * 2] + channels_3 = [self.width, self.width * 2, self.width * 4] + channels_4 = [ + self.width, self.width * 2, self.width * 4, self.width * 8 + ] + + self.conv_layer1_1 = ConvBNLayer( + num_channels=3, + num_filters=64, + filter_size=3, + stride=2, + act="relu") + + self.conv_layer1_2 = ConvBNLayer( + num_channels=64, + num_filters=64, + filter_size=3, + stride=2, + act="relu") + + self.layer1 = nn.Sequential(* [ + BottleneckBlock( + num_channels=64 if i == 0 else 256, + num_filters=64, + has_se=has_se, + stride=1, + downsample=True if i == 0 else False) for i in range(4) + ]) + + self.conv_tr1_1 = ConvBNLayer( + num_channels=256, num_filters=width, filter_size=3) + self.conv_tr1_2 = ConvBNLayer( + num_channels=256, num_filters=width * 2, filter_size=3, stride=2) + + self.st2 = Stage( + num_modules=1, num_filters=channels_2, has_se=self.has_se) + + self.conv_tr2 = ConvBNLayer( + num_channels=width * 2, + num_filters=width * 4, + filter_size=3, + stride=2) + self.st3 = Stage( + num_modules=4, num_filters=channels_3, has_se=self.has_se) + + self.conv_tr3 = ConvBNLayer( + num_channels=width * 4, + num_filters=width * 8, + filter_size=3, + stride=2) + + self.st4 = Stage( + num_modules=3, num_filters=channels_4, has_se=self.has_se) + + # classification + num_filters_list = [32, 64, 128, 256] + self.last_cls = LastClsOut( + num_channel_list=channels_4, + has_se=self.has_se, + num_filters_list=num_filters_list) + + last_num_filters = [256, 512, 1024] + self.cls_head_conv_list = nn.LayerList() + for idx in range(3): + self.cls_head_conv_list.append( + ConvBNLayer( + num_channels=num_filters_list[idx] * 4, + num_filters=last_num_filters[idx], + filter_size=3, + stride=2)) + + self.conv_last = ConvBNLayer( + num_channels=1024, num_filters=2048, filter_size=1, stride=1) + + self.avg_pool = nn.AdaptiveAvgPool2D(1) + + stdv = 1.0 / math.sqrt(2048 * 1.0) + + self.fc = nn.Linear( + 2048, + class_num, + weight_attr=ParamAttr(initializer=Uniform(-stdv, stdv))) + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, x): + x = self.conv_layer1_1(x) + x = self.conv_layer1_2(x) + + x = self.layer1(x) + + tr1_1 = self.conv_tr1_1(x) + tr1_2 = self.conv_tr1_2(x) + x = self.st2([tr1_1, tr1_2]) + + tr2 = self.conv_tr2(x[-1]) + x.append(tr2) + x = self.st3(x) + + tr3 = self.conv_tr3(x[-1]) + x.append(tr3) + x = self.st4(x) + + x = self.last_cls(x) + + y = x[0] + for idx in range(3): + y = paddle.add(x[idx + 1], self.cls_head_conv_list[idx](y)) + + y = self.conv_last(y) + y = self.avg_pool(y) + y = paddle.reshape(y, shape=[-1, y.shape[1]]) + y = self.fc(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def HRNet_W18_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W18_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W18_C` model depends on args. + """ + model = HRNet( + width=18, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W18_C"], use_ssld) + return model + + +def HRNet_W30_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W30_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W30_C` model depends on args. + """ + model = HRNet( + width=30, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W30_C"], use_ssld) + return model + + +def HRNet_W32_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W32_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W32_C` model depends on args. + """ + model = HRNet( + width=32, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W32_C"], use_ssld) + return model + + +def HRNet_W40_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W40_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W40_C` model depends on args. + """ + model = HRNet( + width=40, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W40_C"], use_ssld) + return model + + +def HRNet_W44_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W44_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W44_C` model depends on args. + """ + model = HRNet( + width=44, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W44_C"], use_ssld) + return model + + +def HRNet_W48_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W48_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W48_C` model depends on args. + """ + model = HRNet( + width=48, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W48_C"], use_ssld) + return model + + +def HRNet_W60_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W60_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W60_C` model depends on args. + """ + model = HRNet( + width=60, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W60_C"], use_ssld) + return model + + +def HRNet_W64_C(pretrained=False, use_ssld=False, **kwargs): + """ + HRNet_W64_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `HRNet_W64_C` model depends on args. + """ + model = HRNet( + width=64, stages_pattern=MODEL_STAGES_PATTERN["HRNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HRNet_W64_C"], use_ssld) + return model + + +def SE_HRNet_W18_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W18_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W18_C` model depends on args. + """ + model = HRNet( + width=18, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W18_C"], use_ssld) + return model + + +def SE_HRNet_W30_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W30_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W30_C` model depends on args. + """ + model = HRNet( + width=30, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W30_C"], use_ssld) + return model + + +def SE_HRNet_W32_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W32_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W32_C` model depends on args. + """ + model = HRNet( + width=32, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W32_C"], use_ssld) + return model + + +def SE_HRNet_W40_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W40_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W40_C` model depends on args. + """ + model = HRNet( + width=40, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W40_C"], use_ssld) + return model + + +def SE_HRNet_W44_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W44_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W44_C` model depends on args. + """ + model = HRNet( + width=44, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W44_C"], use_ssld) + return model + + +def SE_HRNet_W48_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W48_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W48_C` model depends on args. + """ + model = HRNet( + width=48, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W48_C"], use_ssld) + return model + + +def SE_HRNet_W60_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W60_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W60_C` model depends on args. + """ + model = HRNet( + width=60, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W60_C"], use_ssld) + return model + + +def SE_HRNet_W64_C(pretrained=False, use_ssld=False, **kwargs): + """ + SE_HRNet_W64_C + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `SE_HRNet_W64_C` model depends on args. + """ + model = HRNet( + width=64, + stages_pattern=MODEL_STAGES_PATTERN["HRNet"], + has_se=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["SE_HRNet_W64_C"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/inception_v3.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/inception_v3.py new file mode 100644 index 00000000..5575f8c9 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/inception_v3.py @@ -0,0 +1,557 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function +import math +import paddle +from paddle import ParamAttr +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "InceptionV3": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/InceptionV3_pretrained.pdparams" +} + +MODEL_STAGES_PATTERN = { + "InceptionV3": [ + "inception_block_list[2]", "inception_block_list[3]", + "inception_block_list[7]", "inception_block_list[8]", + "inception_block_list[10]" + ] +} + +__all__ = MODEL_URLS.keys() +''' +InceptionV3 config: dict. + key: inception blocks of InceptionV3. + values: conv num in different blocks. +''' +NET_CONFIG = { + "inception_a": [[192, 256, 288], [32, 64, 64]], + "inception_b": [288], + "inception_c": [[768, 768, 768, 768], [128, 160, 160, 192]], + "inception_d": [768], + "inception_e": [1280, 2048] +} + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + padding=0, + groups=1, + act="relu"): + super().__init__() + self.act = act + self.conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=padding, + groups=groups, + bias_attr=False) + self.bn = BatchNorm(num_filters) + self.relu = nn.ReLU() + + def forward(self, x): + x = self.conv(x) + x = self.bn(x) + if self.act: + x = self.relu(x) + return x + + +class InceptionStem(TheseusLayer): + def __init__(self): + super().__init__() + self.conv_1a_3x3 = ConvBNLayer( + num_channels=3, + num_filters=32, + filter_size=3, + stride=2, + act="relu") + self.conv_2a_3x3 = ConvBNLayer( + num_channels=32, + num_filters=32, + filter_size=3, + stride=1, + act="relu") + self.conv_2b_3x3 = ConvBNLayer( + num_channels=32, + num_filters=64, + filter_size=3, + padding=1, + act="relu") + + self.max_pool = MaxPool2D(kernel_size=3, stride=2, padding=0) + self.conv_3b_1x1 = ConvBNLayer( + num_channels=64, num_filters=80, filter_size=1, act="relu") + self.conv_4a_3x3 = ConvBNLayer( + num_channels=80, num_filters=192, filter_size=3, act="relu") + + def forward(self, x): + x = self.conv_1a_3x3(x) + x = self.conv_2a_3x3(x) + x = self.conv_2b_3x3(x) + x = self.max_pool(x) + x = self.conv_3b_1x1(x) + x = self.conv_4a_3x3(x) + x = self.max_pool(x) + return x + + +class InceptionA(TheseusLayer): + def __init__(self, num_channels, pool_features): + super().__init__() + self.branch1x1 = ConvBNLayer( + num_channels=num_channels, + num_filters=64, + filter_size=1, + act="relu") + self.branch5x5_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=48, + filter_size=1, + act="relu") + self.branch5x5_2 = ConvBNLayer( + num_channels=48, + num_filters=64, + filter_size=5, + padding=2, + act="relu") + + self.branch3x3dbl_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=64, + filter_size=1, + act="relu") + self.branch3x3dbl_2 = ConvBNLayer( + num_channels=64, + num_filters=96, + filter_size=3, + padding=1, + act="relu") + self.branch3x3dbl_3 = ConvBNLayer( + num_channels=96, + num_filters=96, + filter_size=3, + padding=1, + act="relu") + self.branch_pool = AvgPool2D( + kernel_size=3, stride=1, padding=1, exclusive=False) + self.branch_pool_conv = ConvBNLayer( + num_channels=num_channels, + num_filters=pool_features, + filter_size=1, + act="relu") + + def forward(self, x): + branch1x1 = self.branch1x1(x) + branch5x5 = self.branch5x5_1(x) + branch5x5 = self.branch5x5_2(branch5x5) + + branch3x3dbl = self.branch3x3dbl_1(x) + branch3x3dbl = self.branch3x3dbl_2(branch3x3dbl) + branch3x3dbl = self.branch3x3dbl_3(branch3x3dbl) + + branch_pool = self.branch_pool(x) + branch_pool = self.branch_pool_conv(branch_pool) + x = paddle.concat( + [branch1x1, branch5x5, branch3x3dbl, branch_pool], axis=1) + return x + + +class InceptionB(TheseusLayer): + def __init__(self, num_channels): + super().__init__() + self.branch3x3 = ConvBNLayer( + num_channels=num_channels, + num_filters=384, + filter_size=3, + stride=2, + act="relu") + self.branch3x3dbl_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=64, + filter_size=1, + act="relu") + self.branch3x3dbl_2 = ConvBNLayer( + num_channels=64, + num_filters=96, + filter_size=3, + padding=1, + act="relu") + self.branch3x3dbl_3 = ConvBNLayer( + num_channels=96, + num_filters=96, + filter_size=3, + stride=2, + act="relu") + self.branch_pool = MaxPool2D(kernel_size=3, stride=2) + + def forward(self, x): + branch3x3 = self.branch3x3(x) + + branch3x3dbl = self.branch3x3dbl_1(x) + branch3x3dbl = self.branch3x3dbl_2(branch3x3dbl) + branch3x3dbl = self.branch3x3dbl_3(branch3x3dbl) + + branch_pool = self.branch_pool(x) + + x = paddle.concat([branch3x3, branch3x3dbl, branch_pool], axis=1) + + return x + + +class InceptionC(TheseusLayer): + def __init__(self, num_channels, channels_7x7): + super().__init__() + self.branch1x1 = ConvBNLayer( + num_channels=num_channels, + num_filters=192, + filter_size=1, + act="relu") + + self.branch7x7_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=channels_7x7, + filter_size=1, + stride=1, + act="relu") + self.branch7x7_2 = ConvBNLayer( + num_channels=channels_7x7, + num_filters=channels_7x7, + filter_size=(1, 7), + stride=1, + padding=(0, 3), + act="relu") + self.branch7x7_3 = ConvBNLayer( + num_channels=channels_7x7, + num_filters=192, + filter_size=(7, 1), + stride=1, + padding=(3, 0), + act="relu") + + self.branch7x7dbl_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=channels_7x7, + filter_size=1, + act="relu") + self.branch7x7dbl_2 = ConvBNLayer( + num_channels=channels_7x7, + num_filters=channels_7x7, + filter_size=(7, 1), + padding=(3, 0), + act="relu") + self.branch7x7dbl_3 = ConvBNLayer( + num_channels=channels_7x7, + num_filters=channels_7x7, + filter_size=(1, 7), + padding=(0, 3), + act="relu") + self.branch7x7dbl_4 = ConvBNLayer( + num_channels=channels_7x7, + num_filters=channels_7x7, + filter_size=(7, 1), + padding=(3, 0), + act="relu") + self.branch7x7dbl_5 = ConvBNLayer( + num_channels=channels_7x7, + num_filters=192, + filter_size=(1, 7), + padding=(0, 3), + act="relu") + + self.branch_pool = AvgPool2D( + kernel_size=3, stride=1, padding=1, exclusive=False) + self.branch_pool_conv = ConvBNLayer( + num_channels=num_channels, + num_filters=192, + filter_size=1, + act="relu") + + def forward(self, x): + branch1x1 = self.branch1x1(x) + + branch7x7 = self.branch7x7_1(x) + branch7x7 = self.branch7x7_2(branch7x7) + branch7x7 = self.branch7x7_3(branch7x7) + + branch7x7dbl = self.branch7x7dbl_1(x) + branch7x7dbl = self.branch7x7dbl_2(branch7x7dbl) + branch7x7dbl = self.branch7x7dbl_3(branch7x7dbl) + branch7x7dbl = self.branch7x7dbl_4(branch7x7dbl) + branch7x7dbl = self.branch7x7dbl_5(branch7x7dbl) + + branch_pool = self.branch_pool(x) + branch_pool = self.branch_pool_conv(branch_pool) + + x = paddle.concat( + [branch1x1, branch7x7, branch7x7dbl, branch_pool], axis=1) + + return x + + +class InceptionD(TheseusLayer): + def __init__(self, num_channels): + super().__init__() + self.branch3x3_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=192, + filter_size=1, + act="relu") + self.branch3x3_2 = ConvBNLayer( + num_channels=192, + num_filters=320, + filter_size=3, + stride=2, + act="relu") + self.branch7x7x3_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=192, + filter_size=1, + act="relu") + self.branch7x7x3_2 = ConvBNLayer( + num_channels=192, + num_filters=192, + filter_size=(1, 7), + padding=(0, 3), + act="relu") + self.branch7x7x3_3 = ConvBNLayer( + num_channels=192, + num_filters=192, + filter_size=(7, 1), + padding=(3, 0), + act="relu") + self.branch7x7x3_4 = ConvBNLayer( + num_channels=192, + num_filters=192, + filter_size=3, + stride=2, + act="relu") + self.branch_pool = MaxPool2D(kernel_size=3, stride=2) + + def forward(self, x): + branch3x3 = self.branch3x3_1(x) + branch3x3 = self.branch3x3_2(branch3x3) + + branch7x7x3 = self.branch7x7x3_1(x) + branch7x7x3 = self.branch7x7x3_2(branch7x7x3) + branch7x7x3 = self.branch7x7x3_3(branch7x7x3) + branch7x7x3 = self.branch7x7x3_4(branch7x7x3) + + branch_pool = self.branch_pool(x) + + x = paddle.concat([branch3x3, branch7x7x3, branch_pool], axis=1) + return x + + +class InceptionE(TheseusLayer): + def __init__(self, num_channels): + super().__init__() + self.branch1x1 = ConvBNLayer( + num_channels=num_channels, + num_filters=320, + filter_size=1, + act="relu") + self.branch3x3_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=384, + filter_size=1, + act="relu") + self.branch3x3_2a = ConvBNLayer( + num_channels=384, + num_filters=384, + filter_size=(1, 3), + padding=(0, 1), + act="relu") + self.branch3x3_2b = ConvBNLayer( + num_channels=384, + num_filters=384, + filter_size=(3, 1), + padding=(1, 0), + act="relu") + + self.branch3x3dbl_1 = ConvBNLayer( + num_channels=num_channels, + num_filters=448, + filter_size=1, + act="relu") + self.branch3x3dbl_2 = ConvBNLayer( + num_channels=448, + num_filters=384, + filter_size=3, + padding=1, + act="relu") + self.branch3x3dbl_3a = ConvBNLayer( + num_channels=384, + num_filters=384, + filter_size=(1, 3), + padding=(0, 1), + act="relu") + self.branch3x3dbl_3b = ConvBNLayer( + num_channels=384, + num_filters=384, + filter_size=(3, 1), + padding=(1, 0), + act="relu") + self.branch_pool = AvgPool2D( + kernel_size=3, stride=1, padding=1, exclusive=False) + self.branch_pool_conv = ConvBNLayer( + num_channels=num_channels, + num_filters=192, + filter_size=1, + act="relu") + + def forward(self, x): + branch1x1 = self.branch1x1(x) + + branch3x3 = self.branch3x3_1(x) + branch3x3 = [ + self.branch3x3_2a(branch3x3), + self.branch3x3_2b(branch3x3), + ] + branch3x3 = paddle.concat(branch3x3, axis=1) + + branch3x3dbl = self.branch3x3dbl_1(x) + branch3x3dbl = self.branch3x3dbl_2(branch3x3dbl) + branch3x3dbl = [ + self.branch3x3dbl_3a(branch3x3dbl), + self.branch3x3dbl_3b(branch3x3dbl), + ] + branch3x3dbl = paddle.concat(branch3x3dbl, axis=1) + + branch_pool = self.branch_pool(x) + branch_pool = self.branch_pool_conv(branch_pool) + + x = paddle.concat( + [branch1x1, branch3x3, branch3x3dbl, branch_pool], axis=1) + return x + + +class Inception_V3(TheseusLayer): + """ + Inception_V3 + Args: + config: dict. config of Inception_V3. + class_num: int=1000. The number of classes. + pretrained: (True or False) or path of pretrained_model. Whether to load the pretrained model. + Returns: + model: nn.Layer. Specific Inception_V3 model depends on args. + """ + + def __init__(self, + config, + stages_pattern, + class_num=1000, + return_patterns=None, + return_stages=None): + super().__init__() + + self.inception_a_list = config["inception_a"] + self.inception_c_list = config["inception_c"] + self.inception_b_list = config["inception_b"] + self.inception_d_list = config["inception_d"] + self.inception_e_list = config["inception_e"] + + self.inception_stem = InceptionStem() + + self.inception_block_list = nn.LayerList() + for i in range(len(self.inception_a_list[0])): + inception_a = InceptionA(self.inception_a_list[0][i], + self.inception_a_list[1][i]) + self.inception_block_list.append(inception_a) + + for i in range(len(self.inception_b_list)): + inception_b = InceptionB(self.inception_b_list[i]) + self.inception_block_list.append(inception_b) + + for i in range(len(self.inception_c_list[0])): + inception_c = InceptionC(self.inception_c_list[0][i], + self.inception_c_list[1][i]) + self.inception_block_list.append(inception_c) + + for i in range(len(self.inception_d_list)): + inception_d = InceptionD(self.inception_d_list[i]) + self.inception_block_list.append(inception_d) + + for i in range(len(self.inception_e_list)): + inception_e = InceptionE(self.inception_e_list[i]) + self.inception_block_list.append(inception_e) + + self.avg_pool = AdaptiveAvgPool2D(1) + self.dropout = Dropout(p=0.2, mode="downscale_in_infer") + stdv = 1.0 / math.sqrt(2048 * 1.0) + self.fc = Linear( + 2048, + class_num, + weight_attr=ParamAttr(initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr()) + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, x): + x = self.inception_stem(x) + for inception_block in self.inception_block_list: + x = inception_block(x) + x = self.avg_pool(x) + x = paddle.reshape(x, shape=[-1, 2048]) + x = self.dropout(x) + x = self.fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def InceptionV3(pretrained=False, use_ssld=False, **kwargs): + """ + InceptionV3 + Args: + pretrained: bool=false or str. if `true` load pretrained parameters, `false` otherwise. + if str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `InceptionV3` model + """ + model = Inception_V3( + NET_CONFIG, + stages_pattern=MODEL_STAGES_PATTERN["InceptionV3"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["InceptionV3"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/mobilenet_v1.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/mobilenet_v1.py new file mode 100644 index 00000000..9767d69b --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/mobilenet_v1.py @@ -0,0 +1,257 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +from paddle import ParamAttr +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear, ReLU, Flatten +from paddle.nn import AdaptiveAvgPool2D +from paddle.nn.initializer import KaimingNormal + +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "MobileNetV1_x0_25": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_25_pretrained.pdparams", + "MobileNetV1_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_5_pretrained.pdparams", + "MobileNetV1_x0_75": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_75_pretrained.pdparams", + "MobileNetV1": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_pretrained.pdparams" +} + +MODEL_STAGES_PATTERN = { + "MobileNetV1": ["blocks[0]", "blocks[2]", "blocks[4]", "blocks[10]"] +} + +__all__ = MODEL_URLS.keys() + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + num_channels, + filter_size, + num_filters, + stride, + padding, + num_groups=1): + super().__init__() + + self.conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=padding, + groups=num_groups, + weight_attr=ParamAttr(initializer=KaimingNormal()), + bias_attr=False) + self.bn = BatchNorm(num_filters) + self.relu = ReLU() + + def forward(self, x): + x = self.conv(x) + x = self.bn(x) + x = self.relu(x) + return x + + +class DepthwiseSeparable(TheseusLayer): + def __init__(self, num_channels, num_filters1, num_filters2, num_groups, + stride, scale): + super().__init__() + + self.depthwise_conv = ConvBNLayer( + num_channels=num_channels, + num_filters=int(num_filters1 * scale), + filter_size=3, + stride=stride, + padding=1, + num_groups=int(num_groups * scale)) + + self.pointwise_conv = ConvBNLayer( + num_channels=int(num_filters1 * scale), + filter_size=1, + num_filters=int(num_filters2 * scale), + stride=1, + padding=0) + + def forward(self, x): + x = self.depthwise_conv(x) + x = self.pointwise_conv(x) + return x + + +class MobileNet(TheseusLayer): + """ + MobileNet + Args: + scale: float=1.0. The coefficient that controls the size of network parameters. + class_num: int=1000. The number of classes. + Returns: + model: nn.Layer. Specific MobileNet model depends on args. + """ + + def __init__(self, + stages_pattern, + scale=1.0, + class_num=1000, + return_patterns=None, + return_stages=None): + super().__init__() + self.scale = scale + + self.conv = ConvBNLayer( + num_channels=3, + filter_size=3, + num_filters=int(32 * scale), + stride=2, + padding=1) + + #num_channels, num_filters1, num_filters2, num_groups, stride + self.cfg = [[int(32 * scale), 32, 64, 32, 1], + [int(64 * scale), 64, 128, 64, 2], + [int(128 * scale), 128, 128, 128, 1], + [int(128 * scale), 128, 256, 128, 2], + [int(256 * scale), 256, 256, 256, 1], + [int(256 * scale), 256, 512, 256, 2], + [int(512 * scale), 512, 512, 512, 1], + [int(512 * scale), 512, 512, 512, 1], + [int(512 * scale), 512, 512, 512, 1], + [int(512 * scale), 512, 512, 512, 1], + [int(512 * scale), 512, 512, 512, 1], + [int(512 * scale), 512, 1024, 512, 2], + [int(1024 * scale), 1024, 1024, 1024, 1]] + + self.blocks = nn.Sequential(* [ + DepthwiseSeparable( + num_channels=params[0], + num_filters1=params[1], + num_filters2=params[2], + num_groups=params[3], + stride=params[4], + scale=scale) for params in self.cfg + ]) + + self.avg_pool = AdaptiveAvgPool2D(1) + self.flatten = Flatten(start_axis=1, stop_axis=-1) + + self.fc = Linear( + int(1024 * scale), + class_num, + weight_attr=ParamAttr(initializer=KaimingNormal())) + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, x): + x = self.conv(x) + x = self.blocks(x) + x = self.avg_pool(x) + x = self.flatten(x) + x = self.fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def MobileNetV1_x0_25(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV1_x0_25 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV1_x0_25` model depends on args. + """ + model = MobileNet( + scale=0.25, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV1"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV1_x0_25"], + use_ssld) + return model + + +def MobileNetV1_x0_5(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV1_x0_5 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV1_x0_5` model depends on args. + """ + model = MobileNet( + scale=0.5, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV1"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV1_x0_5"], + use_ssld) + return model + + +def MobileNetV1_x0_75(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV1_x0_75 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV1_x0_75` model depends on args. + """ + model = MobileNet( + scale=0.75, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV1"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV1_x0_75"], + use_ssld) + return model + + +def MobileNetV1(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV1 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV1` model depends on args. + """ + model = MobileNet( + scale=1.0, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV1"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV1"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/mobilenet_v3.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/mobilenet_v3.py new file mode 100644 index 00000000..836c54cd --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/mobilenet_v3.py @@ -0,0 +1,586 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import paddle +import paddle.nn as nn +from paddle import ParamAttr +from paddle.nn import AdaptiveAvgPool2D, BatchNorm, Conv2D, Dropout, Linear +from paddle.regularizer import L2Decay +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "MobileNetV3_small_x0_35": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_35_pretrained.pdparams", + "MobileNetV3_small_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_5_pretrained.pdparams", + "MobileNetV3_small_x0_75": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_75_pretrained.pdparams", + "MobileNetV3_small_x1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_0_pretrained.pdparams", + "MobileNetV3_small_x1_25": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_25_pretrained.pdparams", + "MobileNetV3_large_x0_35": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_35_pretrained.pdparams", + "MobileNetV3_large_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_5_pretrained.pdparams", + "MobileNetV3_large_x0_75": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_75_pretrained.pdparams", + "MobileNetV3_large_x1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_0_pretrained.pdparams", + "MobileNetV3_large_x1_25": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_25_pretrained.pdparams", +} + +MODEL_STAGES_PATTERN = { + "MobileNetV3_small": + ["blocks[0]", "blocks[2]", "blocks[7]", "blocks[10]"], + "MobileNetV3_large": + ["blocks[0]", "blocks[2]", "blocks[5]", "blocks[11]", "blocks[14]"] +} + +__all__ = MODEL_URLS.keys() + +# "large", "small" is just for MobinetV3_large, MobileNetV3_small respectively. +# The type of "large" or "small" config is a list. Each element(list) represents a depthwise block, which is composed of k, exp, se, act, s. +# k: kernel_size +# exp: middle channel number in depthwise block +# c: output channel number in depthwise block +# se: whether to use SE block +# act: which activation to use +# s: stride in depthwise block +NET_CONFIG = { + "large": [ + # k, exp, c, se, act, s + [3, 16, 16, False, "relu", 1], + [3, 64, 24, False, "relu", 2], + [3, 72, 24, False, "relu", 1], + [5, 72, 40, True, "relu", 2], + [5, 120, 40, True, "relu", 1], + [5, 120, 40, True, "relu", 1], + [3, 240, 80, False, "hardswish", 2], + [3, 200, 80, False, "hardswish", 1], + [3, 184, 80, False, "hardswish", 1], + [3, 184, 80, False, "hardswish", 1], + [3, 480, 112, True, "hardswish", 1], + [3, 672, 112, True, "hardswish", 1], + [5, 672, 160, True, "hardswish", 2], + [5, 960, 160, True, "hardswish", 1], + [5, 960, 160, True, "hardswish", 1], + ], + "small": [ + # k, exp, c, se, act, s + [3, 16, 16, True, "relu", 2], + [3, 72, 24, False, "relu", 2], + [3, 88, 24, False, "relu", 1], + [5, 96, 40, True, "hardswish", 2], + [5, 240, 40, True, "hardswish", 1], + [5, 240, 40, True, "hardswish", 1], + [5, 120, 48, True, "hardswish", 1], + [5, 144, 48, True, "hardswish", 1], + [5, 288, 96, True, "hardswish", 2], + [5, 576, 96, True, "hardswish", 1], + [5, 576, 96, True, "hardswish", 1], + ] +} +# first conv output channel number in MobileNetV3 +STEM_CONV_NUMBER = 16 +# last second conv output channel for "small" +LAST_SECOND_CONV_SMALL = 576 +# last second conv output channel for "large" +LAST_SECOND_CONV_LARGE = 960 +# last conv output channel number for "large" and "small" +LAST_CONV = 1280 + + +def _make_divisible(v, divisor=8, min_value=None): + if min_value is None: + min_value = divisor + new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) + if new_v < 0.9 * v: + new_v += divisor + return new_v + + +def _create_act(act): + if act == "hardswish": + return nn.Hardswish() + elif act == "relu": + return nn.ReLU() + elif act is None: + return None + else: + raise RuntimeError( + "The activation function is not supported: {}".format(act)) + + +class MobileNetV3(TheseusLayer): + """ + MobileNetV3 + Args: + config: list. MobileNetV3 depthwise blocks config. + scale: float=1.0. The coefficient that controls the size of network parameters. + class_num: int=1000. The number of classes. + inplanes: int=16. The output channel number of first convolution layer. + class_squeeze: int=960. The output channel number of penultimate convolution layer. + class_expand: int=1280. The output channel number of last convolution layer. + dropout_prob: float=0.2. Probability of setting units to zero. + Returns: + model: nn.Layer. Specific MobileNetV3 model depends on args. + """ + + def __init__(self, + config, + stages_pattern, + scale=1.0, + class_num=1000, + inplanes=STEM_CONV_NUMBER, + class_squeeze=LAST_SECOND_CONV_LARGE, + class_expand=LAST_CONV, + dropout_prob=0.2, + return_patterns=None, + return_stages=None): + super().__init__() + + self.cfg = config + self.scale = scale + self.inplanes = inplanes + self.class_squeeze = class_squeeze + self.class_expand = class_expand + self.class_num = class_num + + self.conv = ConvBNLayer( + in_c=3, + out_c=_make_divisible(self.inplanes * self.scale), + filter_size=3, + stride=2, + padding=1, + num_groups=1, + if_act=True, + act="hardswish") + + self.blocks = nn.Sequential(* [ + ResidualUnit( + in_c=_make_divisible(self.inplanes * self.scale if i == 0 else + self.cfg[i - 1][2] * self.scale), + mid_c=_make_divisible(self.scale * exp), + out_c=_make_divisible(self.scale * c), + filter_size=k, + stride=s, + use_se=se, + act=act) for i, (k, exp, c, se, act, s) in enumerate(self.cfg) + ]) + + self.last_second_conv = ConvBNLayer( + in_c=_make_divisible(self.cfg[-1][2] * self.scale), + out_c=_make_divisible(self.scale * self.class_squeeze), + filter_size=1, + stride=1, + padding=0, + num_groups=1, + if_act=True, + act="hardswish") + + self.avg_pool = AdaptiveAvgPool2D(1) + + self.last_conv = Conv2D( + in_channels=_make_divisible(self.scale * self.class_squeeze), + out_channels=self.class_expand, + kernel_size=1, + stride=1, + padding=0, + bias_attr=False) + + self.hardswish = nn.Hardswish() + if dropout_prob is not None: + self.dropout = Dropout(p=dropout_prob, mode="downscale_in_infer") + else: + self.dropout = None + self.flatten = nn.Flatten(start_axis=1, stop_axis=-1) + + self.fc = Linear(self.class_expand, class_num) + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, x): + x = self.conv(x) + x = self.blocks(x) + x = self.last_second_conv(x) + x = self.avg_pool(x) + x = self.last_conv(x) + x = self.hardswish(x) + if self.dropout is not None: + x = self.dropout(x) + x = self.flatten(x) + x = self.fc(x) + + return x + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + in_c, + out_c, + filter_size, + stride, + padding, + num_groups=1, + if_act=True, + act=None): + super().__init__() + + self.conv = Conv2D( + in_channels=in_c, + out_channels=out_c, + kernel_size=filter_size, + stride=stride, + padding=padding, + groups=num_groups, + bias_attr=False) + self.bn = BatchNorm( + num_channels=out_c, + act=None, + param_attr=ParamAttr(regularizer=L2Decay(0.0)), + bias_attr=ParamAttr(regularizer=L2Decay(0.0))) + self.if_act = if_act + self.act = _create_act(act) + + def forward(self, x): + x = self.conv(x) + x = self.bn(x) + if self.if_act: + x = self.act(x) + return x + + +class ResidualUnit(TheseusLayer): + def __init__(self, + in_c, + mid_c, + out_c, + filter_size, + stride, + use_se, + act=None): + super().__init__() + self.if_shortcut = stride == 1 and in_c == out_c + self.if_se = use_se + + self.expand_conv = ConvBNLayer( + in_c=in_c, + out_c=mid_c, + filter_size=1, + stride=1, + padding=0, + if_act=True, + act=act) + self.bottleneck_conv = ConvBNLayer( + in_c=mid_c, + out_c=mid_c, + filter_size=filter_size, + stride=stride, + padding=int((filter_size - 1) // 2), + num_groups=mid_c, + if_act=True, + act=act) + if self.if_se: + self.mid_se = SEModule(mid_c) + self.linear_conv = ConvBNLayer( + in_c=mid_c, + out_c=out_c, + filter_size=1, + stride=1, + padding=0, + if_act=False, + act=None) + + def forward(self, x): + identity = x + x = self.expand_conv(x) + x = self.bottleneck_conv(x) + if self.if_se: + x = self.mid_se(x) + x = self.linear_conv(x) + if self.if_shortcut: + x = paddle.add(identity, x) + return x + + +# nn.Hardsigmoid can't transfer "slope" and "offset" in nn.functional.hardsigmoid +class Hardsigmoid(TheseusLayer): + def __init__(self, slope=0.2, offset=0.5): + super().__init__() + self.slope = slope + self.offset = offset + + def forward(self, x): + return nn.functional.hardsigmoid( + x, slope=self.slope, offset=self.offset) + + +class SEModule(TheseusLayer): + def __init__(self, channel, reduction=4): + super().__init__() + self.avg_pool = AdaptiveAvgPool2D(1) + self.conv1 = Conv2D( + in_channels=channel, + out_channels=channel // reduction, + kernel_size=1, + stride=1, + padding=0) + self.relu = nn.ReLU() + self.conv2 = Conv2D( + in_channels=channel // reduction, + out_channels=channel, + kernel_size=1, + stride=1, + padding=0) + self.hardsigmoid = Hardsigmoid(slope=0.2, offset=0.5) + + def forward(self, x): + identity = x + x = self.avg_pool(x) + x = self.conv1(x) + x = self.relu(x) + x = self.conv2(x) + x = self.hardsigmoid(x) + return paddle.multiply(x=identity, y=x) + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def MobileNetV3_small_x0_35(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_small_x0_35 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x0_35` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["small"], + scale=0.35, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_small"], + class_squeeze=LAST_SECOND_CONV_SMALL, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_small_x0_35"], + use_ssld) + return model + + +def MobileNetV3_small_x0_5(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_small_x0_5 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x0_5` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["small"], + scale=0.5, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_small"], + class_squeeze=LAST_SECOND_CONV_SMALL, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_small_x0_5"], + use_ssld) + return model + + +def MobileNetV3_small_x0_75(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_small_x0_75 + Args: + pretrained: bool=false or str. if `true` load pretrained parameters, `false` otherwise. + if str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x0_75` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["small"], + scale=0.75, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_small"], + class_squeeze=LAST_SECOND_CONV_SMALL, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_small_x0_75"], + use_ssld) + return model + + +def MobileNetV3_small_x1_0(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_small_x1_0 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x1_0` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["small"], + scale=1.0, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_small"], + class_squeeze=LAST_SECOND_CONV_SMALL, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_small_x1_0"], + use_ssld) + return model + + +def MobileNetV3_small_x1_25(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_small_x1_25 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_small_x1_25` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["small"], + scale=1.25, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_small"], + class_squeeze=LAST_SECOND_CONV_SMALL, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_small_x1_25"], + use_ssld) + return model + + +def MobileNetV3_large_x0_35(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_large_x0_35 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x0_35` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["large"], + scale=0.35, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_small"], + class_squeeze=LAST_SECOND_CONV_LARGE, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_large_x0_35"], + use_ssld) + return model + + +def MobileNetV3_large_x0_5(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_large_x0_5 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x0_5` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["large"], + scale=0.5, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_large"], + class_squeeze=LAST_SECOND_CONV_LARGE, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_large_x0_5"], + use_ssld) + return model + + +def MobileNetV3_large_x0_75(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_large_x0_75 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x0_75` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["large"], + scale=0.75, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_large"], + class_squeeze=LAST_SECOND_CONV_LARGE, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_large_x0_75"], + use_ssld) + return model + + +def MobileNetV3_large_x1_0(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_large_x1_0 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x1_0` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["large"], + scale=1.0, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_large"], + class_squeeze=LAST_SECOND_CONV_LARGE, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_large_x1_0"], + use_ssld) + return model + + +def MobileNetV3_large_x1_25(pretrained=False, use_ssld=False, **kwargs): + """ + MobileNetV3_large_x1_25 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `MobileNetV3_large_x1_25` model depends on args. + """ + model = MobileNetV3( + config=NET_CONFIG["large"], + scale=1.25, + stages_pattern=MODEL_STAGES_PATTERN["MobileNetV3_large"], + class_squeeze=LAST_SECOND_CONV_LARGE, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["MobileNetV3_large_x1_25"], + use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/pp_lcnet.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/pp_lcnet.py new file mode 100644 index 00000000..40174622 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/pp_lcnet.py @@ -0,0 +1,419 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import paddle +import paddle.nn as nn +from paddle import ParamAttr +from paddle.nn import AdaptiveAvgPool2D, BatchNorm, Conv2D, Dropout, Linear +from paddle.regularizer import L2Decay +from paddle.nn.initializer import KaimingNormal +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "PPLCNet_x0_25": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_25_pretrained.pdparams", + "PPLCNet_x0_35": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_35_pretrained.pdparams", + "PPLCNet_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_pretrained.pdparams", + "PPLCNet_x0_75": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_75_pretrained.pdparams", + "PPLCNet_x1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_pretrained.pdparams", + "PPLCNet_x1_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_5_pretrained.pdparams", + "PPLCNet_x2_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_0_pretrained.pdparams", + "PPLCNet_x2_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_pretrained.pdparams" +} + +MODEL_STAGES_PATTERN = { + "PPLCNet": ["blocks2", "blocks3", "blocks4", "blocks5", "blocks6"] +} + +__all__ = list(MODEL_URLS.keys()) + +# Each element(list) represents a depthwise block, which is composed of k, in_c, out_c, s, use_se. +# k: kernel_size +# in_c: input channel number in depthwise block +# out_c: output channel number in depthwise block +# s: stride in depthwise block +# use_se: whether to use SE block + +NET_CONFIG = { + "blocks2": + #k, in_c, out_c, s, use_se + [[3, 16, 32, 1, False]], + "blocks3": [[3, 32, 64, 2, False], [3, 64, 64, 1, False]], + "blocks4": [[3, 64, 128, 2, False], [3, 128, 128, 1, False]], + "blocks5": [[3, 128, 256, 2, False], [5, 256, 256, 1, False], + [5, 256, 256, 1, False], [5, 256, 256, 1, False], + [5, 256, 256, 1, False], [5, 256, 256, 1, False]], + "blocks6": [[5, 256, 512, 2, True], [5, 512, 512, 1, True]] +} + + +def make_divisible(v, divisor=8, min_value=None): + if min_value is None: + min_value = divisor + new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) + if new_v < 0.9 * v: + new_v += divisor + return new_v + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + num_channels, + filter_size, + num_filters, + stride, + num_groups=1): + super().__init__() + + self.conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=num_groups, + weight_attr=ParamAttr(initializer=KaimingNormal()), + bias_attr=False) + + self.bn = BatchNorm( + num_filters, + param_attr=ParamAttr(regularizer=L2Decay(0.0)), + bias_attr=ParamAttr(regularizer=L2Decay(0.0))) + self.hardswish = nn.Hardswish() + + def forward(self, x): + x = self.conv(x) + x = self.bn(x) + x = self.hardswish(x) + return x + + +class DepthwiseSeparable(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + stride, + dw_size=3, + use_se=False): + super().__init__() + self.use_se = use_se + self.dw_conv = ConvBNLayer( + num_channels=num_channels, + num_filters=num_channels, + filter_size=dw_size, + stride=stride, + num_groups=num_channels) + if use_se: + self.se = SEModule(num_channels) + self.pw_conv = ConvBNLayer( + num_channels=num_channels, + filter_size=1, + num_filters=num_filters, + stride=1) + + def forward(self, x): + x = self.dw_conv(x) + if self.use_se: + x = self.se(x) + x = self.pw_conv(x) + return x + + +class SEModule(TheseusLayer): + def __init__(self, channel, reduction=4): + super().__init__() + self.avg_pool = AdaptiveAvgPool2D(1) + self.conv1 = Conv2D( + in_channels=channel, + out_channels=channel // reduction, + kernel_size=1, + stride=1, + padding=0) + self.relu = nn.ReLU() + self.conv2 = Conv2D( + in_channels=channel // reduction, + out_channels=channel, + kernel_size=1, + stride=1, + padding=0) + self.hardsigmoid = nn.Hardsigmoid() + + def forward(self, x): + identity = x + x = self.avg_pool(x) + x = self.conv1(x) + x = self.relu(x) + x = self.conv2(x) + x = self.hardsigmoid(x) + x = paddle.multiply(x=identity, y=x) + return x + + +class PPLCNet(TheseusLayer): + def __init__(self, + stages_pattern, + scale=1.0, + class_num=1000, + dropout_prob=0.2, + class_expand=1280, + return_patterns=None, + return_stages=None): + super().__init__() + self.scale = scale + self.class_expand = class_expand + + self.conv1 = ConvBNLayer( + num_channels=3, + filter_size=3, + num_filters=make_divisible(16 * scale), + stride=2) + + self.blocks2 = nn.Sequential(* [ + DepthwiseSeparable( + num_channels=make_divisible(in_c * scale), + num_filters=make_divisible(out_c * scale), + dw_size=k, + stride=s, + use_se=se) + for i, (k, in_c, out_c, s, se) in enumerate(NET_CONFIG["blocks2"]) + ]) + + self.blocks3 = nn.Sequential(* [ + DepthwiseSeparable( + num_channels=make_divisible(in_c * scale), + num_filters=make_divisible(out_c * scale), + dw_size=k, + stride=s, + use_se=se) + for i, (k, in_c, out_c, s, se) in enumerate(NET_CONFIG["blocks3"]) + ]) + + self.blocks4 = nn.Sequential(* [ + DepthwiseSeparable( + num_channels=make_divisible(in_c * scale), + num_filters=make_divisible(out_c * scale), + dw_size=k, + stride=s, + use_se=se) + for i, (k, in_c, out_c, s, se) in enumerate(NET_CONFIG["blocks4"]) + ]) + + self.blocks5 = nn.Sequential(* [ + DepthwiseSeparable( + num_channels=make_divisible(in_c * scale), + num_filters=make_divisible(out_c * scale), + dw_size=k, + stride=s, + use_se=se) + for i, (k, in_c, out_c, s, se) in enumerate(NET_CONFIG["blocks5"]) + ]) + + self.blocks6 = nn.Sequential(* [ + DepthwiseSeparable( + num_channels=make_divisible(in_c * scale), + num_filters=make_divisible(out_c * scale), + dw_size=k, + stride=s, + use_se=se) + for i, (k, in_c, out_c, s, se) in enumerate(NET_CONFIG["blocks6"]) + ]) + + self.avg_pool = AdaptiveAvgPool2D(1) + + self.last_conv = Conv2D( + in_channels=make_divisible(NET_CONFIG["blocks6"][-1][2] * scale), + out_channels=self.class_expand, + kernel_size=1, + stride=1, + padding=0, + bias_attr=False) + + self.hardswish = nn.Hardswish() + self.dropout = Dropout(p=dropout_prob, mode="downscale_in_infer") + self.flatten = nn.Flatten(start_axis=1, stop_axis=-1) + + self.fc = Linear(self.class_expand, class_num) + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, x): + x = self.conv1(x) + + x = self.blocks2(x) + x = self.blocks3(x) + x = self.blocks4(x) + x = self.blocks5(x) + x = self.blocks6(x) + + x = self.avg_pool(x) + x = self.last_conv(x) + x = self.hardswish(x) + x = self.dropout(x) + x = self.flatten(x) + x = self.fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def PPLCNet_x0_25(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x0_25 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x0_25` model depends on args. + """ + model = PPLCNet( + scale=0.25, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x0_25"], use_ssld) + return model + + +def PPLCNet_x0_35(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x0_35 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x0_35` model depends on args. + """ + model = PPLCNet( + scale=0.35, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x0_35"], use_ssld) + return model + + +def PPLCNet_x0_5(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x0_5 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x0_5` model depends on args. + """ + model = PPLCNet( + scale=0.5, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x0_5"], use_ssld) + return model + + +def PPLCNet_x0_75(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x0_75 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x0_75` model depends on args. + """ + model = PPLCNet( + scale=0.75, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x0_75"], use_ssld) + return model + + +def PPLCNet_x1_0(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x1_0 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x1_0` model depends on args. + """ + model = PPLCNet( + scale=1.0, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x1_0"], use_ssld) + return model + + +def PPLCNet_x1_5(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x1_5 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x1_5` model depends on args. + """ + model = PPLCNet( + scale=1.5, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x1_5"], use_ssld) + return model + + +def PPLCNet_x2_0(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x2_0 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x2_0` model depends on args. + """ + model = PPLCNet( + scale=2.0, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x2_0"], use_ssld) + return model + + +def PPLCNet_x2_5(pretrained=False, use_ssld=False, **kwargs): + """ + PPLCNet_x2_5 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `PPLCNet_x2_5` model depends on args. + """ + model = PPLCNet( + scale=2.5, stages_pattern=MODEL_STAGES_PATTERN["PPLCNet"], **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["PPLCNet_x2_5"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/resnet.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/resnet.py new file mode 100644 index 00000000..74c5c5fa --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/resnet.py @@ -0,0 +1,591 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform +import math + +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ResNet18": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_pretrained.pdparams", + "ResNet18_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_vd_pretrained.pdparams", + "ResNet34": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_pretrained.pdparams", + "ResNet34_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_pretrained.pdparams", + "ResNet50": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_pretrained.pdparams", + "ResNet50_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_pretrained.pdparams", + "ResNet101": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_pretrained.pdparams", + "ResNet101_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_vd_pretrained.pdparams", + "ResNet152": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_pretrained.pdparams", + "ResNet152_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_vd_pretrained.pdparams", + "ResNet200_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet200_vd_pretrained.pdparams", +} + +MODEL_STAGES_PATTERN = { + "ResNet18": ["blocks[1]", "blocks[3]", "blocks[5]", "blocks[7]"], + "ResNet34": ["blocks[2]", "blocks[6]", "blocks[12]", "blocks[15]"], + "ResNet50": ["blocks[2]", "blocks[6]", "blocks[12]", "blocks[15]"], + "ResNet101": ["blocks[2]", "blocks[6]", "blocks[29]", "blocks[32]"], + "ResNet152": ["blocks[2]", "blocks[10]", "blocks[46]", "blocks[49]"], + "ResNet200": ["blocks[2]", "blocks[14]", "blocks[62]", "blocks[65]"] +} + +__all__ = MODEL_URLS.keys() +''' +ResNet config: dict. + key: depth of ResNet. + values: config's dict of specific model. + keys: + block_type: Two different blocks in ResNet, BasicBlock and BottleneckBlock are optional. + block_depth: The number of blocks in different stages in ResNet. + num_channels: The number of channels to enter the next stage. +''' +NET_CONFIG = { + "18": { + "block_type": "BasicBlock", + "block_depth": [2, 2, 2, 2], + "num_channels": [64, 64, 128, 256] + }, + "34": { + "block_type": "BasicBlock", + "block_depth": [3, 4, 6, 3], + "num_channels": [64, 64, 128, 256] + }, + "50": { + "block_type": "BottleneckBlock", + "block_depth": [3, 4, 6, 3], + "num_channels": [64, 256, 512, 1024] + }, + "101": { + "block_type": "BottleneckBlock", + "block_depth": [3, 4, 23, 3], + "num_channels": [64, 256, 512, 1024] + }, + "152": { + "block_type": "BottleneckBlock", + "block_depth": [3, 8, 36, 3], + "num_channels": [64, 256, 512, 1024] + }, + "200": { + "block_type": "BottleneckBlock", + "block_depth": [3, 12, 48, 3], + "num_channels": [64, 256, 512, 1024] + }, +} + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + is_vd_mode=False, + act=None, + lr_mult=1.0, + data_format="NCHW"): + super().__init__() + self.is_vd_mode = is_vd_mode + self.act = act + self.avg_pool = AvgPool2D( + kernel_size=2, stride=2, padding=0, ceil_mode=True) + self.conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(learning_rate=lr_mult), + bias_attr=False, + data_format=data_format) + self.bn = BatchNorm( + num_filters, + param_attr=ParamAttr(learning_rate=lr_mult), + bias_attr=ParamAttr(learning_rate=lr_mult), + data_layout=data_format) + self.relu = nn.ReLU() + + def forward(self, x): + if self.is_vd_mode: + x = self.avg_pool(x) + x = self.conv(x) + x = self.bn(x) + if self.act: + x = self.relu(x) + return x + + +class BottleneckBlock(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + if_first=False, + lr_mult=1.0, + data_format="NCHW"): + super().__init__() + + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act="relu", + lr_mult=lr_mult, + data_format=data_format) + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + stride=stride, + act="relu", + lr_mult=lr_mult, + data_format=data_format) + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 4, + filter_size=1, + act=None, + lr_mult=lr_mult, + data_format=data_format) + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 4, + filter_size=1, + stride=stride if if_first else 1, + is_vd_mode=False if if_first else True, + lr_mult=lr_mult, + data_format=data_format) + self.relu = nn.ReLU() + self.shortcut = shortcut + + def forward(self, x): + identity = x + x = self.conv0(x) + x = self.conv1(x) + x = self.conv2(x) + + if self.shortcut: + short = identity + else: + short = self.short(identity) + x = paddle.add(x=x, y=short) + x = self.relu(x) + return x + + +class BasicBlock(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + if_first=False, + lr_mult=1.0, + data_format="NCHW"): + super().__init__() + + self.stride = stride + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=3, + stride=stride, + act="relu", + lr_mult=lr_mult, + data_format=data_format) + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + act=None, + lr_mult=lr_mult, + data_format=data_format) + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + stride=stride if if_first else 1, + is_vd_mode=False if if_first else True, + lr_mult=lr_mult, + data_format=data_format) + self.shortcut = shortcut + self.relu = nn.ReLU() + + def forward(self, x): + identity = x + x = self.conv0(x) + x = self.conv1(x) + if self.shortcut: + short = identity + else: + short = self.short(identity) + x = paddle.add(x=x, y=short) + x = self.relu(x) + return x + + +class ResNet(TheseusLayer): + """ + ResNet + Args: + config: dict. config of ResNet. + version: str="vb". Different version of ResNet, version vd can perform better. + class_num: int=1000. The number of classes. + lr_mult_list: list. Control the learning rate of different stages. + Returns: + model: nn.Layer. Specific ResNet model depends on args. + """ + + def __init__(self, + config, + stages_pattern, + version="vb", + class_num=1000, + lr_mult_list=[1.0, 1.0, 1.0, 1.0, 1.0], + data_format="NCHW", + input_image_channel=3, + return_patterns=None, + return_stages=None): + super().__init__() + + self.cfg = config + self.lr_mult_list = lr_mult_list + self.is_vd_mode = version == "vd" + self.class_num = class_num + self.num_filters = [64, 128, 256, 512] + self.block_depth = self.cfg["block_depth"] + self.block_type = self.cfg["block_type"] + self.num_channels = self.cfg["num_channels"] + self.channels_mult = 1 if self.num_channels[-1] == 256 else 4 + + assert isinstance(self.lr_mult_list, ( + list, tuple + )), "lr_mult_list should be in (list, tuple) but got {}".format( + type(self.lr_mult_list)) + assert len(self.lr_mult_list + ) == 5, "lr_mult_list length should be 5 but got {}".format( + len(self.lr_mult_list)) + + self.stem_cfg = { + #num_channels, num_filters, filter_size, stride + "vb": [[input_image_channel, 64, 7, 2]], + "vd": + [[input_image_channel, 32, 3, 2], [32, 32, 3, 1], [32, 64, 3, 1]] + } + + self.stem = nn.Sequential(* [ + ConvBNLayer( + num_channels=in_c, + num_filters=out_c, + filter_size=k, + stride=s, + act="relu", + lr_mult=self.lr_mult_list[0], + data_format=data_format) + for in_c, out_c, k, s in self.stem_cfg[version] + ]) + + self.max_pool = MaxPool2D( + kernel_size=3, stride=2, padding=1, data_format=data_format) + block_list = [] + for block_idx in range(len(self.block_depth)): + shortcut = False + for i in range(self.block_depth[block_idx]): + block_list.append(globals()[self.block_type]( + num_channels=self.num_channels[block_idx] if i == 0 else + self.num_filters[block_idx] * self.channels_mult, + num_filters=self.num_filters[block_idx], + stride=2 if i == 0 and block_idx != 0 else 1, + shortcut=shortcut, + if_first=block_idx == i == 0 if version == "vd" else True, + lr_mult=self.lr_mult_list[block_idx + 1], + data_format=data_format)) + shortcut = True + self.blocks = nn.Sequential(*block_list) + + self.avg_pool = AdaptiveAvgPool2D(1, data_format=data_format) + self.flatten = nn.Flatten() + self.avg_pool_channels = self.num_channels[-1] * 2 + stdv = 1.0 / math.sqrt(self.avg_pool_channels * 1.0) + self.fc = Linear( + self.avg_pool_channels, + self.class_num, + weight_attr=ParamAttr(initializer=Uniform(-stdv, stdv))) + + self.data_format = data_format + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, x): + with paddle.static.amp.fp16_guard(): + if self.data_format == "NHWC": + x = paddle.transpose(x, [0, 2, 3, 1]) + x.stop_gradient = True + x = self.stem(x) + x = self.max_pool(x) + x = self.blocks(x) + x = self.avg_pool(x) + x = self.flatten(x) + x = self.fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ResNet18(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet18 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet18` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["18"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet18"], + version="vb", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet18"], use_ssld) + return model + + +def ResNet18_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet18_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet18_vd` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["18"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet18"], + version="vd", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet18_vd"], use_ssld) + return model + + +def ResNet34(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet34 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet34` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["34"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet34"], + version="vb", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet34"], use_ssld) + return model + + +def ResNet34_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet34_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet34_vd` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["34"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet34"], + version="vd", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet34_vd"], use_ssld) + return model + + +def ResNet50(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet50 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet50` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["50"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet50"], + version="vb", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet50"], use_ssld) + return model + + +def ResNet50_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet50_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet50_vd` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["50"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet50"], + version="vd", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet50_vd"], use_ssld) + return model + + +def ResNet101(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet101 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet101` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["101"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet101"], + version="vb", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet101"], use_ssld) + return model + + +def ResNet101_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet101_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet101_vd` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["101"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet101"], + version="vd", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet101_vd"], use_ssld) + return model + + +def ResNet152(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet152 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet152` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["152"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet152"], + version="vb", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet152"], use_ssld) + return model + + +def ResNet152_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet152_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet152_vd` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["152"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet152"], + version="vd", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet152_vd"], use_ssld) + return model + + +def ResNet200_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet200_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet200_vd` model depends on args. + """ + model = ResNet( + config=NET_CONFIG["200"], + stages_pattern=MODEL_STAGES_PATTERN["ResNet200"], + version="vd", + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet200_vd"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/vgg.py b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/vgg.py new file mode 100644 index 00000000..74d5cfad --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/legendary_models/vgg.py @@ -0,0 +1,259 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import MaxPool2D + +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "VGG11": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG11_pretrained.pdparams", + "VGG13": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG13_pretrained.pdparams", + "VGG16": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG16_pretrained.pdparams", + "VGG19": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG19_pretrained.pdparams", +} + +MODEL_STAGES_PATTERN = { + "VGG": [ + "conv_block_1", "conv_block_2", "conv_block_3", "conv_block_4", + "conv_block_5" + ] +} + +__all__ = MODEL_URLS.keys() + +# VGG config +# key: VGG network depth +# value: conv num in different blocks +NET_CONFIG = { + 11: [1, 1, 2, 2, 2], + 13: [2, 2, 2, 2, 2], + 16: [2, 2, 3, 3, 3], + 19: [2, 2, 4, 4, 4] +} + + +class ConvBlock(TheseusLayer): + def __init__(self, input_channels, output_channels, groups): + super().__init__() + + self.groups = groups + self.conv1 = Conv2D( + in_channels=input_channels, + out_channels=output_channels, + kernel_size=3, + stride=1, + padding=1, + bias_attr=False) + if groups == 2 or groups == 3 or groups == 4: + self.conv2 = Conv2D( + in_channels=output_channels, + out_channels=output_channels, + kernel_size=3, + stride=1, + padding=1, + bias_attr=False) + if groups == 3 or groups == 4: + self.conv3 = Conv2D( + in_channels=output_channels, + out_channels=output_channels, + kernel_size=3, + stride=1, + padding=1, + bias_attr=False) + if groups == 4: + self.conv4 = Conv2D( + in_channels=output_channels, + out_channels=output_channels, + kernel_size=3, + stride=1, + padding=1, + bias_attr=False) + + self.max_pool = MaxPool2D(kernel_size=2, stride=2, padding=0) + self.relu = nn.ReLU() + + def forward(self, inputs): + x = self.conv1(inputs) + x = self.relu(x) + if self.groups == 2 or self.groups == 3 or self.groups == 4: + x = self.conv2(x) + x = self.relu(x) + if self.groups == 3 or self.groups == 4: + x = self.conv3(x) + x = self.relu(x) + if self.groups == 4: + x = self.conv4(x) + x = self.relu(x) + x = self.max_pool(x) + return x + + +class VGGNet(TheseusLayer): + """ + VGGNet + Args: + config: list. VGGNet config. + stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` + class_num: int=1000. The number of classes. + Returns: + model: nn.Layer. Specific VGG model depends on args. + """ + + def __init__(self, + config, + stages_pattern, + stop_grad_layers=0, + class_num=1000, + return_patterns=None, + return_stages=None): + super().__init__() + + self.stop_grad_layers = stop_grad_layers + + self.conv_block_1 = ConvBlock(3, 64, config[0]) + self.conv_block_2 = ConvBlock(64, 128, config[1]) + self.conv_block_3 = ConvBlock(128, 256, config[2]) + self.conv_block_4 = ConvBlock(256, 512, config[3]) + self.conv_block_5 = ConvBlock(512, 512, config[4]) + + self.relu = nn.ReLU() + self.flatten = nn.Flatten(start_axis=1, stop_axis=-1) + + for idx, block in enumerate([ + self.conv_block_1, self.conv_block_2, self.conv_block_3, + self.conv_block_4, self.conv_block_5 + ]): + if self.stop_grad_layers >= idx + 1: + for param in block.parameters(): + param.trainable = False + + self.drop = Dropout(p=0.5, mode="downscale_in_infer") + self.fc1 = Linear(7 * 7 * 512, 4096) + self.fc2 = Linear(4096, 4096) + self.fc3 = Linear(4096, class_num) + + super().init_res( + stages_pattern, + return_patterns=return_patterns, + return_stages=return_stages) + + def forward(self, inputs): + x = self.conv_block_1(inputs) + x = self.conv_block_2(x) + x = self.conv_block_3(x) + x = self.conv_block_4(x) + x = self.conv_block_5(x) + x = self.flatten(x) + x = self.fc1(x) + x = self.relu(x) + x = self.drop(x) + x = self.fc2(x) + x = self.relu(x) + x = self.drop(x) + x = self.fc3(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def VGG11(pretrained=False, use_ssld=False, **kwargs): + """ + VGG11 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `VGG11` model depends on args. + """ + model = VGGNet( + config=NET_CONFIG[11], + stages_pattern=MODEL_STAGES_PATTERN["VGG"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["VGG11"], use_ssld) + return model + + +def VGG13(pretrained=False, use_ssld=False, **kwargs): + """ + VGG13 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `VGG13` model depends on args. + """ + model = VGGNet( + config=NET_CONFIG[13], + stages_pattern=MODEL_STAGES_PATTERN["VGG"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["VGG13"], use_ssld) + return model + + +def VGG16(pretrained=False, use_ssld=False, **kwargs): + """ + VGG16 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `VGG16` model depends on args. + """ + model = VGGNet( + config=NET_CONFIG[16], + stages_pattern=MODEL_STAGES_PATTERN["VGG"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["VGG16"], use_ssld) + return model + + +def VGG19(pretrained=False, use_ssld=False, **kwargs): + """ + VGG19 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `VGG19` model depends on args. + """ + model = VGGNet( + config=NET_CONFIG[19], + stages_pattern=MODEL_STAGES_PATTERN["VGG"], + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["VGG19"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/__init__.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/__init__.py new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/alexnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/alexnet.py new file mode 100644 index 00000000..b44901a6 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/alexnet.py @@ -0,0 +1,168 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout, ReLU +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "AlexNet": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/AlexNet_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvPoolLayer(nn.Layer): + def __init__(self, + input_channels, + output_channels, + filter_size, + stride, + padding, + stdv, + groups=1, + act=None, + name=None): + super(ConvPoolLayer, self).__init__() + + self.relu = ReLU() if act == "relu" else None + + self._conv = Conv2D( + in_channels=input_channels, + out_channels=output_channels, + kernel_size=filter_size, + stride=stride, + padding=padding, + groups=groups, + weight_attr=ParamAttr( + name=name + "_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr( + name=name + "_offset", initializer=Uniform(-stdv, stdv))) + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=0) + + def forward(self, inputs): + x = self._conv(inputs) + if self.relu is not None: + x = self.relu(x) + x = self._pool(x) + return x + + +class AlexNetDY(nn.Layer): + def __init__(self, class_num=1000): + super(AlexNetDY, self).__init__() + + stdv = 1.0 / math.sqrt(3 * 11 * 11) + self._conv1 = ConvPoolLayer( + 3, 64, 11, 4, 2, stdv, act="relu", name="conv1") + stdv = 1.0 / math.sqrt(64 * 5 * 5) + self._conv2 = ConvPoolLayer( + 64, 192, 5, 1, 2, stdv, act="relu", name="conv2") + stdv = 1.0 / math.sqrt(192 * 3 * 3) + self._conv3 = Conv2D( + 192, + 384, + 3, + stride=1, + padding=1, + weight_attr=ParamAttr( + name="conv3_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr( + name="conv3_offset", initializer=Uniform(-stdv, stdv))) + stdv = 1.0 / math.sqrt(384 * 3 * 3) + self._conv4 = Conv2D( + 384, + 256, + 3, + stride=1, + padding=1, + weight_attr=ParamAttr( + name="conv4_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr( + name="conv4_offset", initializer=Uniform(-stdv, stdv))) + stdv = 1.0 / math.sqrt(256 * 3 * 3) + self._conv5 = ConvPoolLayer( + 256, 256, 3, 1, 1, stdv, act="relu", name="conv5") + stdv = 1.0 / math.sqrt(256 * 6 * 6) + + self._drop1 = Dropout(p=0.5, mode="downscale_in_infer") + self._fc6 = Linear( + in_features=256 * 6 * 6, + out_features=4096, + weight_attr=ParamAttr( + name="fc6_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr( + name="fc6_offset", initializer=Uniform(-stdv, stdv))) + + self._drop2 = Dropout(p=0.5, mode="downscale_in_infer") + self._fc7 = Linear( + in_features=4096, + out_features=4096, + weight_attr=ParamAttr( + name="fc7_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr( + name="fc7_offset", initializer=Uniform(-stdv, stdv))) + self._fc8 = Linear( + in_features=4096, + out_features=class_num, + weight_attr=ParamAttr( + name="fc8_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr( + name="fc8_offset", initializer=Uniform(-stdv, stdv))) + + def forward(self, inputs): + x = self._conv1(inputs) + x = self._conv2(x) + x = self._conv3(x) + x = F.relu(x) + x = self._conv4(x) + x = F.relu(x) + x = self._conv5(x) + x = paddle.flatten(x, start_axis=1, stop_axis=-1) + x = self._drop1(x) + x = self._fc6(x) + x = F.relu(x) + x = self._drop2(x) + x = self._fc7(x) + x = F.relu(x) + x = self._fc8(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def AlexNet(pretrained=False, use_ssld=False, **kwargs): + model = AlexNetDY(**kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["AlexNet"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/cspnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/cspnet.py new file mode 100644 index 00000000..ab5021fc --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/cspnet.py @@ -0,0 +1,376 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was heavily based on https://github.com/rwightman/pytorch-image-models + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F +from paddle import ParamAttr + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "CSPDarkNet53": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSPDarkNet53_pretrained.pdparams" +} + +MODEL_CFGS = { + "CSPDarkNet53": dict( + stem=dict( + out_chs=32, kernel_size=3, stride=1, pool=''), + stage=dict( + out_chs=(64, 128, 256, 512, 1024), + depth=(1, 2, 8, 8, 4), + stride=(2, ) * 5, + exp_ratio=(2., ) + (1., ) * 4, + bottle_ratio=(0.5, ) + (1.0, ) * 4, + block_ratio=(1., ) + (0.5, ) * 4, + down_growth=True, )) +} + +__all__ = ['CSPDarkNet53' + ] # model_registry will add each entrypoint fn to this + + +class ConvBnAct(nn.Layer): + def __init__(self, + input_channels, + output_channels, + kernel_size=1, + stride=1, + padding=None, + dilation=1, + groups=1, + act_layer=nn.LeakyReLU, + norm_layer=nn.BatchNorm2D): + super().__init__() + if padding is None: + padding = (kernel_size - 1) // 2 + self.conv = nn.Conv2D( + in_channels=input_channels, + out_channels=output_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + dilation=dilation, + groups=groups, + weight_attr=ParamAttr(), + bias_attr=False) + + self.bn = norm_layer(num_features=output_channels) + self.act = act_layer() + + def forward(self, inputs): + x = self.conv(inputs) + x = self.bn(x) + if self.act is not None: + x = self.act(x) + return x + + +def create_stem(in_chans=3, + out_chs=32, + kernel_size=3, + stride=2, + pool='', + act_layer=None, + norm_layer=None): + stem = nn.Sequential() + if not isinstance(out_chs, (tuple, list)): + out_chs = [out_chs] + assert len(out_chs) + in_c = in_chans + for i, out_c in enumerate(out_chs): + conv_name = f'conv{i + 1}' + stem.add_sublayer( + conv_name, + ConvBnAct( + in_c, + out_c, + kernel_size, + stride=stride if i == 0 else 1, + act_layer=act_layer, + norm_layer=norm_layer)) + in_c = out_c + last_conv = conv_name + if pool: + stem.add_sublayer( + 'pool', nn.MaxPool2D( + kernel_size=3, stride=2, padding=1)) + return stem, dict( + num_chs=in_c, reduction=stride, module='.'.join(['stem', last_conv])) + + +class DarkBlock(nn.Layer): + def __init__(self, + in_chs, + out_chs, + dilation=1, + bottle_ratio=0.5, + groups=1, + act_layer=nn.ReLU, + norm_layer=nn.BatchNorm2D, + attn_layer=None, + drop_block=None): + super(DarkBlock, self).__init__() + mid_chs = int(round(out_chs * bottle_ratio)) + ckwargs = dict(act_layer=act_layer, norm_layer=norm_layer) + self.conv1 = ConvBnAct(in_chs, mid_chs, kernel_size=1, **ckwargs) + self.conv2 = ConvBnAct( + mid_chs, + out_chs, + kernel_size=3, + dilation=dilation, + groups=groups, + **ckwargs) + + def forward(self, x): + shortcut = x + x = self.conv1(x) + x = self.conv2(x) + x = x + shortcut + return x + + +class CrossStage(nn.Layer): + def __init__(self, + in_chs, + out_chs, + stride, + dilation, + depth, + block_ratio=1., + bottle_ratio=1., + exp_ratio=1., + groups=1, + first_dilation=None, + down_growth=False, + cross_linear=False, + block_dpr=None, + block_fn=DarkBlock, + **block_kwargs): + super(CrossStage, self).__init__() + first_dilation = first_dilation or dilation + down_chs = out_chs if down_growth else in_chs + exp_chs = int(round(out_chs * exp_ratio)) + block_out_chs = int(round(out_chs * block_ratio)) + conv_kwargs = dict( + act_layer=block_kwargs.get('act_layer'), + norm_layer=block_kwargs.get('norm_layer')) + + if stride != 1 or first_dilation != dilation: + self.conv_down = ConvBnAct( + in_chs, + down_chs, + kernel_size=3, + stride=stride, + dilation=first_dilation, + groups=groups, + **conv_kwargs) + prev_chs = down_chs + else: + self.conv_down = None + prev_chs = in_chs + + self.conv_exp = ConvBnAct( + prev_chs, exp_chs, kernel_size=1, **conv_kwargs) + prev_chs = exp_chs // 2 # output of conv_exp is always split in two + + self.blocks = nn.Sequential() + for i in range(depth): + self.blocks.add_sublayer( + str(i), + block_fn(prev_chs, block_out_chs, dilation, bottle_ratio, + groups, **block_kwargs)) + prev_chs = block_out_chs + + # transition convs + self.conv_transition_b = ConvBnAct( + prev_chs, exp_chs // 2, kernel_size=1, **conv_kwargs) + self.conv_transition = ConvBnAct( + exp_chs, out_chs, kernel_size=1, **conv_kwargs) + + def forward(self, x): + if self.conv_down is not None: + x = self.conv_down(x) + x = self.conv_exp(x) + split = x.shape[1] // 2 + xs, xb = x[:, :split], x[:, split:] + xb = self.blocks(xb) + xb = self.conv_transition_b(xb) + out = self.conv_transition(paddle.concat([xs, xb], axis=1)) + return out + + +class DarkStage(nn.Layer): + def __init__(self, + in_chs, + out_chs, + stride, + dilation, + depth, + block_ratio=1., + bottle_ratio=1., + groups=1, + first_dilation=None, + block_fn=DarkBlock, + block_dpr=None, + **block_kwargs): + super().__init__() + first_dilation = first_dilation or dilation + + self.conv_down = ConvBnAct( + in_chs, + out_chs, + kernel_size=3, + stride=stride, + dilation=first_dilation, + groups=groups, + act_layer=block_kwargs.get('act_layer'), + norm_layer=block_kwargs.get('norm_layer')) + + prev_chs = out_chs + block_out_chs = int(round(out_chs * block_ratio)) + self.blocks = nn.Sequential() + for i in range(depth): + self.blocks.add_sublayer( + str(i), + block_fn(prev_chs, block_out_chs, dilation, bottle_ratio, + groups, **block_kwargs)) + prev_chs = block_out_chs + + def forward(self, x): + x = self.conv_down(x) + x = self.blocks(x) + return x + + +def _cfg_to_stage_args(cfg, curr_stride=2, output_stride=32): + # get per stage args for stage and containing blocks, calculate strides to meet target output_stride + num_stages = len(cfg['depth']) + if 'groups' not in cfg: + cfg['groups'] = (1, ) * num_stages + if 'down_growth' in cfg and not isinstance(cfg['down_growth'], + (list, tuple)): + cfg['down_growth'] = (cfg['down_growth'], ) * num_stages + stage_strides = [] + stage_dilations = [] + stage_first_dilations = [] + dilation = 1 + for cfg_stride in cfg['stride']: + stage_first_dilations.append(dilation) + if curr_stride >= output_stride: + dilation *= cfg_stride + stride = 1 + else: + stride = cfg_stride + curr_stride *= stride + stage_strides.append(stride) + stage_dilations.append(dilation) + cfg['stride'] = stage_strides + cfg['dilation'] = stage_dilations + cfg['first_dilation'] = stage_first_dilations + stage_args = [ + dict(zip(cfg.keys(), values)) for values in zip(*cfg.values()) + ] + return stage_args + + +class CSPNet(nn.Layer): + def __init__(self, + cfg, + in_chans=3, + class_num=1000, + output_stride=32, + global_pool='avg', + drop_rate=0., + act_layer=nn.LeakyReLU, + norm_layer=nn.BatchNorm2D, + zero_init_last_bn=True, + stage_fn=CrossStage, + block_fn=DarkBlock): + super().__init__() + self.class_num = class_num + self.drop_rate = drop_rate + assert output_stride in (8, 16, 32) + layer_args = dict(act_layer=act_layer, norm_layer=norm_layer) + + # Construct the stem + self.stem, stem_feat_info = create_stem(in_chans, **cfg['stem'], + **layer_args) + self.feature_info = [stem_feat_info] + prev_chs = stem_feat_info['num_chs'] + curr_stride = stem_feat_info[ + 'reduction'] # reduction does not include pool + if cfg['stem']['pool']: + curr_stride *= 2 + + # Construct the stages + per_stage_args = _cfg_to_stage_args( + cfg['stage'], curr_stride=curr_stride, output_stride=output_stride) + self.stages = nn.LayerList() + for i, sa in enumerate(per_stage_args): + self.stages.add_sublayer( + str(i), + stage_fn( + prev_chs, **sa, **layer_args, block_fn=block_fn)) + prev_chs = sa['out_chs'] + curr_stride *= sa['stride'] + self.feature_info += [ + dict( + num_chs=prev_chs, + reduction=curr_stride, + module=f'stages.{i}') + ] + + # Construct the head + self.num_features = prev_chs + + self.pool = nn.AdaptiveAvgPool2D(1) + self.flatten = nn.Flatten(1) + self.fc = nn.Linear( + prev_chs, + class_num, + weight_attr=ParamAttr(), + bias_attr=ParamAttr()) + + def forward(self, x): + x = self.stem(x) + for stage in self.stages: + x = stage(x) + x = self.pool(x) + x = self.flatten(x) + x = self.fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def CSPDarkNet53(pretrained=False, use_ssld=False, **kwargs): + model = CSPNet(MODEL_CFGS["CSPDarkNet53"], block_fn=DarkBlock, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["CSPDarkNet53"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/darknet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/darknet.py new file mode 100644 index 00000000..75aafd85 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/darknet.py @@ -0,0 +1,197 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "DarkNet53": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DarkNet53_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + input_channels, + output_channels, + filter_size, + stride, + padding, + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=input_channels, + out_channels=output_channels, + kernel_size=filter_size, + stride=stride, + padding=padding, + weight_attr=ParamAttr(name=name + ".conv.weights"), + bias_attr=False) + + bn_name = name + ".bn" + self._bn = BatchNorm( + num_channels=output_channels, + act="relu", + param_attr=ParamAttr(name=bn_name + ".scale"), + bias_attr=ParamAttr(name=bn_name + ".offset"), + moving_mean_name=bn_name + ".mean", + moving_variance_name=bn_name + ".var") + + def forward(self, inputs): + x = self._conv(inputs) + x = self._bn(x) + return x + + +class BasicBlock(nn.Layer): + def __init__(self, input_channels, output_channels, name=None): + super(BasicBlock, self).__init__() + + self._conv1 = ConvBNLayer( + input_channels, output_channels, 1, 1, 0, name=name + ".0") + self._conv2 = ConvBNLayer( + output_channels, output_channels * 2, 3, 1, 1, name=name + ".1") + + def forward(self, inputs): + x = self._conv1(inputs) + x = self._conv2(x) + return paddle.add(x=inputs, y=x) + + +class DarkNet(nn.Layer): + def __init__(self, class_num=1000): + super(DarkNet, self).__init__() + + self.stages = [1, 2, 8, 8, 4] + self._conv1 = ConvBNLayer(3, 32, 3, 1, 1, name="yolo_input") + self._conv2 = ConvBNLayer( + 32, 64, 3, 2, 1, name="yolo_input.downsample") + + self._basic_block_01 = BasicBlock(64, 32, name="stage.0.0") + self._downsample_0 = ConvBNLayer( + 64, 128, 3, 2, 1, name="stage.0.downsample") + + self._basic_block_11 = BasicBlock(128, 64, name="stage.1.0") + self._basic_block_12 = BasicBlock(128, 64, name="stage.1.1") + self._downsample_1 = ConvBNLayer( + 128, 256, 3, 2, 1, name="stage.1.downsample") + + self._basic_block_21 = BasicBlock(256, 128, name="stage.2.0") + self._basic_block_22 = BasicBlock(256, 128, name="stage.2.1") + self._basic_block_23 = BasicBlock(256, 128, name="stage.2.2") + self._basic_block_24 = BasicBlock(256, 128, name="stage.2.3") + self._basic_block_25 = BasicBlock(256, 128, name="stage.2.4") + self._basic_block_26 = BasicBlock(256, 128, name="stage.2.5") + self._basic_block_27 = BasicBlock(256, 128, name="stage.2.6") + self._basic_block_28 = BasicBlock(256, 128, name="stage.2.7") + self._downsample_2 = ConvBNLayer( + 256, 512, 3, 2, 1, name="stage.2.downsample") + + self._basic_block_31 = BasicBlock(512, 256, name="stage.3.0") + self._basic_block_32 = BasicBlock(512, 256, name="stage.3.1") + self._basic_block_33 = BasicBlock(512, 256, name="stage.3.2") + self._basic_block_34 = BasicBlock(512, 256, name="stage.3.3") + self._basic_block_35 = BasicBlock(512, 256, name="stage.3.4") + self._basic_block_36 = BasicBlock(512, 256, name="stage.3.5") + self._basic_block_37 = BasicBlock(512, 256, name="stage.3.6") + self._basic_block_38 = BasicBlock(512, 256, name="stage.3.7") + self._downsample_3 = ConvBNLayer( + 512, 1024, 3, 2, 1, name="stage.3.downsample") + + self._basic_block_41 = BasicBlock(1024, 512, name="stage.4.0") + self._basic_block_42 = BasicBlock(1024, 512, name="stage.4.1") + self._basic_block_43 = BasicBlock(1024, 512, name="stage.4.2") + self._basic_block_44 = BasicBlock(1024, 512, name="stage.4.3") + + self._pool = AdaptiveAvgPool2D(1) + + stdv = 1.0 / math.sqrt(1024.0) + self._out = Linear( + 1024, + class_num, + weight_attr=ParamAttr( + name="fc_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, inputs): + x = self._conv1(inputs) + x = self._conv2(x) + + x = self._basic_block_01(x) + x = self._downsample_0(x) + + x = self._basic_block_11(x) + x = self._basic_block_12(x) + x = self._downsample_1(x) + + x = self._basic_block_21(x) + x = self._basic_block_22(x) + x = self._basic_block_23(x) + x = self._basic_block_24(x) + x = self._basic_block_25(x) + x = self._basic_block_26(x) + x = self._basic_block_27(x) + x = self._basic_block_28(x) + x = self._downsample_2(x) + + x = self._basic_block_31(x) + x = self._basic_block_32(x) + x = self._basic_block_33(x) + x = self._basic_block_34(x) + x = self._basic_block_35(x) + x = self._basic_block_36(x) + x = self._basic_block_37(x) + x = self._basic_block_38(x) + x = self._downsample_3(x) + + x = self._basic_block_41(x) + x = self._basic_block_42(x) + x = self._basic_block_43(x) + x = self._basic_block_44(x) + + x = self._pool(x) + x = paddle.squeeze(x, axis=[2, 3]) + x = self._out(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def DarkNet53(pretrained=False, use_ssld=False, **kwargs): + model = DarkNet(**kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["DarkNet53"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/densenet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/densenet.py new file mode 100644 index 00000000..7e6e2025 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/densenet.py @@ -0,0 +1,344 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "DenseNet121": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet121_pretrained.pdparams", + "DenseNet161": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet161_pretrained.pdparams", + "DenseNet169": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet169_pretrained.pdparams", + "DenseNet201": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet201_pretrained.pdparams", + "DenseNet264": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet264_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class BNACConvLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + pad=0, + groups=1, + act="relu", + name=None): + super(BNACConvLayer, self).__init__() + + self._batch_norm = BatchNorm( + num_channels, + act=act, + param_attr=ParamAttr(name=name + '_bn_scale'), + bias_attr=ParamAttr(name + '_bn_offset'), + moving_mean_name=name + '_bn_mean', + moving_variance_name=name + '_bn_variance') + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=pad, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + + def forward(self, input): + y = self._batch_norm(input) + y = self._conv(y) + return y + + +class DenseLayer(nn.Layer): + def __init__(self, num_channels, growth_rate, bn_size, dropout, name=None): + super(DenseLayer, self).__init__() + self.dropout = dropout + + self.bn_ac_func1 = BNACConvLayer( + num_channels=num_channels, + num_filters=bn_size * growth_rate, + filter_size=1, + pad=0, + stride=1, + name=name + "_x1") + + self.bn_ac_func2 = BNACConvLayer( + num_channels=bn_size * growth_rate, + num_filters=growth_rate, + filter_size=3, + pad=1, + stride=1, + name=name + "_x2") + + if dropout: + self.dropout_func = Dropout(p=dropout, mode="downscale_in_infer") + + def forward(self, input): + conv = self.bn_ac_func1(input) + conv = self.bn_ac_func2(conv) + if self.dropout: + conv = self.dropout_func(conv) + conv = paddle.concat([input, conv], axis=1) + return conv + + +class DenseBlock(nn.Layer): + def __init__(self, + num_channels, + num_layers, + bn_size, + growth_rate, + dropout, + name=None): + super(DenseBlock, self).__init__() + self.dropout = dropout + + self.dense_layer_func = [] + + pre_channel = num_channels + for layer in range(num_layers): + self.dense_layer_func.append( + self.add_sublayer( + "{}_{}".format(name, layer + 1), + DenseLayer( + num_channels=pre_channel, + growth_rate=growth_rate, + bn_size=bn_size, + dropout=dropout, + name=name + '_' + str(layer + 1)))) + pre_channel = pre_channel + growth_rate + + def forward(self, input): + conv = input + for func in self.dense_layer_func: + conv = func(conv) + return conv + + +class TransitionLayer(nn.Layer): + def __init__(self, num_channels, num_output_features, name=None): + super(TransitionLayer, self).__init__() + + self.conv_ac_func = BNACConvLayer( + num_channels=num_channels, + num_filters=num_output_features, + filter_size=1, + pad=0, + stride=1, + name=name) + + self.pool2d_avg = AvgPool2D(kernel_size=2, stride=2, padding=0) + + def forward(self, input): + y = self.conv_ac_func(input) + y = self.pool2d_avg(y) + return y + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + pad=0, + groups=1, + act="relu", + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=pad, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=name + '_bn_scale'), + bias_attr=ParamAttr(name + '_bn_offset'), + moving_mean_name=name + '_bn_mean', + moving_variance_name=name + '_bn_variance') + + def forward(self, input): + y = self._conv(input) + y = self._batch_norm(y) + return y + + +class DenseNet(nn.Layer): + def __init__(self, layers=60, bn_size=4, dropout=0, class_num=1000): + super(DenseNet, self).__init__() + + supported_layers = [121, 161, 169, 201, 264] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + densenet_spec = { + 121: (64, 32, [6, 12, 24, 16]), + 161: (96, 48, [6, 12, 36, 24]), + 169: (64, 32, [6, 12, 32, 32]), + 201: (64, 32, [6, 12, 48, 32]), + 264: (64, 32, [6, 12, 64, 48]) + } + num_init_features, growth_rate, block_config = densenet_spec[layers] + + self.conv1_func = ConvBNLayer( + num_channels=3, + num_filters=num_init_features, + filter_size=7, + stride=2, + pad=3, + act='relu', + name="conv1") + + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.block_config = block_config + + self.dense_block_func_list = [] + self.transition_func_list = [] + pre_num_channels = num_init_features + num_features = num_init_features + for i, num_layers in enumerate(block_config): + self.dense_block_func_list.append( + self.add_sublayer( + "db_conv_{}".format(i + 2), + DenseBlock( + num_channels=pre_num_channels, + num_layers=num_layers, + bn_size=bn_size, + growth_rate=growth_rate, + dropout=dropout, + name='conv' + str(i + 2)))) + + num_features = num_features + num_layers * growth_rate + pre_num_channels = num_features + + if i != len(block_config) - 1: + self.transition_func_list.append( + self.add_sublayer( + "tr_conv{}_blk".format(i + 2), + TransitionLayer( + num_channels=pre_num_channels, + num_output_features=num_features // 2, + name='conv' + str(i + 2) + "_blk"))) + pre_num_channels = num_features // 2 + num_features = num_features // 2 + + self.batch_norm = BatchNorm( + num_features, + act="relu", + param_attr=ParamAttr(name='conv5_blk_bn_scale'), + bias_attr=ParamAttr(name='conv5_blk_bn_offset'), + moving_mean_name='conv5_blk_bn_mean', + moving_variance_name='conv5_blk_bn_variance') + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + stdv = 1.0 / math.sqrt(num_features * 1.0) + + self.out = Linear( + num_features, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_weights"), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, input): + conv = self.conv1_func(input) + conv = self.pool2d_max(conv) + + for i, num_layers in enumerate(self.block_config): + conv = self.dense_block_func_list[i](conv) + if i != len(self.block_config) - 1: + conv = self.transition_func_list[i](conv) + + conv = self.batch_norm(conv) + y = self.pool2d_avg(conv) + y = paddle.flatten(y, start_axis=1, stop_axis=-1) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def DenseNet121(pretrained=False, use_ssld=False, **kwargs): + model = DenseNet(layers=121, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["DenseNet121"], use_ssld=use_ssld) + return model + + +def DenseNet161(pretrained=False, use_ssld=False, **kwargs): + model = DenseNet(layers=161, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["DenseNet161"], use_ssld=use_ssld) + return model + + +def DenseNet169(pretrained=False, use_ssld=False, **kwargs): + model = DenseNet(layers=169, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["DenseNet169"], use_ssld=use_ssld) + return model + + +def DenseNet201(pretrained=False, use_ssld=False, **kwargs): + model = DenseNet(layers=201, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["DenseNet201"], use_ssld=use_ssld) + return model + + +def DenseNet264(pretrained=False, use_ssld=False, **kwargs): + model = DenseNet(layers=264, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["DenseNet264"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/distilled_vision_transformer.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/distilled_vision_transformer.py new file mode 100644 index 00000000..676a2897 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/distilled_vision_transformer.py @@ -0,0 +1,272 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was heavily based on https://github.com/facebookresearch/deit + +import paddle +import paddle.nn as nn +from .vision_transformer import VisionTransformer, Identity, trunc_normal_, zeros_ + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "DeiT_tiny_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_patch16_224_pretrained.pdparams", + "DeiT_small_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_small_patch16_224_pretrained.pdparams", + "DeiT_base_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_224_pretrained.pdparams", + "DeiT_tiny_distilled_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_distilled_patch16_224_pretrained.pdparams", + "DeiT_small_distilled_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_small_distilled_patch16_224_pretrained.pdparams", + "DeiT_base_distilled_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_distilled_patch16_224_pretrained.pdparams", + "DeiT_base_patch16_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_384_pretrained.pdparams", + "DeiT_base_distilled_patch16_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_distilled_patch16_384_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class DistilledVisionTransformer(VisionTransformer): + def __init__(self, + img_size=224, + patch_size=16, + class_num=1000, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=False, + norm_layer='nn.LayerNorm', + epsilon=1e-5, + **kwargs): + super().__init__( + img_size=img_size, + patch_size=patch_size, + class_num=class_num, + embed_dim=embed_dim, + depth=depth, + num_heads=num_heads, + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + norm_layer=norm_layer, + epsilon=epsilon, + **kwargs) + self.pos_embed = self.create_parameter( + shape=(1, self.patch_embed.num_patches + 2, self.embed_dim), + default_initializer=zeros_) + self.add_parameter("pos_embed", self.pos_embed) + + self.dist_token = self.create_parameter( + shape=(1, 1, self.embed_dim), default_initializer=zeros_) + self.add_parameter("cls_token", self.cls_token) + + self.head_dist = nn.Linear( + self.embed_dim, + self.class_num) if self.class_num > 0 else Identity() + + trunc_normal_(self.dist_token) + trunc_normal_(self.pos_embed) + self.head_dist.apply(self._init_weights) + + def forward_features(self, x): + B = paddle.shape(x)[0] + x = self.patch_embed(x) + + cls_tokens = self.cls_token.expand((B, -1, -1)) + dist_token = self.dist_token.expand((B, -1, -1)) + x = paddle.concat((cls_tokens, dist_token, x), axis=1) + + x = x + self.pos_embed + x = self.pos_drop(x) + + for blk in self.blocks: + x = blk(x) + + x = self.norm(x) + return x[:, 0], x[:, 1] + + def forward(self, x): + x, x_dist = self.forward_features(x) + x = self.head(x) + x_dist = self.head_dist(x_dist) + return (x + x_dist) / 2 + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def DeiT_tiny_patch16_224(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + patch_size=16, + embed_dim=192, + depth=12, + num_heads=3, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_tiny_patch16_224"], + use_ssld=use_ssld) + return model + + +def DeiT_small_patch16_224(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + patch_size=16, + embed_dim=384, + depth=12, + num_heads=6, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_small_patch16_224"], + use_ssld=use_ssld) + return model + + +def DeiT_base_patch16_224(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + patch_size=16, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_base_patch16_224"], + use_ssld=use_ssld) + return model + + +def DeiT_tiny_distilled_patch16_224(pretrained=False, use_ssld=False, + **kwargs): + model = DistilledVisionTransformer( + patch_size=16, + embed_dim=192, + depth=12, + num_heads=3, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_tiny_distilled_patch16_224"], + use_ssld=use_ssld) + return model + + +def DeiT_small_distilled_patch16_224(pretrained=False, + use_ssld=False, + **kwargs): + model = DistilledVisionTransformer( + patch_size=16, + embed_dim=384, + depth=12, + num_heads=6, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_small_distilled_patch16_224"], + use_ssld=use_ssld) + return model + + +def DeiT_base_distilled_patch16_224(pretrained=False, use_ssld=False, + **kwargs): + model = DistilledVisionTransformer( + patch_size=16, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_base_distilled_patch16_224"], + use_ssld=use_ssld) + return model + + +def DeiT_base_patch16_384(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + img_size=384, + patch_size=16, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_base_patch16_384"], + use_ssld=use_ssld) + return model + + +def DeiT_base_distilled_patch16_384(pretrained=False, use_ssld=False, + **kwargs): + model = DistilledVisionTransformer( + img_size=384, + patch_size=16, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["DeiT_base_distilled_patch16_384"], + use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/dla.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/dla.py new file mode 100644 index 00000000..b1c00b2d --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/dla.py @@ -0,0 +1,528 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/ucbdrive/dla + +import math + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + +from paddle.nn.initializer import Normal, Constant + +from ppcls.arch.backbone.base.theseus_layer import Identity +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "DLA34": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA34_pretrained.pdparams", + "DLA46_c": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA46_c_pretrained.pdparams", + "DLA46x_c": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA46x_c_pretrained.pdparams", + "DLA60": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60_pretrained.pdparams", + "DLA60x": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_pretrained.pdparams", + "DLA60x_c": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_c_pretrained.pdparams", + "DLA102": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102_pretrained.pdparams", + "DLA102x": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x_pretrained.pdparams", + "DLA102x2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x2_pretrained.pdparams", + "DLA169": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA169_pretrained.pdparams" +} + +__all__ = MODEL_URLS.keys() + +zeros_ = Constant(value=0.) +ones_ = Constant(value=1.) + + +class DlaBasic(nn.Layer): + def __init__(self, inplanes, planes, stride=1, dilation=1, **cargs): + super(DlaBasic, self).__init__() + self.conv1 = nn.Conv2D( + inplanes, + planes, + kernel_size=3, + stride=stride, + padding=dilation, + bias_attr=False, + dilation=dilation) + self.bn1 = nn.BatchNorm2D(planes) + self.relu = nn.ReLU() + self.conv2 = nn.Conv2D( + planes, + planes, + kernel_size=3, + stride=1, + padding=dilation, + bias_attr=False, + dilation=dilation) + self.bn2 = nn.BatchNorm2D(planes) + self.stride = stride + + def forward(self, x, residual=None): + if residual is None: + residual = x + + out = self.conv1(x) + out = self.bn1(out) + out = self.relu(out) + + out = self.conv2(out) + out = self.bn2(out) + + out += residual + out = self.relu(out) + + return out + + +class DlaBottleneck(nn.Layer): + expansion = 2 + + def __init__(self, + inplanes, + outplanes, + stride=1, + dilation=1, + cardinality=1, + base_width=64): + super(DlaBottleneck, self).__init__() + self.stride = stride + mid_planes = int( + math.floor(outplanes * (base_width / 64)) * cardinality) + mid_planes = mid_planes // self.expansion + + self.conv1 = nn.Conv2D( + inplanes, mid_planes, kernel_size=1, bias_attr=False) + self.bn1 = nn.BatchNorm2D(mid_planes) + self.conv2 = nn.Conv2D( + mid_planes, + mid_planes, + kernel_size=3, + stride=stride, + padding=dilation, + bias_attr=False, + dilation=dilation, + groups=cardinality) + self.bn2 = nn.BatchNorm2D(mid_planes) + self.conv3 = nn.Conv2D( + mid_planes, outplanes, kernel_size=1, bias_attr=False) + self.bn3 = nn.BatchNorm2D(outplanes) + self.relu = nn.ReLU() + + def forward(self, x, residual=None): + if residual is None: + residual = x + + out = self.conv1(x) + out = self.bn1(out) + out = self.relu(out) + + out = self.conv2(out) + out = self.bn2(out) + out = self.relu(out) + + out = self.conv3(out) + out = self.bn3(out) + + out += residual + out = self.relu(out) + + return out + + +class DlaRoot(nn.Layer): + def __init__(self, in_channels, out_channels, kernel_size, residual): + super(DlaRoot, self).__init__() + self.conv = nn.Conv2D( + in_channels, + out_channels, + 1, + stride=1, + bias_attr=False, + padding=(kernel_size - 1) // 2) + self.bn = nn.BatchNorm2D(out_channels) + self.relu = nn.ReLU() + self.residual = residual + + def forward(self, *x): + children = x + x = self.conv(paddle.concat(x, 1)) + x = self.bn(x) + if self.residual: + x += children[0] + x = self.relu(x) + + return x + + +class DlaTree(nn.Layer): + def __init__(self, + levels, + block, + in_channels, + out_channels, + stride=1, + dilation=1, + cardinality=1, + base_width=64, + level_root=False, + root_dim=0, + root_kernel_size=1, + root_residual=False): + super(DlaTree, self).__init__() + if root_dim == 0: + root_dim = 2 * out_channels + if level_root: + root_dim += in_channels + + self.downsample = nn.MaxPool2D( + stride, stride=stride) if stride > 1 else Identity() + self.project = Identity() + cargs = dict( + dilation=dilation, cardinality=cardinality, base_width=base_width) + + if levels == 1: + self.tree1 = block(in_channels, out_channels, stride, **cargs) + self.tree2 = block(out_channels, out_channels, 1, **cargs) + if in_channels != out_channels: + self.project = nn.Sequential( + nn.Conv2D( + in_channels, + out_channels, + kernel_size=1, + stride=1, + bias_attr=False), + nn.BatchNorm2D(out_channels)) + else: + cargs.update( + dict( + root_kernel_size=root_kernel_size, + root_residual=root_residual)) + self.tree1 = DlaTree( + levels - 1, + block, + in_channels, + out_channels, + stride, + root_dim=0, + **cargs) + self.tree2 = DlaTree( + levels - 1, + block, + out_channels, + out_channels, + root_dim=root_dim + out_channels, + **cargs) + + if levels == 1: + self.root = DlaRoot(root_dim, out_channels, root_kernel_size, + root_residual) + + self.level_root = level_root + self.root_dim = root_dim + self.levels = levels + + def forward(self, x, residual=None, children=None): + children = [] if children is None else children + bottom = self.downsample(x) + residual = self.project(bottom) + + if self.level_root: + children.append(bottom) + x1 = self.tree1(x, residual) + + if self.levels == 1: + x2 = self.tree2(x1) + x = self.root(x2, x1, *children) + else: + children.append(x1) + x = self.tree2(x1, children=children) + return x + + +class DLA(nn.Layer): + def __init__(self, + levels, + channels, + in_chans=3, + cardinality=1, + base_width=64, + block=DlaBottleneck, + residual_root=False, + drop_rate=0.0, + class_num=1000, + with_pool=True): + super(DLA, self).__init__() + self.channels = channels + self.class_num = class_num + self.with_pool = with_pool + self.cardinality = cardinality + self.base_width = base_width + self.drop_rate = drop_rate + + self.base_layer = nn.Sequential( + nn.Conv2D( + in_chans, + channels[0], + kernel_size=7, + stride=1, + padding=3, + bias_attr=False), + nn.BatchNorm2D(channels[0]), + nn.ReLU()) + + self.level0 = self._make_conv_level(channels[0], channels[0], + levels[0]) + self.level1 = self._make_conv_level( + channels[0], channels[1], levels[1], stride=2) + + cargs = dict( + cardinality=cardinality, + base_width=base_width, + root_residual=residual_root) + + self.level2 = DlaTree( + levels[2], + block, + channels[1], + channels[2], + 2, + level_root=False, + **cargs) + self.level3 = DlaTree( + levels[3], + block, + channels[2], + channels[3], + 2, + level_root=True, + **cargs) + self.level4 = DlaTree( + levels[4], + block, + channels[3], + channels[4], + 2, + level_root=True, + **cargs) + self.level5 = DlaTree( + levels[5], + block, + channels[4], + channels[5], + 2, + level_root=True, + **cargs) + + self.feature_info = [ + # rare to have a meaningful stride 1 level + dict( + num_chs=channels[0], reduction=1, module='level0'), + dict( + num_chs=channels[1], reduction=2, module='level1'), + dict( + num_chs=channels[2], reduction=4, module='level2'), + dict( + num_chs=channels[3], reduction=8, module='level3'), + dict( + num_chs=channels[4], reduction=16, module='level4'), + dict( + num_chs=channels[5], reduction=32, module='level5'), + ] + + self.num_features = channels[-1] + + if with_pool: + self.global_pool = nn.AdaptiveAvgPool2D(1) + + if class_num > 0: + self.fc = nn.Conv2D(self.num_features, class_num, 1) + + for m in self.sublayers(): + if isinstance(m, nn.Conv2D): + n = m._kernel_size[0] * m._kernel_size[1] * m._out_channels + normal_ = Normal(mean=0.0, std=math.sqrt(2. / n)) + normal_(m.weight) + elif isinstance(m, nn.BatchNorm2D): + ones_(m.weight) + zeros_(m.bias) + + def _make_conv_level(self, inplanes, planes, convs, stride=1, dilation=1): + modules = [] + for i in range(convs): + modules.extend([ + nn.Conv2D( + inplanes, + planes, + kernel_size=3, + stride=stride if i == 0 else 1, + padding=dilation, + bias_attr=False, + dilation=dilation), nn.BatchNorm2D(planes), nn.ReLU() + ]) + inplanes = planes + return nn.Sequential(*modules) + + def forward_features(self, x): + x = self.base_layer(x) + + x = self.level0(x) + x = self.level1(x) + x = self.level2(x) + x = self.level3(x) + x = self.level4(x) + x = self.level5(x) + + return x + + def forward(self, x): + x = self.forward_features(x) + + if self.with_pool: + x = self.global_pool(x) + + if self.drop_rate > 0.: + x = F.dropout(x, p=self.drop_rate, training=self.training) + + if self.class_num > 0: + x = self.fc(x) + x = x.flatten(1) + + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def DLA34(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 2, 2, 1), + channels=(16, 32, 64, 128, 256, 512), + block=DlaBasic, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA34"]) + return model + + +def DLA46_c(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 2, 2, 1), + channels=(16, 32, 64, 64, 128, 256), + block=DlaBottleneck, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA46_c"]) + return model + + +def DLA46x_c(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 2, 2, 1), + channels=(16, 32, 64, 64, 128, 256), + block=DlaBottleneck, + cardinality=32, + base_width=4, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA46x_c"]) + return model + + +def DLA60(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 2, 3, 1), + channels=(16, 32, 128, 256, 512, 1024), + block=DlaBottleneck, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA60"]) + return model + + +def DLA60x(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 2, 3, 1), + channels=(16, 32, 128, 256, 512, 1024), + block=DlaBottleneck, + cardinality=32, + base_width=4, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA60x"]) + return model + + +def DLA60x_c(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 2, 3, 1), + channels=(16, 32, 64, 64, 128, 256), + block=DlaBottleneck, + cardinality=32, + base_width=4, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA60x_c"]) + return model + + +def DLA102(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 3, 4, 1), + channels=(16, 32, 128, 256, 512, 1024), + block=DlaBottleneck, + residual_root=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA102"]) + return model + + +def DLA102x(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 3, 4, 1), + channels=(16, 32, 128, 256, 512, 1024), + block=DlaBottleneck, + cardinality=32, + base_width=4, + residual_root=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA102x"]) + return model + + +def DLA102x2(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 1, 3, 4, 1), + channels=(16, 32, 128, 256, 512, 1024), + block=DlaBottleneck, + cardinality=64, + base_width=4, + residual_root=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA102x2"]) + return model + + +def DLA169(pretrained=False, **kwargs): + model = DLA(levels=(1, 1, 2, 3, 5, 1), + channels=(16, 32, 128, 256, 512, 1024), + block=DlaBottleneck, + residual_root=True, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DLA169"]) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/dpn.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/dpn.py new file mode 100644 index 00000000..55953ed2 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/dpn.py @@ -0,0 +1,451 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import sys +import paddle +from paddle import ParamAttr +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "DPN68": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN68_pretrained.pdparams", + "DPN92": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN92_pretrained.pdparams", + "DPN98": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN98_pretrained.pdparams", + "DPN107": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN107_pretrained.pdparams", + "DPN131": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN131_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + pad=0, + groups=1, + act="relu", + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=pad, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=name + '_bn_scale'), + bias_attr=ParamAttr(name + '_bn_offset'), + moving_mean_name=name + '_bn_mean', + moving_variance_name=name + '_bn_variance') + + def forward(self, input): + y = self._conv(input) + y = self._batch_norm(y) + return y + + +class BNACConvLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + pad=0, + groups=1, + act="relu", + name=None): + super(BNACConvLayer, self).__init__() + self.num_channels = num_channels + + self._batch_norm = BatchNorm( + num_channels, + act=act, + param_attr=ParamAttr(name=name + '_bn_scale'), + bias_attr=ParamAttr(name + '_bn_offset'), + moving_mean_name=name + '_bn_mean', + moving_variance_name=name + '_bn_variance') + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=pad, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + + def forward(self, input): + y = self._batch_norm(input) + y = self._conv(y) + return y + + +class DualPathFactory(nn.Layer): + def __init__(self, + num_channels, + num_1x1_a, + num_3x3_b, + num_1x1_c, + inc, + G, + _type='normal', + name=None): + super(DualPathFactory, self).__init__() + + self.num_1x1_c = num_1x1_c + self.inc = inc + self.name = name + + kw = 3 + kh = 3 + pw = (kw - 1) // 2 + ph = (kh - 1) // 2 + + # type + if _type == 'proj': + key_stride = 1 + self.has_proj = True + elif _type == 'down': + key_stride = 2 + self.has_proj = True + elif _type == 'normal': + key_stride = 1 + self.has_proj = False + else: + print("not implemented now!!!") + sys.exit(1) + + data_in_ch = sum(num_channels) if isinstance(num_channels, + list) else num_channels + + if self.has_proj: + self.c1x1_w_func = BNACConvLayer( + num_channels=data_in_ch, + num_filters=num_1x1_c + 2 * inc, + filter_size=(1, 1), + pad=(0, 0), + stride=(key_stride, key_stride), + name=name + "_match") + + self.c1x1_a_func = BNACConvLayer( + num_channels=data_in_ch, + num_filters=num_1x1_a, + filter_size=(1, 1), + pad=(0, 0), + name=name + "_conv1") + + self.c3x3_b_func = BNACConvLayer( + num_channels=num_1x1_a, + num_filters=num_3x3_b, + filter_size=(kw, kh), + pad=(pw, ph), + stride=(key_stride, key_stride), + groups=G, + name=name + "_conv2") + + self.c1x1_c_func = BNACConvLayer( + num_channels=num_3x3_b, + num_filters=num_1x1_c + inc, + filter_size=(1, 1), + pad=(0, 0), + name=name + "_conv3") + + def forward(self, input): + # PROJ + if isinstance(input, list): + data_in = paddle.concat([input[0], input[1]], axis=1) + else: + data_in = input + + if self.has_proj: + c1x1_w = self.c1x1_w_func(data_in) + data_o1, data_o2 = paddle.split( + c1x1_w, num_or_sections=[self.num_1x1_c, 2 * self.inc], axis=1) + else: + data_o1 = input[0] + data_o2 = input[1] + + c1x1_a = self.c1x1_a_func(data_in) + c3x3_b = self.c3x3_b_func(c1x1_a) + c1x1_c = self.c1x1_c_func(c3x3_b) + + c1x1_c1, c1x1_c2 = paddle.split( + c1x1_c, num_or_sections=[self.num_1x1_c, self.inc], axis=1) + + # OUTPUTS + summ = paddle.add(x=data_o1, y=c1x1_c1) + dense = paddle.concat([data_o2, c1x1_c2], axis=1) + # tensor, channels + return [summ, dense] + + +class DPN(nn.Layer): + def __init__(self, layers=68, class_num=1000): + super(DPN, self).__init__() + + self._class_num = class_num + + args = self.get_net_args(layers) + bws = args['bw'] + inc_sec = args['inc_sec'] + rs = args['r'] + k_r = args['k_r'] + k_sec = args['k_sec'] + G = args['G'] + init_num_filter = args['init_num_filter'] + init_filter_size = args['init_filter_size'] + init_padding = args['init_padding'] + + self.k_sec = k_sec + + self.conv1_x_1_func = ConvBNLayer( + num_channels=3, + num_filters=init_num_filter, + filter_size=init_filter_size, + stride=2, + pad=init_padding, + act='relu', + name="conv1") + + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + num_channel_dpn = init_num_filter + + self.dpn_func_list = [] + #conv2 - conv5 + match_list, num = [], 0 + for gc in range(4): + bw = bws[gc] + inc = inc_sec[gc] + R = (k_r * bw) // rs[gc] + if gc == 0: + _type1 = 'proj' + _type2 = 'normal' + match = 1 + else: + _type1 = 'down' + _type2 = 'normal' + match = match + k_sec[gc - 1] + match_list.append(match) + self.dpn_func_list.append( + self.add_sublayer( + "dpn{}".format(match), + DualPathFactory( + num_channels=num_channel_dpn, + num_1x1_a=R, + num_3x3_b=R, + num_1x1_c=bw, + inc=inc, + G=G, + _type=_type1, + name="dpn" + str(match)))) + num_channel_dpn = [bw, 3 * inc] + + for i_ly in range(2, k_sec[gc] + 1): + num += 1 + if num in match_list: + num += 1 + self.dpn_func_list.append( + self.add_sublayer( + "dpn{}".format(num), + DualPathFactory( + num_channels=num_channel_dpn, + num_1x1_a=R, + num_3x3_b=R, + num_1x1_c=bw, + inc=inc, + G=G, + _type=_type2, + name="dpn" + str(num)))) + + num_channel_dpn = [ + num_channel_dpn[0], num_channel_dpn[1] + inc + ] + + out_channel = sum(num_channel_dpn) + + self.conv5_x_x_bn = BatchNorm( + num_channels=sum(num_channel_dpn), + act="relu", + param_attr=ParamAttr(name='final_concat_bn_scale'), + bias_attr=ParamAttr('final_concat_bn_offset'), + moving_mean_name='final_concat_bn_mean', + moving_variance_name='final_concat_bn_variance') + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + stdv = 0.01 + + self.out = Linear( + out_channel, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_weights"), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, input): + conv1_x_1 = self.conv1_x_1_func(input) + convX_x_x = self.pool2d_max(conv1_x_1) + + dpn_idx = 0 + for gc in range(4): + convX_x_x = self.dpn_func_list[dpn_idx](convX_x_x) + dpn_idx += 1 + for i_ly in range(2, self.k_sec[gc] + 1): + convX_x_x = self.dpn_func_list[dpn_idx](convX_x_x) + dpn_idx += 1 + + conv5_x_x = paddle.concat(convX_x_x, axis=1) + conv5_x_x = self.conv5_x_x_bn(conv5_x_x) + + y = self.pool2d_avg(conv5_x_x) + y = paddle.flatten(y, start_axis=1, stop_axis=-1) + y = self.out(y) + return y + + def get_net_args(self, layers): + if layers == 68: + k_r = 128 + G = 32 + k_sec = [3, 4, 12, 3] + inc_sec = [16, 32, 32, 64] + bw = [64, 128, 256, 512] + r = [64, 64, 64, 64] + init_num_filter = 10 + init_filter_size = 3 + init_padding = 1 + elif layers == 92: + k_r = 96 + G = 32 + k_sec = [3, 4, 20, 3] + inc_sec = [16, 32, 24, 128] + bw = [256, 512, 1024, 2048] + r = [256, 256, 256, 256] + init_num_filter = 64 + init_filter_size = 7 + init_padding = 3 + elif layers == 98: + k_r = 160 + G = 40 + k_sec = [3, 6, 20, 3] + inc_sec = [16, 32, 32, 128] + bw = [256, 512, 1024, 2048] + r = [256, 256, 256, 256] + init_num_filter = 96 + init_filter_size = 7 + init_padding = 3 + elif layers == 107: + k_r = 200 + G = 50 + k_sec = [4, 8, 20, 3] + inc_sec = [20, 64, 64, 128] + bw = [256, 512, 1024, 2048] + r = [256, 256, 256, 256] + init_num_filter = 128 + init_filter_size = 7 + init_padding = 3 + elif layers == 131: + k_r = 160 + G = 40 + k_sec = [4, 8, 28, 3] + inc_sec = [16, 32, 32, 128] + bw = [256, 512, 1024, 2048] + r = [256, 256, 256, 256] + init_num_filter = 128 + init_filter_size = 7 + init_padding = 3 + else: + raise NotImplementedError + net_arg = { + 'k_r': k_r, + 'G': G, + 'k_sec': k_sec, + 'inc_sec': inc_sec, + 'bw': bw, + 'r': r + } + net_arg['init_num_filter'] = init_num_filter + net_arg['init_filter_size'] = init_filter_size + net_arg['init_padding'] = init_padding + + return net_arg + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def DPN68(pretrained=False, use_ssld=False, **kwargs): + model = DPN(layers=68, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DPN68"]) + return model + + +def DPN92(pretrained=False, use_ssld=False, **kwargs): + model = DPN(layers=92, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DPN92"]) + return model + + +def DPN98(pretrained=False, use_ssld=False, **kwargs): + model = DPN(layers=98, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DPN98"]) + return model + + +def DPN107(pretrained=False, use_ssld=False, **kwargs): + model = DPN(layers=107, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DPN107"]) + return model + + +def DPN131(pretrained=False, use_ssld=False, **kwargs): + model = DPN(layers=131, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["DPN131"]) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/efficientnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/efficientnet.py new file mode 100644 index 00000000..bd0cffa6 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/efficientnet.py @@ -0,0 +1,976 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/lukemelas/EfficientNet-PyTorch + +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +import math +import collections +import re +import copy + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "EfficientNetB0_small": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_small_pretrained.pdparams", + "EfficientNetB0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_pretrained.pdparams", + "EfficientNetB1": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB1_pretrained.pdparams", + "EfficientNetB2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB2_pretrained.pdparams", + "EfficientNetB3": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB3_pretrained.pdparams", + "EfficientNetB4": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB4_pretrained.pdparams", + "EfficientNetB5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB5_pretrained.pdparams", + "EfficientNetB6": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB6_pretrained.pdparams", + "EfficientNetB7": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB7_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + +GlobalParams = collections.namedtuple('GlobalParams', [ + 'batch_norm_momentum', + 'batch_norm_epsilon', + 'dropout_rate', + 'num_classes', + 'width_coefficient', + 'depth_coefficient', + 'depth_divisor', + 'min_depth', + 'drop_connect_rate', +]) + +BlockArgs = collections.namedtuple('BlockArgs', [ + 'kernel_size', 'num_repeat', 'input_filters', 'output_filters', + 'expand_ratio', 'id_skip', 'stride', 'se_ratio' +]) + +GlobalParams.__new__.__defaults__ = (None, ) * len(GlobalParams._fields) +BlockArgs.__new__.__defaults__ = (None, ) * len(BlockArgs._fields) + + +def efficientnet_params(model_name): + """ Map EfficientNet model name to parameter coefficients. """ + params_dict = { + # Coefficients: width,depth,resolution,dropout + 'efficientnet-b0': (1.0, 1.0, 224, 0.2), + 'efficientnet-b1': (1.0, 1.1, 240, 0.2), + 'efficientnet-b2': (1.1, 1.2, 260, 0.3), + 'efficientnet-b3': (1.2, 1.4, 300, 0.3), + 'efficientnet-b4': (1.4, 1.8, 380, 0.4), + 'efficientnet-b5': (1.6, 2.2, 456, 0.4), + 'efficientnet-b6': (1.8, 2.6, 528, 0.5), + 'efficientnet-b7': (2.0, 3.1, 600, 0.5), + } + return params_dict[model_name] + + +def efficientnet(width_coefficient=None, + depth_coefficient=None, + dropout_rate=0.2, + drop_connect_rate=0.2): + """ Get block arguments according to parameter and coefficients. """ + blocks_args = [ + 'r1_k3_s11_e1_i32_o16_se0.25', + 'r2_k3_s22_e6_i16_o24_se0.25', + 'r2_k5_s22_e6_i24_o40_se0.25', + 'r3_k3_s22_e6_i40_o80_se0.25', + 'r3_k5_s11_e6_i80_o112_se0.25', + 'r4_k5_s22_e6_i112_o192_se0.25', + 'r1_k3_s11_e6_i192_o320_se0.25', + ] + blocks_args = BlockDecoder.decode(blocks_args) + + global_params = GlobalParams( + batch_norm_momentum=0.99, + batch_norm_epsilon=1e-3, + dropout_rate=dropout_rate, + drop_connect_rate=drop_connect_rate, + num_classes=1000, + width_coefficient=width_coefficient, + depth_coefficient=depth_coefficient, + depth_divisor=8, + min_depth=None) + + return blocks_args, global_params + + +def get_model_params(model_name, override_params): + """ Get the block args and global params for a given model """ + if model_name.startswith('efficientnet'): + w, d, _, p = efficientnet_params(model_name) + blocks_args, global_params = efficientnet( + width_coefficient=w, depth_coefficient=d, dropout_rate=p) + else: + raise NotImplementedError('model name is not pre-defined: %s' % + model_name) + if override_params: + global_params = global_params._replace(**override_params) + return blocks_args, global_params + + +def round_filters(filters, global_params): + """ Calculate and round number of filters based on depth multiplier. """ + multiplier = global_params.width_coefficient + if not multiplier: + return filters + divisor = global_params.depth_divisor + min_depth = global_params.min_depth + filters *= multiplier + min_depth = min_depth or divisor + new_filters = max(min_depth, + int(filters + divisor / 2) // divisor * divisor) + if new_filters < 0.9 * filters: # prevent rounding by more than 10% + new_filters += divisor + return int(new_filters) + + +def round_repeats(repeats, global_params): + """ Round number of filters based on depth multiplier. """ + multiplier = global_params.depth_coefficient + if not multiplier: + return repeats + return int(math.ceil(multiplier * repeats)) + + +class BlockDecoder(object): + """ + Block Decoder, straight from the official TensorFlow repository. + """ + + @staticmethod + def _decode_block_string(block_string): + """ Gets a block through a string notation of arguments. """ + assert isinstance(block_string, str) + + ops = block_string.split('_') + options = {} + for op in ops: + splits = re.split(r'(\d.*)', op) + if len(splits) >= 2: + key, value = splits[:2] + options[key] = value + + # Check stride + cond_1 = ('s' in options and len(options['s']) == 1) + cond_2 = ((len(options['s']) == 2) and + (options['s'][0] == options['s'][1])) + assert (cond_1 or cond_2) + + return BlockArgs( + kernel_size=int(options['k']), + num_repeat=int(options['r']), + input_filters=int(options['i']), + output_filters=int(options['o']), + expand_ratio=int(options['e']), + id_skip=('noskip' not in block_string), + se_ratio=float(options['se']) if 'se' in options else None, + stride=[int(options['s'][0])]) + + @staticmethod + def _encode_block_string(block): + """Encodes a block to a string.""" + args = [ + 'r%d' % block.num_repeat, 'k%d' % block.kernel_size, 's%d%d' % + (block.strides[0], block.strides[1]), 'e%s' % block.expand_ratio, + 'i%d' % block.input_filters, 'o%d' % block.output_filters + ] + if 0 < block.se_ratio <= 1: + args.append('se%s' % block.se_ratio) + if block.id_skip is False: + args.append('noskip') + return '_'.join(args) + + @staticmethod + def decode(string_list): + """ + Decode a list of string notations to specify blocks in the network. + + string_list: list of strings, each string is a notation of block + return + list of BlockArgs namedtuples of block args + """ + assert isinstance(string_list, list) + blocks_args = [] + for block_string in string_list: + blocks_args.append(BlockDecoder._decode_block_string(block_string)) + return blocks_args + + @staticmethod + def encode(blocks_args): + """ + Encodes a list of BlockArgs to a list of strings. + + :param blocks_args: a list of BlockArgs namedtuples of block args + :return: a list of strings, each string is a notation of block + """ + block_strings = [] + for block in blocks_args: + block_strings.append(BlockDecoder._encode_block_string(block)) + return block_strings + + +def initial_type(name, use_bias=False): + param_attr = ParamAttr(name=name + "_weights") + if use_bias: + bias_attr = ParamAttr(name=name + "_offset") + else: + bias_attr = False + return param_attr, bias_attr + + +def init_batch_norm_layer(name="batch_norm"): + param_attr = ParamAttr(name=name + "_scale") + bias_attr = ParamAttr(name=name + "_offset") + return param_attr, bias_attr + + +def init_fc_layer(name="fc"): + param_attr = ParamAttr(name=name + "_weights") + bias_attr = ParamAttr(name=name + "_offset") + return param_attr, bias_attr + + +def cal_padding(img_size, stride, filter_size, dilation=1): + """Calculate padding size.""" + if img_size % stride == 0: + out_size = max(filter_size - stride, 0) + else: + out_size = max(filter_size - (img_size % stride), 0) + return out_size // 2, out_size - out_size // 2 + + +inp_shape = { + "b0_small": [224, 112, 112, 56, 28, 14, 14, 7], + "b0": [224, 112, 112, 56, 28, 14, 14, 7], + "b1": [240, 120, 120, 60, 30, 15, 15, 8], + "b2": [260, 130, 130, 65, 33, 17, 17, 9], + "b3": [300, 150, 150, 75, 38, 19, 19, 10], + "b4": [380, 190, 190, 95, 48, 24, 24, 12], + "b5": [456, 228, 228, 114, 57, 29, 29, 15], + "b6": [528, 264, 264, 132, 66, 33, 33, 17], + "b7": [600, 300, 300, 150, 75, 38, 38, 19] +} + + +def _drop_connect(inputs, prob, is_test): + if is_test: + output = inputs + else: + keep_prob = 1.0 - prob + inputs_shape = paddle.shape(inputs) + random_tensor = keep_prob + paddle.rand( + shape=[inputs_shape[0], 1, 1, 1]) + binary_tensor = paddle.floor(random_tensor) + output = paddle.multiply(inputs, binary_tensor) / keep_prob + return output + + +class Conv2ds(nn.Layer): + def __init__(self, + input_channels, + output_channels, + filter_size, + stride=1, + padding=0, + groups=None, + name="conv2d", + act=None, + use_bias=False, + padding_type=None, + model_name=None, + cur_stage=None): + super(Conv2ds, self).__init__() + assert act in [None, "swish", "sigmoid"] + self.act = act + + param_attr, bias_attr = initial_type(name=name, use_bias=use_bias) + + def get_padding(filter_size, stride=1, dilation=1): + padding = ((stride - 1) + dilation * (filter_size - 1)) // 2 + return padding + + inps = 1 if model_name == None and cur_stage == None else inp_shape[ + model_name][cur_stage] + self.need_crop = False + if padding_type == "SAME": + top_padding, bottom_padding = cal_padding(inps, stride, + filter_size) + left_padding, right_padding = cal_padding(inps, stride, + filter_size) + height_padding = bottom_padding + width_padding = right_padding + if top_padding != bottom_padding or left_padding != right_padding: + height_padding = top_padding + stride + width_padding = left_padding + stride + self.need_crop = True + padding = [height_padding, width_padding] + elif padding_type == "VALID": + height_padding = 0 + width_padding = 0 + padding = [height_padding, width_padding] + elif padding_type == "DYNAMIC": + padding = get_padding(filter_size, stride) + else: + padding = padding_type + + groups = 1 if groups is None else groups + self._conv = Conv2D( + input_channels, + output_channels, + filter_size, + groups=groups, + stride=stride, + # act=act, + padding=padding, + weight_attr=param_attr, + bias_attr=bias_attr) + + def forward(self, inputs): + x = self._conv(inputs) + if self.act == "swish": + x = F.swish(x) + elif self.act == "sigmoid": + x = F.sigmoid(x) + + if self.need_crop: + x = x[:, :, 1:, 1:] + return x + + +class ConvBNLayer(nn.Layer): + def __init__(self, + input_channels, + filter_size, + output_channels, + stride=1, + num_groups=1, + padding_type="SAME", + conv_act=None, + bn_act="swish", + use_bn=True, + use_bias=False, + name=None, + conv_name=None, + bn_name=None, + model_name=None, + cur_stage=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2ds( + input_channels=input_channels, + output_channels=output_channels, + filter_size=filter_size, + stride=stride, + groups=num_groups, + act=conv_act, + padding_type=padding_type, + name=conv_name, + use_bias=use_bias, + model_name=model_name, + cur_stage=cur_stage) + self.use_bn = use_bn + if use_bn is True: + bn_name = name + bn_name + param_attr, bias_attr = init_batch_norm_layer(bn_name) + + self._bn = BatchNorm( + num_channels=output_channels, + act=bn_act, + momentum=0.99, + epsilon=0.001, + moving_mean_name=bn_name + "_mean", + moving_variance_name=bn_name + "_variance", + param_attr=param_attr, + bias_attr=bias_attr) + + def forward(self, inputs): + if self.use_bn: + x = self._conv(inputs) + x = self._bn(x) + return x + else: + return self._conv(inputs) + + +class ExpandConvNorm(nn.Layer): + def __init__(self, + input_channels, + block_args, + padding_type, + name=None, + model_name=None, + cur_stage=None): + super(ExpandConvNorm, self).__init__() + + self.oup = block_args.input_filters * block_args.expand_ratio + self.expand_ratio = block_args.expand_ratio + + if self.expand_ratio != 1: + self._conv = ConvBNLayer( + input_channels, + 1, + self.oup, + bn_act=None, + padding_type=padding_type, + name=name, + conv_name=name + "_expand_conv", + bn_name="_bn0", + model_name=model_name, + cur_stage=cur_stage) + + def forward(self, inputs): + if self.expand_ratio != 1: + return self._conv(inputs) + else: + return inputs + + +class DepthwiseConvNorm(nn.Layer): + def __init__(self, + input_channels, + block_args, + padding_type, + name=None, + model_name=None, + cur_stage=None): + super(DepthwiseConvNorm, self).__init__() + + self.k = block_args.kernel_size + self.s = block_args.stride + if isinstance(self.s, list) or isinstance(self.s, tuple): + self.s = self.s[0] + oup = block_args.input_filters * block_args.expand_ratio + + self._conv = ConvBNLayer( + input_channels, + self.k, + oup, + self.s, + num_groups=input_channels, + bn_act=None, + padding_type=padding_type, + name=name, + conv_name=name + "_depthwise_conv", + bn_name="_bn1", + model_name=model_name, + cur_stage=cur_stage) + + def forward(self, inputs): + return self._conv(inputs) + + +class ProjectConvNorm(nn.Layer): + def __init__(self, + input_channels, + block_args, + padding_type, + name=None, + model_name=None, + cur_stage=None): + super(ProjectConvNorm, self).__init__() + + final_oup = block_args.output_filters + + self._conv = ConvBNLayer( + input_channels, + 1, + final_oup, + bn_act=None, + padding_type=padding_type, + name=name, + conv_name=name + "_project_conv", + bn_name="_bn2", + model_name=model_name, + cur_stage=cur_stage) + + def forward(self, inputs): + return self._conv(inputs) + + +class SEBlock(nn.Layer): + def __init__(self, + input_channels, + num_squeezed_channels, + oup, + padding_type, + name=None, + model_name=None, + cur_stage=None): + super(SEBlock, self).__init__() + + self._pool = AdaptiveAvgPool2D(1) + self._conv1 = Conv2ds( + input_channels, + num_squeezed_channels, + 1, + use_bias=True, + padding_type=padding_type, + act="swish", + name=name + "_se_reduce") + + self._conv2 = Conv2ds( + num_squeezed_channels, + oup, + 1, + act="sigmoid", + use_bias=True, + padding_type=padding_type, + name=name + "_se_expand") + + def forward(self, inputs): + x = self._pool(inputs) + x = self._conv1(x) + x = self._conv2(x) + out = paddle.multiply(inputs, x) + return out + + +class MbConvBlock(nn.Layer): + def __init__(self, + input_channels, + block_args, + padding_type, + use_se, + name=None, + drop_connect_rate=None, + model_name=None, + cur_stage=None): + super(MbConvBlock, self).__init__() + + oup = block_args.input_filters * block_args.expand_ratio + self.block_args = block_args + self.has_se = use_se and (block_args.se_ratio is not None) and ( + 0 < block_args.se_ratio <= 1) + self.id_skip = block_args.id_skip + self.expand_ratio = block_args.expand_ratio + self.drop_connect_rate = drop_connect_rate + + if self.expand_ratio != 1: + self._ecn = ExpandConvNorm( + input_channels, + block_args, + padding_type=padding_type, + name=name, + model_name=model_name, + cur_stage=cur_stage) + + self._dcn = DepthwiseConvNorm( + input_channels * block_args.expand_ratio, + block_args, + padding_type=padding_type, + name=name, + model_name=model_name, + cur_stage=cur_stage) + + if self.has_se: + num_squeezed_channels = max( + 1, int(block_args.input_filters * block_args.se_ratio)) + self._se = SEBlock( + input_channels * block_args.expand_ratio, + num_squeezed_channels, + oup, + padding_type=padding_type, + name=name, + model_name=model_name, + cur_stage=cur_stage) + + self._pcn = ProjectConvNorm( + input_channels * block_args.expand_ratio, + block_args, + padding_type=padding_type, + name=name, + model_name=model_name, + cur_stage=cur_stage) + + def forward(self, inputs): + x = inputs + if self.expand_ratio != 1: + x = self._ecn(x) + x = F.swish(x) + + x = self._dcn(x) + x = F.swish(x) + if self.has_se: + x = self._se(x) + x = self._pcn(x) + + if self.id_skip and \ + self.block_args.stride == 1 and \ + self.block_args.input_filters == self.block_args.output_filters: + if self.drop_connect_rate: + x = _drop_connect(x, self.drop_connect_rate, not self.training) + x = paddle.add(x, inputs) + return x + + +class ConvStemNorm(nn.Layer): + def __init__(self, + input_channels, + padding_type, + _global_params, + name=None, + model_name=None, + cur_stage=None): + super(ConvStemNorm, self).__init__() + + output_channels = round_filters(32, _global_params) + self._conv = ConvBNLayer( + input_channels, + filter_size=3, + output_channels=output_channels, + stride=2, + bn_act=None, + padding_type=padding_type, + name="", + conv_name="_conv_stem", + bn_name="_bn0", + model_name=model_name, + cur_stage=cur_stage) + + def forward(self, inputs): + return self._conv(inputs) + + +class ExtractFeatures(nn.Layer): + def __init__(self, + input_channels, + _block_args, + _global_params, + padding_type, + use_se, + model_name=None): + super(ExtractFeatures, self).__init__() + + self._global_params = _global_params + + self._conv_stem = ConvStemNorm( + input_channels, + padding_type=padding_type, + _global_params=_global_params, + model_name=model_name, + cur_stage=0) + + self.block_args_copy = copy.deepcopy(_block_args) + idx = 0 + block_size = 0 + for block_arg in self.block_args_copy: + block_arg = block_arg._replace( + input_filters=round_filters(block_arg.input_filters, + _global_params), + output_filters=round_filters(block_arg.output_filters, + _global_params), + num_repeat=round_repeats(block_arg.num_repeat, _global_params)) + block_size += 1 + for _ in range(block_arg.num_repeat - 1): + block_size += 1 + + self.conv_seq = [] + cur_stage = 1 + for block_args in _block_args: + block_args = block_args._replace( + input_filters=round_filters(block_args.input_filters, + _global_params), + output_filters=round_filters(block_args.output_filters, + _global_params), + num_repeat=round_repeats(block_args.num_repeat, + _global_params)) + + drop_connect_rate = self._global_params.drop_connect_rate + if drop_connect_rate: + drop_connect_rate *= float(idx) / block_size + + _mc_block = self.add_sublayer( + "_blocks." + str(idx) + ".", + MbConvBlock( + block_args.input_filters, + block_args=block_args, + padding_type=padding_type, + use_se=use_se, + name="_blocks." + str(idx) + ".", + drop_connect_rate=drop_connect_rate, + model_name=model_name, + cur_stage=cur_stage)) + self.conv_seq.append(_mc_block) + idx += 1 + if block_args.num_repeat > 1: + block_args = block_args._replace( + input_filters=block_args.output_filters, stride=1) + for _ in range(block_args.num_repeat - 1): + drop_connect_rate = self._global_params.drop_connect_rate + if drop_connect_rate: + drop_connect_rate *= float(idx) / block_size + _mc_block = self.add_sublayer( + "block." + str(idx) + ".", + MbConvBlock( + block_args.input_filters, + block_args, + padding_type=padding_type, + use_se=use_se, + name="_blocks." + str(idx) + ".", + drop_connect_rate=drop_connect_rate, + model_name=model_name, + cur_stage=cur_stage)) + self.conv_seq.append(_mc_block) + idx += 1 + cur_stage += 1 + + def forward(self, inputs): + x = self._conv_stem(inputs) + x = F.swish(x) + for _mc_block in self.conv_seq: + x = _mc_block(x) + return x + + +class EfficientNet(nn.Layer): + def __init__(self, + name="b0", + padding_type="SAME", + override_params=None, + use_se=True, + class_num=1000): + super(EfficientNet, self).__init__() + + model_name = 'efficientnet-' + name + self.name = name + self._block_args, self._global_params = get_model_params( + model_name, override_params) + self.padding_type = padding_type + self.use_se = use_se + + self._ef = ExtractFeatures( + 3, + self._block_args, + self._global_params, + self.padding_type, + self.use_se, + model_name=self.name) + + output_channels = round_filters(1280, self._global_params) + if name == "b0_small" or name == "b0" or name == "b1": + oup = 320 + elif name == "b2": + oup = 352 + elif name == "b3": + oup = 384 + elif name == "b4": + oup = 448 + elif name == "b5": + oup = 512 + elif name == "b6": + oup = 576 + elif name == "b7": + oup = 640 + self._conv = ConvBNLayer( + oup, + 1, + output_channels, + bn_act="swish", + padding_type=self.padding_type, + name="", + conv_name="_conv_head", + bn_name="_bn1", + model_name=self.name, + cur_stage=7) + self._pool = AdaptiveAvgPool2D(1) + + if self._global_params.dropout_rate: + self._drop = Dropout( + p=self._global_params.dropout_rate, mode="upscale_in_train") + + param_attr, bias_attr = init_fc_layer("_fc") + self._fc = Linear( + output_channels, + class_num, + weight_attr=param_attr, + bias_attr=bias_attr) + + def forward(self, inputs): + x = self._ef(inputs) + x = self._conv(x) + x = self._pool(x) + if self._global_params.dropout_rate: + x = self._drop(x) + x = paddle.squeeze(x, axis=[2, 3]) + x = self._fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def EfficientNetB0_small(padding_type='DYNAMIC', + override_params=None, + use_se=False, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b0', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB0_small"]) + return model + + +def EfficientNetB0(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b0', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB0"]) + return model + + +def EfficientNetB1(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b1', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB1"]) + return model + + +def EfficientNetB2(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b2', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB2"]) + return model + + +def EfficientNetB3(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b3', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB3"]) + return model + + +def EfficientNetB4(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b4', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB4"]) + return model + + +def EfficientNetB5(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b5', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB5"]) + return model + + +def EfficientNetB6(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b6', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB6"]) + return model + + +def EfficientNetB7(padding_type='SAME', + override_params=None, + use_se=True, + pretrained=False, + use_ssld=False, + **kwargs): + model = EfficientNet( + name='b7', + padding_type=padding_type, + override_params=override_params, + use_se=use_se, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["EfficientNetB7"]) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/ghostnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/ghostnet.py new file mode 100644 index 00000000..4d338c1d --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/ghostnet.py @@ -0,0 +1,363 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/huawei-noah/CV-Backbones/tree/master/ghostnet_pytorch + +import math +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, AdaptiveAvgPool2D, Linear +from paddle.regularizer import L2Decay +from paddle.nn.initializer import Uniform, KaimingNormal + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "GhostNet_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x0_5_pretrained.pdparams", + "GhostNet_x1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_0_pretrained.pdparams", + "GhostNet_x1_3": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + in_channels, + out_channels, + kernel_size, + stride=1, + groups=1, + act="relu", + name=None): + super(ConvBNLayer, self).__init__() + self._conv = Conv2D( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=(kernel_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr( + initializer=KaimingNormal(), name=name + "_weights"), + bias_attr=False) + bn_name = name + "_bn" + + self._batch_norm = BatchNorm( + num_channels=out_channels, + act=act, + param_attr=ParamAttr( + name=bn_name + "_scale", regularizer=L2Decay(0.0)), + bias_attr=ParamAttr( + name=bn_name + "_offset", regularizer=L2Decay(0.0)), + moving_mean_name=bn_name + "_mean", + moving_variance_name=bn_name + "_variance") + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class SEBlock(nn.Layer): + def __init__(self, num_channels, reduction_ratio=4, name=None): + super(SEBlock, self).__init__() + self.pool2d_gap = AdaptiveAvgPool2D(1) + self._num_channels = num_channels + stdv = 1.0 / math.sqrt(num_channels * 1.0) + med_ch = num_channels // reduction_ratio + self.squeeze = Linear( + num_channels, + med_ch, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_1_weights"), + bias_attr=ParamAttr(name=name + "_1_offset")) + stdv = 1.0 / math.sqrt(med_ch * 1.0) + self.excitation = Linear( + med_ch, + num_channels, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_2_weights"), + bias_attr=ParamAttr(name=name + "_2_offset")) + + def forward(self, inputs): + pool = self.pool2d_gap(inputs) + pool = paddle.squeeze(pool, axis=[2, 3]) + squeeze = self.squeeze(pool) + squeeze = F.relu(squeeze) + excitation = self.excitation(squeeze) + excitation = paddle.clip(x=excitation, min=0, max=1) + excitation = paddle.unsqueeze(excitation, axis=[2, 3]) + out = paddle.multiply(inputs, excitation) + return out + + +class GhostModule(nn.Layer): + def __init__(self, + in_channels, + output_channels, + kernel_size=1, + ratio=2, + dw_size=3, + stride=1, + relu=True, + name=None): + super(GhostModule, self).__init__() + init_channels = int(math.ceil(output_channels / ratio)) + new_channels = int(init_channels * (ratio - 1)) + self.primary_conv = ConvBNLayer( + in_channels=in_channels, + out_channels=init_channels, + kernel_size=kernel_size, + stride=stride, + groups=1, + act="relu" if relu else None, + name=name + "_primary_conv") + self.cheap_operation = ConvBNLayer( + in_channels=init_channels, + out_channels=new_channels, + kernel_size=dw_size, + stride=1, + groups=init_channels, + act="relu" if relu else None, + name=name + "_cheap_operation") + + def forward(self, inputs): + x = self.primary_conv(inputs) + y = self.cheap_operation(x) + out = paddle.concat([x, y], axis=1) + return out + + +class GhostBottleneck(nn.Layer): + def __init__(self, + in_channels, + hidden_dim, + output_channels, + kernel_size, + stride, + use_se, + name=None): + super(GhostBottleneck, self).__init__() + self._stride = stride + self._use_se = use_se + self._num_channels = in_channels + self._output_channels = output_channels + self.ghost_module_1 = GhostModule( + in_channels=in_channels, + output_channels=hidden_dim, + kernel_size=1, + stride=1, + relu=True, + name=name + "_ghost_module_1") + if stride == 2: + self.depthwise_conv = ConvBNLayer( + in_channels=hidden_dim, + out_channels=hidden_dim, + kernel_size=kernel_size, + stride=stride, + groups=hidden_dim, + act=None, + name=name + + "_depthwise_depthwise" # looks strange due to an old typo, will be fixed later. + ) + if use_se: + self.se_block = SEBlock(num_channels=hidden_dim, name=name + "_se") + self.ghost_module_2 = GhostModule( + in_channels=hidden_dim, + output_channels=output_channels, + kernel_size=1, + relu=False, + name=name + "_ghost_module_2") + if stride != 1 or in_channels != output_channels: + self.shortcut_depthwise = ConvBNLayer( + in_channels=in_channels, + out_channels=in_channels, + kernel_size=kernel_size, + stride=stride, + groups=in_channels, + act=None, + name=name + + "_shortcut_depthwise_depthwise" # looks strange due to an old typo, will be fixed later. + ) + self.shortcut_conv = ConvBNLayer( + in_channels=in_channels, + out_channels=output_channels, + kernel_size=1, + stride=1, + groups=1, + act=None, + name=name + "_shortcut_conv") + + def forward(self, inputs): + x = self.ghost_module_1(inputs) + if self._stride == 2: + x = self.depthwise_conv(x) + if self._use_se: + x = self.se_block(x) + x = self.ghost_module_2(x) + if self._stride == 1 and self._num_channels == self._output_channels: + shortcut = inputs + else: + shortcut = self.shortcut_depthwise(inputs) + shortcut = self.shortcut_conv(shortcut) + return paddle.add(x=x, y=shortcut) + + +class GhostNet(nn.Layer): + def __init__(self, scale, class_num=1000): + super(GhostNet, self).__init__() + self.cfgs = [ + # k, t, c, SE, s + [3, 16, 16, 0, 1], + [3, 48, 24, 0, 2], + [3, 72, 24, 0, 1], + [5, 72, 40, 1, 2], + [5, 120, 40, 1, 1], + [3, 240, 80, 0, 2], + [3, 200, 80, 0, 1], + [3, 184, 80, 0, 1], + [3, 184, 80, 0, 1], + [3, 480, 112, 1, 1], + [3, 672, 112, 1, 1], + [5, 672, 160, 1, 2], + [5, 960, 160, 0, 1], + [5, 960, 160, 1, 1], + [5, 960, 160, 0, 1], + [5, 960, 160, 1, 1] + ] + self.scale = scale + output_channels = int(self._make_divisible(16 * self.scale, 4)) + self.conv1 = ConvBNLayer( + in_channels=3, + out_channels=output_channels, + kernel_size=3, + stride=2, + groups=1, + act="relu", + name="conv1") + # build inverted residual blocks + idx = 0 + self.ghost_bottleneck_list = [] + for k, exp_size, c, use_se, s in self.cfgs: + in_channels = output_channels + output_channels = int(self._make_divisible(c * self.scale, 4)) + hidden_dim = int(self._make_divisible(exp_size * self.scale, 4)) + ghost_bottleneck = self.add_sublayer( + name="_ghostbottleneck_" + str(idx), + sublayer=GhostBottleneck( + in_channels=in_channels, + hidden_dim=hidden_dim, + output_channels=output_channels, + kernel_size=k, + stride=s, + use_se=use_se, + name="_ghostbottleneck_" + str(idx))) + self.ghost_bottleneck_list.append(ghost_bottleneck) + idx += 1 + # build last several layers + in_channels = output_channels + output_channels = int(self._make_divisible(exp_size * self.scale, 4)) + self.conv_last = ConvBNLayer( + in_channels=in_channels, + out_channels=output_channels, + kernel_size=1, + stride=1, + groups=1, + act="relu", + name="conv_last") + self.pool2d_gap = AdaptiveAvgPool2D(1) + in_channels = output_channels + self._fc0_output_channels = 1280 + self.fc_0 = ConvBNLayer( + in_channels=in_channels, + out_channels=self._fc0_output_channels, + kernel_size=1, + stride=1, + act="relu", + name="fc_0") + self.dropout = nn.Dropout(p=0.2) + stdv = 1.0 / math.sqrt(self._fc0_output_channels * 1.0) + self.fc_1 = Linear( + self._fc0_output_channels, + class_num, + weight_attr=ParamAttr( + name="fc_1_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr(name="fc_1_offset")) + + def forward(self, inputs): + x = self.conv1(inputs) + for ghost_bottleneck in self.ghost_bottleneck_list: + x = ghost_bottleneck(x) + x = self.conv_last(x) + x = self.pool2d_gap(x) + x = self.fc_0(x) + x = self.dropout(x) + x = paddle.reshape(x, shape=[-1, self._fc0_output_channels]) + x = self.fc_1(x) + return x + + def _make_divisible(self, v, divisor, min_value=None): + """ + This function is taken from the original tf repo. + It ensures that all layers have a channel number that is divisible by 8 + It can be seen here: + https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py + """ + if min_value is None: + min_value = divisor + new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) + # Make sure that round down does not go down by more than 10%. + if new_v < 0.9 * v: + new_v += divisor + return new_v + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def GhostNet_x0_5(pretrained=False, use_ssld=False, **kwargs): + model = GhostNet(scale=0.5, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["GhostNet_x0_5"], use_ssld=use_ssld) + return model + + +def GhostNet_x1_0(pretrained=False, use_ssld=False, **kwargs): + model = GhostNet(scale=1.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["GhostNet_x1_0"], use_ssld=use_ssld) + return model + + +def GhostNet_x1_3(pretrained=False, use_ssld=False, **kwargs): + model = GhostNet(scale=1.3, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["GhostNet_x1_3"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/googlenet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/googlenet.py new file mode 100644 index 00000000..22528427 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/googlenet.py @@ -0,0 +1,229 @@ +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "GoogLeNet": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GoogLeNet_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +def xavier(channels, filter_size, name): + stdv = (3.0 / (filter_size**2 * channels))**0.5 + param_attr = ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_weights") + return param_attr + + +class ConvLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + act=None, + name=None): + super(ConvLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + + def forward(self, inputs): + y = self._conv(inputs) + return y + + +class Inception(nn.Layer): + def __init__(self, + input_channels, + output_channels, + filter1, + filter3R, + filter3, + filter5R, + filter5, + proj, + name=None): + super(Inception, self).__init__() + + self._conv1 = ConvLayer( + input_channels, filter1, 1, name="inception_" + name + "_1x1") + self._conv3r = ConvLayer( + input_channels, + filter3R, + 1, + name="inception_" + name + "_3x3_reduce") + self._conv3 = ConvLayer( + filter3R, filter3, 3, name="inception_" + name + "_3x3") + self._conv5r = ConvLayer( + input_channels, + filter5R, + 1, + name="inception_" + name + "_5x5_reduce") + self._conv5 = ConvLayer( + filter5R, filter5, 5, name="inception_" + name + "_5x5") + self._pool = MaxPool2D(kernel_size=3, stride=1, padding=1) + + self._convprj = ConvLayer( + input_channels, proj, 1, name="inception_" + name + "_3x3_proj") + + def forward(self, inputs): + conv1 = self._conv1(inputs) + + conv3r = self._conv3r(inputs) + conv3 = self._conv3(conv3r) + + conv5r = self._conv5r(inputs) + conv5 = self._conv5(conv5r) + + pool = self._pool(inputs) + convprj = self._convprj(pool) + + cat = paddle.concat([conv1, conv3, conv5, convprj], axis=1) + cat = F.relu(cat) + return cat + + +class GoogLeNetDY(nn.Layer): + def __init__(self, class_num=1000): + super(GoogLeNetDY, self).__init__() + self._conv = ConvLayer(3, 64, 7, 2, name="conv1") + self._pool = MaxPool2D(kernel_size=3, stride=2) + self._conv_1 = ConvLayer(64, 64, 1, name="conv2_1x1") + self._conv_2 = ConvLayer(64, 192, 3, name="conv2_3x3") + + self._ince3a = Inception( + 192, 192, 64, 96, 128, 16, 32, 32, name="ince3a") + self._ince3b = Inception( + 256, 256, 128, 128, 192, 32, 96, 64, name="ince3b") + + self._ince4a = Inception( + 480, 480, 192, 96, 208, 16, 48, 64, name="ince4a") + self._ince4b = Inception( + 512, 512, 160, 112, 224, 24, 64, 64, name="ince4b") + self._ince4c = Inception( + 512, 512, 128, 128, 256, 24, 64, 64, name="ince4c") + self._ince4d = Inception( + 512, 512, 112, 144, 288, 32, 64, 64, name="ince4d") + self._ince4e = Inception( + 528, 528, 256, 160, 320, 32, 128, 128, name="ince4e") + + self._ince5a = Inception( + 832, 832, 256, 160, 320, 32, 128, 128, name="ince5a") + self._ince5b = Inception( + 832, 832, 384, 192, 384, 48, 128, 128, name="ince5b") + + self._pool_5 = AdaptiveAvgPool2D(1) + + self._drop = Dropout(p=0.4, mode="downscale_in_infer") + self._fc_out = Linear( + 1024, + class_num, + weight_attr=xavier(1024, 1, "out"), + bias_attr=ParamAttr(name="out_offset")) + self._pool_o1 = AvgPool2D(kernel_size=5, stride=3) + self._conv_o1 = ConvLayer(512, 128, 1, name="conv_o1") + self._fc_o1 = Linear( + 1152, + 1024, + weight_attr=xavier(2048, 1, "fc_o1"), + bias_attr=ParamAttr(name="fc_o1_offset")) + self._drop_o1 = Dropout(p=0.7, mode="downscale_in_infer") + self._out1 = Linear( + 1024, + class_num, + weight_attr=xavier(1024, 1, "out1"), + bias_attr=ParamAttr(name="out1_offset")) + self._pool_o2 = AvgPool2D(kernel_size=5, stride=3) + self._conv_o2 = ConvLayer(528, 128, 1, name="conv_o2") + self._fc_o2 = Linear( + 1152, + 1024, + weight_attr=xavier(2048, 1, "fc_o2"), + bias_attr=ParamAttr(name="fc_o2_offset")) + self._drop_o2 = Dropout(p=0.7, mode="downscale_in_infer") + self._out2 = Linear( + 1024, + class_num, + weight_attr=xavier(1024, 1, "out2"), + bias_attr=ParamAttr(name="out2_offset")) + + def forward(self, inputs): + x = self._conv(inputs) + x = self._pool(x) + x = self._conv_1(x) + x = self._conv_2(x) + x = self._pool(x) + + x = self._ince3a(x) + x = self._ince3b(x) + x = self._pool(x) + + ince4a = self._ince4a(x) + x = self._ince4b(ince4a) + x = self._ince4c(x) + ince4d = self._ince4d(x) + x = self._ince4e(ince4d) + x = self._pool(x) + + x = self._ince5a(x) + ince5b = self._ince5b(x) + + x = self._pool_5(ince5b) + x = self._drop(x) + x = paddle.squeeze(x, axis=[2, 3]) + out = self._fc_out(x) + + x = self._pool_o1(ince4a) + x = self._conv_o1(x) + x = paddle.flatten(x, start_axis=1, stop_axis=-1) + x = self._fc_o1(x) + x = F.relu(x) + x = self._drop_o1(x) + out1 = self._out1(x) + + x = self._pool_o2(ince4d) + x = self._conv_o2(x) + x = paddle.flatten(x, start_axis=1, stop_axis=-1) + x = self._fc_o2(x) + x = self._drop_o2(x) + out2 = self._out2(x) + return [out, out1, out2] + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def GoogLeNet(pretrained=False, use_ssld=False, **kwargs): + model = GoogLeNetDY(**kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["GoogLeNet"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/gvt.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/gvt.py new file mode 100644 index 00000000..2af7ccf4 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/gvt.py @@ -0,0 +1,693 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/Meituan-AutoML/Twins + +from functools import partial + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.regularizer import L2Decay + +from .vision_transformer import trunc_normal_, normal_, zeros_, ones_, to_2tuple, DropPath, Identity, Mlp +from .vision_transformer import Block as ViTBlock + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "pcpvt_small": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_small_pretrained.pdparams", + "pcpvt_base": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_base_pretrained.pdparams", + "pcpvt_large": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_large_pretrained.pdparams", + "alt_gvt_small": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_small_pretrained.pdparams", + "alt_gvt_base": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_base_pretrained.pdparams", + "alt_gvt_large": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_large_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +class GroupAttention(nn.Layer): + """LSA: self attention within a group. + """ + + def __init__(self, + dim, + num_heads=8, + qkv_bias=False, + qk_scale=None, + attn_drop=0., + proj_drop=0., + ws=1): + super().__init__() + if ws == 1: + raise Exception("ws {ws} should not be 1") + if dim % num_heads != 0: + raise Exception( + "dim {dim} should be divided by num_heads {num_heads}.") + + self.dim = dim + self.num_heads = num_heads + head_dim = dim // num_heads + self.scale = qk_scale or head_dim**-0.5 + + self.qkv = nn.Linear(dim, dim * 3, bias_attr=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + self.proj_drop = nn.Dropout(proj_drop) + self.ws = ws + + def forward(self, x, H, W): + B, N, C = x.shape + h_group, w_group = H // self.ws, W // self.ws + total_groups = h_group * w_group + x = x.reshape([B, h_group, self.ws, w_group, self.ws, C]).transpose( + [0, 1, 3, 2, 4, 5]) + qkv = self.qkv(x).reshape([ + B, total_groups, self.ws**2, 3, self.num_heads, C // self.num_heads + ]).transpose([3, 0, 1, 4, 2, 5]) + q, k, v = qkv[0], qkv[1], qkv[2] + attn = paddle.matmul(q, k.transpose([0, 1, 2, 4, 3])) * self.scale + + attn = nn.Softmax(axis=-1)(attn) + attn = self.attn_drop(attn) + attn = paddle.matmul(attn, v).transpose([0, 1, 3, 2, 4]).reshape( + [B, h_group, w_group, self.ws, self.ws, C]) + + x = attn.transpose([0, 1, 3, 2, 4, 5]).reshape([B, N, C]) + x = self.proj(x) + x = self.proj_drop(x) + return x + + +class Attention(nn.Layer): + """GSA: using a key to summarize the information for a group to be efficient. + """ + + def __init__(self, + dim, + num_heads=8, + qkv_bias=False, + qk_scale=None, + attn_drop=0., + proj_drop=0., + sr_ratio=1): + super().__init__() + assert dim % num_heads == 0, f"dim {dim} should be divided by num_heads {num_heads}." + + self.dim = dim + self.num_heads = num_heads + head_dim = dim // num_heads + self.scale = qk_scale or head_dim**-0.5 + + self.q = nn.Linear(dim, dim, bias_attr=qkv_bias) + self.kv = nn.Linear(dim, dim * 2, bias_attr=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + self.proj_drop = nn.Dropout(proj_drop) + + self.sr_ratio = sr_ratio + if sr_ratio > 1: + self.sr = nn.Conv2D( + dim, dim, kernel_size=sr_ratio, stride=sr_ratio) + self.norm = nn.LayerNorm(dim) + + def forward(self, x, H, W): + B, N, C = x.shape + q = self.q(x).reshape( + [B, N, self.num_heads, C // self.num_heads]).transpose( + [0, 2, 1, 3]) + + if self.sr_ratio > 1: + x_ = x.transpose([0, 2, 1]).reshape([B, C, H, W]) + tmp_n = H * W // self.sr_ratio**2 + x_ = self.sr(x_).reshape([B, C, tmp_n]).transpose([0, 2, 1]) + x_ = self.norm(x_) + kv = self.kv(x_).reshape( + [B, tmp_n, 2, self.num_heads, C // self.num_heads]).transpose( + [2, 0, 3, 1, 4]) + else: + kv = self.kv(x).reshape( + [B, N, 2, self.num_heads, C // self.num_heads]).transpose( + [2, 0, 3, 1, 4]) + k, v = kv[0], kv[1] + + attn = paddle.matmul(q, k.transpose([0, 1, 3, 2])) * self.scale + attn = nn.Softmax(axis=-1)(attn) + attn = self.attn_drop(attn) + + x = paddle.matmul(attn, v).transpose([0, 2, 1, 3]).reshape([B, N, C]) + x = self.proj(x) + x = self.proj_drop(x) + return x + + +class Block(nn.Layer): + def __init__(self, + dim, + num_heads, + mlp_ratio=4., + qkv_bias=False, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer=nn.LayerNorm, + sr_ratio=1): + super().__init__() + self.norm1 = norm_layer(dim) + self.attn = Attention( + dim, + num_heads=num_heads, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + attn_drop=attn_drop, + proj_drop=drop, + sr_ratio=sr_ratio) + self.drop_path = DropPath(drop_path) if drop_path > 0. else Identity() + self.norm2 = norm_layer(dim) + mlp_hidden_dim = int(dim * mlp_ratio) + self.mlp = Mlp(in_features=dim, + hidden_features=mlp_hidden_dim, + act_layer=act_layer, + drop=drop) + + def forward(self, x, H, W): + x = x + self.drop_path(self.attn(self.norm1(x), H, W)) + x = x + self.drop_path(self.mlp(self.norm2(x))) + return x + + +class SBlock(ViTBlock): + def __init__(self, + dim, + num_heads, + mlp_ratio=4., + qkv_bias=False, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer=nn.LayerNorm, + sr_ratio=1): + super().__init__(dim, num_heads, mlp_ratio, qkv_bias, qk_scale, drop, + attn_drop, drop_path, act_layer, norm_layer) + + def forward(self, x, H, W): + return super().forward(x) + + +class GroupBlock(ViTBlock): + def __init__(self, + dim, + num_heads, + mlp_ratio=4., + qkv_bias=False, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer=nn.LayerNorm, + sr_ratio=1, + ws=1): + super().__init__(dim, num_heads, mlp_ratio, qkv_bias, qk_scale, drop, + attn_drop, drop_path, act_layer, norm_layer) + del self.attn + if ws == 1: + self.attn = Attention(dim, num_heads, qkv_bias, qk_scale, + attn_drop, drop, sr_ratio) + else: + self.attn = GroupAttention(dim, num_heads, qkv_bias, qk_scale, + attn_drop, drop, ws) + + def forward(self, x, H, W): + x = x + self.drop_path(self.attn(self.norm1(x), H, W)) + x = x + self.drop_path(self.mlp(self.norm2(x))) + return x + + +class PatchEmbed(nn.Layer): + """ Image to Patch Embedding. + """ + + def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768): + super().__init__() + if img_size % patch_size != 0: + raise Exception( + f"img_size {img_size} should be divided by patch_size {patch_size}." + ) + + img_size = to_2tuple(img_size) + patch_size = to_2tuple(patch_size) + + self.img_size = img_size + self.patch_size = patch_size + self.H, self.W = img_size[0] // patch_size[0], img_size[ + 1] // patch_size[1] + self.num_patches = self.H * self.W + self.proj = nn.Conv2D( + in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) + self.norm = nn.LayerNorm(embed_dim) + + def forward(self, x): + B, C, H, W = x.shape + x = self.proj(x).flatten(2).transpose([0, 2, 1]) + x = self.norm(x) + H, W = H // self.patch_size[0], W // self.patch_size[1] + return x, (H, W) + + +# borrow from PVT https://github.com/whai362/PVT.git +class PyramidVisionTransformer(nn.Layer): + def __init__(self, + img_size=224, + patch_size=16, + in_chans=3, + class_num=1000, + embed_dims=[64, 128, 256, 512], + num_heads=[1, 2, 4, 8], + mlp_ratios=[4, 4, 4, 4], + qkv_bias=False, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0., + norm_layer=nn.LayerNorm, + depths=[3, 4, 6, 3], + sr_ratios=[8, 4, 2, 1], + block_cls=Block): + super().__init__() + self.class_num = class_num + self.depths = depths + + # patch_embed + self.patch_embeds = nn.LayerList() + self.pos_embeds = nn.ParameterList() + self.pos_drops = nn.LayerList() + self.blocks = nn.LayerList() + + for i in range(len(depths)): + if i == 0: + self.patch_embeds.append( + PatchEmbed(img_size, patch_size, in_chans, embed_dims[i])) + else: + self.patch_embeds.append( + PatchEmbed(img_size // patch_size // 2**(i - 1), 2, + embed_dims[i - 1], embed_dims[i])) + patch_num = self.patch_embeds[i].num_patches + 1 if i == len( + embed_dims) - 1 else self.patch_embeds[i].num_patches + self.pos_embeds.append( + self.create_parameter( + shape=[1, patch_num, embed_dims[i]], + default_initializer=zeros_)) + self.pos_drops.append(nn.Dropout(p=drop_rate)) + + dpr = [ + x.numpy()[0] + for x in paddle.linspace(0, drop_path_rate, sum(depths)) + ] # stochastic depth decay rule + + cur = 0 + for k in range(len(depths)): + _block = nn.LayerList([ + block_cls( + dim=embed_dims[k], + num_heads=num_heads[k], + mlp_ratio=mlp_ratios[k], + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop_rate, + attn_drop=attn_drop_rate, + drop_path=dpr[cur + i], + norm_layer=norm_layer, + sr_ratio=sr_ratios[k]) for i in range(depths[k]) + ]) + self.blocks.append(_block) + cur += depths[k] + + self.norm = norm_layer(embed_dims[-1]) + + # cls_token + self.cls_token = self.create_parameter( + shape=[1, 1, embed_dims[-1]], + default_initializer=zeros_, + attr=paddle.ParamAttr(regularizer=L2Decay(0.0))) + + # classification head + self.head = nn.Linear(embed_dims[-1], + class_num) if class_num > 0 else Identity() + + # init weights + for pos_emb in self.pos_embeds: + trunc_normal_(pos_emb) + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + + def forward_features(self, x): + B = x.shape[0] + for i in range(len(self.depths)): + x, (H, W) = self.patch_embeds[i](x) + if i == len(self.depths) - 1: + cls_tokens = self.cls_token.expand([B, -1, -1]) + x = paddle.concat([cls_tokens, x], dim=1) + x = x + self.pos_embeds[i] + x = self.pos_drops[i](x) + for blk in self.blocks[i]: + x = blk(x, H, W) + if i < len(self.depths) - 1: + x = x.reshape([B, H, W, -1]).transpose( + [0, 3, 1, 2]).contiguous() + x = self.norm(x) + return x[:, 0] + + def forward(self, x): + x = self.forward_features(x) + x = self.head(x) + return x + + +# PEG from https://arxiv.org/abs/2102.10882 +class PosCNN(nn.Layer): + def __init__(self, in_chans, embed_dim=768, s=1): + super().__init__() + self.proj = nn.Sequential( + nn.Conv2D( + in_chans, + embed_dim, + 3, + s, + 1, + bias_attr=paddle.ParamAttr(regularizer=L2Decay(0.0)), + groups=embed_dim, + weight_attr=paddle.ParamAttr(regularizer=L2Decay(0.0)), )) + self.s = s + + def forward(self, x, H, W): + B, N, C = x.shape + feat_token = x + cnn_feat = feat_token.transpose([0, 2, 1]).reshape([B, C, H, W]) + if self.s == 1: + x = self.proj(cnn_feat) + cnn_feat + else: + x = self.proj(cnn_feat) + x = x.flatten(2).transpose([0, 2, 1]) + return x + + +class CPVTV2(PyramidVisionTransformer): + """ + Use useful results from CPVT. PEG and GAP. + Therefore, cls token is no longer required. + PEG is used to encode the absolute position on the fly, which greatly affects the performance when input resolution + changes during the training (such as segmentation, detection) + """ + + def __init__(self, + img_size=224, + patch_size=4, + in_chans=3, + class_num=1000, + embed_dims=[64, 128, 256, 512], + num_heads=[1, 2, 4, 8], + mlp_ratios=[4, 4, 4, 4], + qkv_bias=False, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0., + norm_layer=nn.LayerNorm, + depths=[3, 4, 6, 3], + sr_ratios=[8, 4, 2, 1], + block_cls=Block): + super().__init__(img_size, patch_size, in_chans, class_num, embed_dims, + num_heads, mlp_ratios, qkv_bias, qk_scale, drop_rate, + attn_drop_rate, drop_path_rate, norm_layer, depths, + sr_ratios, block_cls) + del self.pos_embeds + del self.cls_token + self.pos_block = nn.LayerList( + [PosCNN(embed_dim, embed_dim) for embed_dim in embed_dims]) + self.apply(self._init_weights) + + def _init_weights(self, m): + import math + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + elif isinstance(m, nn.Conv2D): + fan_out = m._kernel_size[0] * m._kernel_size[1] * m._out_channels + fan_out //= m._groups + normal_(0, math.sqrt(2.0 / fan_out))(m.weight) + if m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.BatchNorm2D): + m.weight.data.fill_(1.0) + m.bias.data.zero_() + + def forward_features(self, x): + B = x.shape[0] + + for i in range(len(self.depths)): + x, (H, W) = self.patch_embeds[i](x) + x = self.pos_drops[i](x) + + for j, blk in enumerate(self.blocks[i]): + x = blk(x, H, W) + if j == 0: + x = self.pos_block[i](x, H, W) # PEG here + + if i < len(self.depths) - 1: + x = x.reshape([B, H, W, x.shape[-1]]).transpose([0, 3, 1, 2]) + + x = self.norm(x) + return x.mean(axis=1) # GAP here + + +class PCPVT(CPVTV2): + def __init__(self, + img_size=224, + patch_size=4, + in_chans=3, + class_num=1000, + embed_dims=[64, 128, 256], + num_heads=[1, 2, 4], + mlp_ratios=[4, 4, 4], + qkv_bias=False, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0., + norm_layer=nn.LayerNorm, + depths=[4, 4, 4], + sr_ratios=[4, 2, 1], + block_cls=SBlock): + super().__init__(img_size, patch_size, in_chans, class_num, embed_dims, + num_heads, mlp_ratios, qkv_bias, qk_scale, drop_rate, + attn_drop_rate, drop_path_rate, norm_layer, depths, + sr_ratios, block_cls) + + +class ALTGVT(PCPVT): + """ + alias Twins-SVT + """ + + def __init__(self, + img_size=224, + patch_size=4, + in_chans=3, + class_num=1000, + embed_dims=[64, 128, 256], + num_heads=[1, 2, 4], + mlp_ratios=[4, 4, 4], + qkv_bias=False, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0., + norm_layer=nn.LayerNorm, + depths=[4, 4, 4], + sr_ratios=[4, 2, 1], + block_cls=GroupBlock, + wss=[7, 7, 7]): + super().__init__(img_size, patch_size, in_chans, class_num, embed_dims, + num_heads, mlp_ratios, qkv_bias, qk_scale, drop_rate, + attn_drop_rate, drop_path_rate, norm_layer, depths, + sr_ratios, block_cls) + del self.blocks + self.wss = wss + # transformer encoder + dpr = [ + x.numpy()[0] + for x in paddle.linspace(0, drop_path_rate, sum(depths)) + ] # stochastic depth decay rule + cur = 0 + self.blocks = nn.LayerList() + for k in range(len(depths)): + _block = nn.LayerList([ + block_cls( + dim=embed_dims[k], + num_heads=num_heads[k], + mlp_ratio=mlp_ratios[k], + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop_rate, + attn_drop=attn_drop_rate, + drop_path=dpr[cur + i], + norm_layer=norm_layer, + sr_ratio=sr_ratios[k], + ws=1 if i % 2 == 1 else wss[k]) for i in range(depths[k]) + ]) + self.blocks.append(_block) + cur += depths[k] + self.apply(self._init_weights) + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def pcpvt_small(pretrained=False, use_ssld=False, **kwargs): + model = CPVTV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 4, 6, 3], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["pcpvt_small"], use_ssld=use_ssld) + return model + + +def pcpvt_base(pretrained=False, use_ssld=False, **kwargs): + model = CPVTV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 4, 18, 3], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["pcpvt_base"], use_ssld=use_ssld) + return model + + +def pcpvt_large(pretrained=False, use_ssld=False, **kwargs): + model = CPVTV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 8, 27, 3], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["pcpvt_large"], use_ssld=use_ssld) + return model + + +def alt_gvt_small(pretrained=False, use_ssld=False, **kwargs): + model = ALTGVT( + patch_size=4, + embed_dims=[64, 128, 256, 512], + num_heads=[2, 4, 8, 16], + mlp_ratios=[4, 4, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[2, 2, 10, 4], + wss=[7, 7, 7, 7], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["alt_gvt_small"], use_ssld=use_ssld) + return model + + +def alt_gvt_base(pretrained=False, use_ssld=False, **kwargs): + model = ALTGVT( + patch_size=4, + embed_dims=[96, 192, 384, 768], + num_heads=[3, 6, 12, 24], + mlp_ratios=[4, 4, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[2, 2, 18, 2], + wss=[7, 7, 7, 7], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["alt_gvt_base"], use_ssld=use_ssld) + return model + + +def alt_gvt_large(pretrained=False, use_ssld=False, **kwargs): + model = ALTGVT( + patch_size=4, + embed_dims=[128, 256, 512, 1024], + num_heads=[4, 8, 16, 32], + mlp_ratios=[4, 4, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[2, 2, 18, 2], + wss=[7, 7, 7, 7], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["alt_gvt_large"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/hardnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/hardnet.py new file mode 100644 index 00000000..fffd3a42 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/hardnet.py @@ -0,0 +1,293 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/PingoLH/Pytorch-HarDNet + +import paddle +import paddle.nn as nn + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + 'HarDNet39_ds': + 'https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet39_ds_pretrained.pdparams', + 'HarDNet68_ds': + 'https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_ds_pretrained.pdparams', + 'HarDNet68': + 'https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_pretrained.pdparams', + 'HarDNet85': + 'https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet85_pretrained.pdparams' +} + +__all__ = MODEL_URLS.keys() + + +def ConvLayer(in_channels, + out_channels, + kernel_size=3, + stride=1, + bias_attr=False): + layer = nn.Sequential( + ('conv', nn.Conv2D( + in_channels, + out_channels, + kernel_size=kernel_size, + stride=stride, + padding=kernel_size // 2, + groups=1, + bias_attr=bias_attr)), ('norm', nn.BatchNorm2D(out_channels)), + ('relu', nn.ReLU6())) + return layer + + +def DWConvLayer(in_channels, + out_channels, + kernel_size=3, + stride=1, + bias_attr=False): + layer = nn.Sequential( + ('dwconv', nn.Conv2D( + in_channels, + out_channels, + kernel_size=kernel_size, + stride=stride, + padding=1, + groups=out_channels, + bias_attr=bias_attr)), ('norm', nn.BatchNorm2D(out_channels))) + return layer + + +def CombConvLayer(in_channels, out_channels, kernel_size=1, stride=1): + layer = nn.Sequential( + ('layer1', ConvLayer( + in_channels, out_channels, kernel_size=kernel_size)), + ('layer2', DWConvLayer( + out_channels, out_channels, stride=stride))) + return layer + + +class HarDBlock(nn.Layer): + def __init__(self, + in_channels, + growth_rate, + grmul, + n_layers, + keepBase=False, + residual_out=False, + dwconv=False): + super().__init__() + self.keepBase = keepBase + self.links = [] + layers_ = [] + self.out_channels = 0 # if upsample else in_channels + for i in range(n_layers): + outch, inch, link = self.get_link(i + 1, in_channels, growth_rate, + grmul) + self.links.append(link) + if dwconv: + layers_.append(CombConvLayer(inch, outch)) + else: + layers_.append(ConvLayer(inch, outch)) + + if (i % 2 == 0) or (i == n_layers - 1): + self.out_channels += outch + # print("Blk out =",self.out_channels) + self.layers = nn.LayerList(layers_) + + def get_link(self, layer, base_ch, growth_rate, grmul): + if layer == 0: + return base_ch, 0, [] + out_channels = growth_rate + + link = [] + for i in range(10): + dv = 2**i + if layer % dv == 0: + k = layer - dv + link.append(k) + if i > 0: + out_channels *= grmul + + out_channels = int(int(out_channels + 1) / 2) * 2 + in_channels = 0 + + for i in link: + ch, _, _ = self.get_link(i, base_ch, growth_rate, grmul) + in_channels += ch + + return out_channels, in_channels, link + + def forward(self, x): + layers_ = [x] + + for layer in range(len(self.layers)): + link = self.links[layer] + tin = [] + for i in link: + tin.append(layers_[i]) + if len(tin) > 1: + x = paddle.concat(tin, 1) + else: + x = tin[0] + out = self.layers[layer](x) + layers_.append(out) + + t = len(layers_) + out_ = [] + for i in range(t): + if (i == 0 and self.keepBase) or (i == t - 1) or (i % 2 == 1): + out_.append(layers_[i]) + out = paddle.concat(out_, 1) + + return out + + +class HarDNet(nn.Layer): + def __init__(self, + depth_wise=False, + arch=85, + class_num=1000, + with_pool=True): + super().__init__() + first_ch = [32, 64] + second_kernel = 3 + max_pool = True + grmul = 1.7 + drop_rate = 0.1 + + # HarDNet68 + ch_list = [128, 256, 320, 640, 1024] + gr = [14, 16, 20, 40, 160] + n_layers = [8, 16, 16, 16, 4] + downSamp = [1, 0, 1, 1, 0] + + if arch == 85: + # HarDNet85 + first_ch = [48, 96] + ch_list = [192, 256, 320, 480, 720, 1280] + gr = [24, 24, 28, 36, 48, 256] + n_layers = [8, 16, 16, 16, 16, 4] + downSamp = [1, 0, 1, 0, 1, 0] + drop_rate = 0.2 + + elif arch == 39: + # HarDNet39 + first_ch = [24, 48] + ch_list = [96, 320, 640, 1024] + grmul = 1.6 + gr = [16, 20, 64, 160] + n_layers = [4, 16, 8, 4] + downSamp = [1, 1, 1, 0] + + if depth_wise: + second_kernel = 1 + max_pool = False + drop_rate = 0.05 + + blks = len(n_layers) + self.base = nn.LayerList([]) + + # First Layer: Standard Conv3x3, Stride=2 + self.base.append( + ConvLayer( + in_channels=3, + out_channels=first_ch[0], + kernel_size=3, + stride=2, + bias_attr=False)) + + # Second Layer + self.base.append( + ConvLayer( + first_ch[0], first_ch[1], kernel_size=second_kernel)) + + # Maxpooling or DWConv3x3 downsampling + if max_pool: + self.base.append(nn.MaxPool2D(kernel_size=3, stride=2, padding=1)) + else: + self.base.append(DWConvLayer(first_ch[1], first_ch[1], stride=2)) + + # Build all HarDNet blocks + ch = first_ch[1] + for i in range(blks): + blk = HarDBlock(ch, gr[i], grmul, n_layers[i], dwconv=depth_wise) + ch = blk.out_channels + self.base.append(blk) + + if i == blks - 1 and arch == 85: + self.base.append(nn.Dropout(0.1)) + + self.base.append(ConvLayer(ch, ch_list[i], kernel_size=1)) + ch = ch_list[i] + if downSamp[i] == 1: + if max_pool: + self.base.append(nn.MaxPool2D(kernel_size=2, stride=2)) + else: + self.base.append(DWConvLayer(ch, ch, stride=2)) + + ch = ch_list[blks - 1] + + layers = [] + + if with_pool: + layers.append(nn.AdaptiveAvgPool2D((1, 1))) + + if class_num > 0: + layers.append(nn.Flatten()) + layers.append(nn.Dropout(drop_rate)) + layers.append(nn.Linear(ch, class_num)) + + self.base.append(nn.Sequential(*layers)) + + def forward(self, x): + for layer in self.base: + x = layer(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def HarDNet39_ds(pretrained=False, **kwargs): + model = HarDNet(arch=39, depth_wise=True, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HarDNet39_ds"]) + return model + + +def HarDNet68_ds(pretrained=False, **kwargs): + model = HarDNet(arch=68, depth_wise=True, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HarDNet68_ds"]) + return model + + +def HarDNet68(pretrained=False, **kwargs): + model = HarDNet(arch=68, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HarDNet68"]) + return model + + +def HarDNet85(pretrained=False, **kwargs): + model = HarDNet(arch=85, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["HarDNet85"]) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/inception_v4.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/inception_v4.py new file mode 100644 index 00000000..e0460d48 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/inception_v4.py @@ -0,0 +1,477 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "InceptionV4": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV4_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + padding=0, + groups=1, + act='relu', + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=padding, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + bn_name = name + "_bn" + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + "_scale"), + bias_attr=ParamAttr(name=bn_name + "_offset"), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class InceptionStem(nn.Layer): + def __init__(self): + super(InceptionStem, self).__init__() + self._conv_1 = ConvBNLayer( + 3, 32, 3, stride=2, act="relu", name="conv1_3x3_s2") + self._conv_2 = ConvBNLayer(32, 32, 3, act="relu", name="conv2_3x3_s1") + self._conv_3 = ConvBNLayer( + 32, 64, 3, padding=1, act="relu", name="conv3_3x3_s1") + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=0) + self._conv2 = ConvBNLayer( + 64, 96, 3, stride=2, act="relu", name="inception_stem1_3x3_s2") + self._conv1_1 = ConvBNLayer( + 160, 64, 1, act="relu", name="inception_stem2_3x3_reduce") + self._conv1_2 = ConvBNLayer( + 64, 96, 3, act="relu", name="inception_stem2_3x3") + self._conv2_1 = ConvBNLayer( + 160, 64, 1, act="relu", name="inception_stem2_1x7_reduce") + self._conv2_2 = ConvBNLayer( + 64, + 64, (7, 1), + padding=(3, 0), + act="relu", + name="inception_stem2_1x7") + self._conv2_3 = ConvBNLayer( + 64, + 64, (1, 7), + padding=(0, 3), + act="relu", + name="inception_stem2_7x1") + self._conv2_4 = ConvBNLayer( + 64, 96, 3, act="relu", name="inception_stem2_3x3_2") + self._conv3 = ConvBNLayer( + 192, 192, 3, stride=2, act="relu", name="inception_stem3_3x3_s2") + + def forward(self, inputs): + conv = self._conv_1(inputs) + conv = self._conv_2(conv) + conv = self._conv_3(conv) + + pool1 = self._pool(conv) + conv2 = self._conv2(conv) + concat = paddle.concat([pool1, conv2], axis=1) + + conv1 = self._conv1_1(concat) + conv1 = self._conv1_2(conv1) + + conv2 = self._conv2_1(concat) + conv2 = self._conv2_2(conv2) + conv2 = self._conv2_3(conv2) + conv2 = self._conv2_4(conv2) + + concat = paddle.concat([conv1, conv2], axis=1) + + conv1 = self._conv3(concat) + pool1 = self._pool(concat) + + concat = paddle.concat([conv1, pool1], axis=1) + return concat + + +class InceptionA(nn.Layer): + def __init__(self, name): + super(InceptionA, self).__init__() + self._pool = AvgPool2D(kernel_size=3, stride=1, padding=1) + self._conv1 = ConvBNLayer( + 384, 96, 1, act="relu", name="inception_a" + name + "_1x1") + self._conv2 = ConvBNLayer( + 384, 96, 1, act="relu", name="inception_a" + name + "_1x1_2") + self._conv3_1 = ConvBNLayer( + 384, 64, 1, act="relu", name="inception_a" + name + "_3x3_reduce") + self._conv3_2 = ConvBNLayer( + 64, + 96, + 3, + padding=1, + act="relu", + name="inception_a" + name + "_3x3") + self._conv4_1 = ConvBNLayer( + 384, + 64, + 1, + act="relu", + name="inception_a" + name + "_3x3_2_reduce") + self._conv4_2 = ConvBNLayer( + 64, + 96, + 3, + padding=1, + act="relu", + name="inception_a" + name + "_3x3_2") + self._conv4_3 = ConvBNLayer( + 96, + 96, + 3, + padding=1, + act="relu", + name="inception_a" + name + "_3x3_3") + + def forward(self, inputs): + pool1 = self._pool(inputs) + conv1 = self._conv1(pool1) + + conv2 = self._conv2(inputs) + + conv3 = self._conv3_1(inputs) + conv3 = self._conv3_2(conv3) + + conv4 = self._conv4_1(inputs) + conv4 = self._conv4_2(conv4) + conv4 = self._conv4_3(conv4) + + concat = paddle.concat([conv1, conv2, conv3, conv4], axis=1) + return concat + + +class ReductionA(nn.Layer): + def __init__(self): + super(ReductionA, self).__init__() + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=0) + self._conv2 = ConvBNLayer( + 384, 384, 3, stride=2, act="relu", name="reduction_a_3x3") + self._conv3_1 = ConvBNLayer( + 384, 192, 1, act="relu", name="reduction_a_3x3_2_reduce") + self._conv3_2 = ConvBNLayer( + 192, 224, 3, padding=1, act="relu", name="reduction_a_3x3_2") + self._conv3_3 = ConvBNLayer( + 224, 256, 3, stride=2, act="relu", name="reduction_a_3x3_3") + + def forward(self, inputs): + pool1 = self._pool(inputs) + conv2 = self._conv2(inputs) + conv3 = self._conv3_1(inputs) + conv3 = self._conv3_2(conv3) + conv3 = self._conv3_3(conv3) + concat = paddle.concat([pool1, conv2, conv3], axis=1) + return concat + + +class InceptionB(nn.Layer): + def __init__(self, name=None): + super(InceptionB, self).__init__() + self._pool = AvgPool2D(kernel_size=3, stride=1, padding=1) + self._conv1 = ConvBNLayer( + 1024, 128, 1, act="relu", name="inception_b" + name + "_1x1") + self._conv2 = ConvBNLayer( + 1024, 384, 1, act="relu", name="inception_b" + name + "_1x1_2") + self._conv3_1 = ConvBNLayer( + 1024, + 192, + 1, + act="relu", + name="inception_b" + name + "_1x7_reduce") + self._conv3_2 = ConvBNLayer( + 192, + 224, (1, 7), + padding=(0, 3), + act="relu", + name="inception_b" + name + "_1x7") + self._conv3_3 = ConvBNLayer( + 224, + 256, (7, 1), + padding=(3, 0), + act="relu", + name="inception_b" + name + "_7x1") + self._conv4_1 = ConvBNLayer( + 1024, + 192, + 1, + act="relu", + name="inception_b" + name + "_7x1_2_reduce") + self._conv4_2 = ConvBNLayer( + 192, + 192, (1, 7), + padding=(0, 3), + act="relu", + name="inception_b" + name + "_1x7_2") + self._conv4_3 = ConvBNLayer( + 192, + 224, (7, 1), + padding=(3, 0), + act="relu", + name="inception_b" + name + "_7x1_2") + self._conv4_4 = ConvBNLayer( + 224, + 224, (1, 7), + padding=(0, 3), + act="relu", + name="inception_b" + name + "_1x7_3") + self._conv4_5 = ConvBNLayer( + 224, + 256, (7, 1), + padding=(3, 0), + act="relu", + name="inception_b" + name + "_7x1_3") + + def forward(self, inputs): + pool1 = self._pool(inputs) + conv1 = self._conv1(pool1) + + conv2 = self._conv2(inputs) + + conv3 = self._conv3_1(inputs) + conv3 = self._conv3_2(conv3) + conv3 = self._conv3_3(conv3) + + conv4 = self._conv4_1(inputs) + conv4 = self._conv4_2(conv4) + conv4 = self._conv4_3(conv4) + conv4 = self._conv4_4(conv4) + conv4 = self._conv4_5(conv4) + + concat = paddle.concat([conv1, conv2, conv3, conv4], axis=1) + return concat + + +class ReductionB(nn.Layer): + def __init__(self): + super(ReductionB, self).__init__() + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=0) + self._conv2_1 = ConvBNLayer( + 1024, 192, 1, act="relu", name="reduction_b_3x3_reduce") + self._conv2_2 = ConvBNLayer( + 192, 192, 3, stride=2, act="relu", name="reduction_b_3x3") + self._conv3_1 = ConvBNLayer( + 1024, 256, 1, act="relu", name="reduction_b_1x7_reduce") + self._conv3_2 = ConvBNLayer( + 256, + 256, (1, 7), + padding=(0, 3), + act="relu", + name="reduction_b_1x7") + self._conv3_3 = ConvBNLayer( + 256, + 320, (7, 1), + padding=(3, 0), + act="relu", + name="reduction_b_7x1") + self._conv3_4 = ConvBNLayer( + 320, 320, 3, stride=2, act="relu", name="reduction_b_3x3_2") + + def forward(self, inputs): + pool1 = self._pool(inputs) + + conv2 = self._conv2_1(inputs) + conv2 = self._conv2_2(conv2) + + conv3 = self._conv3_1(inputs) + conv3 = self._conv3_2(conv3) + conv3 = self._conv3_3(conv3) + conv3 = self._conv3_4(conv3) + + concat = paddle.concat([pool1, conv2, conv3], axis=1) + + return concat + + +class InceptionC(nn.Layer): + def __init__(self, name=None): + super(InceptionC, self).__init__() + self._pool = AvgPool2D(kernel_size=3, stride=1, padding=1) + self._conv1 = ConvBNLayer( + 1536, 256, 1, act="relu", name="inception_c" + name + "_1x1") + self._conv2 = ConvBNLayer( + 1536, 256, 1, act="relu", name="inception_c" + name + "_1x1_2") + self._conv3_0 = ConvBNLayer( + 1536, 384, 1, act="relu", name="inception_c" + name + "_1x1_3") + self._conv3_1 = ConvBNLayer( + 384, + 256, (1, 3), + padding=(0, 1), + act="relu", + name="inception_c" + name + "_1x3") + self._conv3_2 = ConvBNLayer( + 384, + 256, (3, 1), + padding=(1, 0), + act="relu", + name="inception_c" + name + "_3x1") + self._conv4_0 = ConvBNLayer( + 1536, 384, 1, act="relu", name="inception_c" + name + "_1x1_4") + self._conv4_00 = ConvBNLayer( + 384, + 448, (1, 3), + padding=(0, 1), + act="relu", + name="inception_c" + name + "_1x3_2") + self._conv4_000 = ConvBNLayer( + 448, + 512, (3, 1), + padding=(1, 0), + act="relu", + name="inception_c" + name + "_3x1_2") + self._conv4_1 = ConvBNLayer( + 512, + 256, (1, 3), + padding=(0, 1), + act="relu", + name="inception_c" + name + "_1x3_3") + self._conv4_2 = ConvBNLayer( + 512, + 256, (3, 1), + padding=(1, 0), + act="relu", + name="inception_c" + name + "_3x1_3") + + def forward(self, inputs): + pool1 = self._pool(inputs) + conv1 = self._conv1(pool1) + + conv2 = self._conv2(inputs) + + conv3 = self._conv3_0(inputs) + conv3_1 = self._conv3_1(conv3) + conv3_2 = self._conv3_2(conv3) + + conv4 = self._conv4_0(inputs) + conv4 = self._conv4_00(conv4) + conv4 = self._conv4_000(conv4) + conv4_1 = self._conv4_1(conv4) + conv4_2 = self._conv4_2(conv4) + + concat = paddle.concat( + [conv1, conv2, conv3_1, conv3_2, conv4_1, conv4_2], axis=1) + + return concat + + +class InceptionV4DY(nn.Layer): + def __init__(self, class_num=1000): + super(InceptionV4DY, self).__init__() + self._inception_stem = InceptionStem() + + self._inceptionA_1 = InceptionA(name="1") + self._inceptionA_2 = InceptionA(name="2") + self._inceptionA_3 = InceptionA(name="3") + self._inceptionA_4 = InceptionA(name="4") + self._reductionA = ReductionA() + + self._inceptionB_1 = InceptionB(name="1") + self._inceptionB_2 = InceptionB(name="2") + self._inceptionB_3 = InceptionB(name="3") + self._inceptionB_4 = InceptionB(name="4") + self._inceptionB_5 = InceptionB(name="5") + self._inceptionB_6 = InceptionB(name="6") + self._inceptionB_7 = InceptionB(name="7") + self._reductionB = ReductionB() + + self._inceptionC_1 = InceptionC(name="1") + self._inceptionC_2 = InceptionC(name="2") + self._inceptionC_3 = InceptionC(name="3") + + self.avg_pool = AdaptiveAvgPool2D(1) + self._drop = Dropout(p=0.2, mode="downscale_in_infer") + stdv = 1.0 / math.sqrt(1536 * 1.0) + self.out = Linear( + 1536, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="final_fc_weights"), + bias_attr=ParamAttr(name="final_fc_offset")) + + def forward(self, inputs): + x = self._inception_stem(inputs) + + x = self._inceptionA_1(x) + x = self._inceptionA_2(x) + x = self._inceptionA_3(x) + x = self._inceptionA_4(x) + x = self._reductionA(x) + + x = self._inceptionB_1(x) + x = self._inceptionB_2(x) + x = self._inceptionB_3(x) + x = self._inceptionB_4(x) + x = self._inceptionB_5(x) + x = self._inceptionB_6(x) + x = self._inceptionB_7(x) + x = self._reductionB(x) + + x = self._inceptionC_1(x) + x = self._inceptionC_2(x) + x = self._inceptionC_3(x) + + x = self.avg_pool(x) + x = paddle.squeeze(x, axis=[2, 3]) + x = self._drop(x) + x = self.out(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def InceptionV4(pretrained=False, use_ssld=False, **kwargs): + model = InceptionV4DY(**kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["InceptionV4"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/levit.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/levit.py new file mode 100644 index 00000000..991f832b --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/levit.py @@ -0,0 +1,589 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/facebookresearch/LeViT + +import itertools +import math +import warnings + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn.initializer import TruncatedNormal, Constant +from paddle.regularizer import L2Decay + +from .vision_transformer import trunc_normal_, zeros_, ones_, Identity + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "LeViT_128S": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams", + "LeViT_128": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128_pretrained.pdparams", + "LeViT_192": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_192_pretrained.pdparams", + "LeViT_256": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_256_pretrained.pdparams", + "LeViT_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_384_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +def cal_attention_biases(attention_biases, attention_bias_idxs): + gather_list = [] + attention_bias_t = paddle.transpose(attention_biases, (1, 0)) + nums = attention_bias_idxs.shape[0] + for idx in range(nums): + gather = paddle.gather(attention_bias_t, attention_bias_idxs[idx]) + gather_list.append(gather) + shape0, shape1 = attention_bias_idxs.shape + gather = paddle.concat(gather_list) + return paddle.transpose(gather, (1, 0)).reshape((0, shape0, shape1)) + + +class Conv2d_BN(nn.Sequential): + def __init__(self, + a, + b, + ks=1, + stride=1, + pad=0, + dilation=1, + groups=1, + bn_weight_init=1, + resolution=-10000): + super().__init__() + self.add_sublayer( + 'c', + nn.Conv2D( + a, b, ks, stride, pad, dilation, groups, bias_attr=False)) + bn = nn.BatchNorm2D(b) + ones_(bn.weight) + zeros_(bn.bias) + self.add_sublayer('bn', bn) + + +class Linear_BN(nn.Sequential): + def __init__(self, a, b, bn_weight_init=1): + super().__init__() + self.add_sublayer('c', nn.Linear(a, b, bias_attr=False)) + bn = nn.BatchNorm1D(b) + if bn_weight_init == 0: + zeros_(bn.weight) + else: + ones_(bn.weight) + zeros_(bn.bias) + self.add_sublayer('bn', bn) + + def forward(self, x): + l, bn = self._sub_layers.values() + x = l(x) + return paddle.reshape(bn(x.flatten(0, 1)), x.shape) + + +class BN_Linear(nn.Sequential): + def __init__(self, a, b, bias=True, std=0.02): + super().__init__() + self.add_sublayer('bn', nn.BatchNorm1D(a)) + l = nn.Linear(a, b, bias_attr=bias) + trunc_normal_(l.weight) + if bias: + zeros_(l.bias) + self.add_sublayer('l', l) + + +def b16(n, activation, resolution=224): + return nn.Sequential( + Conv2d_BN( + 3, n // 8, 3, 2, 1, resolution=resolution), + activation(), + Conv2d_BN( + n // 8, n // 4, 3, 2, 1, resolution=resolution // 2), + activation(), + Conv2d_BN( + n // 4, n // 2, 3, 2, 1, resolution=resolution // 4), + activation(), + Conv2d_BN( + n // 2, n, 3, 2, 1, resolution=resolution // 8)) + + +class Residual(nn.Layer): + def __init__(self, m, drop): + super().__init__() + self.m = m + self.drop = drop + + def forward(self, x): + if self.training and self.drop > 0: + y = paddle.rand( + shape=[x.shape[0], 1, 1]).__ge__(self.drop).astype("float32") + y = y.divide(paddle.full_like(y, 1 - self.drop)) + return paddle.add(x, y) + else: + return paddle.add(x, self.m(x)) + + +class Attention(nn.Layer): + def __init__(self, + dim, + key_dim, + num_heads=8, + attn_ratio=4, + activation=None, + resolution=14): + super().__init__() + self.num_heads = num_heads + self.scale = key_dim**-0.5 + self.key_dim = key_dim + self.nh_kd = nh_kd = key_dim * num_heads + self.d = int(attn_ratio * key_dim) + self.dh = int(attn_ratio * key_dim) * num_heads + self.attn_ratio = attn_ratio + self.h = self.dh + nh_kd * 2 + self.qkv = Linear_BN(dim, self.h) + self.proj = nn.Sequential( + activation(), Linear_BN( + self.dh, dim, bn_weight_init=0)) + points = list(itertools.product(range(resolution), range(resolution))) + N = len(points) + attention_offsets = {} + idxs = [] + for p1 in points: + for p2 in points: + offset = (abs(p1[0] - p2[0]), abs(p1[1] - p2[1])) + if offset not in attention_offsets: + attention_offsets[offset] = len(attention_offsets) + idxs.append(attention_offsets[offset]) + self.attention_biases = self.create_parameter( + shape=(num_heads, len(attention_offsets)), + default_initializer=zeros_, + attr=paddle.ParamAttr(regularizer=L2Decay(0.0))) + tensor_idxs = paddle.to_tensor(idxs, dtype='int64') + self.register_buffer('attention_bias_idxs', + paddle.reshape(tensor_idxs, [N, N])) + + @paddle.no_grad() + def train(self, mode=True): + if mode: + super().train() + else: + super().eval() + if mode and hasattr(self, 'ab'): + del self.ab + else: + self.ab = cal_attention_biases(self.attention_biases, + self.attention_bias_idxs) + + def forward(self, x): + self.training = True + B, N, C = x.shape + qkv = self.qkv(x) + qkv = paddle.reshape(qkv, + [B, N, self.num_heads, self.h // self.num_heads]) + q, k, v = paddle.split( + qkv, [self.key_dim, self.key_dim, self.d], axis=3) + q = paddle.transpose(q, perm=[0, 2, 1, 3]) + k = paddle.transpose(k, perm=[0, 2, 1, 3]) + v = paddle.transpose(v, perm=[0, 2, 1, 3]) + k_transpose = paddle.transpose(k, perm=[0, 1, 3, 2]) + + if self.training: + attention_biases = cal_attention_biases(self.attention_biases, + self.attention_bias_idxs) + else: + attention_biases = self.ab + attn = (paddle.matmul(q, k_transpose) * self.scale + attention_biases) + attn = F.softmax(attn) + x = paddle.transpose(paddle.matmul(attn, v), perm=[0, 2, 1, 3]) + x = paddle.reshape(x, [B, N, self.dh]) + x = self.proj(x) + return x + + +class Subsample(nn.Layer): + def __init__(self, stride, resolution): + super().__init__() + self.stride = stride + self.resolution = resolution + + def forward(self, x): + B, N, C = x.shape + x = paddle.reshape(x, [B, self.resolution, self.resolution, C]) + end1, end2 = x.shape[1], x.shape[2] + x = x[:, 0:end1:self.stride, 0:end2:self.stride] + x = paddle.reshape(x, [B, -1, C]) + return x + + +class AttentionSubsample(nn.Layer): + def __init__(self, + in_dim, + out_dim, + key_dim, + num_heads=8, + attn_ratio=2, + activation=None, + stride=2, + resolution=14, + resolution_=7): + super().__init__() + self.num_heads = num_heads + self.scale = key_dim**-0.5 + self.key_dim = key_dim + self.nh_kd = nh_kd = key_dim * num_heads + self.d = int(attn_ratio * key_dim) + self.dh = int(attn_ratio * key_dim) * self.num_heads + self.attn_ratio = attn_ratio + self.resolution_ = resolution_ + self.resolution_2 = resolution_**2 + self.training = True + h = self.dh + nh_kd + self.kv = Linear_BN(in_dim, h) + + self.q = nn.Sequential( + Subsample(stride, resolution), Linear_BN(in_dim, nh_kd)) + self.proj = nn.Sequential(activation(), Linear_BN(self.dh, out_dim)) + + self.stride = stride + self.resolution = resolution + points = list(itertools.product(range(resolution), range(resolution))) + points_ = list( + itertools.product(range(resolution_), range(resolution_))) + + N = len(points) + N_ = len(points_) + attention_offsets = {} + idxs = [] + i = 0 + j = 0 + for p1 in points_: + i += 1 + for p2 in points: + j += 1 + size = 1 + offset = (abs(p1[0] * stride - p2[0] + (size - 1) / 2), + abs(p1[1] * stride - p2[1] + (size - 1) / 2)) + if offset not in attention_offsets: + attention_offsets[offset] = len(attention_offsets) + idxs.append(attention_offsets[offset]) + self.attention_biases = self.create_parameter( + shape=(num_heads, len(attention_offsets)), + default_initializer=zeros_, + attr=paddle.ParamAttr(regularizer=L2Decay(0.0))) + + tensor_idxs_ = paddle.to_tensor(idxs, dtype='int64') + self.register_buffer('attention_bias_idxs', + paddle.reshape(tensor_idxs_, [N_, N])) + + @paddle.no_grad() + def train(self, mode=True): + if mode: + super().train() + else: + super().eval() + if mode and hasattr(self, 'ab'): + del self.ab + else: + self.ab = cal_attention_biases(self.attention_biases, + self.attention_bias_idxs) + + def forward(self, x): + self.training = True + B, N, C = x.shape + kv = self.kv(x) + kv = paddle.reshape(kv, [B, N, self.num_heads, -1]) + k, v = paddle.split(kv, [self.key_dim, self.d], axis=3) + k = paddle.transpose(k, perm=[0, 2, 1, 3]) # BHNC + v = paddle.transpose(v, perm=[0, 2, 1, 3]) + q = paddle.reshape( + self.q(x), [B, self.resolution_2, self.num_heads, self.key_dim]) + q = paddle.transpose(q, perm=[0, 2, 1, 3]) + + if self.training: + attention_biases = cal_attention_biases(self.attention_biases, + self.attention_bias_idxs) + else: + attention_biases = self.ab + + attn = (paddle.matmul( + q, paddle.transpose( + k, perm=[0, 1, 3, 2]))) * self.scale + attention_biases + attn = F.softmax(attn) + + x = paddle.reshape( + paddle.transpose( + paddle.matmul(attn, v), perm=[0, 2, 1, 3]), [B, -1, self.dh]) + x = self.proj(x) + return x + + +class LeViT(nn.Layer): + """ Vision Transformer with support for patch or hybrid CNN input stage + """ + + def __init__(self, + img_size=224, + patch_size=16, + in_chans=3, + class_num=1000, + embed_dim=[192], + key_dim=[64], + depth=[12], + num_heads=[3], + attn_ratio=[2], + mlp_ratio=[2], + hybrid_backbone=None, + down_ops=[], + attention_activation=nn.Hardswish, + mlp_activation=nn.Hardswish, + distillation=True, + drop_path=0): + super().__init__() + + self.class_num = class_num + self.num_features = embed_dim[-1] + self.embed_dim = embed_dim + self.distillation = distillation + + self.patch_embed = hybrid_backbone + + self.blocks = [] + down_ops.append(['']) + resolution = img_size // patch_size + for i, (ed, kd, dpth, nh, ar, mr, do) in enumerate( + zip(embed_dim, key_dim, depth, num_heads, attn_ratio, + mlp_ratio, down_ops)): + for _ in range(dpth): + self.blocks.append( + Residual( + Attention( + ed, + kd, + nh, + attn_ratio=ar, + activation=attention_activation, + resolution=resolution, ), + drop_path)) + if mr > 0: + h = int(ed * mr) + self.blocks.append( + Residual( + nn.Sequential( + Linear_BN(ed, h), + mlp_activation(), + Linear_BN( + h, ed, bn_weight_init=0), ), + drop_path)) + if do[0] == 'Subsample': + #('Subsample',key_dim, num_heads, attn_ratio, mlp_ratio, stride) + resolution_ = (resolution - 1) // do[5] + 1 + self.blocks.append( + AttentionSubsample( + *embed_dim[i:i + 2], + key_dim=do[1], + num_heads=do[2], + attn_ratio=do[3], + activation=attention_activation, + stride=do[5], + resolution=resolution, + resolution_=resolution_)) + resolution = resolution_ + if do[4] > 0: # mlp_ratio + h = int(embed_dim[i + 1] * do[4]) + self.blocks.append( + Residual( + nn.Sequential( + Linear_BN(embed_dim[i + 1], h), + mlp_activation(), + Linear_BN( + h, embed_dim[i + 1], bn_weight_init=0), ), + drop_path)) + self.blocks = nn.Sequential(*self.blocks) + + # Classifier head + self.head = BN_Linear(embed_dim[-1], + class_num) if class_num > 0 else Identity() + if distillation: + self.head_dist = BN_Linear( + embed_dim[-1], class_num) if class_num > 0 else Identity() + + def forward(self, x): + x = self.patch_embed(x) + x = x.flatten(2) + x = paddle.transpose(x, perm=[0, 2, 1]) + x = self.blocks(x) + x = x.mean(1) + + x = paddle.reshape(x, [-1, self.embed_dim[-1]]) + if self.distillation: + x = self.head(x), self.head_dist(x) + if not self.training: + x = (x[0] + x[1]) / 2 + else: + x = self.head(x) + return x + + +def model_factory(C, D, X, N, drop_path, class_num, distillation): + embed_dim = [int(x) for x in C.split('_')] + num_heads = [int(x) for x in N.split('_')] + depth = [int(x) for x in X.split('_')] + act = nn.Hardswish + model = LeViT( + patch_size=16, + embed_dim=embed_dim, + num_heads=num_heads, + key_dim=[D] * 3, + depth=depth, + attn_ratio=[2, 2, 2], + mlp_ratio=[2, 2, 2], + down_ops=[ + #('Subsample',key_dim, num_heads, attn_ratio, mlp_ratio, stride) + ['Subsample', D, embed_dim[0] // D, 4, 2, 2], + ['Subsample', D, embed_dim[1] // D, 4, 2, 2], + ], + attention_activation=act, + mlp_activation=act, + hybrid_backbone=b16(embed_dim[0], activation=act), + class_num=class_num, + drop_path=drop_path, + distillation=distillation) + + return model + + +specification = { + 'LeViT_128S': { + 'C': '128_256_384', + 'D': 16, + 'N': '4_6_8', + 'X': '2_3_4', + 'drop_path': 0 + }, + 'LeViT_128': { + 'C': '128_256_384', + 'D': 16, + 'N': '4_8_12', + 'X': '4_4_4', + 'drop_path': 0 + }, + 'LeViT_192': { + 'C': '192_288_384', + 'D': 32, + 'N': '3_5_6', + 'X': '4_4_4', + 'drop_path': 0 + }, + 'LeViT_256': { + 'C': '256_384_512', + 'D': 32, + 'N': '4_6_8', + 'X': '4_4_4', + 'drop_path': 0 + }, + 'LeViT_384': { + 'C': '384_512_768', + 'D': 32, + 'N': '6_9_12', + 'X': '4_4_4', + 'drop_path': 0.1 + }, +} + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def LeViT_128S(pretrained=False, + use_ssld=False, + class_num=1000, + distillation=False, + **kwargs): + model = model_factory( + **specification['LeViT_128S'], + class_num=class_num, + distillation=distillation) + _load_pretrained( + pretrained, model, MODEL_URLS["LeViT_128S"], use_ssld=use_ssld) + return model + + +def LeViT_128(pretrained=False, + use_ssld=False, + class_num=1000, + distillation=False, + **kwargs): + model = model_factory( + **specification['LeViT_128'], + class_num=class_num, + distillation=distillation) + _load_pretrained( + pretrained, model, MODEL_URLS["LeViT_128"], use_ssld=use_ssld) + return model + + +def LeViT_192(pretrained=False, + use_ssld=False, + class_num=1000, + distillation=False, + **kwargs): + model = model_factory( + **specification['LeViT_192'], + class_num=class_num, + distillation=distillation) + _load_pretrained( + pretrained, model, MODEL_URLS["LeViT_192"], use_ssld=use_ssld) + return model + + +def LeViT_256(pretrained=False, + use_ssld=False, + class_num=1000, + distillation=False, + **kwargs): + model = model_factory( + **specification['LeViT_256'], + class_num=class_num, + distillation=distillation) + _load_pretrained( + pretrained, model, MODEL_URLS["LeViT_256"], use_ssld=use_ssld) + return model + + +def LeViT_384(pretrained=False, + use_ssld=False, + class_num=1000, + distillation=False, + **kwargs): + model = model_factory( + **specification['LeViT_384'], + class_num=class_num, + distillation=distillation) + _load_pretrained( + pretrained, model, MODEL_URLS["LeViT_384"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/mixnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/mixnet.py new file mode 100644 index 00000000..c2a1adb1 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/mixnet.py @@ -0,0 +1,815 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +""" + MixNet for ImageNet-1K, implemented in Paddle. + Original paper: 'MixConv: Mixed Depthwise Convolutional Kernels,' + https://arxiv.org/abs/1907.09595. +""" + +import os +from inspect import isfunction +from functools import reduce +import paddle +import paddle.nn as nn + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "MixNet_S": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_S_pretrained.pdparams", + "MixNet_M": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_M_pretrained.pdparams", + "MixNet_L": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_L_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +class Identity(nn.Layer): + """ + Identity block. + """ + + def __init__(self): + super(Identity, self).__init__() + + def forward(self, x): + return x + + +def round_channels(channels, divisor=8): + """ + Round weighted channel number (make divisible operation). + + Parameters: + ---------- + channels : int or float + Original number of channels. + divisor : int, default 8 + Alignment value. + + Returns: + ------- + int + Weighted number of channels. + """ + rounded_channels = max( + int(channels + divisor / 2.0) // divisor * divisor, divisor) + if float(rounded_channels) < 0.9 * channels: + rounded_channels += divisor + return rounded_channels + + +def get_activation_layer(activation): + """ + Create activation layer from string/function. + + Parameters: + ---------- + activation : function, or str, or nn.Module + Activation function or name of activation function. + + Returns: + ------- + nn.Module + Activation layer. + """ + assert activation is not None + if isfunction(activation): + return activation() + elif isinstance(activation, str): + if activation == "relu": + return nn.ReLU() + elif activation == "relu6": + return nn.ReLU6() + elif activation == "swish": + return nn.Swish() + elif activation == "hswish": + return nn.Hardswish() + elif activation == "sigmoid": + return nn.Sigmoid() + elif activation == "hsigmoid": + return nn.Hardsigmoid() + elif activation == "identity": + return Identity() + else: + raise NotImplementedError() + else: + assert isinstance(activation, nn.Layer) + return activation + + +class ConvBlock(nn.Layer): + """ + Standard convolution block with Batch normalization and activation. + + Parameters: + ---------- + in_channels : int + Number of input channels. + out_channels : int + Number of output channels. + kernel_size : int or tuple/list of 2 int + Convolution window size. + stride : int or tuple/list of 2 int + Strides of the convolution. + padding : int, or tuple/list of 2 int, or tuple/list of 4 int + Padding value for convolution layer. + dilation : int or tuple/list of 2 int, default 1 + Dilation value for convolution layer. + groups : int, default 1 + Number of groups. + bias : bool, default False + Whether the layer uses a bias vector. + use_bn : bool, default True + Whether to use BatchNorm layer. + bn_eps : float, default 1e-5 + Small float added to variance in Batch norm. + activation : function or str or None, default nn.ReLU() + Activation function or name of activation function. + """ + + def __init__(self, + in_channels, + out_channels, + kernel_size, + stride, + padding, + dilation=1, + groups=1, + bias=False, + use_bn=True, + bn_eps=1e-5, + activation=nn.ReLU()): + super(ConvBlock, self).__init__() + self.activate = (activation is not None) + self.use_bn = use_bn + self.use_pad = (isinstance(padding, (list, tuple)) and + (len(padding) == 4)) + + if self.use_pad: + self.pad = padding + self.conv = nn.Conv2D( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + dilation=dilation, + groups=groups, + bias_attr=bias, + weight_attr=None) + if self.use_bn: + self.bn = nn.BatchNorm2D(num_features=out_channels, epsilon=bn_eps) + if self.activate: + self.activ = get_activation_layer(activation) + + def forward(self, x): + x = self.conv(x) + if self.use_bn: + x = self.bn(x) + if self.activate: + x = self.activ(x) + return x + + +class SEBlock(nn.Layer): + def __init__(self, + channels, + reduction=16, + mid_channels=None, + round_mid=False, + use_conv=True, + mid_activation=nn.ReLU(), + out_activation=nn.Sigmoid()): + super(SEBlock, self).__init__() + self.use_conv = use_conv + if mid_channels is None: + mid_channels = channels // reduction if not round_mid else round_channels( + float(channels) / reduction) + + self.pool = nn.AdaptiveAvgPool2D(output_size=1) + if use_conv: + self.conv1 = nn.Conv2D( + in_channels=channels, + out_channels=mid_channels, + kernel_size=1, + stride=1, + groups=1, + bias_attr=True, + weight_attr=None) + + else: + self.fc1 = nn.Linear( + in_features=channels, out_features=mid_channels) + self.activ = get_activation_layer(mid_activation) + if use_conv: + self.conv2 = nn.Conv2D( + in_channels=mid_channels, + out_channels=channels, + kernel_size=1, + stride=1, + groups=1, + bias_attr=True, + weight_attr=None) + else: + self.fc2 = nn.Linear( + in_features=mid_channels, out_features=channels) + self.sigmoid = get_activation_layer(out_activation) + + def forward(self, x): + w = self.pool(x) + if not self.use_conv: + w = w.reshape(shape=[w.shape[0], -1]) + w = self.conv1(w) if self.use_conv else self.fc1(w) + w = self.activ(w) + w = self.conv2(w) if self.use_conv else self.fc2(w) + w = self.sigmoid(w) + if not self.use_conv: + w = w.unsqueeze(2).unsqueeze(3) + x = x * w + return x + + +class MixConv(nn.Layer): + """ + Mixed convolution layer from 'MixConv: Mixed Depthwise Convolutional Kernels,' + https://arxiv.org/abs/1907.09595. + + Parameters: + ---------- + in_channels : int + Number of input channels. + out_channels : int + Number of output channels. + kernel_size : int or tuple/list of int, or tuple/list of tuple/list of 2 int + Convolution window size. + stride : int or tuple/list of 2 int + Strides of the convolution. + padding : int or tuple/list of int, or tuple/list of tuple/list of 2 int + Padding value for convolution layer. + dilation : int or tuple/list of 2 int, default 1 + Dilation value for convolution layer. + groups : int, default 1 + Number of groups. + bias : bool, default False + Whether the layer uses a bias vector. + axis : int, default 1 + The axis on which to concatenate the outputs. + """ + + def __init__(self, + in_channels, + out_channels, + kernel_size, + stride, + padding, + dilation=1, + groups=1, + bias=False, + axis=1): + super(MixConv, self).__init__() + kernel_size = kernel_size if isinstance(kernel_size, + list) else [kernel_size] + padding = padding if isinstance(padding, list) else [padding] + kernel_count = len(kernel_size) + self.splitted_in_channels = self.split_channels(in_channels, + kernel_count) + splitted_out_channels = self.split_channels(out_channels, kernel_count) + for i, kernel_size_i in enumerate(kernel_size): + in_channels_i = self.splitted_in_channels[i] + out_channels_i = splitted_out_channels[i] + padding_i = padding[i] + _ = self.add_sublayer( + name=str(i), + sublayer=nn.Conv2D( + in_channels=in_channels_i, + out_channels=out_channels_i, + kernel_size=kernel_size_i, + stride=stride, + padding=padding_i, + dilation=dilation, + groups=(out_channels_i + if out_channels == groups else groups), + bias_attr=bias, + weight_attr=None)) + self.axis = axis + + def forward(self, x): + xx = paddle.split(x, self.splitted_in_channels, axis=self.axis) + xx = paddle.split(x, self.splitted_in_channels, axis=self.axis) + out = [ + conv_i(x_i) for x_i, conv_i in zip(xx, self._sub_layers.values()) + ] + x = paddle.concat(tuple(out), axis=self.axis) + return x + + @staticmethod + def split_channels(channels, kernel_count): + splitted_channels = [channels // kernel_count] * kernel_count + splitted_channels[0] += channels - sum(splitted_channels) + return splitted_channels + + +class MixConvBlock(nn.Layer): + """ + Mixed convolution block with Batch normalization and activation. + + Parameters: + ---------- + in_channels : int + Number of input channels. + out_channels : int + Number of output channels. + kernel_size : int or tuple/list of int, or tuple/list of tuple/list of 2 int + Convolution window size. + stride : int or tuple/list of 2 int + Strides of the convolution. + padding : int or tuple/list of int, or tuple/list of tuple/list of 2 int + Padding value for convolution layer. + dilation : int or tuple/list of 2 int, default 1 + Dilation value for convolution layer. + groups : int, default 1 + Number of groups. + bias : bool, default False + Whether the layer uses a bias vector. + use_bn : bool, default True + Whether to use BatchNorm layer. + bn_eps : float, default 1e-5 + Small float added to variance in Batch norm. + activation : function or str or None, default nn.ReLU() + Activation function or name of activation function. + activate : bool, default True + Whether activate the convolution block. + """ + + def __init__(self, + in_channels, + out_channels, + kernel_size, + stride, + padding, + dilation=1, + groups=1, + bias=False, + use_bn=True, + bn_eps=1e-5, + activation=nn.ReLU()): + super(MixConvBlock, self).__init__() + self.activate = (activation is not None) + self.use_bn = use_bn + + self.conv = MixConv( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + dilation=dilation, + groups=groups, + bias=bias) + if self.use_bn: + self.bn = nn.BatchNorm2D(num_features=out_channels, epsilon=bn_eps) + if self.activate: + self.activ = get_activation_layer(activation) + + def forward(self, x): + x = self.conv(x) + if self.use_bn: + x = self.bn(x) + if self.activate: + x = self.activ(x) + return x + + +def mixconv1x1_block(in_channels, + out_channels, + kernel_count, + stride=1, + groups=1, + bias=False, + use_bn=True, + bn_eps=1e-5, + activation=nn.ReLU()): + """ + 1x1 version of the mixed convolution block. + + Parameters: + ---------- + in_channels : int + Number of input channels. + out_channels : int + Number of output channels. + kernel_count : int + Kernel count. + stride : int or tuple/list of 2 int, default 1 + Strides of the convolution. + groups : int, default 1 + Number of groups. + bias : bool, default False + Whether the layer uses a bias vector. + use_bn : bool, default True + Whether to use BatchNorm layer. + bn_eps : float, default 1e-5 + Small float added to variance in Batch norm. + activation : function or str, or None, default nn.ReLU() + Activation function or name of activation function. + """ + return MixConvBlock( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=([1] * kernel_count), + stride=stride, + padding=([0] * kernel_count), + groups=groups, + bias=bias, + use_bn=use_bn, + bn_eps=bn_eps, + activation=activation) + + +class MixUnit(nn.Layer): + """ + MixNet unit. + + Parameters: + ---------- + in_channels : int + Number of input channels. + out_channels : int + Number of output channels. exp_channels : int + Number of middle (expanded) channels. + stride : int or tuple/list of 2 int + Strides of the second convolution layer. + exp_kernel_count : int + Expansion convolution kernel count for each unit. + conv1_kernel_count : int + Conv1 kernel count for each unit. + conv2_kernel_count : int + Conv2 kernel count for each unit. + exp_factor : int + Expansion factor for each unit. + se_factor : int + SE reduction factor for each unit. + activation : str + Activation function or name of activation function. + """ + + def __init__(self, in_channels, out_channels, stride, exp_kernel_count, + conv1_kernel_count, conv2_kernel_count, exp_factor, se_factor, + activation): + super(MixUnit, self).__init__() + assert exp_factor >= 1 + assert se_factor >= 0 + self.residual = (in_channels == out_channels) and (stride == 1) + self.use_se = se_factor > 0 + mid_channels = exp_factor * in_channels + self.use_exp_conv = exp_factor > 1 + + if self.use_exp_conv: + if exp_kernel_count == 1: + self.exp_conv = ConvBlock( + in_channels=in_channels, + out_channels=mid_channels, + kernel_size=1, + stride=1, + padding=0, + groups=1, + bias=False, + use_bn=True, + bn_eps=1e-5, + activation=activation) + else: + self.exp_conv = mixconv1x1_block( + in_channels=in_channels, + out_channels=mid_channels, + kernel_count=exp_kernel_count, + activation=activation) + if conv1_kernel_count == 1: + self.conv1 = ConvBlock( + in_channels=mid_channels, + out_channels=mid_channels, + kernel_size=3, + stride=stride, + padding=1, + dilation=1, + groups=mid_channels, + bias=False, + use_bn=True, + bn_eps=1e-5, + activation=activation) + else: + self.conv1 = MixConvBlock( + in_channels=mid_channels, + out_channels=mid_channels, + kernel_size=[3 + 2 * i for i in range(conv1_kernel_count)], + stride=stride, + padding=[1 + i for i in range(conv1_kernel_count)], + groups=mid_channels, + activation=activation) + if self.use_se: + self.se = SEBlock( + channels=mid_channels, + reduction=(exp_factor * se_factor), + round_mid=False, + mid_activation=activation) + if conv2_kernel_count == 1: + self.conv2 = ConvBlock( + in_channels=mid_channels, + out_channels=out_channels, + activation=None, + kernel_size=1, + stride=1, + padding=0, + groups=1, + bias=False, + use_bn=True, + bn_eps=1e-5) + else: + self.conv2 = mixconv1x1_block( + in_channels=mid_channels, + out_channels=out_channels, + kernel_count=conv2_kernel_count, + activation=None) + + def forward(self, x): + if self.residual: + identity = x + if self.use_exp_conv: + x = self.exp_conv(x) + x = self.conv1(x) + if self.use_se: + x = self.se(x) + x = self.conv2(x) + if self.residual: + x = x + identity + return x + + +class MixInitBlock(nn.Layer): + """ + MixNet specific initial block. + + Parameters: + ---------- + in_channels : int + Number of input channels. + out_channels : int + Number of output channels. + """ + + def __init__(self, in_channels, out_channels): + super(MixInitBlock, self).__init__() + self.conv1 = ConvBlock( + in_channels=in_channels, + out_channels=out_channels, + stride=2, + kernel_size=3, + padding=1) + self.conv2 = MixUnit( + in_channels=out_channels, + out_channels=out_channels, + stride=1, + exp_kernel_count=1, + conv1_kernel_count=1, + conv2_kernel_count=1, + exp_factor=1, + se_factor=0, + activation="relu") + + def forward(self, x): + x = self.conv1(x) + x = self.conv2(x) + return x + + +class MixNet(nn.Layer): + """ + MixNet model from 'MixConv: Mixed Depthwise Convolutional Kernels,' + https://arxiv.org/abs/1907.09595. + + Parameters: + ---------- + channels : list of list of int + Number of output channels for each unit. + init_block_channels : int + Number of output channels for the initial unit. + final_block_channels : int + Number of output channels for the final block of the feature extractor. + exp_kernel_counts : list of list of int + Expansion convolution kernel count for each unit. + conv1_kernel_counts : list of list of int + Conv1 kernel count for each unit. + conv2_kernel_counts : list of list of int + Conv2 kernel count for each unit. + exp_factors : list of list of int + Expansion factor for each unit. + se_factors : list of list of int + SE reduction factor for each unit. + in_channels : int, default 3 + Number of input channels. + in_size : tuple of two ints, default (224, 224) + Spatial size of the expected input image. + class_num : int, default 1000 + Number of classification classes. + """ + + def __init__(self, + channels, + init_block_channels, + final_block_channels, + exp_kernel_counts, + conv1_kernel_counts, + conv2_kernel_counts, + exp_factors, + se_factors, + in_channels=3, + in_size=(224, 224), + class_num=1000): + super(MixNet, self).__init__() + self.in_size = in_size + self.class_num = class_num + + self.features = nn.Sequential() + self.features.add_sublayer( + "init_block", + MixInitBlock( + in_channels=in_channels, out_channels=init_block_channels)) + in_channels = init_block_channels + for i, channels_per_stage in enumerate(channels): + stage = nn.Sequential() + for j, out_channels in enumerate(channels_per_stage): + stride = 2 if ((j == 0) and (i != 3)) or ( + (j == len(channels_per_stage) // 2) and (i == 3)) else 1 + exp_kernel_count = exp_kernel_counts[i][j] + conv1_kernel_count = conv1_kernel_counts[i][j] + conv2_kernel_count = conv2_kernel_counts[i][j] + exp_factor = exp_factors[i][j] + se_factor = se_factors[i][j] + activation = "relu" if i == 0 else "swish" + stage.add_sublayer( + "unit{}".format(j + 1), + MixUnit( + in_channels=in_channels, + out_channels=out_channels, + stride=stride, + exp_kernel_count=exp_kernel_count, + conv1_kernel_count=conv1_kernel_count, + conv2_kernel_count=conv2_kernel_count, + exp_factor=exp_factor, + se_factor=se_factor, + activation=activation)) + in_channels = out_channels + self.features.add_sublayer("stage{}".format(i + 1), stage) + self.features.add_sublayer( + "final_block", + ConvBlock( + in_channels=in_channels, + out_channels=final_block_channels, + kernel_size=1, + stride=1, + padding=0, + groups=1, + bias=False, + use_bn=True, + bn_eps=1e-5, + activation=nn.ReLU())) + in_channels = final_block_channels + self.features.add_sublayer( + "final_pool", nn.AvgPool2D( + kernel_size=7, stride=1)) + + self.output = nn.Linear( + in_features=in_channels, out_features=class_num) + + def forward(self, x): + x = self.features(x) + reshape_dim = reduce(lambda x, y: x * y, x.shape[1:]) + x = x.reshape(shape=[x.shape[0], reshape_dim]) + x = self.output(x) + return x + + +def get_mixnet(version, width_scale, model_name=None, **kwargs): + """ + Create MixNet model with specific parameters. + + Parameters: + ---------- + version : str + Version of MobileNetV3 ('s' or 'm'). + width_scale : float + Scale factor for width of layers. + model_name : str or None, default None + Model name. + """ + + if version == "s": + init_block_channels = 16 + channels = [[24, 24], [40, 40, 40, 40], [80, 80, 80], + [120, 120, 120, 200, 200, 200]] + exp_kernel_counts = [[2, 2], [1, 2, 2, 2], [1, 1, 1], + [2, 2, 2, 1, 1, 1]] + conv1_kernel_counts = [[1, 1], [3, 2, 2, 2], [3, 2, 2], + [3, 4, 4, 5, 4, 4]] + conv2_kernel_counts = [[2, 2], [1, 2, 2, 2], [2, 2, 2], + [2, 2, 2, 1, 2, 2]] + exp_factors = [[6, 3], [6, 6, 6, 6], [6, 6, 6], [6, 3, 3, 6, 6, 6]] + se_factors = [[0, 0], [2, 2, 2, 2], [4, 4, 4], [2, 2, 2, 2, 2, 2]] + elif version == "m": + init_block_channels = 24 + channels = [[32, 32], [40, 40, 40, 40], [80, 80, 80, 80], + [120, 120, 120, 120, 200, 200, 200, 200]] + exp_kernel_counts = [[2, 2], [1, 2, 2, 2], [1, 2, 2, 2], + [1, 2, 2, 2, 1, 1, 1, 1]] + conv1_kernel_counts = [[3, 1], [4, 2, 2, 2], [3, 4, 4, 4], + [1, 4, 4, 4, 4, 4, 4, 4]] + conv2_kernel_counts = [[2, 2], [1, 2, 2, 2], [1, 2, 2, 2], + [1, 2, 2, 2, 1, 2, 2, 2]] + exp_factors = [[6, 3], [6, 6, 6, 6], [6, 6, 6, 6], + [6, 3, 3, 3, 6, 6, 6, 6]] + se_factors = [[0, 0], [2, 2, 2, 2], [4, 4, 4, 4], + [2, 2, 2, 2, 2, 2, 2, 2]] + else: + raise ValueError("Unsupported MixNet version {}".format(version)) + + final_block_channels = 1536 + + if width_scale != 1.0: + channels = [[round_channels(cij * width_scale) for cij in ci] + for ci in channels] + init_block_channels = round_channels(init_block_channels * width_scale) + + net = MixNet( + channels=channels, + init_block_channels=init_block_channels, + final_block_channels=final_block_channels, + exp_kernel_counts=exp_kernel_counts, + conv1_kernel_counts=conv1_kernel_counts, + conv2_kernel_counts=conv2_kernel_counts, + exp_factors=exp_factors, + se_factors=se_factors, + **kwargs) + + return net + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def MixNet_S(pretrained=False, use_ssld=False, **kwargs): + """ + MixNet-S model from 'MixConv: Mixed Depthwise Convolutional Kernels,' + https://arxiv.org/abs/1907.09595. + """ + model = get_mixnet( + version="s", width_scale=1.0, model_name="MixNet_S", **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MixNet_S"], use_ssld=use_ssld) + return model + + +def MixNet_M(pretrained=False, use_ssld=False, **kwargs): + """ + MixNet-M model from 'MixConv: Mixed Depthwise Convolutional Kernels,' + https://arxiv.org/abs/1907.09595. + """ + model = get_mixnet( + version="m", width_scale=1.0, model_name="MixNet_M", **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MixNet_M"], use_ssld=use_ssld) + return model + + +def MixNet_L(pretrained=False, use_ssld=False, **kwargs): + """ + MixNet-S model from 'MixConv: Mixed Depthwise Convolutional Kernels,' + https://arxiv.org/abs/1907.09595. + """ + model = get_mixnet( + version="m", width_scale=1.3, model_name="MixNet_L", **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MixNet_L"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/mobilenet_v2.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/mobilenet_v2.py new file mode 100644 index 00000000..b32c0250 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/mobilenet_v2.py @@ -0,0 +1,287 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "MobileNetV2_x0_25": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_25_pretrained.pdparams", + "MobileNetV2_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_5_pretrained.pdparams", + "MobileNetV2_x0_75": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_75_pretrained.pdparams", + "MobileNetV2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_pretrained.pdparams", + "MobileNetV2_x1_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x1_5_pretrained.pdparams", + "MobileNetV2_x2_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x2_0_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + filter_size, + num_filters, + stride, + padding, + channels=None, + num_groups=1, + name=None, + use_cudnn=True): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=padding, + groups=num_groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + + self._batch_norm = BatchNorm( + num_filters, + param_attr=ParamAttr(name=name + "_bn_scale"), + bias_attr=ParamAttr(name=name + "_bn_offset"), + moving_mean_name=name + "_bn_mean", + moving_variance_name=name + "_bn_variance") + + def forward(self, inputs, if_act=True): + y = self._conv(inputs) + y = self._batch_norm(y) + if if_act: + y = F.relu6(y) + return y + + +class InvertedResidualUnit(nn.Layer): + def __init__(self, num_channels, num_in_filter, num_filters, stride, + filter_size, padding, expansion_factor, name): + super(InvertedResidualUnit, self).__init__() + num_expfilter = int(round(num_in_filter * expansion_factor)) + self._expand_conv = ConvBNLayer( + num_channels=num_channels, + num_filters=num_expfilter, + filter_size=1, + stride=1, + padding=0, + num_groups=1, + name=name + "_expand") + + self._bottleneck_conv = ConvBNLayer( + num_channels=num_expfilter, + num_filters=num_expfilter, + filter_size=filter_size, + stride=stride, + padding=padding, + num_groups=num_expfilter, + use_cudnn=False, + name=name + "_dwise") + + self._linear_conv = ConvBNLayer( + num_channels=num_expfilter, + num_filters=num_filters, + filter_size=1, + stride=1, + padding=0, + num_groups=1, + name=name + "_linear") + + def forward(self, inputs, ifshortcut): + y = self._expand_conv(inputs, if_act=True) + y = self._bottleneck_conv(y, if_act=True) + y = self._linear_conv(y, if_act=False) + if ifshortcut: + y = paddle.add(inputs, y) + return y + + +class InvresiBlocks(nn.Layer): + def __init__(self, in_c, t, c, n, s, name): + super(InvresiBlocks, self).__init__() + + self._first_block = InvertedResidualUnit( + num_channels=in_c, + num_in_filter=in_c, + num_filters=c, + stride=s, + filter_size=3, + padding=1, + expansion_factor=t, + name=name + "_1") + + self._block_list = [] + for i in range(1, n): + block = self.add_sublayer( + name + "_" + str(i + 1), + sublayer=InvertedResidualUnit( + num_channels=c, + num_in_filter=c, + num_filters=c, + stride=1, + filter_size=3, + padding=1, + expansion_factor=t, + name=name + "_" + str(i + 1))) + self._block_list.append(block) + + def forward(self, inputs): + y = self._first_block(inputs, ifshortcut=False) + for block in self._block_list: + y = block(y, ifshortcut=True) + return y + + +class MobileNet(nn.Layer): + def __init__(self, class_num=1000, scale=1.0, prefix_name=""): + super(MobileNet, self).__init__() + self.scale = scale + self.class_num = class_num + + bottleneck_params_list = [ + (1, 16, 1, 1), + (6, 24, 2, 2), + (6, 32, 3, 2), + (6, 64, 4, 2), + (6, 96, 3, 1), + (6, 160, 3, 2), + (6, 320, 1, 1), + ] + + self.conv1 = ConvBNLayer( + num_channels=3, + num_filters=int(32 * scale), + filter_size=3, + stride=2, + padding=1, + name=prefix_name + "conv1_1") + + self.block_list = [] + i = 1 + in_c = int(32 * scale) + for layer_setting in bottleneck_params_list: + t, c, n, s = layer_setting + i += 1 + block = self.add_sublayer( + prefix_name + "conv" + str(i), + sublayer=InvresiBlocks( + in_c=in_c, + t=t, + c=int(c * scale), + n=n, + s=s, + name=prefix_name + "conv" + str(i))) + self.block_list.append(block) + in_c = int(c * scale) + + self.out_c = int(1280 * scale) if scale > 1.0 else 1280 + self.conv9 = ConvBNLayer( + num_channels=in_c, + num_filters=self.out_c, + filter_size=1, + stride=1, + padding=0, + name=prefix_name + "conv9") + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.out = Linear( + self.out_c, + class_num, + weight_attr=ParamAttr(name=prefix_name + "fc10_weights"), + bias_attr=ParamAttr(name=prefix_name + "fc10_offset")) + + def forward(self, inputs): + y = self.conv1(inputs, if_act=True) + for block in self.block_list: + y = block(y) + y = self.conv9(y, if_act=True) + y = self.pool2d_avg(y) + y = paddle.flatten(y, start_axis=1, stop_axis=-1) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def MobileNetV2_x0_25(pretrained=False, use_ssld=False, **kwargs): + model = MobileNet(scale=0.25, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileNetV2_x0_25"], use_ssld=use_ssld) + return model + + +def MobileNetV2_x0_5(pretrained=False, use_ssld=False, **kwargs): + model = MobileNet(scale=0.5, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileNetV2_x0_5"], use_ssld=use_ssld) + return model + + +def MobileNetV2_x0_75(pretrained=False, use_ssld=False, **kwargs): + model = MobileNet(scale=0.75, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileNetV2_x0_75"], use_ssld=use_ssld) + return model + + +def MobileNetV2(pretrained=False, use_ssld=False, **kwargs): + model = MobileNet(scale=1.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileNetV2"], use_ssld=use_ssld) + return model + + +def MobileNetV2_x1_5(pretrained=False, use_ssld=False, **kwargs): + model = MobileNet(scale=1.5, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileNetV2_x1_5"], use_ssld=use_ssld) + return model + + +def MobileNetV2_x2_0(pretrained=False, use_ssld=False, **kwargs): + model = MobileNet(scale=2.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileNetV2_x2_0"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/pvt_v2.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/pvt_v2.py new file mode 100644 index 00000000..e2fdfd4f --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/pvt_v2.py @@ -0,0 +1,492 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was heavily based on https://github.com/whai362/PVT + +from functools import partial +import math +import paddle +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn.initializer import TruncatedNormal, Constant + +from .vision_transformer import trunc_normal_, zeros_, ones_, to_2tuple, DropPath, Identity, drop_path + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "PVT_V2_B0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B0_pretrained.pdparams", + "PVT_V2_B1": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B1_pretrained.pdparams", + "PVT_V2_B2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2_pretrained.pdparams", + "PVT_V2_B2_Linear": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2_Linear_pretrained.pdparams", + "PVT_V2_B3": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B3_pretrained.pdparams", + "PVT_V2_B4": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B4_pretrained.pdparams", + "PVT_V2_B5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B5_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +@paddle.jit.not_to_static +def swapdim(x, dim1, dim2): + a = list(range(len(x.shape))) + a[dim1], a[dim2] = a[dim2], a[dim1] + return x.transpose(a) + + +class Mlp(nn.Layer): + def __init__(self, + in_features, + hidden_features=None, + out_features=None, + act_layer=nn.GELU, + drop=0., + linear=False): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Linear(in_features, hidden_features) + self.dwconv = DWConv(hidden_features) + self.act = act_layer() + self.fc2 = nn.Linear(hidden_features, out_features) + self.drop = nn.Dropout(drop) + self.linear = linear + if self.linear: + self.relu = nn.ReLU() + + def forward(self, x, H, W): + x = self.fc1(x) + if self.linear: + x = self.relu(x) + x = self.dwconv(x, H, W) + x = self.act(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +class Attention(nn.Layer): + def __init__(self, + dim, + num_heads=8, + qkv_bias=False, + qk_scale=None, + attn_drop=0., + proj_drop=0., + sr_ratio=1, + linear=False): + super().__init__() + assert dim % num_heads == 0 + + self.dim = dim + self.num_heads = num_heads + head_dim = dim // num_heads + self.scale = qk_scale or head_dim**-0.5 + + self.q = nn.Linear(dim, dim, bias_attr=qkv_bias) + self.kv = nn.Linear(dim, dim * 2, bias_attr=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + self.proj_drop = nn.Dropout(proj_drop) + + self.linear = linear + self.sr_ratio = sr_ratio + if not linear: + if sr_ratio > 1: + self.sr = nn.Conv2D( + dim, dim, kernel_size=sr_ratio, stride=sr_ratio) + self.norm = nn.LayerNorm(dim) + else: + self.pool = nn.AdaptiveAvgPool2D(7) + self.sr = nn.Conv2D(dim, dim, kernel_size=1, stride=1) + self.norm = nn.LayerNorm(dim) + self.act = nn.GELU() + + def forward(self, x, H, W): + B, N, C = x.shape + q = self.q(x).reshape( + [B, N, self.num_heads, C // self.num_heads]).transpose( + [0, 2, 1, 3]) + + if not self.linear: + if self.sr_ratio > 1: + x_ = x.transpose([0, 2, 1]).reshape([B, C, H, W]) + x_ = self.sr(x_) + h_, w_ = x_.shape[-2:] + x_ = x_.reshape([B, C, h_ * w_]).transpose([0, 2, 1]) + x_ = self.norm(x_) + kv = self.kv(x_) + kv = kv.reshape([ + B, kv.shape[2] * kv.shape[1] // 2 // C, 2, self.num_heads, + C // self.num_heads + ]).transpose([2, 0, 3, 1, 4]) + else: + kv = self.kv(x) + kv = kv.reshape([ + B, kv.shape[2] * kv.shape[1] // 2 // C, 2, self.num_heads, + C // self.num_heads + ]).transpose([2, 0, 3, 1, 4]) + else: + x_ = x.transpose([0, 2, 1]).reshape([B, C, H, W]) + x_ = self.sr(self.pool(x_)) + x_ = x_.reshape([B, C, x_.shape[2] * x_.shape[3]]).transpose( + [0, 2, 1]) + x_ = self.norm(x_) + x_ = self.act(x_) + kv = self.kv(x_) + kv = kv.reshape([ + B, kv.shape[2] * kv.shape[1] // 2 // C, 2, self.num_heads, + C // self.num_heads + ]).transpose([2, 0, 3, 1, 4]) + k, v = kv[0], kv[1] + + attn = (q @swapdim(k, -2, -1)) * self.scale + attn = F.softmax(attn, axis=-1) + attn = self.attn_drop(attn) + + x = swapdim((attn @v), 1, 2).reshape([B, N, C]) + x = self.proj(x) + x = self.proj_drop(x) + + return x + + +class Block(nn.Layer): + def __init__(self, + dim, + num_heads, + mlp_ratio=4., + qkv_bias=False, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer=nn.LayerNorm, + sr_ratio=1, + linear=False): + super().__init__() + self.norm1 = norm_layer(dim) + self.attn = Attention( + dim, + num_heads=num_heads, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + attn_drop=attn_drop, + proj_drop=drop, + sr_ratio=sr_ratio, + linear=linear) + self.drop_path = DropPath(drop_path) if drop_path > 0. else Identity() + self.norm2 = norm_layer(dim) + mlp_hidden_dim = int(dim * mlp_ratio) + self.mlp = Mlp(in_features=dim, + hidden_features=mlp_hidden_dim, + act_layer=act_layer, + drop=drop, + linear=linear) + + def forward(self, x, H, W): + x = x + self.drop_path(self.attn(self.norm1(x), H, W)) + x = x + self.drop_path(self.mlp(self.norm2(x), H, W)) + + return x + + +class OverlapPatchEmbed(nn.Layer): + """ Image to Patch Embedding + """ + + def __init__(self, + img_size=224, + patch_size=7, + stride=4, + in_chans=3, + embed_dim=768): + super().__init__() + img_size = to_2tuple(img_size) + patch_size = to_2tuple(patch_size) + + self.img_size = img_size + self.patch_size = patch_size + self.H, self.W = img_size[0] // patch_size[0], img_size[ + 1] // patch_size[1] + self.num_patches = self.H * self.W + self.proj = nn.Conv2D( + in_chans, + embed_dim, + kernel_size=patch_size, + stride=stride, + padding=(patch_size[0] // 2, patch_size[1] // 2)) + self.norm = nn.LayerNorm(embed_dim) + + def forward(self, x): + x = self.proj(x) + _, _, H, W = x.shape + x = x.flatten(2) + x = swapdim(x, 1, 2) + x = self.norm(x) + + return x, H, W + + +class PyramidVisionTransformerV2(nn.Layer): + def __init__(self, + img_size=224, + patch_size=16, + in_chans=3, + class_num=1000, + embed_dims=[64, 128, 256, 512], + num_heads=[1, 2, 4, 8], + mlp_ratios=[4, 4, 4, 4], + qkv_bias=False, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0., + norm_layer=nn.LayerNorm, + depths=[3, 4, 6, 3], + sr_ratios=[8, 4, 2, 1], + num_stages=4, + linear=False): + super().__init__() + self.class_num = class_num + self.depths = depths + self.num_stages = num_stages + + dpr = [x for x in paddle.linspace(0, drop_path_rate, sum(depths)) + ] # stochastic depth decay rule + cur = 0 + + for i in range(num_stages): + patch_embed = OverlapPatchEmbed( + img_size=img_size if i == 0 else img_size // (2**(i + 1)), + patch_size=7 if i == 0 else 3, + stride=4 if i == 0 else 2, + in_chans=in_chans if i == 0 else embed_dims[i - 1], + embed_dim=embed_dims[i]) + + block = nn.LayerList([ + Block( + dim=embed_dims[i], + num_heads=num_heads[i], + mlp_ratio=mlp_ratios[i], + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop_rate, + attn_drop=attn_drop_rate, + drop_path=dpr[cur + j], + norm_layer=norm_layer, + sr_ratio=sr_ratios[i], + linear=linear) for j in range(depths[i]) + ]) + norm = norm_layer(embed_dims[i]) + cur += depths[i] + + setattr(self, f"patch_embed{i + 1}", patch_embed) + setattr(self, f"block{i + 1}", block) + setattr(self, f"norm{i + 1}", norm) + + # classification head + self.head = nn.Linear(embed_dims[3], + class_num) if class_num > 0 else Identity() + + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + + def forward_features(self, x): + B = x.shape[0] + + for i in range(self.num_stages): + patch_embed = getattr(self, f"patch_embed{i + 1}") + block = getattr(self, f"block{i + 1}") + norm = getattr(self, f"norm{i + 1}") + x, H, W = patch_embed(x) + for blk in block: + x = blk(x, H, W) + x = norm(x) + if i != self.num_stages - 1: + x = x.reshape([B, H, W, x.shape[2]]).transpose([0, 3, 1, 2]) + + return x.mean(axis=1) + + def forward(self, x): + x = self.forward_features(x) + x = self.head(x) + + return x + + +class DWConv(nn.Layer): + def __init__(self, dim=768): + super().__init__() + self.dwconv = nn.Conv2D(dim, dim, 3, 1, 1, bias_attr=True, groups=dim) + + def forward(self, x, H, W): + B, N, C = x.shape + x = swapdim(x, 1, 2) + x = x.reshape([B, C, H, W]) + x = self.dwconv(x) + x = x.flatten(2) + x = swapdim(x, 1, 2) + + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def PVT_V2_B0(pretrained=False, use_ssld=False, **kwargs): + model = PyramidVisionTransformerV2( + patch_size=4, + embed_dims=[32, 64, 160, 256], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[2, 2, 2, 2], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["PVT_V2_B0"], use_ssld=use_ssld) + return model + + +def PVT_V2_B1(pretrained=False, use_ssld=False, **kwargs): + model = PyramidVisionTransformerV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[2, 2, 2, 2], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["PVT_V2_B1"], use_ssld=use_ssld) + return model + + +def PVT_V2_B2(pretrained=False, use_ssld=False, **kwargs): + model = PyramidVisionTransformerV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 4, 6, 3], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["PVT_V2_B2"], use_ssld=use_ssld) + return model + + +def PVT_V2_B3(pretrained=False, use_ssld=False, **kwargs): + model = PyramidVisionTransformerV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 4, 18, 3], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["PVT_V2_B3"], use_ssld=use_ssld) + return model + + +def PVT_V2_B4(pretrained=False, use_ssld=False, **kwargs): + model = PyramidVisionTransformerV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 8, 27, 3], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["PVT_V2_B4"], use_ssld=use_ssld) + return model + + +def PVT_V2_B5(pretrained=False, use_ssld=False, **kwargs): + model = PyramidVisionTransformerV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[4, 4, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 6, 40, 3], + sr_ratios=[8, 4, 2, 1], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["PVT_V2_B5"], use_ssld=use_ssld) + return model + + +def PVT_V2_B2_Linear(pretrained=False, use_ssld=False, **kwargs): + model = PyramidVisionTransformerV2( + patch_size=4, + embed_dims=[64, 128, 320, 512], + num_heads=[1, 2, 5, 8], + mlp_ratios=[8, 8, 4, 4], + qkv_bias=True, + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 4, 6, 3], + sr_ratios=[8, 4, 2, 1], + linear=True, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["PVT_V2_B2_Linear"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/rednet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/rednet.py new file mode 100644 index 00000000..be84da1e --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/rednet.py @@ -0,0 +1,203 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/d-li14/involution + +import paddle +import paddle.nn as nn + +from paddle.vision.models import resnet + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "RedNet26": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet26_pretrained.pdparams", + "RedNet38": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet38_pretrained.pdparams", + "RedNet50": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet50_pretrained.pdparams", + "RedNet101": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet101_pretrained.pdparams", + "RedNet152": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet152_pretrained.pdparams" +} + +__all__ = MODEL_URLS.keys() + + +class Involution(nn.Layer): + def __init__(self, channels, kernel_size, stride): + super(Involution, self).__init__() + self.kernel_size = kernel_size + self.stride = stride + self.channels = channels + reduction_ratio = 4 + self.group_channels = 16 + self.groups = self.channels // self.group_channels + self.conv1 = nn.Sequential( + ('conv', nn.Conv2D( + in_channels=channels, + out_channels=channels // reduction_ratio, + kernel_size=1, + bias_attr=False)), + ('bn', nn.BatchNorm2D(channels // reduction_ratio)), + ('activate', nn.ReLU())) + self.conv2 = nn.Sequential(('conv', nn.Conv2D( + in_channels=channels // reduction_ratio, + out_channels=kernel_size**2 * self.groups, + kernel_size=1, + stride=1))) + if stride > 1: + self.avgpool = nn.AvgPool2D(stride, stride) + + def forward(self, x): + weight = self.conv2( + self.conv1(x if self.stride == 1 else self.avgpool(x))) + b, c, h, w = weight.shape + weight = weight.reshape( + (b, self.groups, self.kernel_size**2, h, w)).unsqueeze(2) + + out = nn.functional.unfold(x, self.kernel_size, self.stride, + (self.kernel_size - 1) // 2, 1) + out = out.reshape( + (b, self.groups, self.group_channels, self.kernel_size**2, h, w)) + out = (weight * out).sum(axis=3).reshape((b, self.channels, h, w)) + return out + + +class BottleneckBlock(resnet.BottleneckBlock): + def __init__(self, + inplanes, + planes, + stride=1, + downsample=None, + groups=1, + base_width=64, + dilation=1, + norm_layer=None): + super(BottleneckBlock, self).__init__(inplanes, planes, stride, + downsample, groups, base_width, + dilation, norm_layer) + width = int(planes * (base_width / 64.)) * groups + self.conv2 = Involution(width, 7, stride) + + +class RedNet(resnet.ResNet): + def __init__(self, block, depth, class_num=1000, with_pool=True): + super(RedNet, self).__init__( + block=block, depth=50, num_classes=class_num, with_pool=with_pool) + layer_cfg = { + 26: [1, 2, 4, 1], + 38: [2, 3, 5, 2], + 50: [3, 4, 6, 3], + 101: [3, 4, 23, 3], + 152: [3, 8, 36, 3] + } + layers = layer_cfg[depth] + + self.conv1 = None + self.bn1 = None + self.relu = None + self.inplanes = 64 + self.class_num = class_num + self.stem = nn.Sequential( + nn.Sequential( + ('conv', nn.Conv2D( + in_channels=3, + out_channels=self.inplanes // 2, + kernel_size=3, + stride=2, + padding=1, + bias_attr=False)), + ('bn', nn.BatchNorm2D(self.inplanes // 2)), + ('activate', nn.ReLU())), + Involution(self.inplanes // 2, 3, 1), + nn.BatchNorm2D(self.inplanes // 2), + nn.ReLU(), + nn.Sequential( + ('conv', nn.Conv2D( + in_channels=self.inplanes // 2, + out_channels=self.inplanes, + kernel_size=3, + stride=1, + padding=1, + bias_attr=False)), ('bn', nn.BatchNorm2D(self.inplanes)), + ('activate', nn.ReLU()))) + + self.layer1 = self._make_layer(block, 64, layers[0]) + self.layer2 = self._make_layer(block, 128, layers[1], stride=2) + self.layer3 = self._make_layer(block, 256, layers[2], stride=2) + self.layer4 = self._make_layer(block, 512, layers[3], stride=2) + + def forward(self, x): + x = self.stem(x) + x = self.maxpool(x) + + x = self.layer1(x) + x = self.layer2(x) + x = self.layer3(x) + x = self.layer4(x) + + if self.with_pool: + x = self.avgpool(x) + + if self.class_num > 0: + x = paddle.flatten(x, 1) + x = self.fc(x) + + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def RedNet26(pretrained=False, **kwargs): + model = RedNet(BottleneckBlock, 26, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["RedNet26"]) + return model + + +def RedNet38(pretrained=False, **kwargs): + model = RedNet(BottleneckBlock, 38, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["RedNet38"]) + return model + + +def RedNet50(pretrained=False, **kwargs): + model = RedNet(BottleneckBlock, 50, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["RedNet50"]) + return model + + +def RedNet101(pretrained=False, **kwargs): + model = RedNet(BottleneckBlock, 101, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["RedNet101"]) + return model + + +def RedNet152(pretrained=False, **kwargs): + model = RedNet(BottleneckBlock, 152, **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["RedNet152"]) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/regnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/regnet.py new file mode 100644 index 00000000..dc381cbc --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/regnet.py @@ -0,0 +1,431 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/facebookresearch/pycls + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "RegNetX_200MF": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetX_200MF_pretrained.pdparams", + "RegNetX_4GF": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetX_4GF_pretrained.pdparams", + "RegNetX_32GF": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetX_32GF_pretrained.pdparams", + "RegNetY_200MF": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetY_200MF_pretrained.pdparams", + "RegNetY_4GF": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetY_4GF_pretrained.pdparams", + "RegNetY_32GF": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetY_32GF_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +def quantize_float(f, q): + """Converts a float to closest non-zero int divisible by q.""" + return int(round(f / q) * q) + + +def adjust_ws_gs_comp(ws, bms, gs): + """Adjusts the compatibility of widths and groups.""" + ws_bot = [int(w * b) for w, b in zip(ws, bms)] + gs = [min(g, w_bot) for g, w_bot in zip(gs, ws_bot)] + ws_bot = [quantize_float(w_bot, g) for w_bot, g in zip(ws_bot, gs)] + ws = [int(w_bot / b) for w_bot, b in zip(ws_bot, bms)] + return ws, gs + + +def get_stages_from_blocks(ws, rs): + """Gets ws/ds of network at each stage from per block values.""" + ts = [ + w != wp or r != rp + for w, wp, r, rp in zip(ws + [0], [0] + ws, rs + [0], [0] + rs) + ] + s_ws = [w for w, t in zip(ws, ts[:-1]) if t] + s_ds = np.diff([d for d, t in zip(range(len(ts)), ts) if t]).tolist() + return s_ws, s_ds + + +def generate_regnet(w_a, w_0, w_m, d, q=8): + """Generates per block ws from RegNet parameters.""" + assert w_a >= 0 and w_0 > 0 and w_m > 1 and w_0 % q == 0 + ws_cont = np.arange(d) * w_a + w_0 + ks = np.round(np.log(ws_cont / w_0) / np.log(w_m)) + ws = w_0 * np.power(w_m, ks) + ws = np.round(np.divide(ws, q)) * q + num_stages, max_stage = len(np.unique(ws)), ks.max() + 1 + ws, ws_cont = ws.astype(int).tolist(), ws_cont.tolist() + return ws, num_stages, max_stage, ws_cont + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + padding=0, + act=None, + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=padding, + groups=groups, + weight_attr=ParamAttr(name=name + ".conv2d.output.1.w_0"), + bias_attr=ParamAttr(name=name + ".conv2d.output.1.b_0")) + bn_name = name + "_bn" + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + ".output.1.w_0"), + bias_attr=ParamAttr(bn_name + ".output.1.b_0"), + moving_mean_name=bn_name + "_mean", + moving_variance_name=bn_name + "_variance") + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + bm, + gw, + se_on, + se_r, + shortcut=True, + name=None): + super(BottleneckBlock, self).__init__() + + # Compute the bottleneck width + w_b = int(round(num_filters * bm)) + # Compute the number of groups + num_gs = w_b // gw + self.se_on = se_on + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=w_b, + filter_size=1, + padding=0, + act="relu", + name=name + "_branch2a") + self.conv1 = ConvBNLayer( + num_channels=w_b, + num_filters=w_b, + filter_size=3, + stride=stride, + padding=1, + groups=num_gs, + act="relu", + name=name + "_branch2b") + if se_on: + w_se = int(round(num_channels * se_r)) + self.se_block = SELayer( + num_channels=w_b, + num_filters=w_b, + reduction_ratio=w_se, + name=name + "_branch2se") + self.conv2 = ConvBNLayer( + num_channels=w_b, + num_filters=num_filters, + filter_size=1, + act=None, + name=name + "_branch2c") + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + stride=stride, + name=name + "_branch1") + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + if self.se_on: + conv1 = self.se_block(conv1) + conv2 = self.conv2(conv1) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + + y = paddle.add(x=short, y=conv2) + y = F.relu(y) + return y + + +class SELayer(nn.Layer): + def __init__(self, num_channels, num_filters, reduction_ratio, name=None): + super(SELayer, self).__init__() + + self.pool2d_gap = AdaptiveAvgPool2D(1) + + self._num_channels = num_channels + + med_ch = int(num_channels / reduction_ratio) + stdv = 1.0 / math.sqrt(num_channels * 1.0) + self.squeeze = Linear( + num_channels, + med_ch, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_sqz_weights"), + bias_attr=ParamAttr(name=name + "_sqz_offset")) + + stdv = 1.0 / math.sqrt(med_ch * 1.0) + self.excitation = Linear( + med_ch, + num_filters, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_exc_weights"), + bias_attr=ParamAttr(name=name + "_exc_offset")) + + def forward(self, input): + pool = self.pool2d_gap(input) + pool = paddle.reshape(pool, shape=[-1, self._num_channels]) + squeeze = self.squeeze(pool) + squeeze = F.relu(squeeze) + excitation = self.excitation(squeeze) + excitation = F.sigmoid(excitation) + excitation = paddle.reshape( + excitation, shape=[-1, self._num_channels, 1, 1]) + out = input * excitation + return out + + +class RegNet(nn.Layer): + def __init__(self, + w_a, + w_0, + w_m, + d, + group_w, + bot_mul, + q=8, + se_on=False, + class_num=1000): + super(RegNet, self).__init__() + + # Generate RegNet ws per block + b_ws, num_s, max_s, ws_cont = generate_regnet(w_a, w_0, w_m, d, q) + # Convert to per stage format + ws, ds = get_stages_from_blocks(b_ws, b_ws) + # Generate group widths and bot muls + gws = [group_w for _ in range(num_s)] + bms = [bot_mul for _ in range(num_s)] + # Adjust the compatibility of ws and gws + ws, gws = adjust_ws_gs_comp(ws, bms, gws) + # Use the same stride for each stage + ss = [2 for _ in range(num_s)] + # Use SE for RegNetY + se_r = 0.25 + # Construct the model + # Group params by stage + stage_params = list(zip(ds, ws, ss, bms, gws)) + # Construct the stem + stem_type = "simple_stem_in" + stem_w = 32 + block_type = "res_bottleneck_block" + + self.conv = ConvBNLayer( + num_channels=3, + num_filters=stem_w, + filter_size=3, + stride=2, + padding=1, + act="relu", + name="stem_conv") + + self.block_list = [] + for block, (d, w_out, stride, bm, gw) in enumerate(stage_params): + shortcut = False + for i in range(d): + num_channels = stem_w if block == i == 0 else in_channels + # Stride apply to the first block of the stage + b_stride = stride if i == 0 else 1 + conv_name = "s" + str(block + 1) + "_b" + str(i + + 1) # chr(97 + i) + bottleneck_block = self.add_sublayer( + conv_name, + BottleneckBlock( + num_channels=num_channels, + num_filters=w_out, + stride=b_stride, + bm=bm, + gw=gw, + se_on=se_on, + se_r=se_r, + shortcut=shortcut, + name=conv_name)) + in_channels = w_out + self.block_list.append(bottleneck_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.pool2d_avg_channels = w_out + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_0.w_0"), + bias_attr=ParamAttr(name="fc_0.b_0")) + + def forward(self, inputs): + y = self.conv(inputs) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def RegNetX_200MF(pretrained=False, use_ssld=False, **kwargs): + model = RegNet( + w_a=36.44, + w_0=24, + w_m=2.49, + d=13, + group_w=8, + bot_mul=1.0, + q=8, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RegNetX_200MF"], use_ssld=use_ssld) + return model + + +def RegNetX_4GF(pretrained=False, use_ssld=False, **kwargs): + model = RegNet( + w_a=38.65, + w_0=96, + w_m=2.43, + d=23, + group_w=40, + bot_mul=1.0, + q=8, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RegNetX_4GF"], use_ssld=use_ssld) + return model + + +def RegNetX_32GF(pretrained=False, use_ssld=False, **kwargs): + model = RegNet( + w_a=69.86, + w_0=320, + w_m=2.0, + d=23, + group_w=168, + bot_mul=1.0, + q=8, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RegNetX_32GF"], use_ssld=use_ssld) + return model + + +def RegNetY_200MF(pretrained=False, use_ssld=False, **kwargs): + model = RegNet( + w_a=36.44, + w_0=24, + w_m=2.49, + d=13, + group_w=8, + bot_mul=1.0, + q=8, + se_on=True, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RegNetX_32GF"], use_ssld=use_ssld) + return model + + +def RegNetY_4GF(pretrained=False, use_ssld=False, **kwargs): + model = RegNet( + w_a=31.41, + w_0=96, + w_m=2.24, + d=22, + group_w=64, + bot_mul=1.0, + q=8, + se_on=True, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RegNetX_32GF"], use_ssld=use_ssld) + return model + + +def RegNetY_32GF(pretrained=False, use_ssld=False, **kwargs): + model = RegNet( + w_a=115.89, + w_0=232, + w_m=2.53, + d=20, + group_w=232, + bot_mul=1.0, + q=8, + se_on=True, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RegNetX_32GF"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/repvgg.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/repvgg.py new file mode 100644 index 00000000..e919a59f --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/repvgg.py @@ -0,0 +1,422 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/DingXiaoH/RepVGG + +import paddle.nn as nn +import paddle +import numpy as np + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "RepVGG_A0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A0_pretrained.pdparams", + "RepVGG_A1": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A1_pretrained.pdparams", + "RepVGG_A2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A2_pretrained.pdparams", + "RepVGG_B0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B0_pretrained.pdparams", + "RepVGG_B1": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1_pretrained.pdparams", + "RepVGG_B2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2_pretrained.pdparams", + "RepVGG_B3": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3_pretrained.pdparams", + "RepVGG_B1g2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g2_pretrained.pdparams", + "RepVGG_B1g4": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g4_pretrained.pdparams", + "RepVGG_B2g2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2g2_pretrained.pdparams", + "RepVGG_B2g4": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2g4_pretrained.pdparams", + "RepVGG_B3g2": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3g2_pretrained.pdparams", + "RepVGG_B3g4": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3g4_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + +optional_groupwise_layers = [2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26] +g2_map = {l: 2 for l in optional_groupwise_layers} +g4_map = {l: 4 for l in optional_groupwise_layers} + + +class ConvBN(nn.Layer): + def __init__(self, + in_channels, + out_channels, + kernel_size, + stride, + padding, + groups=1): + super(ConvBN, self).__init__() + self.conv = nn.Conv2D( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + groups=groups, + bias_attr=False) + self.bn = nn.BatchNorm2D(num_features=out_channels) + + def forward(self, x): + y = self.conv(x) + y = self.bn(y) + return y + + +class RepVGGBlock(nn.Layer): + def __init__(self, + in_channels, + out_channels, + kernel_size, + stride=1, + padding=0, + dilation=1, + groups=1, + padding_mode='zeros'): + super(RepVGGBlock, self).__init__() + self.in_channels = in_channels + self.out_channels = out_channels + self.kernel_size = kernel_size + self.stride = stride + self.padding = padding + self.dilation = dilation + self.groups = groups + self.padding_mode = padding_mode + + assert kernel_size == 3 + assert padding == 1 + + padding_11 = padding - kernel_size // 2 + + self.nonlinearity = nn.ReLU() + + self.rbr_identity = nn.BatchNorm2D( + num_features=in_channels + ) if out_channels == in_channels and stride == 1 else None + self.rbr_dense = ConvBN( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + groups=groups) + self.rbr_1x1 = ConvBN( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=1, + stride=stride, + padding=padding_11, + groups=groups) + + def forward(self, inputs): + if not self.training: + return self.nonlinearity(self.rbr_reparam(inputs)) + + if self.rbr_identity is None: + id_out = 0 + else: + id_out = self.rbr_identity(inputs) + return self.nonlinearity( + self.rbr_dense(inputs) + self.rbr_1x1(inputs) + id_out) + + def eval(self): + if not hasattr(self, 'rbr_reparam'): + self.rbr_reparam = nn.Conv2D( + in_channels=self.in_channels, + out_channels=self.out_channels, + kernel_size=self.kernel_size, + stride=self.stride, + padding=self.padding, + dilation=self.dilation, + groups=self.groups, + padding_mode=self.padding_mode) + self.training = False + kernel, bias = self.get_equivalent_kernel_bias() + self.rbr_reparam.weight.set_value(kernel) + self.rbr_reparam.bias.set_value(bias) + for layer in self.sublayers(): + layer.eval() + + def get_equivalent_kernel_bias(self): + kernel3x3, bias3x3 = self._fuse_bn_tensor(self.rbr_dense) + kernel1x1, bias1x1 = self._fuse_bn_tensor(self.rbr_1x1) + kernelid, biasid = self._fuse_bn_tensor(self.rbr_identity) + return kernel3x3 + self._pad_1x1_to_3x3_tensor( + kernel1x1) + kernelid, bias3x3 + bias1x1 + biasid + + def _pad_1x1_to_3x3_tensor(self, kernel1x1): + if kernel1x1 is None: + return 0 + else: + return nn.functional.pad(kernel1x1, [1, 1, 1, 1]) + + def _fuse_bn_tensor(self, branch): + if branch is None: + return 0, 0 + if isinstance(branch, ConvBN): + kernel = branch.conv.weight + running_mean = branch.bn._mean + running_var = branch.bn._variance + gamma = branch.bn.weight + beta = branch.bn.bias + eps = branch.bn._epsilon + else: + assert isinstance(branch, nn.BatchNorm2D) + if not hasattr(self, 'id_tensor'): + input_dim = self.in_channels // self.groups + kernel_value = np.zeros( + (self.in_channels, input_dim, 3, 3), dtype=np.float32) + for i in range(self.in_channels): + kernel_value[i, i % input_dim, 1, 1] = 1 + self.id_tensor = paddle.to_tensor(kernel_value) + kernel = self.id_tensor + running_mean = branch._mean + running_var = branch._variance + gamma = branch.weight + beta = branch.bias + eps = branch._epsilon + std = (running_var + eps).sqrt() + t = (gamma / std).reshape((-1, 1, 1, 1)) + return kernel * t, beta - running_mean * gamma / std + + +class RepVGG(nn.Layer): + def __init__(self, + num_blocks, + width_multiplier=None, + override_groups_map=None, + class_num=1000): + super(RepVGG, self).__init__() + + assert len(width_multiplier) == 4 + self.override_groups_map = override_groups_map or dict() + + assert 0 not in self.override_groups_map + + self.in_planes = min(64, int(64 * width_multiplier[0])) + + self.stage0 = RepVGGBlock( + in_channels=3, + out_channels=self.in_planes, + kernel_size=3, + stride=2, + padding=1) + self.cur_layer_idx = 1 + self.stage1 = self._make_stage( + int(64 * width_multiplier[0]), num_blocks[0], stride=2) + self.stage2 = self._make_stage( + int(128 * width_multiplier[1]), num_blocks[1], stride=2) + self.stage3 = self._make_stage( + int(256 * width_multiplier[2]), num_blocks[2], stride=2) + self.stage4 = self._make_stage( + int(512 * width_multiplier[3]), num_blocks[3], stride=2) + self.gap = nn.AdaptiveAvgPool2D(output_size=1) + self.linear = nn.Linear(int(512 * width_multiplier[3]), class_num) + + def _make_stage(self, planes, num_blocks, stride): + strides = [stride] + [1] * (num_blocks - 1) + blocks = [] + for stride in strides: + cur_groups = self.override_groups_map.get(self.cur_layer_idx, 1) + blocks.append( + RepVGGBlock( + in_channels=self.in_planes, + out_channels=planes, + kernel_size=3, + stride=stride, + padding=1, + groups=cur_groups)) + self.in_planes = planes + self.cur_layer_idx += 1 + return nn.Sequential(*blocks) + + def eval(self): + self.training = False + for layer in self.sublayers(): + layer.training = False + layer.eval() + + def forward(self, x): + out = self.stage0(x) + out = self.stage1(out) + out = self.stage2(out) + out = self.stage3(out) + out = self.stage4(out) + out = self.gap(out) + out = paddle.flatten(out, start_axis=1) + out = self.linear(out) + return out + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def RepVGG_A0(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[2, 4, 14, 1], + width_multiplier=[0.75, 0.75, 0.75, 2.5], + override_groups_map=None, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_A0"], use_ssld=use_ssld) + return model + + +def RepVGG_A1(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[2, 4, 14, 1], + width_multiplier=[1, 1, 1, 2.5], + override_groups_map=None, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_A1"], use_ssld=use_ssld) + return model + + +def RepVGG_A2(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[2, 4, 14, 1], + width_multiplier=[1.5, 1.5, 1.5, 2.75], + override_groups_map=None, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_A2"], use_ssld=use_ssld) + return model + + +def RepVGG_B0(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[1, 1, 1, 2.5], + override_groups_map=None, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B0"], use_ssld=use_ssld) + return model + + +def RepVGG_B1(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[2, 2, 2, 4], + override_groups_map=None, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B1"], use_ssld=use_ssld) + return model + + +def RepVGG_B1g2(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[2, 2, 2, 4], + override_groups_map=g2_map, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B1g2"], use_ssld=use_ssld) + return model + + +def RepVGG_B1g4(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[2, 2, 2, 4], + override_groups_map=g4_map, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B1g4"], use_ssld=use_ssld) + return model + + +def RepVGG_B2(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[2.5, 2.5, 2.5, 5], + override_groups_map=None, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B2"], use_ssld=use_ssld) + return model + + +def RepVGG_B2g2(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[2.5, 2.5, 2.5, 5], + override_groups_map=g2_map, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B2g2"], use_ssld=use_ssld) + return model + + +def RepVGG_B2g4(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[2.5, 2.5, 2.5, 5], + override_groups_map=g4_map, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B2g4"], use_ssld=use_ssld) + return model + + +def RepVGG_B3(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[3, 3, 3, 5], + override_groups_map=None, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B3"], use_ssld=use_ssld) + return model + + +def RepVGG_B3g2(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[3, 3, 3, 5], + override_groups_map=g2_map, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B3g2"], use_ssld=use_ssld) + return model + + +def RepVGG_B3g4(pretrained=False, use_ssld=False, **kwargs): + model = RepVGG( + num_blocks=[4, 6, 16, 1], + width_multiplier=[3, 3, 3, 5], + override_groups_map=g4_map, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["RepVGG_B3g4"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/res2net.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/res2net.py new file mode 100644 index 00000000..191cc849 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/res2net.py @@ -0,0 +1,264 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "Res2Net50_26w_4s": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_26w_4s_pretrained.pdparams", + "Res2Net50_14w_8s": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_14w_8s_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__( + self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + act=None, + name=None, ): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + if name == "conv1": + bn_name = "bn_" + name + else: + bn_name = "bn" + name[3:] + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels1, + num_channels2, + num_filters, + stride, + scales, + shortcut=True, + if_first=False, + name=None): + super(BottleneckBlock, self).__init__() + self.stride = stride + self.scales = scales + self.conv0 = ConvBNLayer( + num_channels=num_channels1, + num_filters=num_filters, + filter_size=1, + act='relu', + name=name + "_branch2a") + self.conv1_list = [] + for s in range(scales - 1): + conv1 = self.add_sublayer( + name + '_branch2b_' + str(s + 1), + ConvBNLayer( + num_channels=num_filters // scales, + num_filters=num_filters // scales, + filter_size=3, + stride=stride, + act='relu', + name=name + '_branch2b_' + str(s + 1))) + self.conv1_list.append(conv1) + self.pool2d_avg = AvgPool2D(kernel_size=3, stride=stride, padding=1) + + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_channels2, + filter_size=1, + act=None, + name=name + "_branch2c") + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels1, + num_filters=num_channels2, + filter_size=1, + stride=stride, + name=name + "_branch1") + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + xs = paddle.split(y, self.scales, 1) + ys = [] + for s, conv1 in enumerate(self.conv1_list): + if s == 0 or self.stride == 2: + ys.append(conv1(xs[s])) + else: + ys.append(conv1(paddle.add(xs[s], ys[-1]))) + if self.stride == 1: + ys.append(xs[-1]) + else: + ys.append(self.pool2d_avg(xs[-1])) + conv1 = paddle.concat(ys, axis=1) + conv2 = self.conv2(conv1) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + y = paddle.add(x=short, y=conv2) + y = F.relu(y) + return y + + +class Res2Net(nn.Layer): + def __init__(self, layers=50, scales=4, width=26, class_num=1000): + super(Res2Net, self).__init__() + + self.layers = layers + self.scales = scales + self.width = width + basic_width = self.width * self.scales + supported_layers = [50, 101, 152, 200] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + + if layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + elif layers == 200: + depth = [3, 12, 48, 3] + num_channels = [64, 256, 512, 1024] + num_channels2 = [256, 512, 1024, 2048] + num_filters = [basic_width * t for t in [1, 2, 4, 8]] + + self.conv1 = ConvBNLayer( + num_channels=3, + num_filters=64, + filter_size=7, + stride=2, + act='relu', + name="conv1") + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.block_list = [] + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + if layers in [101, 152] and block == 2: + if i == 0: + conv_name = "res" + str(block + 2) + "a" + else: + conv_name = "res" + str(block + 2) + "b" + str(i) + else: + conv_name = "res" + str(block + 2) + chr(97 + i) + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels1=num_channels[block] + if i == 0 else num_channels2[block], + num_channels2=num_channels2[block], + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + scales=scales, + shortcut=shortcut, + if_first=block == i == 0, + name=conv_name)) + self.block_list.append(bottleneck_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_weights"), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, inputs): + y = self.conv1(inputs) + y = self.pool2d_max(y) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def Res2Net50_26w_4s(pretrained=False, use_ssld=False, **kwargs): + model = Res2Net(layers=50, scales=4, width=26, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["Res2Net50_26w_4s"], use_ssld=use_ssld) + return model + + +def Res2Net50_14w_8s(pretrained=False, use_ssld=False, **kwargs): + model = Res2Net(layers=50, scales=8, width=14, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["Res2Net50_14w_8s"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/res2net_vd.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/res2net_vd.py new file mode 100644 index 00000000..a3756798 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/res2net_vd.py @@ -0,0 +1,305 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "Res2Net50_vd_26w_4s": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_pretrained.pdparams", + "Res2Net101_vd_26w_4s": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net101_vd_26w_4s_pretrained.pdparams", + "Res2Net200_vd_26w_4s": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net200_vd_26w_4s_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__( + self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + is_vd_mode=False, + act=None, + name=None, ): + super(ConvBNLayer, self).__init__() + + self.is_vd_mode = is_vd_mode + self._pool2d_avg = AvgPool2D( + kernel_size=2, stride=2, padding=0, ceil_mode=True) + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + if name == "conv1": + bn_name = "bn_" + name + else: + bn_name = "bn" + name[3:] + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + if self.is_vd_mode: + inputs = self._pool2d_avg(inputs) + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels1, + num_channels2, + num_filters, + stride, + scales, + shortcut=True, + if_first=False, + name=None): + super(BottleneckBlock, self).__init__() + self.stride = stride + self.scales = scales + self.conv0 = ConvBNLayer( + num_channels=num_channels1, + num_filters=num_filters, + filter_size=1, + act='relu', + name=name + "_branch2a") + self.conv1_list = [] + for s in range(scales - 1): + conv1 = self.add_sublayer( + name + '_branch2b_' + str(s + 1), + ConvBNLayer( + num_channels=num_filters // scales, + num_filters=num_filters // scales, + filter_size=3, + stride=stride, + act='relu', + name=name + '_branch2b_' + str(s + 1))) + self.conv1_list.append(conv1) + self.pool2d_avg = AvgPool2D(kernel_size=3, stride=stride, padding=1) + + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_channels2, + filter_size=1, + act=None, + name=name + "_branch2c") + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels1, + num_filters=num_channels2, + filter_size=1, + stride=1, + is_vd_mode=False if if_first else True, + name=name + "_branch1") + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + xs = paddle.split(y, self.scales, 1) + ys = [] + for s, conv1 in enumerate(self.conv1_list): + if s == 0 or self.stride == 2: + ys.append(conv1(xs[s])) + else: + ys.append(conv1(xs[s] + ys[-1])) + if self.stride == 1: + ys.append(xs[-1]) + else: + ys.append(self.pool2d_avg(xs[-1])) + conv1 = paddle.concat(ys, axis=1) + conv2 = self.conv2(conv1) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + y = paddle.add(x=short, y=conv2) + y = F.relu(y) + return y + + +class Res2Net_vd(nn.Layer): + def __init__(self, layers=50, scales=4, width=26, class_num=1000): + super(Res2Net_vd, self).__init__() + + self.layers = layers + self.scales = scales + self.width = width + basic_width = self.width * self.scales + supported_layers = [50, 101, 152, 200] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + + if layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + elif layers == 200: + depth = [3, 12, 48, 3] + num_channels = [64, 256, 512, 1024] + num_channels2 = [256, 512, 1024, 2048] + num_filters = [basic_width * t for t in [1, 2, 4, 8]] + + self.conv1_1 = ConvBNLayer( + num_channels=3, + num_filters=32, + filter_size=3, + stride=2, + act='relu', + name="conv1_1") + self.conv1_2 = ConvBNLayer( + num_channels=32, + num_filters=32, + filter_size=3, + stride=1, + act='relu', + name="conv1_2") + self.conv1_3 = ConvBNLayer( + num_channels=32, + num_filters=64, + filter_size=3, + stride=1, + act='relu', + name="conv1_3") + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.block_list = [] + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + if layers in [101, 152, 200] and block == 2: + if i == 0: + conv_name = "res" + str(block + 2) + "a" + else: + conv_name = "res" + str(block + 2) + "b" + str(i) + else: + conv_name = "res" + str(block + 2) + chr(97 + i) + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels1=num_channels[block] + if i == 0 else num_channels2[block], + num_channels2=num_channels2[block], + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + scales=scales, + shortcut=shortcut, + if_first=block == i == 0, + name=conv_name)) + self.block_list.append(bottleneck_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_weights"), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, inputs): + y = self.conv1_1(inputs) + y = self.conv1_2(y) + y = self.conv1_3(y) + y = self.pool2d_max(y) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def Res2Net50_vd_26w_4s(pretrained=False, use_ssld=False, **kwargs): + model = Res2Net_vd(layers=50, scales=4, width=26, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["Res2Net50_vd_26w_4s"], + use_ssld=use_ssld) + return model + + +def Res2Net101_vd_26w_4s(pretrained=False, use_ssld=False, **kwargs): + model = Res2Net_vd(layers=101, scales=4, width=26, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["Res2Net101_vd_26w_4s"], + use_ssld=use_ssld) + return model + + +def Res2Net200_vd_26w_4s(pretrained=False, use_ssld=False, **kwargs): + model = Res2Net_vd(layers=200, scales=4, width=26, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["Res2Net200_vd_26w_4s"], + use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnest.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnest.py new file mode 100644 index 00000000..88eee8aa --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnest.py @@ -0,0 +1,740 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/zhanghang1989/ResNeSt + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +import math +import paddle.nn as nn +import paddle.nn.functional as F +from paddle import ParamAttr +from paddle.nn.initializer import KaimingNormal +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.regularizer import L2Decay + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ResNeSt50_fast_1s1x64d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_fast_1s1x64d_pretrained.pdparams", + "ResNeSt50": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_pretrained.pdparams", + "ResNeSt101": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt101_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + dilation=1, + groups=1, + act=None, + name=None): + super(ConvBNLayer, self).__init__() + + bn_decay = 0.0 + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + dilation=dilation, + groups=groups, + weight_attr=ParamAttr(name=name + "_weight"), + bias_attr=False) + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr( + name=name + "_scale", regularizer=L2Decay(bn_decay)), + bias_attr=ParamAttr( + name + "_offset", regularizer=L2Decay(bn_decay)), + moving_mean_name=name + "_mean", + moving_variance_name=name + "_variance") + + def forward(self, x): + x = self._conv(x) + x = self._batch_norm(x) + return x + + +class rSoftmax(nn.Layer): + def __init__(self, radix, cardinality): + super(rSoftmax, self).__init__() + self.radix = radix + self.cardinality = cardinality + + def forward(self, x): + cardinality = self.cardinality + radix = self.radix + + batch, r, h, w = x.shape + if self.radix > 1: + x = paddle.reshape( + x=x, + shape=[ + batch, cardinality, radix, + int(r * h * w / cardinality / radix) + ]) + x = paddle.transpose(x=x, perm=[0, 2, 1, 3]) + x = nn.functional.softmax(x, axis=1) + x = paddle.reshape(x=x, shape=[batch, r * h * w, 1, 1]) + else: + x = nn.functional.sigmoid(x) + return x + + +class SplatConv(nn.Layer): + def __init__(self, + in_channels, + channels, + kernel_size, + stride=1, + padding=0, + dilation=1, + groups=1, + bias=True, + radix=2, + reduction_factor=4, + rectify_avg=False, + name=None): + super(SplatConv, self).__init__() + + self.radix = radix + + self.conv1 = ConvBNLayer( + num_channels=in_channels, + num_filters=channels * radix, + filter_size=kernel_size, + stride=stride, + groups=groups * radix, + act="relu", + name=name + "_1_weights") + + self.avg_pool2d = AdaptiveAvgPool2D(1) + + inter_channels = int(max(in_channels * radix // reduction_factor, 32)) + + # to calc gap + self.conv2 = ConvBNLayer( + num_channels=channels, + num_filters=inter_channels, + filter_size=1, + stride=1, + groups=groups, + act="relu", + name=name + "_2_weights") + + # to calc atten + self.conv3 = Conv2D( + in_channels=inter_channels, + out_channels=channels * radix, + kernel_size=1, + stride=1, + padding=0, + groups=groups, + weight_attr=ParamAttr( + name=name + "_weights", initializer=KaimingNormal()), + bias_attr=False) + + self.rsoftmax = rSoftmax(radix=radix, cardinality=groups) + + def forward(self, x): + x = self.conv1(x) + + if self.radix > 1: + splited = paddle.split(x, num_or_sections=self.radix, axis=1) + gap = paddle.add_n(splited) + else: + gap = x + + gap = self.avg_pool2d(gap) + gap = self.conv2(gap) + + atten = self.conv3(gap) + atten = self.rsoftmax(atten) + + if self.radix > 1: + attens = paddle.split(atten, num_or_sections=self.radix, axis=1) + y = paddle.add_n([ + paddle.multiply(split, att) + for (att, split) in zip(attens, splited) + ]) + else: + y = paddle.multiply(x, atten) + + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + inplanes, + planes, + stride=1, + radix=1, + cardinality=1, + bottleneck_width=64, + avd=False, + avd_first=False, + dilation=1, + is_first=False, + rectify_avg=False, + last_gamma=False, + avg_down=False, + name=None): + super(BottleneckBlock, self).__init__() + self.inplanes = inplanes + self.planes = planes + self.stride = stride + self.radix = radix + self.cardinality = cardinality + self.avd = avd + self.avd_first = avd_first + self.dilation = dilation + self.is_first = is_first + self.rectify_avg = rectify_avg + self.last_gamma = last_gamma + self.avg_down = avg_down + + group_width = int(planes * (bottleneck_width / 64.)) * cardinality + + self.conv1 = ConvBNLayer( + num_channels=self.inplanes, + num_filters=group_width, + filter_size=1, + stride=1, + groups=1, + act="relu", + name=name + "_conv1") + + if avd and avd_first and (stride > 1 or is_first): + self.avg_pool2d_1 = AvgPool2D( + kernel_size=3, stride=stride, padding=1) + + if radix >= 1: + self.conv2 = SplatConv( + in_channels=group_width, + channels=group_width, + kernel_size=3, + stride=1, + padding=dilation, + dilation=dilation, + groups=cardinality, + bias=False, + radix=radix, + rectify_avg=rectify_avg, + name=name + "_splat") + else: + self.conv2 = ConvBNLayer( + num_channels=group_width, + num_filters=group_width, + filter_size=3, + stride=1, + dilation=dilation, + groups=cardinality, + act="relu", + name=name + "_conv2") + + if avd and avd_first == False and (stride > 1 or is_first): + self.avg_pool2d_2 = AvgPool2D( + kernel_size=3, stride=stride, padding=1) + + self.conv3 = ConvBNLayer( + num_channels=group_width, + num_filters=planes * 4, + filter_size=1, + stride=1, + groups=1, + act=None, + name=name + "_conv3") + + if stride != 1 or self.inplanes != self.planes * 4: + if avg_down: + if dilation == 1: + self.avg_pool2d_3 = AvgPool2D( + kernel_size=stride, stride=stride, padding=0) + else: + self.avg_pool2d_3 = AvgPool2D( + kernel_size=1, stride=1, padding=0, ceil_mode=True) + + self.conv4 = Conv2D( + in_channels=self.inplanes, + out_channels=planes * 4, + kernel_size=1, + stride=1, + padding=0, + groups=1, + weight_attr=ParamAttr( + name=name + "_weights", initializer=KaimingNormal()), + bias_attr=False) + else: + self.conv4 = Conv2D( + in_channels=self.inplanes, + out_channels=planes * 4, + kernel_size=1, + stride=stride, + padding=0, + groups=1, + weight_attr=ParamAttr( + name=name + "_shortcut_weights", + initializer=KaimingNormal()), + bias_attr=False) + + bn_decay = 0.0 + self._batch_norm = BatchNorm( + planes * 4, + act=None, + param_attr=ParamAttr( + name=name + "_shortcut_scale", + regularizer=L2Decay(bn_decay)), + bias_attr=ParamAttr( + name + "_shortcut_offset", regularizer=L2Decay(bn_decay)), + moving_mean_name=name + "_shortcut_mean", + moving_variance_name=name + "_shortcut_variance") + + def forward(self, x): + short = x + + x = self.conv1(x) + if self.avd and self.avd_first and (self.stride > 1 or self.is_first): + x = self.avg_pool2d_1(x) + + x = self.conv2(x) + + if self.avd and self.avd_first == False and (self.stride > 1 or + self.is_first): + x = self.avg_pool2d_2(x) + + x = self.conv3(x) + + if self.stride != 1 or self.inplanes != self.planes * 4: + if self.avg_down: + short = self.avg_pool2d_3(short) + + short = self.conv4(short) + + short = self._batch_norm(short) + + y = paddle.add(x=short, y=x) + y = F.relu(y) + return y + + +class ResNeStLayer(nn.Layer): + def __init__(self, + inplanes, + planes, + blocks, + radix, + cardinality, + bottleneck_width, + avg_down, + avd, + avd_first, + rectify_avg, + last_gamma, + stride=1, + dilation=1, + is_first=True, + name=None): + super(ResNeStLayer, self).__init__() + self.inplanes = inplanes + self.planes = planes + self.blocks = blocks + self.radix = radix + self.cardinality = cardinality + self.bottleneck_width = bottleneck_width + self.avg_down = avg_down + self.avd = avd + self.avd_first = avd_first + self.rectify_avg = rectify_avg + self.last_gamma = last_gamma + self.is_first = is_first + + if dilation == 1 or dilation == 2: + bottleneck_func = self.add_sublayer( + name + "_bottleneck_0", + BottleneckBlock( + inplanes=self.inplanes, + planes=planes, + stride=stride, + radix=radix, + cardinality=cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + dilation=1, + is_first=is_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + name=name + "_bottleneck_0")) + elif dilation == 4: + bottleneck_func = self.add_sublayer( + name + "_bottleneck_0", + BottleneckBlock( + inplanes=self.inplanes, + planes=planes, + stride=stride, + radix=radix, + cardinality=cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + dilation=2, + is_first=is_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + name=name + "_bottleneck_0")) + else: + raise RuntimeError("=>unknown dilation size") + + self.inplanes = planes * 4 + self.bottleneck_block_list = [bottleneck_func] + for i in range(1, blocks): + curr_name = name + "_bottleneck_" + str(i) + + bottleneck_func = self.add_sublayer( + curr_name, + BottleneckBlock( + inplanes=self.inplanes, + planes=planes, + radix=radix, + cardinality=cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + dilation=dilation, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + name=curr_name)) + self.bottleneck_block_list.append(bottleneck_func) + + def forward(self, x): + for bottleneck_block in self.bottleneck_block_list: + x = bottleneck_block(x) + return x + + +class ResNeSt(nn.Layer): + def __init__(self, + layers, + radix=1, + groups=1, + bottleneck_width=64, + dilated=False, + dilation=1, + deep_stem=False, + stem_width=64, + avg_down=False, + rectify_avg=False, + avd=False, + avd_first=False, + final_drop=0.0, + last_gamma=False, + class_num=1000): + super(ResNeSt, self).__init__() + + self.cardinality = groups + self.bottleneck_width = bottleneck_width + # ResNet-D params + self.inplanes = stem_width * 2 if deep_stem else 64 + self.avg_down = avg_down + self.last_gamma = last_gamma + # ResNeSt params + self.radix = radix + self.avd = avd + self.avd_first = avd_first + + self.deep_stem = deep_stem + self.stem_width = stem_width + self.layers = layers + self.final_drop = final_drop + self.dilated = dilated + self.dilation = dilation + + self.rectify_avg = rectify_avg + + if self.deep_stem: + self.stem = nn.Sequential( + ("conv1", ConvBNLayer( + num_channels=3, + num_filters=stem_width, + filter_size=3, + stride=2, + act="relu", + name="conv1")), ("conv2", ConvBNLayer( + num_channels=stem_width, + num_filters=stem_width, + filter_size=3, + stride=1, + act="relu", + name="conv2")), ("conv3", ConvBNLayer( + num_channels=stem_width, + num_filters=stem_width * 2, + filter_size=3, + stride=1, + act="relu", + name="conv3"))) + else: + self.stem = ConvBNLayer( + num_channels=3, + num_filters=stem_width, + filter_size=7, + stride=2, + act="relu", + name="conv1") + + self.max_pool2d = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.layer1 = ResNeStLayer( + inplanes=self.stem_width * 2 + if self.deep_stem else self.stem_width, + planes=64, + blocks=self.layers[0], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=1, + dilation=1, + is_first=False, + name="layer1") + + # return + + self.layer2 = ResNeStLayer( + inplanes=256, + planes=128, + blocks=self.layers[1], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=2, + name="layer2") + + if self.dilated or self.dilation == 4: + self.layer3 = ResNeStLayer( + inplanes=512, + planes=256, + blocks=self.layers[2], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=1, + dilation=2, + name="layer3") + self.layer4 = ResNeStLayer( + inplanes=1024, + planes=512, + blocks=self.layers[3], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=1, + dilation=4, + name="layer4") + elif self.dilation == 2: + self.layer3 = ResNeStLayer( + inplanes=512, + planes=256, + blocks=self.layers[2], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=2, + dilation=1, + name="layer3") + self.layer4 = ResNeStLayer( + inplanes=1024, + planes=512, + blocks=self.layers[3], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=1, + dilation=2, + name="layer4") + else: + self.layer3 = ResNeStLayer( + inplanes=512, + planes=256, + blocks=self.layers[2], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=2, + name="layer3") + self.layer4 = ResNeStLayer( + inplanes=1024, + planes=512, + blocks=self.layers[3], + radix=radix, + cardinality=self.cardinality, + bottleneck_width=bottleneck_width, + avg_down=self.avg_down, + avd=avd, + avd_first=avd_first, + rectify_avg=rectify_avg, + last_gamma=last_gamma, + stride=2, + name="layer4") + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.out_channels = 2048 + + stdv = 1.0 / math.sqrt(self.out_channels * 1.0) + + self.out = Linear( + self.out_channels, + class_num, + weight_attr=ParamAttr( + initializer=nn.initializer.Uniform(-stdv, stdv), + name="fc_weights"), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, x): + x = self.stem(x) + x = self.max_pool2d(x) + x = self.layer1(x) + x = self.layer2(x) + + x = self.layer3(x) + + x = self.layer4(x) + x = self.pool2d_avg(x) + x = paddle.reshape(x, shape=[-1, self.out_channels]) + x = self.out(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ResNeSt50_fast_1s1x64d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeSt( + layers=[3, 4, 6, 3], + radix=1, + groups=1, + bottleneck_width=64, + deep_stem=True, + stem_width=32, + avg_down=True, + avd=True, + avd_first=True, + final_drop=0.0, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeSt50_fast_1s1x64d"], + use_ssld=use_ssld) + return model + + +def ResNeSt50(pretrained=False, use_ssld=False, **kwargs): + model = ResNeSt( + layers=[3, 4, 6, 3], + radix=2, + groups=1, + bottleneck_width=64, + deep_stem=True, + stem_width=32, + avg_down=True, + avd=True, + avd_first=False, + final_drop=0.0, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeSt50"], use_ssld=use_ssld) + return model + + +def ResNeSt101(pretrained=False, use_ssld=False, **kwargs): + model = ResNeSt( + layers=[3, 4, 23, 3], + radix=2, + groups=1, + bottleneck_width=64, + deep_stem=True, + stem_width=64, + avg_down=True, + avd=True, + avd_first=False, + final_drop=0.0, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeSt101"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnet_vc.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnet_vc.py new file mode 100644 index 00000000..6b972dc7 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnet_vc.py @@ -0,0 +1,309 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ResNet50_vc": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vc_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + act=None, + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + if name == "conv1": + bn_name = "bn_" + name + else: + bn_name = "bn" + name[3:] + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + name=None): + super(BottleneckBlock, self).__init__() + + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act='relu', + name=name + "_branch2a") + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + stride=stride, + act='relu', + name=name + "_branch2b") + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 4, + filter_size=1, + act=None, + name=name + "_branch2c") + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 4, + filter_size=1, + stride=stride, + name=name + "_branch1") + + self.shortcut = shortcut + + self._num_channels_out = num_filters * 4 + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + conv2 = self.conv2(conv1) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + + y = paddle.add(x=short, y=conv2) + y = F.relu(y) + return y + + +class BasicBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + name=None): + super(BasicBlock, self).__init__() + self.stride = stride + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=3, + stride=stride, + act='relu', + name=name + "_branch2a") + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + act=None, + name=name + "_branch2b") + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + stride=stride, + name=name + "_branch1") + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + y = paddle.add(x=short, y=conv1) + y = F.relu(y) + return y + + +class ResNet_vc(nn.Layer): + def __init__(self, layers=50, class_num=1000): + super(ResNet_vc, self).__init__() + + self.layers = layers + supported_layers = [18, 34, 50, 101, 152] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + + if layers == 18: + depth = [2, 2, 2, 2] + elif layers == 34 or layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + num_channels = [64, 256, 512, + 1024] if layers >= 50 else [64, 64, 128, 256] + num_filters = [64, 128, 256, 512] + + self.conv1_1 = ConvBNLayer( + num_channels=3, + num_filters=32, + filter_size=3, + stride=2, + act='relu', + name="conv1_1") + self.conv1_2 = ConvBNLayer( + num_channels=32, + num_filters=32, + filter_size=3, + stride=1, + act='relu', + name="conv1_2") + self.conv1_3 = ConvBNLayer( + num_channels=32, + num_filters=64, + filter_size=3, + stride=1, + act='relu', + name="conv1_3") + + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.block_list = [] + if layers >= 50: + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + if layers in [101, 152] and block == 2: + if i == 0: + conv_name = "res" + str(block + 2) + "a" + else: + conv_name = "res" + str(block + 2) + "b" + str(i) + else: + conv_name = "res" + str(block + 2) + chr(97 + i) + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels=num_channels[block] + if i == 0 else num_filters[block] * 4, + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + shortcut=shortcut, + name=conv_name)) + self.block_list.append(bottleneck_block) + shortcut = True + else: + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + conv_name = "res" + str(block + 2) + chr(97 + i) + basic_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BasicBlock( + num_channels=num_channels[block] + if i == 0 else num_filters[block], + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + shortcut=shortcut, + name=conv_name)) + self.block_list.append(basic_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_0.w_0"), + bias_attr=ParamAttr(name="fc_0.b_0")) + + def forward(self, inputs): + y = self.conv1_1(inputs) + y = self.conv1_2(y) + y = self.conv1_3(y) + y = self.pool2d_max(y) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ResNet50_vc(pretrained=False, use_ssld=False, **kwargs): + model = ResNet_vc(layers=50, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNet50_vc"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext.py new file mode 100644 index 00000000..1aef8114 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext.py @@ -0,0 +1,298 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ResNeXt50_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_32x4d_pretrained.pdparams", + "ResNeXt50_64x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_64x4d_pretrained.pdparams", + "ResNeXt101_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x4d_pretrained.pdparams", + "ResNeXt101_64x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_64x4d_pretrained.pdparams", + "ResNeXt152_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_32x4d_pretrained.pdparams", + "ResNeXt152_64x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_64x4d_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + act=None, + name=None, + data_format="NCHW"): + super(ConvBNLayer, self).__init__() + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False, + data_format=data_format) + if name == "conv1": + bn_name = "bn_" + name + else: + bn_name = "bn" + name[3:] + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance', + data_layout=data_format) + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + cardinality, + shortcut=True, + name=None, + data_format="NCHW"): + super(BottleneckBlock, self).__init__() + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act='relu', + name=name + "_branch2a", + data_format=data_format) + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + groups=cardinality, + stride=stride, + act='relu', + name=name + "_branch2b", + data_format=data_format) + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 2 if cardinality == 32 else num_filters, + filter_size=1, + act=None, + name=name + "_branch2c", + data_format=data_format) + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 2 + if cardinality == 32 else num_filters, + filter_size=1, + stride=stride, + name=name + "_branch1", + data_format=data_format) + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + conv2 = self.conv2(conv1) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + + y = paddle.add(x=short, y=conv2) + y = F.relu(y) + return y + + +class ResNeXt(nn.Layer): + def __init__(self, + layers=50, + class_num=1000, + cardinality=32, + input_image_channel=3, + data_format="NCHW"): + super(ResNeXt, self).__init__() + + self.layers = layers + self.data_format = data_format + self.input_image_channel = input_image_channel + self.cardinality = cardinality + supported_layers = [50, 101, 152] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + supported_cardinality = [32, 64] + assert cardinality in supported_cardinality, \ + "supported cardinality is {} but input cardinality is {}" \ + .format(supported_cardinality, cardinality) + if layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + num_channels = [64, 256, 512, 1024] + num_filters = [128, 256, 512, + 1024] if cardinality == 32 else [256, 512, 1024, 2048] + + self.conv = ConvBNLayer( + num_channels=self.input_image_channel, + num_filters=64, + filter_size=7, + stride=2, + act='relu', + name="res_conv1", + data_format=self.data_format) + self.pool2d_max = MaxPool2D( + kernel_size=3, stride=2, padding=1, data_format=self.data_format) + + self.block_list = [] + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + if layers in [101, 152] and block == 2: + if i == 0: + conv_name = "res" + str(block + 2) + "a" + else: + conv_name = "res" + str(block + 2) + "b" + str(i) + else: + conv_name = "res" + str(block + 2) + chr(97 + i) + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels=num_channels[block] if i == 0 else + num_filters[block] * int(64 // self.cardinality), + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + cardinality=self.cardinality, + shortcut=shortcut, + name=conv_name, + data_format=self.data_format)) + self.block_list.append(bottleneck_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1, data_format=self.data_format) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_weights"), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, inputs): + with paddle.static.amp.fp16_guard(): + if self.data_format == "NHWC": + inputs = paddle.tensor.transpose(inputs, [0, 2, 3, 1]) + inputs.stop_gradient = True + y = self.conv(inputs) + y = self.pool2d_max(y) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ResNeXt50_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=50, cardinality=32, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt50_32x4d"], use_ssld=use_ssld) + return model + + +def ResNeXt50_64x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=50, cardinality=64, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt50_64x4d"], use_ssld=use_ssld) + return model + + +def ResNeXt101_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=101, cardinality=32, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt101_32x4d"], use_ssld=use_ssld) + return model + + +def ResNeXt101_64x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=101, cardinality=64, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt101_64x4d"], use_ssld=use_ssld) + return model + + +def ResNeXt152_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=152, cardinality=32, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt152_32x4d"], use_ssld=use_ssld) + return model + + +def ResNeXt152_64x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=152, cardinality=64, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt152_64x4d"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext101_wsl.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext101_wsl.py new file mode 100644 index 00000000..e85e1338 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext101_wsl.py @@ -0,0 +1,490 @@ +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ResNeXt101_32x8d_wsl": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x8d_wsl_pretrained.pdparams", + "ResNeXt101_32x16d_wsl": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x16_wsl_pretrained.pdparams", + "ResNeXt101_32x32d_wsl": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x32d_wsl_pretrained.pdparams", + "ResNeXt101_32x48d_wsl": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x48d_wsl_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + input_channels, + output_channels, + filter_size, + stride=1, + groups=1, + act=None, + name=None): + super(ConvBNLayer, self).__init__() + if "downsample" in name: + conv_name = name + ".0" + else: + conv_name = name + self._conv = Conv2D( + in_channels=input_channels, + out_channels=output_channels, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=conv_name + ".weight"), + bias_attr=False) + if "downsample" in name: + bn_name = name[:9] + "downsample.1" + else: + if "conv1" == name: + bn_name = "bn" + name[-1] + else: + bn_name = (name[:10] if name[7:9].isdigit() else name[:9] + ) + "bn" + name[-1] + self._bn = BatchNorm( + num_channels=output_channels, + act=act, + param_attr=ParamAttr(name=bn_name + ".weight"), + bias_attr=ParamAttr(name=bn_name + ".bias"), + moving_mean_name=bn_name + ".running_mean", + moving_variance_name=bn_name + ".running_var") + + def forward(self, inputs): + x = self._conv(inputs) + x = self._bn(x) + return x + + +class ShortCut(nn.Layer): + def __init__(self, input_channels, output_channels, stride, name=None): + super(ShortCut, self).__init__() + + self.input_channels = input_channels + self.output_channels = output_channels + self.stride = stride + if input_channels != output_channels or stride != 1: + self._conv = ConvBNLayer( + input_channels, + output_channels, + filter_size=1, + stride=stride, + name=name) + + def forward(self, inputs): + if self.input_channels != self.output_channels or self.stride != 1: + return self._conv(inputs) + return inputs + + +class BottleneckBlock(nn.Layer): + def __init__(self, input_channels, output_channels, stride, cardinality, + width, name): + super(BottleneckBlock, self).__init__() + + self._conv0 = ConvBNLayer( + input_channels, + output_channels, + filter_size=1, + act="relu", + name=name + ".conv1") + self._conv1 = ConvBNLayer( + output_channels, + output_channels, + filter_size=3, + act="relu", + stride=stride, + groups=cardinality, + name=name + ".conv2") + self._conv2 = ConvBNLayer( + output_channels, + output_channels // (width // 8), + filter_size=1, + act=None, + name=name + ".conv3") + self._short = ShortCut( + input_channels, + output_channels // (width // 8), + stride=stride, + name=name + ".downsample") + + def forward(self, inputs): + x = self._conv0(inputs) + x = self._conv1(x) + x = self._conv2(x) + y = self._short(inputs) + y = paddle.add(x, y) + y = F.relu(y) + return y + + +class ResNeXt101WSL(nn.Layer): + def __init__(self, layers=101, cardinality=32, width=48, class_num=1000): + super(ResNeXt101WSL, self).__init__() + + self.class_num = class_num + + self.layers = layers + self.cardinality = cardinality + self.width = width + self.scale = width // 8 + + self.depth = [3, 4, 23, 3] + self.base_width = cardinality * width + num_filters = [self.base_width * i + for i in [1, 2, 4, 8]] # [256, 512, 1024, 2048] + self._conv_stem = ConvBNLayer( + 3, 64, 7, stride=2, act="relu", name="conv1") + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self._conv1_0 = BottleneckBlock( + 64, + num_filters[0], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer1.0") + self._conv1_1 = BottleneckBlock( + num_filters[0] // (width // 8), + num_filters[0], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer1.1") + self._conv1_2 = BottleneckBlock( + num_filters[0] // (width // 8), + num_filters[0], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer1.2") + + self._conv2_0 = BottleneckBlock( + num_filters[0] // (width // 8), + num_filters[1], + stride=2, + cardinality=self.cardinality, + width=self.width, + name="layer2.0") + self._conv2_1 = BottleneckBlock( + num_filters[1] // (width // 8), + num_filters[1], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer2.1") + self._conv2_2 = BottleneckBlock( + num_filters[1] // (width // 8), + num_filters[1], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer2.2") + self._conv2_3 = BottleneckBlock( + num_filters[1] // (width // 8), + num_filters[1], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer2.3") + + self._conv3_0 = BottleneckBlock( + num_filters[1] // (width // 8), + num_filters[2], + stride=2, + cardinality=self.cardinality, + width=self.width, + name="layer3.0") + self._conv3_1 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.1") + self._conv3_2 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.2") + self._conv3_3 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.3") + self._conv3_4 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.4") + self._conv3_5 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.5") + self._conv3_6 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.6") + self._conv3_7 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.7") + self._conv3_8 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.8") + self._conv3_9 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.9") + self._conv3_10 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.10") + self._conv3_11 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.11") + self._conv3_12 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.12") + self._conv3_13 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.13") + self._conv3_14 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.14") + self._conv3_15 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.15") + self._conv3_16 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.16") + self._conv3_17 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.17") + self._conv3_18 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.18") + self._conv3_19 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.19") + self._conv3_20 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.20") + self._conv3_21 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.21") + self._conv3_22 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[2], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer3.22") + + self._conv4_0 = BottleneckBlock( + num_filters[2] // (width // 8), + num_filters[3], + stride=2, + cardinality=self.cardinality, + width=self.width, + name="layer4.0") + self._conv4_1 = BottleneckBlock( + num_filters[3] // (width // 8), + num_filters[3], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer4.1") + self._conv4_2 = BottleneckBlock( + num_filters[3] // (width // 8), + num_filters[3], + stride=1, + cardinality=self.cardinality, + width=self.width, + name="layer4.2") + + self._avg_pool = AdaptiveAvgPool2D(1) + self._out = Linear( + num_filters[3] // (width // 8), + class_num, + weight_attr=ParamAttr(name="fc.weight"), + bias_attr=ParamAttr(name="fc.bias")) + + def forward(self, inputs): + x = self._conv_stem(inputs) + x = self._pool(x) + + x = self._conv1_0(x) + x = self._conv1_1(x) + x = self._conv1_2(x) + + x = self._conv2_0(x) + x = self._conv2_1(x) + x = self._conv2_2(x) + x = self._conv2_3(x) + + x = self._conv3_0(x) + x = self._conv3_1(x) + x = self._conv3_2(x) + x = self._conv3_3(x) + x = self._conv3_4(x) + x = self._conv3_5(x) + x = self._conv3_6(x) + x = self._conv3_7(x) + x = self._conv3_8(x) + x = self._conv3_9(x) + x = self._conv3_10(x) + x = self._conv3_11(x) + x = self._conv3_12(x) + x = self._conv3_13(x) + x = self._conv3_14(x) + x = self._conv3_15(x) + x = self._conv3_16(x) + x = self._conv3_17(x) + x = self._conv3_18(x) + x = self._conv3_19(x) + x = self._conv3_20(x) + x = self._conv3_21(x) + x = self._conv3_22(x) + + x = self._conv4_0(x) + x = self._conv4_1(x) + x = self._conv4_2(x) + + x = self._avg_pool(x) + x = paddle.squeeze(x, axis=[2, 3]) + x = self._out(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ResNeXt101_32x8d_wsl(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt101WSL(cardinality=32, width=8, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt101_32x8d_wsl"], + use_ssld=use_ssld) + return model + + +def ResNeXt101_32x16d_wsl(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt101WSL(cardinality=32, width=16, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt101_32x16d_wsl"], + use_ssld=use_ssld) + return model + + +def ResNeXt101_32x32d_wsl(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt101WSL(cardinality=32, width=32, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt101_32x32d_wsl"], + use_ssld=use_ssld) + return model + + +def ResNeXt101_32x48d_wsl(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt101WSL(cardinality=32, width=48, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt101_32x48d_wsl"], + use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext_vd.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext_vd.py new file mode 100644 index 00000000..b2bd484f --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/resnext_vd.py @@ -0,0 +1,317 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ResNeXt50_vd_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_32x4d_pretrained.pdparams", + "ResNeXt50_vd_64x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_64x4d_pretrained.pdparams", + "ResNeXt101_vd_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_32x4d_pretrained.pdparams", + "ResNeXt101_vd_64x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_64x4d_pretrained.pdparams", + "ResNeXt152_vd_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_32x4d_pretrained.pdparams", + "ResNeXt152_vd_64x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_64x4d_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__( + self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + is_vd_mode=False, + act=None, + name=None, ): + super(ConvBNLayer, self).__init__() + + self.is_vd_mode = is_vd_mode + self._pool2d_avg = AvgPool2D( + kernel_size=2, stride=2, padding=0, ceil_mode=True) + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + if name == "conv1": + bn_name = "bn_" + name + else: + bn_name = "bn" + name[3:] + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + if self.is_vd_mode: + inputs = self._pool2d_avg(inputs) + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + cardinality, + shortcut=True, + if_first=False, + name=None): + super(BottleneckBlock, self).__init__() + + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act='relu', + name=name + "_branch2a") + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + groups=cardinality, + stride=stride, + act='relu', + name=name + "_branch2b") + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 2 if cardinality == 32 else num_filters, + filter_size=1, + act=None, + name=name + "_branch2c") + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 2 + if cardinality == 32 else num_filters, + filter_size=1, + stride=1, + is_vd_mode=False if if_first else True, + name=name + "_branch1") + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + conv2 = self.conv2(conv1) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + + y = paddle.add(x=short, y=conv2) + y = F.relu(y) + return y + + +class ResNeXt(nn.Layer): + def __init__(self, layers=50, class_num=1000, cardinality=32): + super(ResNeXt, self).__init__() + + self.layers = layers + self.cardinality = cardinality + supported_layers = [50, 101, 152] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + supported_cardinality = [32, 64] + assert cardinality in supported_cardinality, \ + "supported cardinality is {} but input cardinality is {}" \ + .format(supported_cardinality, cardinality) + if layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + num_channels = [64, 256, 512, 1024] + num_filters = [128, 256, 512, + 1024] if cardinality == 32 else [256, 512, 1024, 2048] + + self.conv1_1 = ConvBNLayer( + num_channels=3, + num_filters=32, + filter_size=3, + stride=2, + act='relu', + name="conv1_1") + self.conv1_2 = ConvBNLayer( + num_channels=32, + num_filters=32, + filter_size=3, + stride=1, + act='relu', + name="conv1_2") + self.conv1_3 = ConvBNLayer( + num_channels=32, + num_filters=64, + filter_size=3, + stride=1, + act='relu', + name="conv1_3") + + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.block_list = [] + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + if layers in [101, 152] and block == 2: + if i == 0: + conv_name = "res" + str(block + 2) + "a" + else: + conv_name = "res" + str(block + 2) + "b" + str(i) + else: + conv_name = "res" + str(block + 2) + chr(97 + i) + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels=num_channels[block] if i == 0 else + num_filters[block] * int(64 // self.cardinality), + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + cardinality=self.cardinality, + shortcut=shortcut, + if_first=block == i == 0, + name=conv_name)) + self.block_list.append(bottleneck_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc_weights"), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, inputs): + y = self.conv1_1(inputs) + y = self.conv1_2(y) + y = self.conv1_3(y) + y = self.pool2d_max(y) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ResNeXt50_vd_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=50, cardinality=32, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt50_vd_32x4d"], use_ssld=use_ssld) + return model + + +def ResNeXt50_vd_64x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=50, cardinality=64, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ResNeXt50_vd_64x4d"], use_ssld=use_ssld) + return model + + +def ResNeXt101_vd_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=101, cardinality=32, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt101_vd_32x4d"], + use_ssld=use_ssld) + return model + + +def ResNeXt101_vd_64x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=101, cardinality=64, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt101_vd_64x4d"], + use_ssld=use_ssld) + return model + + +def ResNeXt152_vd_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=152, cardinality=32, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt152_vd_32x4d"], + use_ssld=use_ssld) + return model + + +def ResNeXt152_vd_64x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=152, cardinality=64, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ResNeXt152_vd_64x4d"], + use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/rexnet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/rexnet.py new file mode 100644 index 00000000..1556a01c --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/rexnet.py @@ -0,0 +1,281 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +from math import ceil + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ReXNet_1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_0_pretrained.pdparams", + "ReXNet_1_3": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_3_pretrained.pdparams", + "ReXNet_1_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_5_pretrained.pdparams", + "ReXNet_2_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_2_0_pretrained.pdparams", + "ReXNet_3_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_3_0_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +def conv_bn_act(out, + in_channels, + channels, + kernel=1, + stride=1, + pad=0, + num_group=1, + active=True, + relu6=False): + out.append( + nn.Conv2D( + in_channels, + channels, + kernel, + stride, + pad, + groups=num_group, + bias_attr=False)) + out.append(nn.BatchNorm2D(channels)) + if active: + out.append(nn.ReLU6() if relu6 else nn.ReLU()) + + +def conv_bn_swish(out, + in_channels, + channels, + kernel=1, + stride=1, + pad=0, + num_group=1): + out.append( + nn.Conv2D( + in_channels, + channels, + kernel, + stride, + pad, + groups=num_group, + bias_attr=False)) + out.append(nn.BatchNorm2D(channels)) + out.append(nn.Swish()) + + +class SE(nn.Layer): + def __init__(self, in_channels, channels, se_ratio=12): + super(SE, self).__init__() + self.avg_pool = nn.AdaptiveAvgPool2D(1) + self.fc = nn.Sequential( + nn.Conv2D( + in_channels, channels // se_ratio, kernel_size=1, padding=0), + nn.BatchNorm2D(channels // se_ratio), + nn.ReLU(), + nn.Conv2D( + channels // se_ratio, channels, kernel_size=1, padding=0), + nn.Sigmoid()) + + def forward(self, x): + y = self.avg_pool(x) + y = self.fc(y) + return x * y + + +class LinearBottleneck(nn.Layer): + def __init__(self, + in_channels, + channels, + t, + stride, + use_se=True, + se_ratio=12, + **kwargs): + super(LinearBottleneck, self).__init__(**kwargs) + self.use_shortcut = stride == 1 and in_channels <= channels + self.in_channels = in_channels + self.out_channels = channels + + out = [] + if t != 1: + dw_channels = in_channels * t + conv_bn_swish(out, in_channels=in_channels, channels=dw_channels) + else: + dw_channels = in_channels + + conv_bn_act( + out, + in_channels=dw_channels, + channels=dw_channels, + kernel=3, + stride=stride, + pad=1, + num_group=dw_channels, + active=False) + + if use_se: + out.append(SE(dw_channels, dw_channels, se_ratio)) + + out.append(nn.ReLU6()) + conv_bn_act( + out, + in_channels=dw_channels, + channels=channels, + active=False, + relu6=True) + self.out = nn.Sequential(*out) + + def forward(self, x): + out = self.out(x) + if self.use_shortcut: + out[:, 0:self.in_channels] += x + + return out + + +class ReXNetV1(nn.Layer): + def __init__(self, + input_ch=16, + final_ch=180, + width_mult=1.0, + depth_mult=1.0, + class_num=1000, + use_se=True, + se_ratio=12, + dropout_ratio=0.2, + bn_momentum=0.9): + super(ReXNetV1, self).__init__() + + layers = [1, 2, 2, 3, 3, 5] + strides = [1, 2, 2, 2, 1, 2] + use_ses = [False, False, True, True, True, True] + + layers = [ceil(element * depth_mult) for element in layers] + strides = sum([[element] + [1] * (layers[idx] - 1) + for idx, element in enumerate(strides)], []) + if use_se: + use_ses = sum([[element] * layers[idx] + for idx, element in enumerate(use_ses)], []) + else: + use_ses = [False] * sum(layers[:]) + ts = [1] * layers[0] + [6] * sum(layers[1:]) + + self.depth = sum(layers[:]) * 3 + stem_channel = 32 / width_mult if width_mult < 1.0 else 32 + inplanes = input_ch / width_mult if width_mult < 1.0 else input_ch + + features = [] + in_channels_group = [] + channels_group = [] + + # The following channel configuration is a simple instance to make each layer become an expand layer. + for i in range(self.depth // 3): + if i == 0: + in_channels_group.append(int(round(stem_channel * width_mult))) + channels_group.append(int(round(inplanes * width_mult))) + else: + in_channels_group.append(int(round(inplanes * width_mult))) + inplanes += final_ch / (self.depth // 3 * 1.0) + channels_group.append(int(round(inplanes * width_mult))) + + conv_bn_swish( + features, + 3, + int(round(stem_channel * width_mult)), + kernel=3, + stride=2, + pad=1) + + for block_idx, (in_c, c, t, s, se) in enumerate( + zip(in_channels_group, channels_group, ts, strides, use_ses)): + features.append( + LinearBottleneck( + in_channels=in_c, + channels=c, + t=t, + stride=s, + use_se=se, + se_ratio=se_ratio)) + + pen_channels = int(1280 * width_mult) + conv_bn_swish(features, c, pen_channels) + + features.append(nn.AdaptiveAvgPool2D(1)) + self.features = nn.Sequential(*features) + self.output = nn.Sequential( + nn.Dropout(dropout_ratio), + nn.Conv2D( + pen_channels, class_num, 1, bias_attr=True)) + + def forward(self, x): + x = self.features(x) + x = self.output(x).squeeze(axis=-1).squeeze(axis=-1) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ReXNet_1_0(pretrained=False, use_ssld=False, **kwargs): + model = ReXNetV1(width_mult=1.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ReXNet_1_0"], use_ssld=use_ssld) + return model + + +def ReXNet_1_3(pretrained=False, use_ssld=False, **kwargs): + model = ReXNetV1(width_mult=1.3, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ReXNet_1_3"], use_ssld=use_ssld) + return model + + +def ReXNet_1_5(pretrained=False, use_ssld=False, **kwargs): + model = ReXNetV1(width_mult=1.5, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ReXNet_1_5"], use_ssld=use_ssld) + return model + + +def ReXNet_2_0(pretrained=False, use_ssld=False, **kwargs): + model = ReXNetV1(width_mult=2.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ReXNet_2_0"], use_ssld=use_ssld) + return model + + +def ReXNet_3_0(pretrained=False, use_ssld=False, **kwargs): + model = ReXNetV1(width_mult=3.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ReXNet_3_0"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnet_vd.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnet_vd.py new file mode 100644 index 00000000..205feec5 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnet_vd.py @@ -0,0 +1,390 @@ +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "SE_ResNet18_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet18_vd_pretrained.pdparams", + "SE_ResNet34_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet34_vd_pretrained.pdparams", + "SE_ResNet50_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet50_vd_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__( + self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + is_vd_mode=False, + act=None, + name=None, ): + super(ConvBNLayer, self).__init__() + + self.is_vd_mode = is_vd_mode + self._pool2d_avg = AvgPool2D( + kernel_size=2, stride=2, padding=0, ceil_mode=True) + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + if name == "conv1": + bn_name = "bn_" + name + else: + bn_name = "bn" + name[3:] + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + if self.is_vd_mode: + inputs = self._pool2d_avg(inputs) + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + if_first=False, + reduction_ratio=16, + name=None): + super(BottleneckBlock, self).__init__() + + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act='relu', + name=name + "_branch2a") + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + stride=stride, + act='relu', + name=name + "_branch2b") + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 4, + filter_size=1, + act=None, + name=name + "_branch2c") + self.scale = SELayer( + num_channels=num_filters * 4, + num_filters=num_filters * 4, + reduction_ratio=reduction_ratio, + name='fc_' + name) + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 4, + filter_size=1, + stride=1, + is_vd_mode=False if if_first else True, + name=name + "_branch1") + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + conv2 = self.conv2(conv1) + scale = self.scale(conv2) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + y = paddle.add(x=short, y=scale) + y = F.relu(y) + return y + + +class BasicBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + if_first=False, + reduction_ratio=16, + name=None): + super(BasicBlock, self).__init__() + self.stride = stride + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=3, + stride=stride, + act='relu', + name=name + "_branch2a") + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + act=None, + name=name + "_branch2b") + + self.scale = SELayer( + num_channels=num_filters, + num_filters=num_filters, + reduction_ratio=reduction_ratio, + name='fc_' + name) + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + stride=1, + is_vd_mode=False if if_first else True, + name=name + "_branch1") + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + scale = self.scale(conv1) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + y = paddle.add(x=short, y=scale) + y = F.relu(y) + return y + + +class SELayer(nn.Layer): + def __init__(self, num_channels, num_filters, reduction_ratio, name=None): + super(SELayer, self).__init__() + + self.pool2d_gap = AdaptiveAvgPool2D(1) + + self._num_channels = num_channels + + med_ch = int(num_channels / reduction_ratio) + stdv = 1.0 / math.sqrt(num_channels * 1.0) + self.squeeze = Linear( + num_channels, + med_ch, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_sqz_weights"), + bias_attr=ParamAttr(name=name + '_sqz_offset')) + + stdv = 1.0 / math.sqrt(med_ch * 1.0) + self.excitation = Linear( + med_ch, + num_filters, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_exc_weights"), + bias_attr=ParamAttr(name=name + '_exc_offset')) + + def forward(self, input): + pool = self.pool2d_gap(input) + pool = paddle.squeeze(pool, axis=[2, 3]) + squeeze = self.squeeze(pool) + squeeze = F.relu(squeeze) + excitation = self.excitation(squeeze) + excitation = F.sigmoid(excitation) + excitation = paddle.unsqueeze(excitation, axis=[2, 3]) + out = input * excitation + return out + + +class SE_ResNet_vd(nn.Layer): + def __init__(self, layers=50, class_num=1000): + super(SE_ResNet_vd, self).__init__() + + self.layers = layers + supported_layers = [18, 34, 50, 101, 152, 200] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + + if layers == 18: + depth = [2, 2, 2, 2] + elif layers == 34 or layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + elif layers == 200: + depth = [3, 12, 48, 3] + num_channels = [64, 256, 512, + 1024] if layers >= 50 else [64, 64, 128, 256] + num_filters = [64, 128, 256, 512] + + self.conv1_1 = ConvBNLayer( + num_channels=3, + num_filters=32, + filter_size=3, + stride=2, + act='relu', + name="conv1_1") + self.conv1_2 = ConvBNLayer( + num_channels=32, + num_filters=32, + filter_size=3, + stride=1, + act='relu', + name="conv1_2") + self.conv1_3 = ConvBNLayer( + num_channels=32, + num_filters=64, + filter_size=3, + stride=1, + act='relu', + name="conv1_3") + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.block_list = [] + if layers >= 50: + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + if layers in [101, 152] and block == 2: + if i == 0: + conv_name = "res" + str(block + 2) + "a" + else: + conv_name = "res" + str(block + 2) + "b" + str(i) + else: + conv_name = "res" + str(block + 2) + chr(97 + i) + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels=num_channels[block] + if i == 0 else num_filters[block] * 4, + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + shortcut=shortcut, + if_first=block == i == 0, + name=conv_name)) + self.block_list.append(bottleneck_block) + shortcut = True + else: + for block in range(len(depth)): + shortcut = False + for i in range(depth[block]): + conv_name = "res" + str(block + 2) + chr(97 + i) + basic_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BasicBlock( + num_channels=num_channels[block] + if i == 0 else num_filters[block], + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + shortcut=shortcut, + if_first=block == i == 0, + name=conv_name)) + self.block_list.append(basic_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc6_weights"), + bias_attr=ParamAttr(name="fc6_offset")) + + def forward(self, inputs): + y = self.conv1_1(inputs) + y = self.conv1_2(y) + y = self.conv1_3(y) + y = self.pool2d_max(y) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def SE_ResNet18_vd(pretrained=False, use_ssld=False, **kwargs): + model = SE_ResNet_vd(layers=18, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["SE_ResNet18_vd"], use_ssld=use_ssld) + return model + + +def SE_ResNet34_vd(pretrained=False, use_ssld=False, **kwargs): + model = SE_ResNet_vd(layers=34, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["SE_ResNet34_vd"], use_ssld=use_ssld) + return model + + +def SE_ResNet50_vd(pretrained=False, use_ssld=False, **kwargs): + model = SE_ResNet_vd(layers=50, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["SE_ResNet50_vd"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnext.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnext.py new file mode 100644 index 00000000..8b7149e2 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnext.py @@ -0,0 +1,364 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "SE_ResNeXt50_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_32x4d_pretrained.pdparams", + "SE_ResNeXt101_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt101_32x4d_pretrained.pdparams", + "SE_ResNeXt152_64x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt152_64x4d_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + act=None, + name=None, + data_format='NCHW'): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False, + data_format=data_format) + bn_name = name + '_bn' + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance', + data_layout=data_format) + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + cardinality, + reduction_ratio, + shortcut=True, + if_first=False, + name=None, + data_format="NCHW"): + super(BottleneckBlock, self).__init__() + + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act='relu', + name='conv' + name + '_x1', + data_format=data_format) + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + groups=cardinality, + stride=stride, + act='relu', + name='conv' + name + '_x2', + data_format=data_format) + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 2 if cardinality == 32 else num_filters, + filter_size=1, + act=None, + name='conv' + name + '_x3', + data_format=data_format) + self.scale = SELayer( + num_channels=num_filters * 2 if cardinality == 32 else num_filters, + num_filters=num_filters * 2 if cardinality == 32 else num_filters, + reduction_ratio=reduction_ratio, + name='fc' + name, + data_format=data_format) + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 2 + if cardinality == 32 else num_filters, + filter_size=1, + stride=stride, + name='conv' + name + '_prj', + data_format=data_format) + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + conv2 = self.conv2(conv1) + scale = self.scale(conv2) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + y = paddle.add(x=short, y=scale) + y = F.relu(y) + return y + + +class SELayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + reduction_ratio, + name=None, + data_format="NCHW"): + super(SELayer, self).__init__() + + self.data_format = data_format + self.pool2d_gap = AdaptiveAvgPool2D(1, data_format=self.data_format) + + self._num_channels = num_channels + + med_ch = int(num_channels / reduction_ratio) + stdv = 1.0 / math.sqrt(num_channels * 1.0) + self.squeeze = Linear( + num_channels, + med_ch, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_sqz_weights"), + bias_attr=ParamAttr(name=name + '_sqz_offset')) + self.relu = nn.ReLU() + stdv = 1.0 / math.sqrt(med_ch * 1.0) + self.excitation = Linear( + med_ch, + num_filters, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_exc_weights"), + bias_attr=ParamAttr(name=name + '_exc_offset')) + self.sigmoid = nn.Sigmoid() + + def forward(self, input): + pool = self.pool2d_gap(input) + if self.data_format == "NHWC": + pool = paddle.squeeze(pool, axis=[1, 2]) + else: + pool = paddle.squeeze(pool, axis=[2, 3]) + squeeze = self.squeeze(pool) + squeeze = self.relu(squeeze) + excitation = self.excitation(squeeze) + excitation = self.sigmoid(excitation) + if self.data_format == "NHWC": + excitation = paddle.unsqueeze(excitation, axis=[1, 2]) + else: + excitation = paddle.unsqueeze(excitation, axis=[2, 3]) + out = input * excitation + return out + + +class ResNeXt(nn.Layer): + def __init__(self, + layers=50, + class_num=1000, + cardinality=32, + input_image_channel=3, + data_format="NCHW"): + super(ResNeXt, self).__init__() + + self.layers = layers + self.cardinality = cardinality + self.reduction_ratio = 16 + self.data_format = data_format + self.input_image_channel = input_image_channel + + supported_layers = [50, 101, 152] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + supported_cardinality = [32, 64] + assert cardinality in supported_cardinality, \ + "supported cardinality is {} but input cardinality is {}" \ + .format(supported_cardinality, cardinality) + if layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + num_channels = [64, 256, 512, 1024] + num_filters = [128, 256, 512, + 1024] if cardinality == 32 else [256, 512, 1024, 2048] + if layers < 152: + self.conv = ConvBNLayer( + num_channels=self.input_image_channel, + num_filters=64, + filter_size=7, + stride=2, + act='relu', + name="conv1", + data_format=self.data_format) + else: + self.conv1_1 = ConvBNLayer( + num_channels=self.input_image_channel, + num_filters=64, + filter_size=3, + stride=2, + act='relu', + name="conv1", + data_format=self.data_format) + self.conv1_2 = ConvBNLayer( + num_channels=64, + num_filters=64, + filter_size=3, + stride=1, + act='relu', + name="conv2", + data_format=self.data_format) + self.conv1_3 = ConvBNLayer( + num_channels=64, + num_filters=128, + filter_size=3, + stride=1, + act='relu', + name="conv3", + data_format=self.data_format) + + self.pool2d_max = MaxPool2D( + kernel_size=3, stride=2, padding=1, data_format=self.data_format) + + self.block_list = [] + n = 1 if layers == 50 or layers == 101 else 3 + for block in range(len(depth)): + n += 1 + shortcut = False + for i in range(depth[block]): + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels=num_channels[block] if i == 0 else + num_filters[block] * int(64 // self.cardinality), + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + cardinality=self.cardinality, + reduction_ratio=self.reduction_ratio, + shortcut=shortcut, + if_first=block == 0, + name=str(n) + '_' + str(i + 1), + data_format=self.data_format)) + self.block_list.append(bottleneck_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1, data_format=self.data_format) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc6_weights"), + bias_attr=ParamAttr(name="fc6_offset")) + + def forward(self, inputs): + with paddle.static.amp.fp16_guard(): + if self.data_format == "NHWC": + inputs = paddle.tensor.transpose(inputs, [0, 2, 3, 1]) + inputs.stop_gradient = True + if self.layers < 152: + y = self.conv(inputs) + else: + y = self.conv1_1(inputs) + y = self.conv1_2(y) + y = self.conv1_3(y) + y = self.pool2d_max(y) + for i, block in enumerate(self.block_list): + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def SE_ResNeXt50_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=50, cardinality=32, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["SE_ResNeXt50_32x4d"], use_ssld=use_ssld) + return model + + +def SE_ResNeXt101_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=101, cardinality=32, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SE_ResNeXt101_32x4d"], + use_ssld=use_ssld) + return model + + +def SE_ResNeXt152_64x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=152, cardinality=64, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SE_ResNeXt152_64x4d"], + use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnext_vd.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnext_vd.py new file mode 100644 index 00000000..ef630259 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/se_resnext_vd.py @@ -0,0 +1,309 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform + +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "SE_ResNeXt50_vd_32x4d": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_vd_32x4d_pretrained.pdparams", + "SENet154_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SENet154_vd_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + is_vd_mode=False, + act=None, + name=None): + super(ConvBNLayer, self).__init__() + + self.is_vd_mode = is_vd_mode + self._pool2d_avg = AvgPool2D( + kernel_size=2, stride=2, padding=0, ceil_mode=True) + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + bn_name = name + '_bn' + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + '_scale'), + bias_attr=ParamAttr(bn_name + '_offset'), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + if self.is_vd_mode: + inputs = self._pool2d_avg(inputs) + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class BottleneckBlock(nn.Layer): + def __init__(self, + num_channels, + num_filters, + stride, + cardinality, + reduction_ratio, + shortcut=True, + if_first=False, + name=None): + super(BottleneckBlock, self).__init__() + + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act='relu', + name='conv' + name + '_x1') + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + groups=cardinality, + stride=stride, + act='relu', + name='conv' + name + '_x2') + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 2 if cardinality == 32 else num_filters, + filter_size=1, + act=None, + name='conv' + name + '_x3') + self.scale = SELayer( + num_channels=num_filters * 2 if cardinality == 32 else num_filters, + num_filters=num_filters * 2 if cardinality == 32 else num_filters, + reduction_ratio=reduction_ratio, + name='fc' + name) + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 2 + if cardinality == 32 else num_filters, + filter_size=1, + stride=1, + is_vd_mode=False if if_first else True, + name='conv' + name + '_prj') + + self.shortcut = shortcut + + def forward(self, inputs): + y = self.conv0(inputs) + conv1 = self.conv1(y) + conv2 = self.conv2(conv1) + scale = self.scale(conv2) + + if self.shortcut: + short = inputs + else: + short = self.short(inputs) + y = paddle.add(x=short, y=scale) + y = F.relu(y) + return y + + +class SELayer(nn.Layer): + def __init__(self, num_channels, num_filters, reduction_ratio, name=None): + super(SELayer, self).__init__() + + self.pool2d_gap = AdaptiveAvgPool2D(1) + + self._num_channels = num_channels + + med_ch = int(num_channels / reduction_ratio) + stdv = 1.0 / math.sqrt(num_channels * 1.0) + self.squeeze = Linear( + num_channels, + med_ch, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_sqz_weights"), + bias_attr=ParamAttr(name=name + '_sqz_offset')) + self.relu = nn.ReLU() + stdv = 1.0 / math.sqrt(med_ch * 1.0) + self.excitation = Linear( + med_ch, + num_filters, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name=name + "_exc_weights"), + bias_attr=ParamAttr(name=name + '_exc_offset')) + self.sigmoid = nn.Sigmoid() + + def forward(self, input): + pool = self.pool2d_gap(input) + pool = paddle.squeeze(pool, axis=[2, 3]) + squeeze = self.squeeze(pool) + squeeze = self.relu(squeeze) + excitation = self.excitation(squeeze) + excitation = self.sigmoid(excitation) + excitation = paddle.unsqueeze(excitation, axis=[2, 3]) + out = paddle.multiply(input, excitation) + return out + + +class ResNeXt(nn.Layer): + def __init__(self, layers=50, class_num=1000, cardinality=32): + super(ResNeXt, self).__init__() + + self.layers = layers + self.cardinality = cardinality + self.reduction_ratio = 16 + supported_layers = [50, 101, 152] + assert layers in supported_layers, \ + "supported layers are {} but input layer is {}".format( + supported_layers, layers) + supported_cardinality = [32, 64] + assert cardinality in supported_cardinality, \ + "supported cardinality is {} but input cardinality is {}" \ + .format(supported_cardinality, cardinality) + if layers == 50: + depth = [3, 4, 6, 3] + elif layers == 101: + depth = [3, 4, 23, 3] + elif layers == 152: + depth = [3, 8, 36, 3] + num_channels = [128, 256, 512, 1024] + num_filters = [128, 256, 512, + 1024] if cardinality == 32 else [256, 512, 1024, 2048] + + self.conv1_1 = ConvBNLayer( + num_channels=3, + num_filters=64, + filter_size=3, + stride=2, + act='relu', + name="conv1_1") + self.conv1_2 = ConvBNLayer( + num_channels=64, + num_filters=64, + filter_size=3, + stride=1, + act='relu', + name="conv1_2") + self.conv1_3 = ConvBNLayer( + num_channels=64, + num_filters=128, + filter_size=3, + stride=1, + act='relu', + name="conv1_3") + + self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1) + + self.block_list = [] + n = 1 if layers == 50 or layers == 101 else 3 + for block in range(len(depth)): + n += 1 + shortcut = False + for i in range(depth[block]): + bottleneck_block = self.add_sublayer( + 'bb_%d_%d' % (block, i), + BottleneckBlock( + num_channels=num_channels[block] if i == 0 else + num_filters[block] * int(64 // self.cardinality), + num_filters=num_filters[block], + stride=2 if i == 0 and block != 0 else 1, + cardinality=self.cardinality, + reduction_ratio=self.reduction_ratio, + shortcut=shortcut, + if_first=block == 0, + name=str(n) + '_' + str(i + 1))) + self.block_list.append(bottleneck_block) + shortcut = True + + self.pool2d_avg = AdaptiveAvgPool2D(1) + + self.pool2d_avg_channels = num_channels[-1] * 2 + + stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0) + + self.out = Linear( + self.pool2d_avg_channels, + class_num, + weight_attr=ParamAttr( + initializer=Uniform(-stdv, stdv), name="fc6_weights"), + bias_attr=ParamAttr(name="fc6_offset")) + + def forward(self, inputs): + y = self.conv1_1(inputs) + y = self.conv1_2(y) + y = self.conv1_3(y) + y = self.pool2d_max(y) + for block in self.block_list: + y = block(y) + y = self.pool2d_avg(y) + y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels]) + y = self.out(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def SE_ResNeXt50_vd_32x4d(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=50, cardinality=32, **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SE_ResNeXt50_vd_32x4d"], + use_ssld=use_ssld) + return model + + +def SENet154_vd(pretrained=False, use_ssld=False, **kwargs): + model = ResNeXt(layers=152, cardinality=64, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["SENet154_vd"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/shufflenet_v2.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/shufflenet_v2.py new file mode 100644 index 00000000..d8bb69ff --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/shufflenet_v2.py @@ -0,0 +1,362 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import paddle +from paddle import ParamAttr, reshape, transpose, concat, split +from paddle.nn import Layer, Conv2D, MaxPool2D, AdaptiveAvgPool2D, BatchNorm, Linear +from paddle.nn.initializer import KaimingNormal +from paddle.nn.functional import swish + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ShuffleNetV2_x0_25": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams", + "ShuffleNetV2_x0_33": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams", + "ShuffleNetV2_x0_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams", + "ShuffleNetV2_x1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams", + "ShuffleNetV2_x1_5": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams", + "ShuffleNetV2_x2_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams", + "ShuffleNetV2_swish": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +def channel_shuffle(x, groups): + batch_size, num_channels, height, width = x.shape[0:4] + channels_per_group = num_channels // groups + + # reshape + x = reshape( + x=x, shape=[batch_size, groups, channels_per_group, height, width]) + + # transpose + x = transpose(x=x, perm=[0, 2, 1, 3, 4]) + + # flatten + x = reshape(x=x, shape=[batch_size, num_channels, height, width]) + return x + + +class ConvBNLayer(Layer): + def __init__( + self, + in_channels, + out_channels, + kernel_size, + stride, + padding, + groups=1, + act=None, + name=None, ): + super(ConvBNLayer, self).__init__() + self._conv = Conv2D( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + groups=groups, + weight_attr=ParamAttr( + initializer=KaimingNormal(), name=name + "_weights"), + bias_attr=False) + + self._batch_norm = BatchNorm( + out_channels, + param_attr=ParamAttr(name=name + "_bn_scale"), + bias_attr=ParamAttr(name=name + "_bn_offset"), + act=act, + moving_mean_name=name + "_bn_mean", + moving_variance_name=name + "_bn_variance") + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class InvertedResidual(Layer): + def __init__(self, + in_channels, + out_channels, + stride, + act="relu", + name=None): + super(InvertedResidual, self).__init__() + self._conv_pw = ConvBNLayer( + in_channels=in_channels // 2, + out_channels=out_channels // 2, + kernel_size=1, + stride=1, + padding=0, + groups=1, + act=act, + name='stage_' + name + '_conv1') + self._conv_dw = ConvBNLayer( + in_channels=out_channels // 2, + out_channels=out_channels // 2, + kernel_size=3, + stride=stride, + padding=1, + groups=out_channels // 2, + act=None, + name='stage_' + name + '_conv2') + self._conv_linear = ConvBNLayer( + in_channels=out_channels // 2, + out_channels=out_channels // 2, + kernel_size=1, + stride=1, + padding=0, + groups=1, + act=act, + name='stage_' + name + '_conv3') + + def forward(self, inputs): + x1, x2 = split( + inputs, + num_or_sections=[inputs.shape[1] // 2, inputs.shape[1] // 2], + axis=1) + x2 = self._conv_pw(x2) + x2 = self._conv_dw(x2) + x2 = self._conv_linear(x2) + out = concat([x1, x2], axis=1) + return channel_shuffle(out, 2) + + +class InvertedResidualDS(Layer): + def __init__(self, + in_channels, + out_channels, + stride, + act="relu", + name=None): + super(InvertedResidualDS, self).__init__() + + # branch1 + self._conv_dw_1 = ConvBNLayer( + in_channels=in_channels, + out_channels=in_channels, + kernel_size=3, + stride=stride, + padding=1, + groups=in_channels, + act=None, + name='stage_' + name + '_conv4') + self._conv_linear_1 = ConvBNLayer( + in_channels=in_channels, + out_channels=out_channels // 2, + kernel_size=1, + stride=1, + padding=0, + groups=1, + act=act, + name='stage_' + name + '_conv5') + # branch2 + self._conv_pw_2 = ConvBNLayer( + in_channels=in_channels, + out_channels=out_channels // 2, + kernel_size=1, + stride=1, + padding=0, + groups=1, + act=act, + name='stage_' + name + '_conv1') + self._conv_dw_2 = ConvBNLayer( + in_channels=out_channels // 2, + out_channels=out_channels // 2, + kernel_size=3, + stride=stride, + padding=1, + groups=out_channels // 2, + act=None, + name='stage_' + name + '_conv2') + self._conv_linear_2 = ConvBNLayer( + in_channels=out_channels // 2, + out_channels=out_channels // 2, + kernel_size=1, + stride=1, + padding=0, + groups=1, + act=act, + name='stage_' + name + '_conv3') + + def forward(self, inputs): + x1 = self._conv_dw_1(inputs) + x1 = self._conv_linear_1(x1) + x2 = self._conv_pw_2(inputs) + x2 = self._conv_dw_2(x2) + x2 = self._conv_linear_2(x2) + out = concat([x1, x2], axis=1) + + return channel_shuffle(out, 2) + + +class ShuffleNet(Layer): + def __init__(self, class_num=1000, scale=1.0, act="relu"): + super(ShuffleNet, self).__init__() + self.scale = scale + self.class_num = class_num + stage_repeats = [4, 8, 4] + + if scale == 0.25: + stage_out_channels = [-1, 24, 24, 48, 96, 512] + elif scale == 0.33: + stage_out_channels = [-1, 24, 32, 64, 128, 512] + elif scale == 0.5: + stage_out_channels = [-1, 24, 48, 96, 192, 1024] + elif scale == 1.0: + stage_out_channels = [-1, 24, 116, 232, 464, 1024] + elif scale == 1.5: + stage_out_channels = [-1, 24, 176, 352, 704, 1024] + elif scale == 2.0: + stage_out_channels = [-1, 24, 224, 488, 976, 2048] + else: + raise NotImplementedError("This scale size:[" + str(scale) + + "] is not implemented!") + # 1. conv1 + self._conv1 = ConvBNLayer( + in_channels=3, + out_channels=stage_out_channels[1], + kernel_size=3, + stride=2, + padding=1, + act=act, + name='stage1_conv') + self._max_pool = MaxPool2D(kernel_size=3, stride=2, padding=1) + + # 2. bottleneck sequences + self._block_list = [] + for stage_id, num_repeat in enumerate(stage_repeats): + for i in range(num_repeat): + if i == 0: + block = self.add_sublayer( + name=str(stage_id + 2) + '_' + str(i + 1), + sublayer=InvertedResidualDS( + in_channels=stage_out_channels[stage_id + 1], + out_channels=stage_out_channels[stage_id + 2], + stride=2, + act=act, + name=str(stage_id + 2) + '_' + str(i + 1))) + else: + block = self.add_sublayer( + name=str(stage_id + 2) + '_' + str(i + 1), + sublayer=InvertedResidual( + in_channels=stage_out_channels[stage_id + 2], + out_channels=stage_out_channels[stage_id + 2], + stride=1, + act=act, + name=str(stage_id + 2) + '_' + str(i + 1))) + self._block_list.append(block) + # 3. last_conv + self._last_conv = ConvBNLayer( + in_channels=stage_out_channels[-2], + out_channels=stage_out_channels[-1], + kernel_size=1, + stride=1, + padding=0, + act=act, + name='conv5') + # 4. pool + self._pool2d_avg = AdaptiveAvgPool2D(1) + self._out_c = stage_out_channels[-1] + # 5. fc + self._fc = Linear( + stage_out_channels[-1], + class_num, + weight_attr=ParamAttr(name='fc6_weights'), + bias_attr=ParamAttr(name='fc6_offset')) + + def forward(self, inputs): + y = self._conv1(inputs) + y = self._max_pool(y) + for inv in self._block_list: + y = inv(y) + y = self._last_conv(y) + y = self._pool2d_avg(y) + y = paddle.flatten(y, start_axis=1, stop_axis=-1) + y = self._fc(y) + return y + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ShuffleNetV2_x0_25(pretrained=False, use_ssld=False, **kwargs): + model = ShuffleNet(scale=0.25, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ShuffleNetV2_x0_25"], use_ssld=use_ssld) + return model + + +def ShuffleNetV2_x0_33(pretrained=False, use_ssld=False, **kwargs): + model = ShuffleNet(scale=0.33, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ShuffleNetV2_x0_33"], use_ssld=use_ssld) + return model + + +def ShuffleNetV2_x0_5(pretrained=False, use_ssld=False, **kwargs): + model = ShuffleNet(scale=0.5, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ShuffleNetV2_x0_5"], use_ssld=use_ssld) + return model + + +def ShuffleNetV2_x1_0(pretrained=False, use_ssld=False, **kwargs): + model = ShuffleNet(scale=1.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ShuffleNetV2_x1_0"], use_ssld=use_ssld) + return model + + +def ShuffleNetV2_x1_5(pretrained=False, use_ssld=False, **kwargs): + model = ShuffleNet(scale=1.5, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ShuffleNetV2_x1_5"], use_ssld=use_ssld) + return model + + +def ShuffleNetV2_x2_0(pretrained=False, use_ssld=False, **kwargs): + model = ShuffleNet(scale=2.0, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ShuffleNetV2_x2_0"], use_ssld=use_ssld) + return model + + +def ShuffleNetV2_swish(pretrained=False, use_ssld=False, **kwargs): + model = ShuffleNet(scale=1.0, act="swish", **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["ShuffleNetV2_swish"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/squeezenet.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/squeezenet.py new file mode 100644 index 00000000..647cd2ea --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/squeezenet.py @@ -0,0 +1,194 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "SqueezeNet1_0": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_0_pretrained.pdparams", + "SqueezeNet1_1": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_1_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class MakeFireConv(nn.Layer): + def __init__(self, + input_channels, + output_channels, + filter_size, + padding=0, + name=None): + super(MakeFireConv, self).__init__() + self._conv = Conv2D( + input_channels, + output_channels, + filter_size, + padding=padding, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=ParamAttr(name=name + "_offset")) + + def forward(self, x): + x = self._conv(x) + x = F.relu(x) + return x + + +class MakeFire(nn.Layer): + def __init__(self, + input_channels, + squeeze_channels, + expand1x1_channels, + expand3x3_channels, + name=None): + super(MakeFire, self).__init__() + self._conv = MakeFireConv( + input_channels, squeeze_channels, 1, name=name + "_squeeze1x1") + self._conv_path1 = MakeFireConv( + squeeze_channels, expand1x1_channels, 1, name=name + "_expand1x1") + self._conv_path2 = MakeFireConv( + squeeze_channels, + expand3x3_channels, + 3, + padding=1, + name=name + "_expand3x3") + + def forward(self, inputs): + x = self._conv(inputs) + x1 = self._conv_path1(x) + x2 = self._conv_path2(x) + return paddle.concat([x1, x2], axis=1) + + +class SqueezeNet(nn.Layer): + def __init__(self, version, class_num=1000): + super(SqueezeNet, self).__init__() + self.version = version + + if self.version == "1.0": + self._conv = Conv2D( + 3, + 96, + 7, + stride=2, + weight_attr=ParamAttr(name="conv1_weights"), + bias_attr=ParamAttr(name="conv1_offset")) + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=0) + self._conv1 = MakeFire(96, 16, 64, 64, name="fire2") + self._conv2 = MakeFire(128, 16, 64, 64, name="fire3") + self._conv3 = MakeFire(128, 32, 128, 128, name="fire4") + + self._conv4 = MakeFire(256, 32, 128, 128, name="fire5") + self._conv5 = MakeFire(256, 48, 192, 192, name="fire6") + self._conv6 = MakeFire(384, 48, 192, 192, name="fire7") + self._conv7 = MakeFire(384, 64, 256, 256, name="fire8") + + self._conv8 = MakeFire(512, 64, 256, 256, name="fire9") + else: + self._conv = Conv2D( + 3, + 64, + 3, + stride=2, + padding=1, + weight_attr=ParamAttr(name="conv1_weights"), + bias_attr=ParamAttr(name="conv1_offset")) + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=0) + self._conv1 = MakeFire(64, 16, 64, 64, name="fire2") + self._conv2 = MakeFire(128, 16, 64, 64, name="fire3") + + self._conv3 = MakeFire(128, 32, 128, 128, name="fire4") + self._conv4 = MakeFire(256, 32, 128, 128, name="fire5") + + self._conv5 = MakeFire(256, 48, 192, 192, name="fire6") + self._conv6 = MakeFire(384, 48, 192, 192, name="fire7") + self._conv7 = MakeFire(384, 64, 256, 256, name="fire8") + self._conv8 = MakeFire(512, 64, 256, 256, name="fire9") + + self._drop = Dropout(p=0.5, mode="downscale_in_infer") + self._conv9 = Conv2D( + 512, + class_num, + 1, + weight_attr=ParamAttr(name="conv10_weights"), + bias_attr=ParamAttr(name="conv10_offset")) + self._avg_pool = AdaptiveAvgPool2D(1) + + def forward(self, inputs): + x = self._conv(inputs) + x = F.relu(x) + x = self._pool(x) + if self.version == "1.0": + x = self._conv1(x) + x = self._conv2(x) + x = self._conv3(x) + x = self._pool(x) + x = self._conv4(x) + x = self._conv5(x) + x = self._conv6(x) + x = self._conv7(x) + x = self._pool(x) + x = self._conv8(x) + else: + x = self._conv1(x) + x = self._conv2(x) + x = self._pool(x) + x = self._conv3(x) + x = self._conv4(x) + x = self._pool(x) + x = self._conv5(x) + x = self._conv6(x) + x = self._conv7(x) + x = self._conv8(x) + x = self._drop(x) + x = self._conv9(x) + x = F.relu(x) + x = self._avg_pool(x) + x = paddle.squeeze(x, axis=[2, 3]) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def SqueezeNet1_0(pretrained=False, use_ssld=False, **kwargs): + model = SqueezeNet(version="1.0", **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["SqueezeNet1_0"], use_ssld=use_ssld) + return model + + +def SqueezeNet1_1(pretrained=False, use_ssld=False, **kwargs): + model = SqueezeNet(version="1.1", **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["SqueezeNet1_1"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/swin_transformer.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/swin_transformer.py new file mode 100644 index 00000000..c783ec68 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/swin_transformer.py @@ -0,0 +1,857 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/microsoft/Swin-Transformer + +import numpy as np +import paddle +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn.initializer import TruncatedNormal, Constant + +from .vision_transformer import trunc_normal_, zeros_, ones_, to_2tuple, DropPath, Identity + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "SwinTransformer_tiny_patch4_window7_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_tiny_patch4_window7_224_pretrained.pdparams", + "SwinTransformer_small_patch4_window7_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_small_patch4_window7_224_pretrained.pdparams", + "SwinTransformer_base_patch4_window7_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_pretrained.pdparams", + "SwinTransformer_base_patch4_window12_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_pretrained.pdparams", + "SwinTransformer_large_patch4_window7_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams", + "SwinTransformer_large_patch4_window12_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class Mlp(nn.Layer): + def __init__(self, + in_features, + hidden_features=None, + out_features=None, + act_layer=nn.GELU, + drop=0.): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Linear(in_features, hidden_features) + self.act = act_layer() + self.fc2 = nn.Linear(hidden_features, out_features) + self.drop = nn.Dropout(drop) + + def forward(self, x): + x = self.fc1(x) + x = self.act(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +def window_partition(x, window_size): + """ + Args: + x: (B, H, W, C) + window_size (int): window size + + Returns: + windows: (num_windows*B, window_size, window_size, C) + """ + B, H, W, C = x.shape + x = x.reshape( + [B, H // window_size, window_size, W // window_size, window_size, C]) + windows = x.transpose([0, 1, 3, 2, 4, 5]).reshape( + [-1, window_size, window_size, C]) + return windows + + +def window_reverse(windows, window_size, H, W, C): + """ + Args: + windows: (num_windows*B, window_size, window_size, C) + window_size (int): Window size + H (int): Height of image + W (int): Width of image + + Returns: + x: (B, H, W, C) + """ + x = windows.reshape( + [-1, H // window_size, W // window_size, window_size, window_size, C]) + x = x.transpose([0, 1, 3, 2, 4, 5]).reshape([-1, H, W, C]) + return x + + +class WindowAttention(nn.Layer): + r""" Window based multi-head self attention (W-MSA) module with relative position bias. + It supports both of shifted and non-shifted window. + + Args: + dim (int): Number of input channels. + window_size (tuple[int]): The height and width of the window. + num_heads (int): Number of attention heads. + qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True + qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set + attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0 + proj_drop (float, optional): Dropout ratio of output. Default: 0.0 + """ + + def __init__(self, + dim, + window_size, + num_heads, + qkv_bias=True, + qk_scale=None, + attn_drop=0., + proj_drop=0.): + super().__init__() + self.dim = dim + self.window_size = window_size # Wh, Ww + self.num_heads = num_heads + head_dim = dim // num_heads + self.scale = qk_scale or head_dim**-0.5 + + # define a parameter table of relative position bias + # 2*Wh-1 * 2*Ww-1, nH + self.relative_position_bias_table = self.create_parameter( + shape=((2 * window_size[0] - 1) * (2 * window_size[1] - 1), + num_heads), + default_initializer=zeros_) + self.add_parameter("relative_position_bias_table", + self.relative_position_bias_table) + + # get pair-wise relative position index for each token inside the window + coords_h = paddle.arange(self.window_size[0]) + coords_w = paddle.arange(self.window_size[1]) + coords = paddle.stack(paddle.meshgrid( + [coords_h, coords_w])) # 2, Wh, Ww + coords_flatten = paddle.flatten(coords, 1) # 2, Wh*Ww + + coords_flatten_1 = coords_flatten.unsqueeze(axis=2) + coords_flatten_2 = coords_flatten.unsqueeze(axis=1) + relative_coords = coords_flatten_1 - coords_flatten_2 + + relative_coords = relative_coords.transpose( + [1, 2, 0]) # Wh*Ww, Wh*Ww, 2 + relative_coords[:, :, 0] += self.window_size[ + 0] - 1 # shift to start from 0 + relative_coords[:, :, 1] += self.window_size[1] - 1 + relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1 + relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww + self.register_buffer("relative_position_index", + relative_position_index) + + self.qkv = nn.Linear(dim, dim * 3, bias_attr=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + self.proj_drop = nn.Dropout(proj_drop) + + trunc_normal_(self.relative_position_bias_table) + self.softmax = nn.Softmax(axis=-1) + + def forward(self, x, mask=None): + """ + Args: + x: input features with shape of (num_windows*B, N, C) + mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None + """ + B_, N, C = x.shape + qkv = self.qkv(x).reshape( + [B_, N, 3, self.num_heads, C // self.num_heads]).transpose( + [2, 0, 3, 1, 4]) + q, k, v = qkv[0], qkv[1], qkv[2] + + q = q * self.scale + attn = paddle.mm(q, k.transpose([0, 1, 3, 2])) + + index = self.relative_position_index.reshape([-1]) + + relative_position_bias = paddle.index_select( + self.relative_position_bias_table, index) + relative_position_bias = relative_position_bias.reshape([ + self.window_size[0] * self.window_size[1], + self.window_size[0] * self.window_size[1], -1 + ]) # Wh*Ww,Wh*Ww,nH + + relative_position_bias = relative_position_bias.transpose( + [2, 0, 1]) # nH, Wh*Ww, Wh*Ww + attn = attn + relative_position_bias.unsqueeze(0) + + if mask is not None: + nW = mask.shape[0] + attn = attn.reshape([B_ // nW, nW, self.num_heads, N, N + ]) + mask.unsqueeze(1).unsqueeze(0) + attn = attn.reshape([-1, self.num_heads, N, N]) + attn = self.softmax(attn) + else: + attn = self.softmax(attn) + + attn = self.attn_drop(attn) + + # x = (attn @ v).transpose(1, 2).reshape([B_, N, C]) + x = paddle.mm(attn, v).transpose([0, 2, 1, 3]).reshape([B_, N, C]) + x = self.proj(x) + x = self.proj_drop(x) + return x + + def extra_repr(self): + return "dim={}, window_size={}, num_heads={}".format( + self.dim, self.window_size, self.num_heads) + + def flops(self, N): + # calculate flops for 1 window with token length of N + flops = 0 + # qkv = self.qkv(x) + flops += N * self.dim * 3 * self.dim + # attn = (q @ k.transpose(-2, -1)) + flops += self.num_heads * N * (self.dim // self.num_heads) * N + # x = (attn @ v) + flops += self.num_heads * N * N * (self.dim // self.num_heads) + # x = self.proj(x) + flops += N * self.dim * self.dim + return flops + + +class SwinTransformerBlock(nn.Layer): + r""" Swin Transformer Block. + + Args: + dim (int): Number of input channels. + input_resolution (tuple[int]): Input resulotion. + num_heads (int): Number of attention heads. + window_size (int): Window size. + shift_size (int): Shift size for SW-MSA. + mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. + qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True + qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. + drop (float, optional): Dropout rate. Default: 0.0 + attn_drop (float, optional): Attention dropout rate. Default: 0.0 + drop_path (float, optional): Stochastic depth rate. Default: 0.0 + act_layer (nn.Layer, optional): Activation layer. Default: nn.GELU + norm_layer (nn.Layer, optional): Normalization layer. Default: nn.LayerNorm + """ + + def __init__(self, + dim, + input_resolution, + num_heads, + window_size=7, + shift_size=0, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer=nn.LayerNorm): + super().__init__() + self.dim = dim + self.input_resolution = input_resolution + self.num_heads = num_heads + self.window_size = window_size + self.shift_size = shift_size + self.mlp_ratio = mlp_ratio + if min(self.input_resolution) <= self.window_size: + # if window size is larger than input resolution, we don't partition windows + self.shift_size = 0 + self.window_size = min(self.input_resolution) + assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size" + + self.norm1 = norm_layer(dim) + self.attn = WindowAttention( + dim, + window_size=to_2tuple(self.window_size), + num_heads=num_heads, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + attn_drop=attn_drop, + proj_drop=drop) + + self.drop_path = DropPath(drop_path) if drop_path > 0. else Identity() + self.norm2 = norm_layer(dim) + mlp_hidden_dim = int(dim * mlp_ratio) + self.mlp = Mlp(in_features=dim, + hidden_features=mlp_hidden_dim, + act_layer=act_layer, + drop=drop) + + if self.shift_size > 0: + # calculate attention mask for SW-MSA + H, W = self.input_resolution + img_mask = paddle.zeros((1, H, W, 1)) # 1 H W 1 + h_slices = (slice(0, -self.window_size), + slice(-self.window_size, -self.shift_size), + slice(-self.shift_size, None)) + w_slices = (slice(0, -self.window_size), + slice(-self.window_size, -self.shift_size), + slice(-self.shift_size, None)) + cnt = 0 + for h in h_slices: + for w in w_slices: + img_mask[:, h, w, :] = cnt + cnt += 1 + + mask_windows = window_partition( + img_mask, self.window_size) # nW, window_size, window_size, 1 + mask_windows = mask_windows.reshape( + [-1, self.window_size * self.window_size]) + attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) + + huns = -100.0 * paddle.ones_like(attn_mask) + attn_mask = huns * (attn_mask != 0).astype("float32") + else: + attn_mask = None + + self.register_buffer("attn_mask", attn_mask) + + def forward(self, x): + H, W = self.input_resolution + B, L, C = x.shape + assert L == H * W, "input feature has wrong size" + + shortcut = x + x = self.norm1(x) + x = x.reshape([B, H, W, C]) + + # cyclic shift + if self.shift_size > 0: + shifted_x = paddle.roll( + x, shifts=(-self.shift_size, -self.shift_size), axis=(1, 2)) + else: + shifted_x = x + + # partition windows + x_windows = window_partition( + shifted_x, self.window_size) # nW*B, window_size, window_size, C + x_windows = x_windows.reshape( + [-1, self.window_size * self.window_size, + C]) # nW*B, window_size*window_size, C + + # W-MSA/SW-MSA + attn_windows = self.attn( + x_windows, mask=self.attn_mask) # nW*B, window_size*window_size, C + + # merge windows + attn_windows = attn_windows.reshape( + [-1, self.window_size, self.window_size, C]) + shifted_x = window_reverse(attn_windows, self.window_size, H, W, + C) # B H' W' C + + # reverse cyclic shift + if self.shift_size > 0: + x = paddle.roll( + shifted_x, + shifts=(self.shift_size, self.shift_size), + axis=(1, 2)) + else: + x = shifted_x + x = x.reshape([B, H * W, C]) + + # FFN + x = shortcut + self.drop_path(x) + x = x + self.drop_path(self.mlp(self.norm2(x))) + + return x + + def extra_repr(self): + return "dim={}, input_resolution={}, num_heads={}, window_size={}, shift_size={}, mlp_ratio={}".format( + self.dim, self.input_resolution, self.num_heads, self.window_size, + self.shift_size, self.mlp_ratio) + + def flops(self): + flops = 0 + H, W = self.input_resolution + # norm1 + flops += self.dim * H * W + # W-MSA/SW-MSA + nW = H * W / self.window_size / self.window_size + flops += nW * self.attn.flops(self.window_size * self.window_size) + # mlp + flops += 2 * H * W * self.dim * self.dim * self.mlp_ratio + # norm2 + flops += self.dim * H * W + return flops + + +class PatchMerging(nn.Layer): + r""" Patch Merging Layer. + + Args: + input_resolution (tuple[int]): Resolution of input feature. + dim (int): Number of input channels. + norm_layer (nn.Layer, optional): Normalization layer. Default: nn.LayerNorm + """ + + def __init__(self, input_resolution, dim, norm_layer=nn.LayerNorm): + super().__init__() + self.input_resolution = input_resolution + self.dim = dim + self.reduction = nn.Linear(4 * dim, 2 * dim, bias_attr=False) + self.norm = norm_layer(4 * dim) + + def forward(self, x): + """ + x: B, H*W, C + """ + H, W = self.input_resolution + B, L, C = x.shape + assert L == H * W, "input feature has wrong size" + assert H % 2 == 0 and W % 2 == 0, "x size ({}*{}) are not even.".format( + H, W) + + x = x.reshape([B, H, W, C]) + + x0 = x[:, 0::2, 0::2, :] # B H/2 W/2 C + x1 = x[:, 1::2, 0::2, :] # B H/2 W/2 C + x2 = x[:, 0::2, 1::2, :] # B H/2 W/2 C + x3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C + x = paddle.concat([x0, x1, x2, x3], -1) # B H/2 W/2 4*C + x = x.reshape([B, H * W // 4, 4 * C]) # B H/2*W/2 4*C + + x = self.norm(x) + x = self.reduction(x) + + return x + + def extra_repr(self): + return "input_resolution={}, dim={}".format(self.input_resolution, + self.dim) + + def flops(self): + H, W = self.input_resolution + flops = H * W * self.dim + flops += (H // 2) * (W // 2) * 4 * self.dim * 2 * self.dim + return flops + + +class BasicLayer(nn.Layer): + """ A basic Swin Transformer layer for one stage. + + Args: + dim (int): Number of input channels. + input_resolution (tuple[int]): Input resolution. + depth (int): Number of blocks. + num_heads (int): Number of attention heads. + window_size (int): Local window size. + mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. + qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True + qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. + drop (float, optional): Dropout rate. Default: 0.0 + attn_drop (float, optional): Attention dropout rate. Default: 0.0 + drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 + norm_layer (nn.Layer, optional): Normalization layer. Default: nn.LayerNorm + downsample (nn.Layer | None, optional): Downsample layer at the end of the layer. Default: None + use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. + """ + + def __init__(self, + dim, + input_resolution, + depth, + num_heads, + window_size, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + norm_layer=nn.LayerNorm, + downsample=None, + use_checkpoint=False): + + super().__init__() + self.dim = dim + self.input_resolution = input_resolution + self.depth = depth + self.use_checkpoint = use_checkpoint + + # build blocks + self.blocks = nn.LayerList([ + SwinTransformerBlock( + dim=dim, + input_resolution=input_resolution, + num_heads=num_heads, + window_size=window_size, + shift_size=0 if (i % 2 == 0) else window_size // 2, + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop, + attn_drop=attn_drop, + drop_path=drop_path[i] + if isinstance(drop_path, list) else drop_path, + norm_layer=norm_layer) for i in range(depth) + ]) + + # patch merging layer + if downsample is not None: + self.downsample = downsample( + input_resolution, dim=dim, norm_layer=norm_layer) + else: + self.downsample = None + + def forward(self, x): + for blk in self.blocks: + x = blk(x) + if self.downsample is not None: + x = self.downsample(x) + return x + + def extra_repr(self): + return "dim={}, input_resolution={}, depth={}".format( + self.dim, self.input_resolution, self.depth) + + def flops(self): + flops = 0 + for blk in self.blocks: + flops += blk.flops() + if self.downsample is not None: + flops += self.downsample.flops() + return flops + + +class PatchEmbed(nn.Layer): + """ Image to Patch Embedding + + Args: + img_size (int): Image size. Default: 224. + patch_size (int): Patch token size. Default: 4. + in_chans (int): Number of input image channels. Default: 3. + embed_dim (int): Number of linear projection output channels. Default: 96. + norm_layer (nn.Layer, optional): Normalization layer. Default: None + """ + + def __init__(self, + img_size=224, + patch_size=4, + in_chans=3, + embed_dim=96, + norm_layer=None): + super().__init__() + img_size = to_2tuple(img_size) + patch_size = to_2tuple(patch_size) + patches_resolution = [ + img_size[0] // patch_size[0], img_size[1] // patch_size[1] + ] + self.img_size = img_size + self.patch_size = patch_size + self.patches_resolution = patches_resolution + self.num_patches = patches_resolution[0] * patches_resolution[1] + + self.in_chans = in_chans + self.embed_dim = embed_dim + + self.proj = nn.Conv2D( + in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) + if norm_layer is not None: + self.norm = norm_layer(embed_dim) + else: + self.norm = None + + def forward(self, x): + B, C, H, W = x.shape + # TODO (littletomatodonkey), uncomment the line will cause failure of jit.save + # assert [H, W] == self.img_size[:2], "Input image size ({H}*{W}) doesn't match model ({}*{}).".format(H, W, self.img_size[0], self.img_size[1]) + x = self.proj(x) + + x = x.flatten(2).transpose([0, 2, 1]) # B Ph*Pw C + if self.norm is not None: + x = self.norm(x) + return x + + def flops(self): + Ho, Wo = self.patches_resolution + flops = Ho * Wo * self.embed_dim * self.in_chans * ( + self.patch_size[0] * self.patch_size[1]) + if self.norm is not None: + flops += Ho * Wo * self.embed_dim + return flops + + +class SwinTransformer(nn.Layer): + """ Swin Transformer + A PaddlePaddle impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` - + https://arxiv.org/pdf/2103.14030 + + Args: + img_size (int | tuple(int)): Input image size. Default 224 + patch_size (int | tuple(int)): Patch size. Default: 4 + in_chans (int): Number of input image channels. Default: 3 + num_classes (int): Number of classes for classification head. Default: 1000 + embed_dim (int): Patch embedding dimension. Default: 96 + depths (tuple(int)): Depth of each Swin Transformer layer. + num_heads (tuple(int)): Number of attention heads in different layers. + window_size (int): Window size. Default: 7 + mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4 + qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True + qk_scale (float): Override default qk scale of head_dim ** -0.5 if set. Default: None + drop_rate (float): Dropout rate. Default: 0 + attn_drop_rate (float): Attention dropout rate. Default: 0 + drop_path_rate (float): Stochastic depth rate. Default: 0.1 + norm_layer (nn.Layer): Normalization layer. Default: nn.LayerNorm. + ape (bool): If True, add absolute position embedding to the patch embedding. Default: False + patch_norm (bool): If True, add normalization after patch embedding. Default: True + use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False + """ + + def __init__(self, + img_size=224, + patch_size=4, + in_chans=3, + class_num=1000, + embed_dim=96, + depths=[2, 2, 6, 2], + num_heads=[3, 6, 12, 24], + window_size=7, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0.1, + norm_layer=nn.LayerNorm, + ape=False, + patch_norm=True, + use_checkpoint=False, + **kwargs): + super(SwinTransformer, self).__init__() + + self.num_classes = num_classes = class_num + self.num_layers = len(depths) + self.embed_dim = embed_dim + self.ape = ape + self.patch_norm = patch_norm + self.num_features = int(embed_dim * 2**(self.num_layers - 1)) + self.mlp_ratio = mlp_ratio + + # split image into non-overlapping patches + self.patch_embed = PatchEmbed( + img_size=img_size, + patch_size=patch_size, + in_chans=in_chans, + embed_dim=embed_dim, + norm_layer=norm_layer if self.patch_norm else None) + num_patches = self.patch_embed.num_patches + patches_resolution = self.patch_embed.patches_resolution + self.patches_resolution = patches_resolution + + # absolute position embedding + if self.ape: + self.absolute_pos_embed = self.create_parameter( + shape=(1, num_patches, embed_dim), default_initializer=zeros_) + self.add_parameter("absolute_pos_embed", self.absolute_pos_embed) + trunc_normal_(self.absolute_pos_embed) + + self.pos_drop = nn.Dropout(p=drop_rate) + + # stochastic depth + dpr = np.linspace(0, drop_path_rate, + sum(depths)).tolist() # stochastic depth decay rule + + # build layers + self.layers = nn.LayerList() + for i_layer in range(self.num_layers): + layer = BasicLayer( + dim=int(embed_dim * 2**i_layer), + input_resolution=(patches_resolution[0] // (2**i_layer), + patches_resolution[1] // (2**i_layer)), + depth=depths[i_layer], + num_heads=num_heads[i_layer], + window_size=window_size, + mlp_ratio=self.mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop_rate, + attn_drop=attn_drop_rate, + drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], + norm_layer=norm_layer, + downsample=PatchMerging + if (i_layer < self.num_layers - 1) else None, + use_checkpoint=use_checkpoint) + self.layers.append(layer) + + self.norm = norm_layer(self.num_features) + self.avgpool = nn.AdaptiveAvgPool1D(1) + self.head = nn.Linear( + self.num_features, + num_classes) if self.num_classes > 0 else nn.Identity() + + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + + def forward_features(self, x): + x = self.patch_embed(x) + if self.ape: + x = x + self.absolute_pos_embed + x = self.pos_drop(x) + + for layer in self.layers: + x = layer(x) + + x = self.norm(x) # B L C + x = self.avgpool(x.transpose([0, 2, 1])) # B C 1 + x = paddle.flatten(x, 1) + return x + + def forward(self, x): + x = self.forward_features(x) + x = self.head(x) + return x + + def flops(self): + flops = 0 + flops += self.patch_embed.flops() + for _, layer in enumerate(self.layers): + flops += layer.flops() + flops += self.num_features * self.patches_resolution[ + 0] * self.patches_resolution[1] // (2**self.num_layers) + flops += self.num_features * self.num_classes + return flops + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def SwinTransformer_tiny_patch4_window7_224(pretrained=False, + use_ssld=False, + **kwargs): + model = SwinTransformer( + embed_dim=96, + depths=[2, 2, 6, 2], + num_heads=[3, 6, 12, 24], + window_size=7, + drop_path_rate=0.2, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SwinTransformer_tiny_patch4_window7_224"], + use_ssld=use_ssld) + return model + + +def SwinTransformer_small_patch4_window7_224(pretrained=False, + use_ssld=False, + **kwargs): + model = SwinTransformer( + embed_dim=96, + depths=[2, 2, 18, 2], + num_heads=[3, 6, 12, 24], + window_size=7, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SwinTransformer_small_patch4_window7_224"], + use_ssld=use_ssld) + return model + + +def SwinTransformer_base_patch4_window7_224(pretrained=False, + use_ssld=False, + **kwargs): + model = SwinTransformer( + embed_dim=128, + depths=[2, 2, 18, 2], + num_heads=[4, 8, 16, 32], + window_size=7, + drop_path_rate=0.5, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SwinTransformer_base_patch4_window7_224"], + use_ssld=use_ssld) + return model + + +def SwinTransformer_base_patch4_window12_384(pretrained=False, + use_ssld=False, + **kwargs): + model = SwinTransformer( + img_size=384, + embed_dim=128, + depths=[2, 2, 18, 2], + num_heads=[4, 8, 16, 32], + window_size=12, + drop_path_rate=0.5, # NOTE: do not appear in offical code + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SwinTransformer_base_patch4_window12_384"], + use_ssld=use_ssld) + return model + + +def SwinTransformer_large_patch4_window7_224(pretrained=False, + use_ssld=False, + **kwargs): + model = SwinTransformer( + embed_dim=192, + depths=[2, 2, 18, 2], + num_heads=[6, 12, 24, 48], + window_size=7, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SwinTransformer_large_patch4_window7_224"], + use_ssld=use_ssld) + return model + + +def SwinTransformer_large_patch4_window12_384(pretrained=False, + use_ssld=False, + **kwargs): + model = SwinTransformer( + img_size=384, + embed_dim=192, + depths=[2, 2, 18, 2], + num_heads=[6, 12, 24, 48], + window_size=12, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["SwinTransformer_large_patch4_window12_384"], + use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/tnt.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/tnt.py new file mode 100644 index 00000000..3ccb7575 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/tnt.py @@ -0,0 +1,385 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/huawei-noah/CV-Backbones/tree/master/tnt_pytorch + +import math +import numpy as np + +import paddle +import paddle.nn as nn + +from paddle.nn.initializer import TruncatedNormal, Constant + +from ppcls.arch.backbone.base.theseus_layer import Identity +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "TNT_small": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/TNT_small_pretrained.pdparams" +} + +__all__ = MODEL_URLS.keys() + +trunc_normal_ = TruncatedNormal(std=.02) +zeros_ = Constant(value=0.) +ones_ = Constant(value=1.) + + +def drop_path(x, drop_prob=0., training=False): + """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). + the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper... + See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... + """ + if drop_prob == 0. or not training: + return x + keep_prob = paddle.to_tensor(1 - drop_prob) + shape = (paddle.shape(x)[0], ) + (1, ) * (x.ndim - 1) + random_tensor = paddle.add(keep_prob, paddle.rand(shape, dtype=x.dtype)) + random_tensor = paddle.floor(random_tensor) # binarize + output = x.divide(keep_prob) * random_tensor + return output + + +class DropPath(nn.Layer): + """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). + """ + + def __init__(self, drop_prob=None): + super(DropPath, self).__init__() + self.drop_prob = drop_prob + + def forward(self, x): + return drop_path(x, self.drop_prob, self.training) + + +class Mlp(nn.Layer): + def __init__(self, + in_features, + hidden_features=None, + out_features=None, + act_layer=nn.GELU, + drop=0.): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Linear(in_features, hidden_features) + self.act = act_layer() + self.fc2 = nn.Linear(hidden_features, out_features) + self.drop = nn.Dropout(drop) + + def forward(self, x): + x = self.fc1(x) + x = self.act(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +class Attention(nn.Layer): + def __init__(self, + dim, + hidden_dim, + num_heads=8, + qkv_bias=False, + attn_drop=0., + proj_drop=0.): + super().__init__() + self.hidden_dim = hidden_dim + self.num_heads = num_heads + head_dim = hidden_dim // num_heads + self.head_dim = head_dim + self.scale = head_dim**-0.5 + + self.qk = nn.Linear(dim, hidden_dim * 2, bias_attr=qkv_bias) + self.v = nn.Linear(dim, dim, bias_attr=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + self.proj_drop = nn.Dropout(proj_drop) + + def forward(self, x): + B, N, C = x.shape + qk = self.qk(x).reshape( + (B, N, 2, self.num_heads, self.head_dim)).transpose( + (2, 0, 3, 1, 4)) + + q, k = qk[0], qk[1] + v = self.v(x).reshape( + (B, N, self.num_heads, x.shape[-1] // self.num_heads)).transpose( + (0, 2, 1, 3)) + + attn = paddle.matmul(q, k.transpose((0, 1, 3, 2))) * self.scale + attn = nn.functional.softmax(attn, axis=-1) + attn = self.attn_drop(attn) + + x = paddle.matmul(attn, v) + x = x.transpose((0, 2, 1, 3)).reshape( + (B, N, x.shape[-1] * x.shape[-3])) + x = self.proj(x) + x = self.proj_drop(x) + return x + + +class Block(nn.Layer): + def __init__(self, + dim, + in_dim, + num_pixel, + num_heads=12, + in_num_head=4, + mlp_ratio=4., + qkv_bias=False, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer=nn.LayerNorm): + super().__init__() + # Inner transformer + self.norm_in = norm_layer(in_dim) + self.attn_in = Attention( + in_dim, + in_dim, + num_heads=in_num_head, + qkv_bias=qkv_bias, + attn_drop=attn_drop, + proj_drop=drop) + + self.norm_mlp_in = norm_layer(in_dim) + self.mlp_in = Mlp(in_features=in_dim, + hidden_features=int(in_dim * 4), + out_features=in_dim, + act_layer=act_layer, + drop=drop) + + self.norm1_proj = norm_layer(in_dim) + self.proj = nn.Linear(in_dim * num_pixel, dim) + # Outer transformer + self.norm_out = norm_layer(dim) + self.attn_out = Attention( + dim, + dim, + num_heads=num_heads, + qkv_bias=qkv_bias, + attn_drop=attn_drop, + proj_drop=drop) + + self.drop_path = DropPath(drop_path) if drop_path > 0. else Identity() + + self.norm_mlp = norm_layer(dim) + self.mlp = Mlp(in_features=dim, + hidden_features=int(dim * mlp_ratio), + out_features=dim, + act_layer=act_layer, + drop=drop) + + def forward(self, pixel_embed, patch_embed): + # inner + pixel_embed = paddle.add( + pixel_embed, + self.drop_path(self.attn_in(self.norm_in(pixel_embed)))) + pixel_embed = paddle.add( + pixel_embed, + self.drop_path(self.mlp_in(self.norm_mlp_in(pixel_embed)))) + # outer + B, N, C = patch_embed.shape + norm1_proj = self.norm1_proj(pixel_embed) + norm1_proj = norm1_proj.reshape( + (B, N - 1, norm1_proj.shape[1] * norm1_proj.shape[2])) + patch_embed[:, 1:] = paddle.add(patch_embed[:, 1:], + self.proj(norm1_proj)) + patch_embed = paddle.add( + patch_embed, + self.drop_path(self.attn_out(self.norm_out(patch_embed)))) + patch_embed = paddle.add( + patch_embed, self.drop_path(self.mlp(self.norm_mlp(patch_embed)))) + return pixel_embed, patch_embed + + +class PixelEmbed(nn.Layer): + def __init__(self, + img_size=224, + patch_size=16, + in_chans=3, + in_dim=48, + stride=4): + super().__init__() + num_patches = (img_size // patch_size)**2 + self.img_size = img_size + self.num_patches = num_patches + self.in_dim = in_dim + new_patch_size = math.ceil(patch_size / stride) + self.new_patch_size = new_patch_size + + self.proj = nn.Conv2D( + in_chans, self.in_dim, kernel_size=7, padding=3, stride=stride) + + def forward(self, x, pixel_pos): + B, C, H, W = x.shape + assert H == self.img_size and W == self.img_size, f"Input image size ({H}*{W}) doesn't match model ({self.img_size}*{self.img_size})." + + x = self.proj(x) + x = nn.functional.unfold(x, self.new_patch_size, self.new_patch_size) + x = x.transpose((0, 2, 1)).reshape( + (-1, self.in_dim, self.new_patch_size, self.new_patch_size)) + x = x + pixel_pos + x = x.reshape((-1, self.in_dim, self.new_patch_size * + self.new_patch_size)).transpose((0, 2, 1)) + return x + + +class TNT(nn.Layer): + def __init__(self, + img_size=224, + patch_size=16, + in_chans=3, + embed_dim=768, + in_dim=48, + depth=12, + num_heads=12, + in_num_head=4, + mlp_ratio=4., + qkv_bias=False, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0., + norm_layer=nn.LayerNorm, + first_stride=4, + class_num=1000): + super().__init__() + self.class_num = class_num + # num_features for consistency with other models + self.num_features = self.embed_dim = embed_dim + + self.pixel_embed = PixelEmbed( + img_size=img_size, + patch_size=patch_size, + in_chans=in_chans, + in_dim=in_dim, + stride=first_stride) + num_patches = self.pixel_embed.num_patches + self.num_patches = num_patches + new_patch_size = self.pixel_embed.new_patch_size + num_pixel = new_patch_size**2 + + self.norm1_proj = norm_layer(num_pixel * in_dim) + self.proj = nn.Linear(num_pixel * in_dim, embed_dim) + self.norm2_proj = norm_layer(embed_dim) + + self.cls_token = self.create_parameter( + shape=(1, 1, embed_dim), default_initializer=zeros_) + self.add_parameter("cls_token", self.cls_token) + + self.patch_pos = self.create_parameter( + shape=(1, num_patches + 1, embed_dim), default_initializer=zeros_) + self.add_parameter("patch_pos", self.patch_pos) + + self.pixel_pos = self.create_parameter( + shape=(1, in_dim, new_patch_size, new_patch_size), + default_initializer=zeros_) + self.add_parameter("pixel_pos", self.pixel_pos) + + self.pos_drop = nn.Dropout(p=drop_rate) + + # stochastic depth decay rule + dpr = np.linspace(0, drop_path_rate, depth) + + blocks = [] + for i in range(depth): + blocks.append( + Block( + dim=embed_dim, + in_dim=in_dim, + num_pixel=num_pixel, + num_heads=num_heads, + in_num_head=in_num_head, + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + drop=drop_rate, + attn_drop=attn_drop_rate, + drop_path=dpr[i], + norm_layer=norm_layer)) + self.blocks = nn.LayerList(blocks) + self.norm = norm_layer(embed_dim) + + if class_num > 0: + self.head = nn.Linear(embed_dim, class_num) + + trunc_normal_(self.cls_token) + trunc_normal_(self.patch_pos) + trunc_normal_(self.pixel_pos) + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + + def forward_features(self, x): + B = paddle.shape(x)[0] + pixel_embed = self.pixel_embed(x, self.pixel_pos) + + patch_embed = self.norm2_proj( + self.proj( + self.norm1_proj( + pixel_embed.reshape((-1, self.num_patches, pixel_embed. + shape[-1] * pixel_embed.shape[-2]))))) + patch_embed = paddle.concat( + (self.cls_token.expand((B, -1, -1)), patch_embed), axis=1) + patch_embed = patch_embed + self.patch_pos + patch_embed = self.pos_drop(patch_embed) + + for blk in self.blocks: + pixel_embed, patch_embed = blk(pixel_embed, patch_embed) + + patch_embed = self.norm(patch_embed) + return patch_embed[:, 0] + + def forward(self, x): + x = self.forward_features(x) + + if self.class_num > 0: + x = self.head(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def TNT_small(pretrained=False, **kwargs): + model = TNT(patch_size=16, + embed_dim=384, + in_dim=24, + depth=12, + num_heads=6, + in_num_head=4, + qkv_bias=False, + **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["TNT_small"]) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/vision_transformer.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/vision_transformer.py new file mode 100644 index 00000000..c71c0262 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/vision_transformer.py @@ -0,0 +1,458 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py + +from collections.abc import Callable + +import numpy as np +import paddle +import paddle.nn as nn +from paddle.nn.initializer import TruncatedNormal, Constant, Normal + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ViT_small_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_small_patch16_224_pretrained.pdparams", + "ViT_base_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_224_pretrained.pdparams", + "ViT_base_patch16_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_384_pretrained.pdparams", + "ViT_base_patch32_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch32_384_pretrained.pdparams", + "ViT_large_patch16_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_224_pretrained.pdparams", + "ViT_large_patch16_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_384_pretrained.pdparams", + "ViT_large_patch32_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch32_384_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + +trunc_normal_ = TruncatedNormal(std=.02) +normal_ = Normal +zeros_ = Constant(value=0.) +ones_ = Constant(value=1.) + + +def to_2tuple(x): + return tuple([x] * 2) + + +def drop_path(x, drop_prob=0., training=False): + """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). + the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper... + See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... + """ + if drop_prob == 0. or not training: + return x + keep_prob = paddle.to_tensor(1 - drop_prob) + shape = (paddle.shape(x)[0], ) + (1, ) * (x.ndim - 1) + random_tensor = keep_prob + paddle.rand(shape, dtype=x.dtype) + random_tensor = paddle.floor(random_tensor) # binarize + output = x.divide(keep_prob) * random_tensor + return output + + +class DropPath(nn.Layer): + """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). + """ + + def __init__(self, drop_prob=None): + super(DropPath, self).__init__() + self.drop_prob = drop_prob + + def forward(self, x): + return drop_path(x, self.drop_prob, self.training) + + +class Identity(nn.Layer): + def __init__(self): + super(Identity, self).__init__() + + def forward(self, input): + return input + + +class Mlp(nn.Layer): + def __init__(self, + in_features, + hidden_features=None, + out_features=None, + act_layer=nn.GELU, + drop=0.): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Linear(in_features, hidden_features) + self.act = act_layer() + self.fc2 = nn.Linear(hidden_features, out_features) + self.drop = nn.Dropout(drop) + + def forward(self, x): + x = self.fc1(x) + x = self.act(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +class Attention(nn.Layer): + def __init__(self, + dim, + num_heads=8, + qkv_bias=False, + qk_scale=None, + attn_drop=0., + proj_drop=0.): + super().__init__() + self.num_heads = num_heads + head_dim = dim // num_heads + self.scale = qk_scale or head_dim**-0.5 + + self.qkv = nn.Linear(dim, dim * 3, bias_attr=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + self.proj_drop = nn.Dropout(proj_drop) + + def forward(self, x): + # B= paddle.shape(x)[0] + N, C = x.shape[1:] + qkv = self.qkv(x).reshape((-1, N, 3, self.num_heads, C // + self.num_heads)).transpose((2, 0, 3, 1, 4)) + q, k, v = qkv[0], qkv[1], qkv[2] + + attn = (q.matmul(k.transpose((0, 1, 3, 2)))) * self.scale + attn = nn.functional.softmax(attn, axis=-1) + attn = self.attn_drop(attn) + + x = (attn.matmul(v)).transpose((0, 2, 1, 3)).reshape((-1, N, C)) + x = self.proj(x) + x = self.proj_drop(x) + return x + + +class Block(nn.Layer): + def __init__(self, + dim, + num_heads, + mlp_ratio=4., + qkv_bias=False, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer='nn.LayerNorm', + epsilon=1e-5): + super().__init__() + if isinstance(norm_layer, str): + self.norm1 = eval(norm_layer)(dim, epsilon=epsilon) + elif isinstance(norm_layer, Callable): + self.norm1 = norm_layer(dim) + else: + raise TypeError( + "The norm_layer must be str or paddle.nn.layer.Layer class") + self.attn = Attention( + dim, + num_heads=num_heads, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + attn_drop=attn_drop, + proj_drop=drop) + # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here + self.drop_path = DropPath(drop_path) if drop_path > 0. else Identity() + if isinstance(norm_layer, str): + self.norm2 = eval(norm_layer)(dim, epsilon=epsilon) + elif isinstance(norm_layer, Callable): + self.norm2 = norm_layer(dim) + else: + raise TypeError( + "The norm_layer must be str or paddle.nn.layer.Layer class") + mlp_hidden_dim = int(dim * mlp_ratio) + self.mlp = Mlp(in_features=dim, + hidden_features=mlp_hidden_dim, + act_layer=act_layer, + drop=drop) + + def forward(self, x): + x = x + self.drop_path(self.attn(self.norm1(x))) + x = x + self.drop_path(self.mlp(self.norm2(x))) + return x + + +class PatchEmbed(nn.Layer): + """ Image to Patch Embedding + """ + + def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768): + super().__init__() + img_size = to_2tuple(img_size) + patch_size = to_2tuple(patch_size) + num_patches = (img_size[1] // patch_size[1]) * \ + (img_size[0] // patch_size[0]) + self.img_size = img_size + self.patch_size = patch_size + self.num_patches = num_patches + + self.proj = nn.Conv2D( + in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) + + def forward(self, x): + B, C, H, W = x.shape + assert H == self.img_size[0] and W == self.img_size[1], \ + f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})." + + x = self.proj(x).flatten(2).transpose((0, 2, 1)) + return x + + +class VisionTransformer(nn.Layer): + """ Vision Transformer with support for patch input + """ + + def __init__(self, + img_size=224, + patch_size=16, + in_chans=3, + class_num=1000, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=False, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0., + norm_layer='nn.LayerNorm', + epsilon=1e-5, + **kwargs): + super().__init__() + self.class_num = class_num + + self.num_features = self.embed_dim = embed_dim + + self.patch_embed = PatchEmbed( + img_size=img_size, + patch_size=patch_size, + in_chans=in_chans, + embed_dim=embed_dim) + num_patches = self.patch_embed.num_patches + + self.pos_embed = self.create_parameter( + shape=(1, num_patches + 1, embed_dim), default_initializer=zeros_) + self.add_parameter("pos_embed", self.pos_embed) + self.cls_token = self.create_parameter( + shape=(1, 1, embed_dim), default_initializer=zeros_) + self.add_parameter("cls_token", self.cls_token) + self.pos_drop = nn.Dropout(p=drop_rate) + + dpr = np.linspace(0, drop_path_rate, depth) + + self.blocks = nn.LayerList([ + Block( + dim=embed_dim, + num_heads=num_heads, + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop_rate, + attn_drop=attn_drop_rate, + drop_path=dpr[i], + norm_layer=norm_layer, + epsilon=epsilon) for i in range(depth) + ]) + + self.norm = eval(norm_layer)(embed_dim, epsilon=epsilon) + + # Classifier head + self.head = nn.Linear(embed_dim, + class_num) if class_num > 0 else Identity() + + trunc_normal_(self.pos_embed) + trunc_normal_(self.cls_token) + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + + def forward_features(self, x): + # B = x.shape[0] + B = paddle.shape(x)[0] + x = self.patch_embed(x) + cls_tokens = self.cls_token.expand((B, -1, -1)) + x = paddle.concat((cls_tokens, x), axis=1) + x = x + self.pos_embed + x = self.pos_drop(x) + for blk in self.blocks: + x = blk(x) + x = self.norm(x) + return x[:, 0] + + def forward(self, x): + x = self.forward_features(x) + x = self.head(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ViT_small_patch16_224(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + patch_size=16, + embed_dim=768, + depth=8, + num_heads=8, + mlp_ratio=3, + qk_scale=768**-0.5, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ViT_small_patch16_224"], + use_ssld=use_ssld) + return model + + +def ViT_base_patch16_224(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + patch_size=16, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ViT_base_patch16_224"], + use_ssld=use_ssld) + return model + + +def ViT_base_patch16_384(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + img_size=384, + patch_size=16, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ViT_base_patch16_384"], + use_ssld=use_ssld) + return model + + +def ViT_base_patch32_384(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + img_size=384, + patch_size=32, + embed_dim=768, + depth=12, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ViT_base_patch32_384"], + use_ssld=use_ssld) + return model + + +def ViT_large_patch16_224(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + patch_size=16, + embed_dim=1024, + depth=24, + num_heads=16, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ViT_large_patch16_224"], + use_ssld=use_ssld) + return model + + +def ViT_large_patch16_384(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + img_size=384, + patch_size=16, + embed_dim=1024, + depth=24, + num_heads=16, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ViT_large_patch16_384"], + use_ssld=use_ssld) + return model + + +def ViT_large_patch32_384(pretrained=False, use_ssld=False, **kwargs): + model = VisionTransformer( + img_size=384, + patch_size=32, + embed_dim=1024, + depth=24, + num_heads=16, + mlp_ratio=4, + qkv_bias=True, + epsilon=1e-6, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["ViT_large_patch32_384"], + use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/xception.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/xception.py new file mode 100644 index 00000000..2b843788 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/xception.py @@ -0,0 +1,377 @@ +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform +import math +import sys + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "Xception41": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_pretrained.pdparams", + "Xception65": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_pretrained.pdparams", + "Xception71": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception71_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +class ConvBNLayer(nn.Layer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + act=None, + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(name=name + "_weights"), + bias_attr=False) + bn_name = "bn_" + name + self._batch_norm = BatchNorm( + num_filters, + act=act, + param_attr=ParamAttr(name=bn_name + "_scale"), + bias_attr=ParamAttr(name=bn_name + "_offset"), + moving_mean_name=bn_name + '_mean', + moving_variance_name=bn_name + '_variance') + + def forward(self, inputs): + y = self._conv(inputs) + y = self._batch_norm(y) + return y + + +class SeparableConv(nn.Layer): + def __init__(self, input_channels, output_channels, stride=1, name=None): + super(SeparableConv, self).__init__() + + self._pointwise_conv = ConvBNLayer( + input_channels, output_channels, 1, name=name + "_sep") + self._depthwise_conv = ConvBNLayer( + output_channels, + output_channels, + 3, + stride=stride, + groups=output_channels, + name=name + "_dw") + + def forward(self, inputs): + x = self._pointwise_conv(inputs) + x = self._depthwise_conv(x) + return x + + +class EntryFlowBottleneckBlock(nn.Layer): + def __init__(self, + input_channels, + output_channels, + stride=2, + name=None, + relu_first=False): + super(EntryFlowBottleneckBlock, self).__init__() + self.relu_first = relu_first + + self._short = Conv2D( + in_channels=input_channels, + out_channels=output_channels, + kernel_size=1, + stride=stride, + padding=0, + weight_attr=ParamAttr(name + "_branch1_weights"), + bias_attr=False) + self._conv1 = SeparableConv( + input_channels, + output_channels, + stride=1, + name=name + "_branch2a_weights") + self._conv2 = SeparableConv( + output_channels, + output_channels, + stride=1, + name=name + "_branch2b_weights") + self._pool = MaxPool2D(kernel_size=3, stride=stride, padding=1) + + def forward(self, inputs): + conv0 = inputs + short = self._short(inputs) + if self.relu_first: + conv0 = F.relu(conv0) + conv1 = self._conv1(conv0) + conv2 = F.relu(conv1) + conv2 = self._conv2(conv2) + pool = self._pool(conv2) + return paddle.add(x=short, y=pool) + + +class EntryFlow(nn.Layer): + def __init__(self, block_num=3): + super(EntryFlow, self).__init__() + + name = "entry_flow" + self.block_num = block_num + self._conv1 = ConvBNLayer( + 3, 32, 3, stride=2, act="relu", name=name + "_conv1") + self._conv2 = ConvBNLayer(32, 64, 3, act="relu", name=name + "_conv2") + if block_num == 3: + self._conv_0 = EntryFlowBottleneckBlock( + 64, 128, stride=2, name=name + "_0", relu_first=False) + self._conv_1 = EntryFlowBottleneckBlock( + 128, 256, stride=2, name=name + "_1", relu_first=True) + self._conv_2 = EntryFlowBottleneckBlock( + 256, 728, stride=2, name=name + "_2", relu_first=True) + elif block_num == 5: + self._conv_0 = EntryFlowBottleneckBlock( + 64, 128, stride=2, name=name + "_0", relu_first=False) + self._conv_1 = EntryFlowBottleneckBlock( + 128, 256, stride=1, name=name + "_1", relu_first=True) + self._conv_2 = EntryFlowBottleneckBlock( + 256, 256, stride=2, name=name + "_2", relu_first=True) + self._conv_3 = EntryFlowBottleneckBlock( + 256, 728, stride=1, name=name + "_3", relu_first=True) + self._conv_4 = EntryFlowBottleneckBlock( + 728, 728, stride=2, name=name + "_4", relu_first=True) + else: + sys.exit(-1) + + def forward(self, inputs): + x = self._conv1(inputs) + x = self._conv2(x) + + if self.block_num == 3: + x = self._conv_0(x) + x = self._conv_1(x) + x = self._conv_2(x) + elif self.block_num == 5: + x = self._conv_0(x) + x = self._conv_1(x) + x = self._conv_2(x) + x = self._conv_3(x) + x = self._conv_4(x) + return x + + +class MiddleFlowBottleneckBlock(nn.Layer): + def __init__(self, input_channels, output_channels, name): + super(MiddleFlowBottleneckBlock, self).__init__() + + self._conv_0 = SeparableConv( + input_channels, + output_channels, + stride=1, + name=name + "_branch2a_weights") + self._conv_1 = SeparableConv( + output_channels, + output_channels, + stride=1, + name=name + "_branch2b_weights") + self._conv_2 = SeparableConv( + output_channels, + output_channels, + stride=1, + name=name + "_branch2c_weights") + + def forward(self, inputs): + conv0 = F.relu(inputs) + conv0 = self._conv_0(conv0) + conv1 = F.relu(conv0) + conv1 = self._conv_1(conv1) + conv2 = F.relu(conv1) + conv2 = self._conv_2(conv2) + return paddle.add(x=inputs, y=conv2) + + +class MiddleFlow(nn.Layer): + def __init__(self, block_num=8): + super(MiddleFlow, self).__init__() + + self.block_num = block_num + self._conv_0 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_0") + self._conv_1 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_1") + self._conv_2 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_2") + self._conv_3 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_3") + self._conv_4 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_4") + self._conv_5 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_5") + self._conv_6 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_6") + self._conv_7 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_7") + if block_num == 16: + self._conv_8 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_8") + self._conv_9 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_9") + self._conv_10 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_10") + self._conv_11 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_11") + self._conv_12 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_12") + self._conv_13 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_13") + self._conv_14 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_14") + self._conv_15 = MiddleFlowBottleneckBlock( + 728, 728, name="middle_flow_15") + + def forward(self, inputs): + x = self._conv_0(inputs) + x = self._conv_1(x) + x = self._conv_2(x) + x = self._conv_3(x) + x = self._conv_4(x) + x = self._conv_5(x) + x = self._conv_6(x) + x = self._conv_7(x) + if self.block_num == 16: + x = self._conv_8(x) + x = self._conv_9(x) + x = self._conv_10(x) + x = self._conv_11(x) + x = self._conv_12(x) + x = self._conv_13(x) + x = self._conv_14(x) + x = self._conv_15(x) + return x + + +class ExitFlowBottleneckBlock(nn.Layer): + def __init__(self, input_channels, output_channels1, output_channels2, + name): + super(ExitFlowBottleneckBlock, self).__init__() + + self._short = Conv2D( + in_channels=input_channels, + out_channels=output_channels2, + kernel_size=1, + stride=2, + padding=0, + weight_attr=ParamAttr(name + "_branch1_weights"), + bias_attr=False) + self._conv_1 = SeparableConv( + input_channels, + output_channels1, + stride=1, + name=name + "_branch2a_weights") + self._conv_2 = SeparableConv( + output_channels1, + output_channels2, + stride=1, + name=name + "_branch2b_weights") + self._pool = MaxPool2D(kernel_size=3, stride=2, padding=1) + + def forward(self, inputs): + short = self._short(inputs) + conv0 = F.relu(inputs) + conv1 = self._conv_1(conv0) + conv2 = F.relu(conv1) + conv2 = self._conv_2(conv2) + pool = self._pool(conv2) + return paddle.add(x=short, y=pool) + + +class ExitFlow(nn.Layer): + def __init__(self, class_num): + super(ExitFlow, self).__init__() + + name = "exit_flow" + + self._conv_0 = ExitFlowBottleneckBlock( + 728, 728, 1024, name=name + "_1") + self._conv_1 = SeparableConv(1024, 1536, stride=1, name=name + "_2") + self._conv_2 = SeparableConv(1536, 2048, stride=1, name=name + "_3") + self._pool = AdaptiveAvgPool2D(1) + stdv = 1.0 / math.sqrt(2048 * 1.0) + self._out = Linear( + 2048, + class_num, + weight_attr=ParamAttr( + name="fc_weights", initializer=Uniform(-stdv, stdv)), + bias_attr=ParamAttr(name="fc_offset")) + + def forward(self, inputs): + conv0 = self._conv_0(inputs) + conv1 = self._conv_1(conv0) + conv1 = F.relu(conv1) + conv2 = self._conv_2(conv1) + conv2 = F.relu(conv2) + pool = self._pool(conv2) + pool = paddle.flatten(pool, start_axis=1, stop_axis=-1) + out = self._out(pool) + return out + + +class Xception(nn.Layer): + def __init__(self, + entry_flow_block_num=3, + middle_flow_block_num=8, + class_num=1000): + super(Xception, self).__init__() + self.entry_flow_block_num = entry_flow_block_num + self.middle_flow_block_num = middle_flow_block_num + self._entry_flow = EntryFlow(entry_flow_block_num) + self._middle_flow = MiddleFlow(middle_flow_block_num) + self._exit_flow = ExitFlow(class_num) + + def forward(self, inputs): + x = self._entry_flow(inputs) + x = self._middle_flow(x) + x = self._exit_flow(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def Xception41(pretrained=False, use_ssld=False, **kwargs): + model = Xception(entry_flow_block_num=3, middle_flow_block_num=8, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["Xception41"], use_ssld=use_ssld) + return model + + +def Xception65(pretrained=False, use_ssld=False, **kwargs): + model = Xception( + entry_flow_block_num=3, middle_flow_block_num=16, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["Xception65"], use_ssld=use_ssld) + return model + + +def Xception71(pretrained=False, use_ssld=False, **kwargs): + model = Xception( + entry_flow_block_num=5, middle_flow_block_num=16, **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["Xception71"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/xception_deeplab.py b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/xception_deeplab.py new file mode 100644 index 00000000..c52769b3 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/model_zoo/xception_deeplab.py @@ -0,0 +1,421 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn import Conv2D, BatchNorm, Linear, Dropout +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "Xception41_deeplab": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_deeplab_pretrained.pdparams", + "Xception65_deeplab": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_deeplab_pretrained.pdparams" +} + +__all__ = list(MODEL_URLS.keys()) + + +def check_data(data, number): + if type(data) == int: + return [data] * number + assert len(data) == number + return data + + +def check_stride(s, os): + if s <= os: + return True + else: + return False + + +def check_points(count, points): + if points is None: + return False + else: + if isinstance(points, list): + return (True if count in points else False) + else: + return (True if count == points else False) + + +def gen_bottleneck_params(backbone='xception_65'): + if backbone == 'xception_65': + bottleneck_params = { + "entry_flow": (3, [2, 2, 2], [128, 256, 728]), + "middle_flow": (16, 1, 728), + "exit_flow": (2, [2, 1], [[728, 1024, 1024], [1536, 1536, 2048]]) + } + elif backbone == 'xception_41': + bottleneck_params = { + "entry_flow": (3, [2, 2, 2], [128, 256, 728]), + "middle_flow": (8, 1, 728), + "exit_flow": (2, [2, 1], [[728, 1024, 1024], [1536, 1536, 2048]]) + } + elif backbone == 'xception_71': + bottleneck_params = { + "entry_flow": (5, [2, 1, 2, 1, 2], [128, 256, 256, 728, 728]), + "middle_flow": (16, 1, 728), + "exit_flow": (2, [2, 1], [[728, 1024, 1024], [1536, 1536, 2048]]) + } + else: + raise Exception( + "xception backbont only support xception_41/xception_65/xception_71" + ) + return bottleneck_params + + +class ConvBNLayer(nn.Layer): + def __init__(self, + input_channels, + output_channels, + filter_size, + stride=1, + padding=0, + act=None, + name=None): + super(ConvBNLayer, self).__init__() + + self._conv = Conv2D( + in_channels=input_channels, + out_channels=output_channels, + kernel_size=filter_size, + stride=stride, + padding=padding, + weight_attr=ParamAttr(name=name + "/weights"), + bias_attr=False) + self._bn = BatchNorm( + num_channels=output_channels, + act=act, + epsilon=1e-3, + momentum=0.99, + param_attr=ParamAttr(name=name + "/BatchNorm/gamma"), + bias_attr=ParamAttr(name=name + "/BatchNorm/beta"), + moving_mean_name=name + "/BatchNorm/moving_mean", + moving_variance_name=name + "/BatchNorm/moving_variance") + + def forward(self, inputs): + return self._bn(self._conv(inputs)) + + +class Seperate_Conv(nn.Layer): + def __init__(self, + input_channels, + output_channels, + stride, + filter, + dilation=1, + act=None, + name=None): + super(Seperate_Conv, self).__init__() + + self._conv1 = Conv2D( + in_channels=input_channels, + out_channels=input_channels, + kernel_size=filter, + stride=stride, + groups=input_channels, + padding=(filter) // 2 * dilation, + dilation=dilation, + weight_attr=ParamAttr(name=name + "/depthwise/weights"), + bias_attr=False) + self._bn1 = BatchNorm( + input_channels, + act=act, + epsilon=1e-3, + momentum=0.99, + param_attr=ParamAttr(name=name + "/depthwise/BatchNorm/gamma"), + bias_attr=ParamAttr(name=name + "/depthwise/BatchNorm/beta"), + moving_mean_name=name + "/depthwise/BatchNorm/moving_mean", + moving_variance_name=name + "/depthwise/BatchNorm/moving_variance") + self._conv2 = Conv2D( + input_channels, + output_channels, + 1, + stride=1, + groups=1, + padding=0, + weight_attr=ParamAttr(name=name + "/pointwise/weights"), + bias_attr=False) + self._bn2 = BatchNorm( + output_channels, + act=act, + epsilon=1e-3, + momentum=0.99, + param_attr=ParamAttr(name=name + "/pointwise/BatchNorm/gamma"), + bias_attr=ParamAttr(name=name + "/pointwise/BatchNorm/beta"), + moving_mean_name=name + "/pointwise/BatchNorm/moving_mean", + moving_variance_name=name + "/pointwise/BatchNorm/moving_variance") + + def forward(self, inputs): + x = self._conv1(inputs) + x = self._bn1(x) + x = self._conv2(x) + x = self._bn2(x) + return x + + +class Xception_Block(nn.Layer): + def __init__(self, + input_channels, + output_channels, + strides=1, + filter_size=3, + dilation=1, + skip_conv=True, + has_skip=True, + activation_fn_in_separable_conv=False, + name=None): + super(Xception_Block, self).__init__() + + repeat_number = 3 + output_channels = check_data(output_channels, repeat_number) + filter_size = check_data(filter_size, repeat_number) + strides = check_data(strides, repeat_number) + + self.has_skip = has_skip + self.skip_conv = skip_conv + self.activation_fn_in_separable_conv = activation_fn_in_separable_conv + if not activation_fn_in_separable_conv: + self._conv1 = Seperate_Conv( + input_channels, + output_channels[0], + stride=strides[0], + filter=filter_size[0], + dilation=dilation, + name=name + "/separable_conv1") + self._conv2 = Seperate_Conv( + output_channels[0], + output_channels[1], + stride=strides[1], + filter=filter_size[1], + dilation=dilation, + name=name + "/separable_conv2") + self._conv3 = Seperate_Conv( + output_channels[1], + output_channels[2], + stride=strides[2], + filter=filter_size[2], + dilation=dilation, + name=name + "/separable_conv3") + else: + self._conv1 = Seperate_Conv( + input_channels, + output_channels[0], + stride=strides[0], + filter=filter_size[0], + act="relu", + dilation=dilation, + name=name + "/separable_conv1") + self._conv2 = Seperate_Conv( + output_channels[0], + output_channels[1], + stride=strides[1], + filter=filter_size[1], + act="relu", + dilation=dilation, + name=name + "/separable_conv2") + self._conv3 = Seperate_Conv( + output_channels[1], + output_channels[2], + stride=strides[2], + filter=filter_size[2], + act="relu", + dilation=dilation, + name=name + "/separable_conv3") + + if has_skip and skip_conv: + self._short = ConvBNLayer( + input_channels, + output_channels[-1], + 1, + stride=strides[-1], + padding=0, + name=name + "/shortcut") + + def forward(self, inputs): + if not self.activation_fn_in_separable_conv: + x = F.relu(inputs) + x = self._conv1(x) + x = F.relu(x) + x = self._conv2(x) + x = F.relu(x) + x = self._conv3(x) + else: + x = self._conv1(inputs) + x = self._conv2(x) + x = self._conv3(x) + if self.has_skip: + if self.skip_conv: + skip = self._short(inputs) + else: + skip = inputs + return paddle.add(x, skip) + else: + return x + + +class XceptionDeeplab(nn.Layer): + def __init__(self, backbone, class_num=1000): + super(XceptionDeeplab, self).__init__() + + bottleneck_params = gen_bottleneck_params(backbone) + self.backbone = backbone + + self._conv1 = ConvBNLayer( + 3, + 32, + 3, + stride=2, + padding=1, + act="relu", + name=self.backbone + "/entry_flow/conv1") + self._conv2 = ConvBNLayer( + 32, + 64, + 3, + stride=1, + padding=1, + act="relu", + name=self.backbone + "/entry_flow/conv2") + + self.block_num = bottleneck_params["entry_flow"][0] + self.strides = bottleneck_params["entry_flow"][1] + self.chns = bottleneck_params["entry_flow"][2] + self.strides = check_data(self.strides, self.block_num) + self.chns = check_data(self.chns, self.block_num) + + self.entry_flow = [] + self.middle_flow = [] + + self.stride = 2 + self.output_stride = 32 + s = self.stride + + for i in range(self.block_num): + stride = self.strides[i] if check_stride(s * self.strides[i], + self.output_stride) else 1 + xception_block = self.add_sublayer( + self.backbone + "/entry_flow/block" + str(i + 1), + Xception_Block( + input_channels=64 if i == 0 else self.chns[i - 1], + output_channels=self.chns[i], + strides=[1, 1, self.stride], + name=self.backbone + "/entry_flow/block" + str(i + 1))) + self.entry_flow.append(xception_block) + s = s * stride + self.stride = s + + self.block_num = bottleneck_params["middle_flow"][0] + self.strides = bottleneck_params["middle_flow"][1] + self.chns = bottleneck_params["middle_flow"][2] + self.strides = check_data(self.strides, self.block_num) + self.chns = check_data(self.chns, self.block_num) + s = self.stride + + for i in range(self.block_num): + stride = self.strides[i] if check_stride(s * self.strides[i], + self.output_stride) else 1 + xception_block = self.add_sublayer( + self.backbone + "/middle_flow/block" + str(i + 1), + Xception_Block( + input_channels=728, + output_channels=728, + strides=[1, 1, self.strides[i]], + skip_conv=False, + name=self.backbone + "/middle_flow/block" + str(i + 1))) + self.middle_flow.append(xception_block) + s = s * stride + self.stride = s + + self.block_num = bottleneck_params["exit_flow"][0] + self.strides = bottleneck_params["exit_flow"][1] + self.chns = bottleneck_params["exit_flow"][2] + self.strides = check_data(self.strides, self.block_num) + self.chns = check_data(self.chns, self.block_num) + s = self.stride + stride = self.strides[0] if check_stride(s * self.strides[0], + self.output_stride) else 1 + self._exit_flow_1 = Xception_Block( + 728, + self.chns[0], [1, 1, stride], + name=self.backbone + "/exit_flow/block1") + s = s * stride + stride = self.strides[1] if check_stride(s * self.strides[1], + self.output_stride) else 1 + self._exit_flow_2 = Xception_Block( + self.chns[0][-1], + self.chns[1], [1, 1, stride], + dilation=2, + has_skip=False, + activation_fn_in_separable_conv=True, + name=self.backbone + "/exit_flow/block2") + s = s * stride + + self.stride = s + + self._drop = Dropout(p=0.5, mode="downscale_in_infer") + self._pool = AdaptiveAvgPool2D(1) + self._fc = Linear( + self.chns[1][-1], + class_num, + weight_attr=ParamAttr(name="fc_weights"), + bias_attr=ParamAttr(name="fc_bias")) + + def forward(self, inputs): + x = self._conv1(inputs) + x = self._conv2(x) + for ef in self.entry_flow: + x = ef(x) + for mf in self.middle_flow: + x = mf(x) + x = self._exit_flow_1(x) + x = self._exit_flow_2(x) + x = self._drop(x) + x = self._pool(x) + x = paddle.squeeze(x, axis=[2, 3]) + x = self._fc(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def Xception41_deeplab(pretrained=False, use_ssld=False, **kwargs): + model = XceptionDeeplab('xception_41', **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["Xception41_deeplab"], use_ssld=use_ssld) + return model + + +def Xception65_deeplab(pretrained=False, use_ssld=False, **kwargs): + model = XceptionDeeplab("xception_65", **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["Xception65_deeplab"], use_ssld=use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/variant_models/__init__.py b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/__init__.py new file mode 100644 index 00000000..75cf29ff --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/__init__.py @@ -0,0 +1,3 @@ +from .resnet_variant import ResNet50_last_stage_stride1 +from .vgg_variant import VGG19Sigmoid +from .pp_lcnet_variant import PPLCNet_x2_5_Tanh diff --git a/exprements/PaddleClas/ppcls/arch/backbone/variant_models/pp_lcnet_variant.py b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/pp_lcnet_variant.py new file mode 100644 index 00000000..dc9747af --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/pp_lcnet_variant.py @@ -0,0 +1,29 @@ +import paddle +from paddle.nn import Sigmoid +from paddle.nn import Tanh +from ppcls.arch.backbone.legendary_models.pp_lcnet import PPLCNet_x2_5 + +__all__ = ["PPLCNet_x2_5_Tanh"] + + +class TanhSuffix(paddle.nn.Layer): + def __init__(self, origin_layer): + super(TanhSuffix, self).__init__() + self.origin_layer = origin_layer + self.tanh = Tanh() + + def forward(self, input, res_dict=None, **kwargs): + x = self.origin_layer(input) + x = self.tanh(x) + return x + + +def PPLCNet_x2_5_Tanh(pretrained=False, use_ssld=False, **kwargs): + def replace_function(origin_layer, pattern): + new_layer = TanhSuffix(origin_layer) + return new_layer + + pattern = "fc" + model = PPLCNet_x2_5(pretrained=pretrained, use_ssld=use_ssld, **kwargs) + model.upgrade_sublayer(pattern, replace_function) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/variant_models/resnet_variant.py b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/resnet_variant.py new file mode 100644 index 00000000..0219344b --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/resnet_variant.py @@ -0,0 +1,23 @@ +from paddle.nn import Conv2D +from ppcls.arch.backbone.legendary_models.resnet import ResNet50, MODEL_URLS, _load_pretrained + +__all__ = ["ResNet50_last_stage_stride1"] + + +def ResNet50_last_stage_stride1(pretrained=False, use_ssld=False, **kwargs): + def replace_function(conv, pattern): + new_conv = Conv2D( + in_channels=conv._in_channels, + out_channels=conv._out_channels, + kernel_size=conv._kernel_size, + stride=1, + padding=conv._padding, + groups=conv._groups, + bias_attr=conv._bias_attr) + return new_conv + + pattern = ["blocks[13].conv1.conv", "blocks[13].short.conv"] + model = ResNet50(pretrained=False, use_ssld=use_ssld, **kwargs) + model.upgrade_sublayer(pattern, replace_function) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet50"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/arch/backbone/variant_models/vgg_variant.py b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/vgg_variant.py new file mode 100644 index 00000000..c1f75ba9 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/backbone/variant_models/vgg_variant.py @@ -0,0 +1,28 @@ +import paddle +from paddle.nn import Sigmoid +from ppcls.arch.backbone.legendary_models.vgg import VGG19 + +__all__ = ["VGG19Sigmoid"] + + +class SigmoidSuffix(paddle.nn.Layer): + def __init__(self, origin_layer): + super().__init__() + self.origin_layer = origin_layer + self.sigmoid = Sigmoid() + + def forward(self, input, res_dict=None, **kwargs): + x = self.origin_layer(input) + x = self.sigmoid(x) + return x + + +def VGG19Sigmoid(pretrained=False, use_ssld=False, **kwargs): + def replace_function(origin_layer, pattern): + new_layer = SigmoidSuffix(origin_layer) + return new_layer + + pattern = "fc2" + model = VGG19(pretrained=pretrained, use_ssld=use_ssld, **kwargs) + model.upgrade_sublayer(pattern, replace_function) + return model diff --git a/exprements/PaddleClas/ppcls/arch/gears/__init__.py b/exprements/PaddleClas/ppcls/arch/gears/__init__.py new file mode 100644 index 00000000..75ca41d8 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/gears/__init__.py @@ -0,0 +1,32 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from .arcmargin import ArcMargin +from .cosmargin import CosMargin +from .circlemargin import CircleMargin +from .fc import FC +from .vehicle_neck import VehicleNeck + +__all__ = ['build_gear'] + + +def build_gear(config): + support_dict = [ + 'ArcMargin', 'CosMargin', 'CircleMargin', 'FC', 'VehicleNeck' + ] + module_name = config.pop('name') + assert module_name in support_dict, Exception( + 'head only support {}'.format(support_dict)) + module_class = eval(module_name)(**config) + return module_class diff --git a/exprements/PaddleClas/ppcls/arch/gears/arcmargin.py b/exprements/PaddleClas/ppcls/arch/gears/arcmargin.py new file mode 100644 index 00000000..22cc76e1 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/gears/arcmargin.py @@ -0,0 +1,72 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +import paddle.nn as nn +import math + + +class ArcMargin(nn.Layer): + def __init__(self, + embedding_size, + class_num, + margin=0.5, + scale=80.0, + easy_margin=False): + super().__init__() + self.embedding_size = embedding_size + self.class_num = class_num + self.margin = margin + self.scale = scale + self.easy_margin = easy_margin + self.weight = self.create_parameter( + shape=[self.embedding_size, self.class_num], + is_bias=False, + default_initializer=paddle.nn.initializer.XavierNormal()) + + def forward(self, input, label=None): + input_norm = paddle.sqrt( + paddle.sum(paddle.square(input), axis=1, keepdim=True)) + input = paddle.divide(input, input_norm) + + weight_norm = paddle.sqrt( + paddle.sum(paddle.square(self.weight), axis=0, keepdim=True)) + weight = paddle.divide(self.weight, weight_norm) + + cos = paddle.matmul(input, weight) + if not self.training or label is None: + return cos + sin = paddle.sqrt(1.0 - paddle.square(cos) + 1e-6) + cos_m = math.cos(self.margin) + sin_m = math.sin(self.margin) + phi = cos * cos_m - sin * sin_m + + th = math.cos(self.margin) * (-1) + mm = math.sin(self.margin) * self.margin + if self.easy_margin: + phi = self._paddle_where_more_than(cos, 0, phi, cos) + else: + phi = self._paddle_where_more_than(cos, th, phi, cos - mm) + + one_hot = paddle.nn.functional.one_hot(label, self.class_num) + one_hot = paddle.squeeze(one_hot, axis=[1]) + output = paddle.multiply(one_hot, phi) + paddle.multiply( + (1.0 - one_hot), cos) + output = output * self.scale + return output + + def _paddle_where_more_than(self, target, limit, x, y): + mask = paddle.cast(x=(target > limit), dtype='float32') + output = paddle.multiply(mask, x) + paddle.multiply((1.0 - mask), y) + return output diff --git a/exprements/PaddleClas/ppcls/arch/gears/circlemargin.py b/exprements/PaddleClas/ppcls/arch/gears/circlemargin.py new file mode 100644 index 00000000..d1bce83c --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/gears/circlemargin.py @@ -0,0 +1,59 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class CircleMargin(nn.Layer): + def __init__(self, embedding_size, class_num, margin, scale): + super(CircleMargin, self).__init__() + self.scale = scale + self.margin = margin + self.embedding_size = embedding_size + self.class_num = class_num + + self.weight = self.create_parameter( + shape=[self.embedding_size, self.class_num], + is_bias=False, + default_initializer=paddle.nn.initializer.XavierNormal()) + + def forward(self, input, label): + feat_norm = paddle.sqrt( + paddle.sum(paddle.square(input), axis=1, keepdim=True)) + input = paddle.divide(input, feat_norm) + + weight_norm = paddle.sqrt( + paddle.sum(paddle.square(self.weight), axis=0, keepdim=True)) + weight = paddle.divide(self.weight, weight_norm) + + logits = paddle.matmul(input, weight) + if not self.training or label is None: + return logits + + alpha_p = paddle.clip(-logits.detach() + 1 + self.margin, min=0.) + alpha_n = paddle.clip(logits.detach() + self.margin, min=0.) + delta_p = 1 - self.margin + delta_n = self.margin + + m_hot = F.one_hot(label.reshape([-1]), num_classes=logits.shape[1]) + + logits_p = alpha_p * (logits - delta_p) + logits_n = alpha_n * (logits - delta_n) + pre_logits = logits_p * m_hot + logits_n * (1 - m_hot) + pre_logits = self.scale * pre_logits + + return pre_logits diff --git a/exprements/PaddleClas/ppcls/arch/gears/cosmargin.py b/exprements/PaddleClas/ppcls/arch/gears/cosmargin.py new file mode 100644 index 00000000..578b64c2 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/gears/cosmargin.py @@ -0,0 +1,55 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +import math +import paddle.nn as nn + + +class CosMargin(paddle.nn.Layer): + def __init__(self, embedding_size, class_num, margin=0.35, scale=64.0): + super(CosMargin, self).__init__() + self.scale = scale + self.margin = margin + self.embedding_size = embedding_size + self.class_num = class_num + + self.weight = self.create_parameter( + shape=[self.embedding_size, self.class_num], + is_bias=False, + default_initializer=paddle.nn.initializer.XavierNormal()) + + def forward(self, input, label): + label.stop_gradient = True + + input_norm = paddle.sqrt( + paddle.sum(paddle.square(input), axis=1, keepdim=True)) + input = paddle.divide(input, input_norm) + + weight_norm = paddle.sqrt( + paddle.sum(paddle.square(self.weight), axis=0, keepdim=True)) + weight = paddle.divide(self.weight, weight_norm) + + cos = paddle.matmul(input, weight) + if not self.training or label is None: + return cos + + cos_m = cos - self.margin + + one_hot = paddle.nn.functional.one_hot(label, self.class_num) + one_hot = paddle.squeeze(one_hot, axis=[1]) + output = paddle.multiply(one_hot, cos_m) + paddle.multiply( + (1.0 - one_hot), cos) + output = output * self.scale + return output diff --git a/exprements/PaddleClas/ppcls/arch/gears/fc.py b/exprements/PaddleClas/ppcls/arch/gears/fc.py new file mode 100644 index 00000000..b3247419 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/gears/fc.py @@ -0,0 +1,35 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import paddle +import paddle.nn as nn + + +class FC(nn.Layer): + def __init__(self, embedding_size, class_num): + super(FC, self).__init__() + self.embedding_size = embedding_size + self.class_num = class_num + weight_attr = paddle.ParamAttr( + initializer=paddle.nn.initializer.XavierNormal()) + self.fc = paddle.nn.Linear( + self.embedding_size, self.class_num, weight_attr=weight_attr) + + def forward(self, input, label=None): + out = self.fc(input) + return out diff --git a/exprements/PaddleClas/ppcls/arch/gears/identity_head.py b/exprements/PaddleClas/ppcls/arch/gears/identity_head.py new file mode 100644 index 00000000..7d11e574 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/gears/identity_head.py @@ -0,0 +1,9 @@ +from paddle import nn + + +class IdentityHead(nn.Layer): + def __init__(self): + super(IdentityHead, self).__init__() + + def forward(self, x, label=None): + return {"features": x, "logits": None} diff --git a/exprements/PaddleClas/ppcls/arch/gears/vehicle_neck.py b/exprements/PaddleClas/ppcls/arch/gears/vehicle_neck.py new file mode 100644 index 00000000..05f4e333 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/gears/vehicle_neck.py @@ -0,0 +1,52 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import paddle +import paddle.nn as nn + + +class VehicleNeck(nn.Layer): + def __init__(self, + in_channels, + out_channels, + kernel_size=1, + stride=1, + padding=0, + dilation=1, + groups=1, + padding_mode='zeros', + weight_attr=None, + bias_attr=None, + data_format='NCHW'): + super().__init__() + self.conv = nn.Conv2D( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + dilation=dilation, + groups=groups, + padding_mode=padding_mode, + weight_attr=weight_attr, + bias_attr=weight_attr, + data_format=data_format) + self.flatten = nn.Flatten() + + def forward(self, x): + x = self.conv(x) + x = self.flatten(x) + return x diff --git a/exprements/PaddleClas/ppcls/arch/slim/__init__.py b/exprements/PaddleClas/ppcls/arch/slim/__init__.py new file mode 100644 index 00000000..3733059c --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/slim/__init__.py @@ -0,0 +1,16 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from ppcls.arch.slim.prune import prune_model +from ppcls.arch.slim.quant import quantize_model diff --git a/exprements/PaddleClas/ppcls/arch/slim/prune.py b/exprements/PaddleClas/ppcls/arch/slim/prune.py new file mode 100644 index 00000000..c0c9d220 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/slim/prune.py @@ -0,0 +1,65 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function +import paddle +from ppcls.utils import logger + + +def prune_model(config, model): + if config.get("Slim", False) and config["Slim"].get("prune", False): + import paddleslim + prune_method_name = config["Slim"]["prune"]["name"].lower() + assert prune_method_name in [ + "fpgm", "l1_norm" + ], "The prune methods only support 'fpgm' and 'l1_norm'" + if prune_method_name == "fpgm": + model.pruner = paddleslim.dygraph.FPGMFilterPruner( + model, [1] + config["Global"]["image_shape"]) + else: + model.pruner = paddleslim.dygraph.L1NormFilterPruner( + model, [1] + config["Global"]["image_shape"]) + + # prune model + _prune_model(config, model) + else: + model.pruner = None + + + +def _prune_model(config, model): + from paddleslim.analysis import dygraph_flops as flops + logger.info("FLOPs before pruning: {}GFLOPs".format( + flops(model, [1] + config["Global"]["image_shape"]) / 1e9)) + model.eval() + + params = [] + for sublayer in model.sublayers(): + for param in sublayer.parameters(include_sublayers=False): + if isinstance(sublayer, paddle.nn.Conv2D): + params.append(param.name) + ratios = {} + for param in params: + ratios[param] = config["Slim"]["prune"]["pruned_ratio"] + plan = model.pruner.prune_vars(ratios, [0]) + + logger.info("FLOPs after pruning: {}GFLOPs; pruned ratio: {}".format( + flops(model, [1] + config["Global"]["image_shape"]) / 1e9, + plan.pruned_flops)) + + for param in model.parameters(): + if "conv2d" in param.name: + logger.info("{}\t{}".format(param.name, param.shape)) + + model.train() diff --git a/exprements/PaddleClas/ppcls/arch/slim/quant.py b/exprements/PaddleClas/ppcls/arch/slim/quant.py new file mode 100644 index 00000000..b8f59a78 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/slim/quant.py @@ -0,0 +1,55 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function +import paddle +from ppcls.utils import logger + +QUANT_CONFIG = { + # weight preprocess type, default is None and no preprocessing is performed. + 'weight_preprocess_type': None, + # activation preprocess type, default is None and no preprocessing is performed. + 'activation_preprocess_type': None, + # weight quantize type, default is 'channel_wise_abs_max' + 'weight_quantize_type': 'channel_wise_abs_max', + # activation quantize type, default is 'moving_average_abs_max' + 'activation_quantize_type': 'moving_average_abs_max', + # weight quantize bit num, default is 8 + 'weight_bits': 8, + # activation quantize bit num, default is 8 + 'activation_bits': 8, + # data type after quantization, such as 'uint8', 'int8', etc. default is 'int8' + 'dtype': 'int8', + # window size for 'range_abs_max' quantization. default is 10000 + 'window_size': 10000, + # The decay coefficient of moving average, default is 0.9 + 'moving_rate': 0.9, + # for dygraph quantization, layers of type in quantizable_layer_type will be quantized + 'quantizable_layer_type': ['Conv2D', 'Linear'], +} + + +def quantize_model(config, model): + if config.get("Slim", False) and config["Slim"].get("quant", False): + from paddleslim.dygraph.quant import QAT + assert config["Slim"]["quant"]["name"].lower( + ) == 'pact', 'Only PACT quantization method is supported now' + QUANT_CONFIG["activation_preprocess_type"] = "PACT" + model.quanter = QAT(config=QUANT_CONFIG) + model.quanter.quantize(model) + logger.info("QAT model summary:") + paddle.summary(model, (1, 3, 224, 224)) + else: + model.quanter = None + return diff --git a/exprements/PaddleClas/ppcls/arch/utils.py b/exprements/PaddleClas/ppcls/arch/utils.py new file mode 100644 index 00000000..308475d7 --- /dev/null +++ b/exprements/PaddleClas/ppcls/arch/utils.py @@ -0,0 +1,53 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import six +import types +from difflib import SequenceMatcher + +from . import backbone + + +def get_architectures(): + """ + get all of model architectures + """ + names = [] + for k, v in backbone.__dict__.items(): + if isinstance(v, (types.FunctionType, six.class_types)): + names.append(k) + return names + + +def get_blacklist_model_in_static_mode(): + from ppcls.arch.backbone import distilled_vision_transformer + from ppcls.arch.backbone import vision_transformer + blacklist = distilled_vision_transformer.__all__ + vision_transformer.__all__ + return blacklist + + +def similar_architectures(name='', names=[], thresh=0.1, topk=10): + """ + inferred similar architectures + """ + scores = [] + for idx, n in enumerate(names): + if n.startswith('__'): + continue + score = SequenceMatcher(None, n.lower(), name.lower()).quick_ratio() + if score > thresh: + scores.append((idx, score)) + scores.sort(key=lambda x: x[1], reverse=True) + similar_names = [names[s[0]] for s in scores[:min(topk, len(scores))]] + return similar_names diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_005.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_005.yaml new file mode 100644 index 00000000..aaf8e915 --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_005.yaml @@ -0,0 +1,149 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: RecModel + infer_output_key: features + infer_add_softmax: False + + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.005 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + + sampler: + name: DistributedBatchSampler + batch_size: 256 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_005_cutout.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_005_cutout.yaml new file mode 100644 index 00000000..4a1c517f --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_005_cutout.yaml @@ -0,0 +1,151 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: RecModel + infer_output_key: features + infer_add_softmax: False + + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.005 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - Cutout: + n_holes: 1 + length: 112 + + sampler: + name: DistributedBatchSampler + batch_size: 256 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_01.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_01.yaml new file mode 100644 index 00000000..0339ae5a --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_01.yaml @@ -0,0 +1,149 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: RecModel + infer_output_key: features + infer_add_softmax: False + + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.01 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + + sampler: + name: DistributedBatchSampler + batch_size: 256 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_01_cutout.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_01_cutout.yaml new file mode 100644 index 00000000..6a468a6f --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_01_cutout.yaml @@ -0,0 +1,153 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: RecModel + infer_output_key: features + infer_add_softmax: False + + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.01 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + + sampler: + name: DistributedBatchSampler + batch_size: 256 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - Cutout: + n_holes: 1 + length: 112 + + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset/ + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_002.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_002.yaml new file mode 100644 index 00000000..c305335c --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_002.yaml @@ -0,0 +1,189 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: true + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: "DistillationModel" + infer_output_key: features + infer_add_softmax: False + is_rec: True + infer_model_name: "Student" + # if not null, its lengths should be same as models + pretrained_list: + # if not null, its lengths should be same as models + freeze_params_list: + - False + - False + models: + - Teacher: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + - Student: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - DistillationGTCELoss: + weight: 1.0 + key: "logits" + model_names: ["Student", "Teacher"] + - DistillationDMLLoss: + weight: 1.0 + key: "logits" + model_name_pairs: + - ["Student", "Teacher"] + Eval: + - DistillationGTCELoss: + weight: 1.0 + model_names: ["Student"] + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.002 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_002_cutout.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_002_cutout.yaml new file mode 100644 index 00000000..b8dad879 --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_002_cutout.yaml @@ -0,0 +1,191 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: true + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: "DistillationModel" + infer_output_key: features + infer_add_softmax: False + is_rec: True + infer_model_name: "Student" + # if not null, its lengths should be same as models + pretrained_list: + # if not null, its lengths should be same as models + freeze_params_list: + - False + - False + models: + - Teacher: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + - Student: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - DistillationGTCELoss: + weight: 1.0 + key: "logits" + model_names: ["Student", "Teacher"] + - DistillationDMLLoss: + weight: 1.0 + key: "logits" + model_name_pairs: + - ["Student", "Teacher"] + Eval: + - DistillationGTCELoss: + weight: 1.0 + model_names: ["Student"] + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.002 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - Cutout: + n_holes: 1 + length: 112 + + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_004.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_004.yaml new file mode 100644 index 00000000..ee85bef0 --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_004.yaml @@ -0,0 +1,189 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: true + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: "DistillationModel" + infer_output_key: features + infer_add_softmax: False + is_rec: True + infer_model_name: "Student" + # if not null, its lengths should be same as models + pretrained_list: + # if not null, its lengths should be same as models + freeze_params_list: + - False + - False + models: + - Teacher: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + - Student: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - DistillationGTCELoss: + weight: 1.0 + key: "logits" + model_names: ["Student", "Teacher"] + - DistillationDMLLoss: + weight: 1.0 + key: "logits" + model_name_pairs: + - ["Student", "Teacher"] + Eval: + - DistillationGTCELoss: + weight: 1.0 + model_names: ["Student"] + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.004 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_004_cutout.yaml b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_004_cutout.yaml new file mode 100644 index 00000000..0b176b94 --- /dev/null +++ b/exprements/PaddleClas/ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_dml_004_cutout.yaml @@ -0,0 +1,192 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: true + eval_interval: 1 + epochs: 400 + print_batch_step: 10 + use_visualdl: True + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + +# model architecture +Arch: + name: "DistillationModel" + infer_output_key: features + infer_add_softmax: False + is_rec: True + infer_model_name: "Student" + # if not null, its lengths should be same as models + pretrained_list: + # if not null, its lengths should be same as models + freeze_params_list: + - False + - False + models: + - Teacher: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + - Student: + name: RecModel + infer_output_key: features + infer_add_softmax: False + Backbone: + name: PPLCNet_x2_5 + pretrained: True + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 358 + margin: 0.2 + scale: 30 + +# loss function config for traing/eval process +Loss: + Train: + - DistillationGTCELoss: + weight: 1.0 + key: "logits" + model_names: ["Student", "Teacher"] + - DistillationDMLLoss: + weight: 1.0 + key: "logits" + model_name_pairs: + - ["Student", "Teacher"] + Eval: + - DistillationGTCELoss: + weight: 1.0 + model_names: ["Student"] + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.004 + warmup_epoch: 5 + regularizer: + name: 'L2' + coeff: 0.00001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - Cutout: + n_holes: 1 + length: 112 + + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: VeriWild + image_root: /root/paddlejob/workspace/code/PaddleClas/dataset + cls_label_path: /root/paddlejob/workspace/code/PaddleClas/dataset/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/data/__init__.py b/exprements/PaddleClas/ppcls/data/__init__.py new file mode 100644 index 00000000..cffac812 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/__init__.py @@ -0,0 +1,144 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import inspect +import copy +import paddle +import numpy as np +from paddle.io import DistributedBatchSampler, BatchSampler, DataLoader +from ppcls.utils import logger + +from ppcls.data import dataloader +# dataset +from ppcls.data.dataloader.imagenet_dataset import ImageNetDataset +from ppcls.data.dataloader.multilabel_dataset import MultiLabelDataset +from ppcls.data.dataloader.common_dataset import create_operators +from ppcls.data.dataloader.vehicle_dataset import CompCars, VeriWild +from ppcls.data.dataloader.logo_dataset import LogoDataset +from ppcls.data.dataloader.icartoon_dataset import ICartoonDataset +from ppcls.data.dataloader.mix_dataset import MixDataset + +# sampler +from ppcls.data.dataloader.DistributedRandomIdentitySampler import DistributedRandomIdentitySampler +from ppcls.data.dataloader.pk_sampler import PKSampler +from ppcls.data.dataloader.mix_sampler import MixSampler +from ppcls.data import preprocess +from ppcls.data.preprocess import transform + + +def create_operators(params, class_num=None): + """ + create operators based on the config + + Args: + params(list): a dict list, used to create some operators + """ + assert isinstance(params, list), ('operator config should be a list') + ops = [] + for operator in params: + assert isinstance(operator, + dict) and len(operator) == 1, "yaml format error" + op_name = list(operator)[0] + param = {} if operator[op_name] is None else operator[op_name] + op_func = getattr(preprocess, op_name) + if "class_num" in inspect.getfullargspec(op_func).args: + param.update({"class_num": class_num}) + op = op_func(**param) + ops.append(op) + + return ops + + +def build_dataloader(config, mode, device, use_dali=False, seed=None): + assert mode in [ + 'Train', 'Eval', 'Test', 'Gallery', 'Query' + ], "Dataset mode should be Train, Eval, Test, Gallery, Query" + # build dataset + if use_dali: + from ppcls.data.dataloader.dali import dali_dataloader + return dali_dataloader(config, mode, paddle.device.get_device(), seed) + + class_num = config.get("class_num", None) + config_dataset = config[mode]['dataset'] + config_dataset = copy.deepcopy(config_dataset) + dataset_name = config_dataset.pop('name') + if 'batch_transform_ops' in config_dataset: + batch_transform = config_dataset.pop('batch_transform_ops') + else: + batch_transform = None + + dataset = eval(dataset_name)(**config_dataset) + + logger.debug("build dataset({}) success...".format(dataset)) + + # build sampler + config_sampler = config[mode]['sampler'] + if "name" not in config_sampler: + batch_sampler = None + batch_size = config_sampler["batch_size"] + drop_last = config_sampler["drop_last"] + shuffle = config_sampler["shuffle"] + else: + sampler_name = config_sampler.pop("name") + batch_sampler = eval(sampler_name)(dataset, **config_sampler) + + logger.debug("build batch_sampler({}) success...".format(batch_sampler)) + + # build batch operator + def mix_collate_fn(batch): + batch = transform(batch, batch_ops) + # batch each field + slots = [] + for items in batch: + for i, item in enumerate(items): + if len(slots) < len(items): + slots.append([item]) + else: + slots[i].append(item) + return [np.stack(slot, axis=0) for slot in slots] + + if isinstance(batch_transform, list): + batch_ops = create_operators(batch_transform, class_num) + batch_collate_fn = mix_collate_fn + else: + batch_collate_fn = None + + # build dataloader + config_loader = config[mode]['loader'] + num_workers = config_loader["num_workers"] + use_shared_memory = config_loader["use_shared_memory"] + + if batch_sampler is None: + data_loader = DataLoader( + dataset=dataset, + places=device, + num_workers=num_workers, + return_list=True, + use_shared_memory=use_shared_memory, + batch_size=batch_size, + shuffle=shuffle, + drop_last=drop_last, + collate_fn=batch_collate_fn) + else: + data_loader = DataLoader( + dataset=dataset, + places=device, + num_workers=num_workers, + return_list=True, + use_shared_memory=use_shared_memory, + batch_sampler=batch_sampler, + collate_fn=batch_collate_fn) + + logger.debug("build data_loader({}) success...".format(data_loader)) + return data_loader diff --git a/exprements/PaddleClas/ppcls/data/dataloader/DistributedRandomIdentitySampler.py b/exprements/PaddleClas/ppcls/data/dataloader/DistributedRandomIdentitySampler.py new file mode 100644 index 00000000..1203803f --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/DistributedRandomIdentitySampler.py @@ -0,0 +1,90 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from collections import defaultdict +import numpy as np +import copy +import random +from paddle.io import DistributedBatchSampler, Sampler + + +class DistributedRandomIdentitySampler(DistributedBatchSampler): + """ + Randomly sample N identities, then for each identity, + randomly sample K instances, therefore batch size is N*K. + Args: + - data_source (list): list of (img_path, pid, camid). + - num_instances (int): number of instances per identity in a batch. + - batch_size (int): number of examples in a batch. + """ + + def __init__(self, dataset, batch_size, num_instances, drop_last, **args): + self.dataset = dataset + self.batch_size = batch_size + self.num_instances = num_instances + self.drop_last = drop_last + self.num_pids_per_batch = self.batch_size // self.num_instances + self.index_dic = defaultdict(list) + for index, pid in enumerate(self.dataset.labels): + self.index_dic[pid].append(index) + self.pids = list(self.index_dic.keys()) + # estimate number of examples in an epoch + self.length = 0 + for pid in self.pids: + idxs = self.index_dic[pid] + num = len(idxs) + if num < self.num_instances: + num = self.num_instances + self.length += num - num % self.num_instances + + def __iter__(self): + batch_idxs_dict = defaultdict(list) + for pid in self.pids: + idxs = copy.deepcopy(self.index_dic[pid]) + if len(idxs) < self.num_instances: + idxs = np.random.choice( + idxs, size=self.num_instances, replace=True) + random.shuffle(idxs) + batch_idxs = [] + for idx in idxs: + batch_idxs.append(idx) + if len(batch_idxs) == self.num_instances: + batch_idxs_dict[pid].append(batch_idxs) + batch_idxs = [] + avai_pids = copy.deepcopy(self.pids) + final_idxs = [] + while len(avai_pids) >= self.num_pids_per_batch: + selected_pids = random.sample(avai_pids, self.num_pids_per_batch) + for pid in selected_pids: + batch_idxs = batch_idxs_dict[pid].pop(0) + final_idxs.extend(batch_idxs) + if len(batch_idxs_dict[pid]) == 0: + avai_pids.remove(pid) + _sample_iter = iter(final_idxs) + batch_indices = [] + for idx in _sample_iter: + batch_indices.append(idx) + if len(batch_indices) == self.batch_size: + yield batch_indices + batch_indices = [] + if not self.drop_last and len(batch_indices) > 0: + yield batch_indices + + def __len__(self): + if self.drop_last: + return self.length // self.batch_size + else: + return (self.length + self.batch_size - 1) // self.batch_size diff --git a/exprements/PaddleClas/ppcls/data/dataloader/__init__.py b/exprements/PaddleClas/ppcls/data/dataloader/__init__.py new file mode 100644 index 00000000..8f819210 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/__init__.py @@ -0,0 +1,9 @@ +from ppcls.data.dataloader.imagenet_dataset import ImageNetDataset +from ppcls.data.dataloader.multilabel_dataset import MultiLabelDataset +from ppcls.data.dataloader.common_dataset import create_operators +from ppcls.data.dataloader.vehicle_dataset import CompCars, VeriWild +from ppcls.data.dataloader.logo_dataset import LogoDataset +from ppcls.data.dataloader.icartoon_dataset import ICartoonDataset +from ppcls.data.dataloader.mix_dataset import MixDataset +from ppcls.data.dataloader.mix_sampler import MixSampler +from ppcls.data.dataloader.pk_sampler import PKSampler diff --git a/exprements/PaddleClas/ppcls/data/dataloader/common_dataset.py b/exprements/PaddleClas/ppcls/data/dataloader/common_dataset.py new file mode 100644 index 00000000..b7b03d8b --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/common_dataset.py @@ -0,0 +1,84 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import numpy as np + +from paddle.io import Dataset +import cv2 + +from ppcls.data import preprocess +from ppcls.data.preprocess import transform +from ppcls.utils import logger + + +def create_operators(params): + """ + create operators based on the config + Args: + params(list): a dict list, used to create some operators + """ + assert isinstance(params, list), ('operator config should be a list') + ops = [] + for operator in params: + assert isinstance(operator, + dict) and len(operator) == 1, "yaml format error" + op_name = list(operator)[0] + param = {} if operator[op_name] is None else operator[op_name] + op = getattr(preprocess, op_name)(**param) + ops.append(op) + + return ops + + +class CommonDataset(Dataset): + def __init__( + self, + image_root, + cls_label_path, + transform_ops=None, ): + self._img_root = image_root + self._cls_path = cls_label_path + if transform_ops: + self._transform_ops = create_operators(transform_ops) + + self.images = [] + self.labels = [] + self._load_anno() + + def _load_anno(self): + pass + + def __getitem__(self, idx): + try: + with open(self.images[idx], 'rb') as f: + img = f.read() + if self._transform_ops: + img = transform(img, self._transform_ops) + img = img.transpose((2, 0, 1)) + return (img, self.labels[idx]) + + except Exception as ex: + logger.error("Exception occured when parse line: {} with msg: {}". + format(self.images[idx], ex)) + rnd_idx = np.random.randint(self.__len__()) + return self.__getitem__(rnd_idx) + + def __len__(self): + return len(self.images) + + @property + def class_num(self): + return len(set(self.labels)) diff --git a/exprements/PaddleClas/ppcls/data/dataloader/dali.py b/exprements/PaddleClas/ppcls/data/dataloader/dali.py new file mode 100644 index 00000000..a15c2315 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/dali.py @@ -0,0 +1,319 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import division + +import copy +import os + +import numpy as np +import nvidia.dali.ops as ops +import nvidia.dali.types as types +import paddle +from nvidia.dali import fn +from nvidia.dali.pipeline import Pipeline +from nvidia.dali.plugin.base_iterator import LastBatchPolicy +from nvidia.dali.plugin.paddle import DALIGenericIterator + + +class HybridTrainPipe(Pipeline): + def __init__(self, + file_root, + file_list, + batch_size, + resize_shorter, + crop, + min_area, + lower, + upper, + interp, + mean, + std, + device_id, + shard_id=0, + num_shards=1, + random_shuffle=True, + num_threads=4, + seed=42, + pad_output=False, + output_dtype=types.FLOAT, + dataset='Train'): + super(HybridTrainPipe, self).__init__( + batch_size, num_threads, device_id, seed=seed) + self.input = ops.readers.File( + file_root=file_root, + file_list=file_list, + shard_id=shard_id, + num_shards=num_shards, + random_shuffle=random_shuffle) + # set internal nvJPEG buffers size to handle full-sized ImageNet images + # without additional reallocations + device_memory_padding = 211025920 + host_memory_padding = 140544512 + self.decode = ops.decoders.ImageRandomCrop( + device='mixed', + output_type=types.DALIImageType.RGB, + device_memory_padding=device_memory_padding, + host_memory_padding=host_memory_padding, + random_aspect_ratio=[lower, upper], + random_area=[min_area, 1.0], + num_attempts=100) + self.res = ops.Resize( + device='gpu', resize_x=crop, resize_y=crop, interp_type=interp) + self.cmnp = ops.CropMirrorNormalize( + device="gpu", + dtype=output_dtype, + output_layout='CHW', + crop=(crop, crop), + mean=mean, + std=std, + pad_output=pad_output) + self.coin = ops.random.CoinFlip(probability=0.5) + self.to_int64 = ops.Cast(dtype=types.DALIDataType.INT64, device="gpu") + + def define_graph(self): + rng = self.coin() + jpegs, labels = self.input(name="Reader") + images = self.decode(jpegs) + images = self.res(images) + output = self.cmnp(images.gpu(), mirror=rng) + return [output, self.to_int64(labels.gpu())] + + def __len__(self): + return self.epoch_size("Reader") + + +class HybridValPipe(Pipeline): + def __init__(self, + file_root, + file_list, + batch_size, + resize_shorter, + crop, + interp, + mean, + std, + device_id, + shard_id=0, + num_shards=1, + random_shuffle=False, + num_threads=4, + seed=42, + pad_output=False, + output_dtype=types.FLOAT): + super(HybridValPipe, self).__init__( + batch_size, num_threads, device_id, seed=seed) + self.input = ops.readers.File( + file_root=file_root, + file_list=file_list, + shard_id=shard_id, + num_shards=num_shards, + random_shuffle=random_shuffle) + self.decode = ops.decoders.Image(device="mixed") + self.res = ops.Resize( + device="gpu", resize_shorter=resize_shorter, interp_type=interp) + self.cmnp = ops.CropMirrorNormalize( + device="gpu", + dtype=output_dtype, + output_layout='CHW', + crop=(crop, crop), + mean=mean, + std=std, + pad_output=pad_output) + self.to_int64 = ops.Cast(dtype=types.DALIDataType.INT64, device="gpu") + + def define_graph(self): + jpegs, labels = self.input(name="Reader") + images = self.decode(jpegs) + images = self.res(images) + output = self.cmnp(images) + return [output, self.to_int64(labels.gpu())] + + def __len__(self): + return self.epoch_size("Reader") + + +def dali_dataloader(config, mode, device, seed=None): + assert "gpu" in device, "gpu training is required for DALI" + device_id = int(device.split(':')[1]) + config_dataloader = config[mode] + seed = 42 if seed is None else seed + ops = [ + list(x.keys())[0] + for x in config_dataloader["dataset"]["transform_ops"] + ] + support_ops_train = [ + "DecodeImage", "NormalizeImage", "RandFlipImage", "RandCropImage" + ] + support_ops_eval = [ + "DecodeImage", "ResizeImage", "CropImage", "NormalizeImage" + ] + + if mode.lower() == 'train': + assert set(ops) == set( + support_ops_train + ), "The supported trasform_ops for train_dataset in dali is : {}".format( + ",".join(support_ops_train)) + else: + assert set(ops) == set( + support_ops_eval + ), "The supported trasform_ops for eval_dataset in dali is : {}".format( + ",".join(support_ops_eval)) + + normalize_ops = [ + op for op in config_dataloader["dataset"]["transform_ops"] + if "NormalizeImage" in op + ][0]["NormalizeImage"] + channel_num = normalize_ops.get("channel_num", 3) + output_dtype = types.FLOAT16 if normalize_ops.get("output_fp16", + False) else types.FLOAT + + env = os.environ + # assert float(env.get('FLAGS_fraction_of_gpu_memory_to_use', 0.92)) < 0.9, \ + # "Please leave enough GPU memory for DALI workspace, e.g., by setting" \ + # " `export FLAGS_fraction_of_gpu_memory_to_use=0.8`" + + gpu_num = paddle.distributed.get_world_size() + + batch_size = config_dataloader["sampler"]["batch_size"] + + file_root = config_dataloader["dataset"]["image_root"] + file_list = config_dataloader["dataset"]["cls_label_path"] + + interp = 1 # settings.interpolation or 1 # default to linear + interp_map = { + 0: types.DALIInterpType.INTERP_NN, # cv2.INTER_NEAREST + 1: types.DALIInterpType.INTERP_LINEAR, # cv2.INTER_LINEAR + 2: types.DALIInterpType.INTERP_CUBIC, # cv2.INTER_CUBIC + 3: types.DALIInterpType. + INTERP_LANCZOS3, # XXX use LANCZOS3 for cv2.INTER_LANCZOS4 + } + + assert interp in interp_map, "interpolation method not supported by DALI" + interp = interp_map[interp] + pad_output = channel_num == 4 + + transforms = { + k: v + for d in config_dataloader["dataset"]["transform_ops"] + for k, v in d.items() + } + + scale = transforms["NormalizeImage"].get("scale", 1.0 / 255) + scale = eval(scale) if isinstance(scale, str) else scale + mean = transforms["NormalizeImage"].get("mean", [0.485, 0.456, 0.406]) + std = transforms["NormalizeImage"].get("std", [0.229, 0.224, 0.225]) + mean = [v / scale for v in mean] + std = [v / scale for v in std] + + sampler_name = config_dataloader["sampler"].get("name", + "DistributedBatchSampler") + assert sampler_name in ["DistributedBatchSampler", "BatchSampler"] + + if mode.lower() == "train": + resize_shorter = 256 + crop = transforms["RandCropImage"]["size"] + scale = transforms["RandCropImage"].get("scale", [0.08, 1.]) + ratio = transforms["RandCropImage"].get("ratio", [3.0 / 4, 4.0 / 3]) + min_area = scale[0] + lower = ratio[0] + upper = ratio[1] + + if 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env: + shard_id = int(env['PADDLE_TRAINER_ID']) + num_shards = int(env['PADDLE_TRAINERS_NUM']) + device_id = int(env['FLAGS_selected_gpus']) + pipe = HybridTrainPipe( + file_root, + file_list, + batch_size, + resize_shorter, + crop, + min_area, + lower, + upper, + interp, + mean, + std, + device_id, + shard_id, + num_shards, + seed=seed + shard_id, + pad_output=pad_output, + output_dtype=output_dtype) + pipe.build() + pipelines = [pipe] + # sample_per_shard = len(pipe) // num_shards + else: + pipe = HybridTrainPipe( + file_root, + file_list, + batch_size, + resize_shorter, + crop, + min_area, + lower, + upper, + interp, + mean, + std, + device_id=device_id, + shard_id=0, + num_shards=1, + seed=seed, + pad_output=pad_output, + output_dtype=output_dtype) + pipe.build() + pipelines = [pipe] + # sample_per_shard = len(pipelines[0]) + return DALIGenericIterator( + pipelines, ['data', 'label'], reader_name='Reader') + else: + resize_shorter = transforms["ResizeImage"].get("resize_short", 256) + crop = transforms["CropImage"]["size"] + if 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env and sampler_name == "DistributedBatchSampler": + shard_id = int(env['PADDLE_TRAINER_ID']) + num_shards = int(env['PADDLE_TRAINERS_NUM']) + device_id = int(env['FLAGS_selected_gpus']) + + pipe = HybridValPipe( + file_root, + file_list, + batch_size, + resize_shorter, + crop, + interp, + mean, + std, + device_id=device_id, + shard_id=shard_id, + num_shards=num_shards, + pad_output=pad_output, + output_dtype=output_dtype) + else: + pipe = HybridValPipe( + file_root, + file_list, + batch_size, + resize_shorter, + crop, + interp, + mean, + std, + device_id=device_id, + pad_output=pad_output, + output_dtype=output_dtype) + pipe.build() + return DALIGenericIterator( + [pipe], ['data', 'label'], reader_name="Reader") diff --git a/exprements/PaddleClas/ppcls/data/dataloader/icartoon_dataset.py b/exprements/PaddleClas/ppcls/data/dataloader/icartoon_dataset.py new file mode 100644 index 00000000..18e3b4b7 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/icartoon_dataset.py @@ -0,0 +1,36 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import numpy as np +import os + +from .common_dataset import CommonDataset + + +class ICartoonDataset(CommonDataset): + def _load_anno(self, seed=None): + assert os.path.exists(self._cls_path) + assert os.path.exists(self._img_root) + self.images = [] + self.labels = [] + + with open(self._cls_path) as fd: + lines = fd.readlines() + for l in lines: + l = l.strip().split("\t") + self.images.append(os.path.join(self._img_root, l[0])) + self.labels.append(np.int64(l[1])) + assert os.path.exists(self.images[-1]) diff --git a/exprements/PaddleClas/ppcls/data/dataloader/imagenet_dataset.py b/exprements/PaddleClas/ppcls/data/dataloader/imagenet_dataset.py new file mode 100644 index 00000000..1166ab38 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/imagenet_dataset.py @@ -0,0 +1,38 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import numpy as np +import os + +from .common_dataset import CommonDataset + + +class ImageNetDataset(CommonDataset): + def _load_anno(self, seed=None): + assert os.path.exists(self._cls_path) + assert os.path.exists(self._img_root) + self.images = [] + self.labels = [] + + with open(self._cls_path) as fd: + lines = fd.readlines() + if seed is not None: + np.random.RandomState(seed).shuffle(lines) + for l in lines: + l = l.strip().split(" ") + self.images.append(os.path.join(self._img_root, l[0])) + self.labels.append(np.int64(l[1])) + assert os.path.exists(self.images[-1]) diff --git a/exprements/PaddleClas/ppcls/data/dataloader/logo_dataset.py b/exprements/PaddleClas/ppcls/data/dataloader/logo_dataset.py new file mode 100644 index 00000000..132ead98 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/logo_dataset.py @@ -0,0 +1,46 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import io +import tarfile +import numpy as np +from PIL import Image #all use default backend + +import paddle +from paddle.io import Dataset +import pickle +import os +import cv2 +import random + +from .common_dataset import CommonDataset + + +class LogoDataset(CommonDataset): + def _load_anno(self): + assert os.path.exists(self._cls_path) + assert os.path.exists(self._img_root) + self.images = [] + self.labels = [] + with open(self._cls_path) as fd: + lines = fd.readlines() + for l in lines: + l = l.strip().split("\t") + if l[0] == 'image_id': + continue + self.images.append(os.path.join(self._img_root, l[3])) + self.labels.append(np.int64(l[1]) - 1) + assert os.path.exists(self.images[-1]) diff --git a/exprements/PaddleClas/ppcls/data/dataloader/mix_dataset.py b/exprements/PaddleClas/ppcls/data/dataloader/mix_dataset.py new file mode 100644 index 00000000..cbf4b402 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/mix_dataset.py @@ -0,0 +1,49 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import numpy as np +import os + +from paddle.io import Dataset +from .. import dataloader + + +class MixDataset(Dataset): + def __init__(self, datasets_config): + super().__init__() + self.dataset_list = [] + start_idx = 0 + end_idx = 0 + for config_i in datasets_config: + dataset_name = config_i.pop('name') + dataset = getattr(dataloader, dataset_name)(**config_i) + end_idx += len(dataset) + self.dataset_list.append([end_idx, start_idx, dataset]) + start_idx = end_idx + + self.length = end_idx + + def __getitem__(self, idx): + for dataset_i in self.dataset_list: + if dataset_i[0] > idx: + dataset_i_idx = idx - dataset_i[1] + return dataset_i[2][dataset_i_idx] + + def __len__(self): + return self.length + + def get_dataset_list(self): + return self.dataset_list diff --git a/exprements/PaddleClas/ppcls/data/dataloader/mix_sampler.py b/exprements/PaddleClas/ppcls/data/dataloader/mix_sampler.py new file mode 100644 index 00000000..2df3109c --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/mix_sampler.py @@ -0,0 +1,79 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division + +from paddle.io import DistributedBatchSampler, Sampler + +from ppcls.utils import logger +from ppcls.data.dataloader.mix_dataset import MixDataset +from ppcls.data import dataloader + + +class MixSampler(DistributedBatchSampler): + def __init__(self, dataset, batch_size, sample_configs, iter_per_epoch): + super().__init__(dataset, batch_size) + assert isinstance(dataset, + MixDataset), "MixSampler only support MixDataset" + self.sampler_list = [] + self.batch_size = batch_size + self.start_list = [] + self.length = iter_per_epoch + dataset_list = dataset.get_dataset_list() + batch_size_left = self.batch_size + self.iter_list = [] + for i, config_i in enumerate(sample_configs): + self.start_list.append(dataset_list[i][1]) + sample_method = config_i.pop("name") + ratio_i = config_i.pop("ratio") + if i < len(sample_configs) - 1: + batch_size_i = int(self.batch_size * ratio_i) + batch_size_left -= batch_size_i + else: + batch_size_i = batch_size_left + assert batch_size_i <= len(dataset_list[i][2]) + config_i["batch_size"] = batch_size_i + if sample_method == "DistributedBatchSampler": + sampler_i = DistributedBatchSampler(dataset_list[i][2], + **config_i) + else: + sampler_i = getattr(dataloader, sample_method)( + dataset_list[i][2], **config_i) + self.sampler_list.append(sampler_i) + self.iter_list.append(iter(sampler_i)) + self.length += len(dataset_list[i][2]) * ratio_i + self.iter_counter = 0 + + def __iter__(self): + while self.iter_counter < self.length: + batch = [] + for i, iter_i in enumerate(self.iter_list): + batch_i = next(iter_i, None) + if batch_i is None: + iter_i = iter(self.sampler_list[i]) + self.iter_list[i] = iter_i + batch_i = next(iter_i, None) + assert batch_i is not None, "dataset {} return None".format( + i) + batch += [idx + self.start_list[i] for idx in batch_i] + if len(batch) == self.batch_size: + self.iter_counter += 1 + yield batch + else: + logger.info("Some dataset reaches end") + self.iter_counter = 0 + + def __len__(self): + return self.length diff --git a/exprements/PaddleClas/ppcls/data/dataloader/multilabel_dataset.py b/exprements/PaddleClas/ppcls/data/dataloader/multilabel_dataset.py new file mode 100644 index 00000000..2c1ed770 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/multilabel_dataset.py @@ -0,0 +1,59 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import numpy as np +import os +import cv2 + +from ppcls.data.preprocess import transform +from ppcls.utils import logger + +from .common_dataset import CommonDataset + + +class MultiLabelDataset(CommonDataset): + def _load_anno(self): + assert os.path.exists(self._cls_path) + assert os.path.exists(self._img_root) + self.images = [] + self.labels = [] + with open(self._cls_path) as fd: + lines = fd.readlines() + for l in lines: + l = l.strip().split("\t") + self.images.append(os.path.join(self._img_root, l[0])) + + labels = l[1].split(',') + labels = [np.int64(i) for i in labels] + + self.labels.append(labels) + assert os.path.exists(self.images[-1]) + + def __getitem__(self, idx): + try: + with open(self.images[idx], 'rb') as f: + img = f.read() + if self._transform_ops: + img = transform(img, self._transform_ops) + img = img.transpose((2, 0, 1)) + label = np.array(self.labels[idx]).astype("float32") + return (img, label) + + except Exception as ex: + logger.error("Exception occured when parse line: {} with msg: {}". + format(self.images[idx], ex)) + rnd_idx = np.random.randint(self.__len__()) + return self.__getitem__(rnd_idx) diff --git a/exprements/PaddleClas/ppcls/data/dataloader/pk_sampler.py b/exprements/PaddleClas/ppcls/data/dataloader/pk_sampler.py new file mode 100644 index 00000000..bf563a6c --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/pk_sampler.py @@ -0,0 +1,105 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from collections import defaultdict +import numpy as np +import random +from paddle.io import DistributedBatchSampler + +from ppcls.utils import logger + + +class PKSampler(DistributedBatchSampler): + """ + First, randomly sample P identities. + Then for each identity randomly sample K instances. + Therefore batch size is P*K, and the sampler called PKSampler. + Args: + dataset (paddle.io.Dataset): list of (img_path, pid, cam_id). + sample_per_id(int): number of instances per identity in a batch. + batch_size (int): number of examples in a batch. + shuffle(bool): whether to shuffle indices order before generating + batch indices. Default False. + """ + + def __init__(self, + dataset, + batch_size, + sample_per_id, + shuffle=True, + drop_last=True, + sample_method="sample_avg_prob"): + super().__init__( + dataset, batch_size, shuffle=shuffle, drop_last=drop_last) + assert batch_size % sample_per_id == 0, \ + "PKSampler configs error, Sample_per_id must be a divisor of batch_size." + assert hasattr(self.dataset, + "labels"), "Dataset must have labels attribute." + self.sample_per_label = sample_per_id + self.label_dict = defaultdict(list) + self.sample_method = sample_method + for idx, label in enumerate(self.dataset.labels): + self.label_dict[label].append(idx) + self.label_list = list(self.label_dict) + assert len(self.label_list) * self.sample_per_label > self.batch_size, \ + "batch size should be smaller than " + if self.sample_method == "id_avg_prob": + self.prob_list = np.array([1 / len(self.label_list)] * + len(self.label_list)) + elif self.sample_method == "sample_avg_prob": + counter = [] + for label_i in self.label_list: + counter.append(len(self.label_dict[label_i])) + self.prob_list = np.array(counter) / sum(counter) + else: + logger.error( + "PKSampler only support id_avg_prob and sample_avg_prob sample method, " + "but receive {}.".format(self.sample_method)) + diff = np.abs(sum(self.prob_list) - 1) + if diff > 0.00000001: + self.prob_list[-1] = 1 - sum(self.prob_list[:-1]) + if self.prob_list[-1] > 1 or self.prob_list[-1] < 0: + logger.error("PKSampler prob list error") + else: + logger.info( + "PKSampler: sum of prob list not equal to 1, diff is {}, change the last prob".format(diff) + ) + + def __iter__(self): + label_per_batch = self.batch_size // self.sample_per_label + for _ in range(len(self)): + batch_index = [] + batch_label_list = np.random.choice( + self.label_list, + size=label_per_batch, + replace=False, + p=self.prob_list) + for label_i in batch_label_list: + label_i_indexes = self.label_dict[label_i] + if self.sample_per_label <= len(label_i_indexes): + batch_index.extend( + np.random.choice( + label_i_indexes, + size=self.sample_per_label, + replace=False)) + else: + batch_index.extend( + np.random.choice( + label_i_indexes, + size=self.sample_per_label, + replace=True)) + if not self.drop_last or len(batch_index) == self.batch_size: + yield batch_index diff --git a/exprements/PaddleClas/ppcls/data/dataloader/vehicle_dataset.py b/exprements/PaddleClas/ppcls/data/dataloader/vehicle_dataset.py new file mode 100644 index 00000000..2981a57a --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/dataloader/vehicle_dataset.py @@ -0,0 +1,138 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import numpy as np +import paddle +from paddle.io import Dataset +import os +import cv2 + +from ppcls.data import preprocess +from ppcls.data.preprocess import transform +from ppcls.utils import logger +from .common_dataset import create_operators + + +class CompCars(Dataset): + def __init__(self, + image_root, + cls_label_path, + label_root=None, + transform_ops=None, + bbox_crop=False): + self._img_root = image_root + self._cls_path = cls_label_path + self._label_root = label_root + if transform_ops: + self._transform_ops = create_operators(transform_ops) + self._bbox_crop = bbox_crop + self._dtype = paddle.get_default_dtype() + self._load_anno() + + def _load_anno(self): + assert os.path.exists(self._cls_path) + assert os.path.exists(self._img_root) + if self._bbox_crop: + assert os.path.exists(self._label_root) + self.images = [] + self.labels = [] + self.bboxes = [] + with open(self._cls_path) as fd: + lines = fd.readlines() + for l in lines: + l = l.strip().split() + if not self._bbox_crop: + self.images.append(os.path.join(self._img_root, l[0])) + self.labels.append(int(l[1])) + else: + label_path = os.path.join(self._label_root, + l[0].split('.')[0] + '.txt') + assert os.path.exists(label_path) + with open(label_path) as f: + bbox = f.readlines()[-1].strip().split() + bbox = [int(x) for x in bbox] + self.images.append(os.path.join(self._img_root, l[0])) + self.labels.append(int(l[1])) + self.bboxes.append(bbox) + assert os.path.exists(self.images[-1]) + + def __getitem__(self, idx): + img = cv2.imread(self.images[idx]) + img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) + if self._bbox_crop: + bbox = self.bboxes[idx] + img = img[bbox[1]:bbox[3], bbox[0]:bbox[2], :] + if self._transform_ops: + img = transform(img, self._transform_ops) + img = img.transpose((2, 0, 1)) + return (img, self.labels[idx]) + + def __len__(self): + return len(self.images) + + @property + def class_num(self): + return len(set(self.labels)) + + +class VeriWild(Dataset): + def __init__( + self, + image_root, + cls_label_path, + transform_ops=None, ): + self._img_root = image_root + self._cls_path = cls_label_path + if transform_ops: + self._transform_ops = create_operators(transform_ops) + self._dtype = paddle.get_default_dtype() + self._load_anno() + + def _load_anno(self): + assert os.path.exists(self._cls_path) + assert os.path.exists(self._img_root) + self.images = [] + self.labels = [] + self.cameras = [] + with open(self._cls_path) as fd: + lines = fd.readlines() + for l in lines: + l = l.strip().split() + self.images.append(os.path.join(self._img_root, l[0])) + self.labels.append(np.int64(l[1])) + self.cameras.append(np.int64(l[2])) + assert os.path.exists(self.images[-1]) + + def __getitem__(self, idx): + try: + with open(self.images[idx], 'rb') as f: + img = f.read() + if self._transform_ops: + img = transform(img, self._transform_ops) + img = img.transpose((2, 0, 1)) + return (img, self.labels[idx], self.cameras[idx]) + except Exception as ex: + logger.error("Exception occured when parse line: {} with msg: {}". + format(self.images[idx], ex)) + rnd_idx = np.random.randint(self.__len__()) + return self.__getitem__(rnd_idx) + + def __len__(self): + return len(self.images) + + @property + def class_num(self): + return len(set(self.labels)) diff --git a/exprements/PaddleClas/ppcls/data/postprocess/__init__.py b/exprements/PaddleClas/ppcls/data/postprocess/__init__.py new file mode 100644 index 00000000..831a4da0 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/postprocess/__init__.py @@ -0,0 +1,41 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import copy +import importlib + +from . import topk + +from .topk import Topk, MultiLabelTopk + + +def build_postprocess(config): + config = copy.deepcopy(config) + model_name = config.pop("name") + mod = importlib.import_module(__name__) + postprocess_func = getattr(mod, model_name)(**config) + return postprocess_func + + +class DistillationPostProcess(object): + def __init__(self, model_name="Student", key=None, func="Topk", **kargs): + super().__init__() + self.func = eval(func)(**kargs) + self.model_name = model_name + self.key = key + + def __call__(self, x, file_names=None): + x = x[self.model_name] + if self.key is not None: + x = x[self.key] + return self.func(x, file_names=file_names) diff --git a/exprements/PaddleClas/ppcls/data/postprocess/topk.py b/exprements/PaddleClas/ppcls/data/postprocess/topk.py new file mode 100644 index 00000000..9c1371bf --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/postprocess/topk.py @@ -0,0 +1,85 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import numpy as np +import paddle +import paddle.nn.functional as F + + +class Topk(object): + def __init__(self, topk=1, class_id_map_file=None): + assert isinstance(topk, (int, )) + self.class_id_map = self.parse_class_id_map(class_id_map_file) + self.topk = topk + + def parse_class_id_map(self, class_id_map_file): + if class_id_map_file is None: + return None + if not os.path.exists(class_id_map_file): + print( + "Warning: If want to use your own label_dict, please input legal path!\nOtherwise label_names will be empty!" + ) + return None + + try: + class_id_map = {} + with open(class_id_map_file, "r") as fin: + lines = fin.readlines() + for line in lines: + partition = line.split("\n")[0].partition(" ") + class_id_map[int(partition[0])] = str(partition[-1]) + except Exception as ex: + print(ex) + class_id_map = None + return class_id_map + + def __call__(self, x, file_names=None, multilabel=False): + assert isinstance(x, paddle.Tensor) + if file_names is not None: + assert x.shape[0] == len(file_names) + x = F.softmax(x, axis=-1) if not multilabel else F.sigmoid(x) + x = x.numpy() + y = [] + for idx, probs in enumerate(x): + index = probs.argsort(axis=0)[-self.topk:][::-1].astype( + "int32") if not multilabel else np.where( + probs >= 0.5)[0].astype("int32") + clas_id_list = [] + score_list = [] + label_name_list = [] + for i in index: + clas_id_list.append(i.item()) + score_list.append(probs[i].item()) + if self.class_id_map is not None: + label_name_list.append(self.class_id_map[i.item()]) + result = { + "class_ids": clas_id_list, + "scores": np.around( + score_list, decimals=5).tolist(), + } + if file_names is not None: + result["file_name"] = file_names[idx] + if label_name_list is not None: + result["label_names"] = label_name_list + y.append(result) + return y + + +class MultiLabelTopk(Topk): + def __init__(self, topk=1, class_id_map_file=None): + super().__init__() + + def __call__(self, x, file_names=None): + return super().__call__(x, file_names, multilabel=True) diff --git a/exprements/PaddleClas/ppcls/data/preprocess/__init__.py b/exprements/PaddleClas/ppcls/data/preprocess/__init__.py new file mode 100644 index 00000000..075ee892 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/__init__.py @@ -0,0 +1,100 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from ppcls.data.preprocess.ops.autoaugment import ImageNetPolicy as RawImageNetPolicy +from ppcls.data.preprocess.ops.randaugment import RandAugment as RawRandAugment +from ppcls.data.preprocess.ops.timm_autoaugment import RawTimmAutoAugment +from ppcls.data.preprocess.ops.cutout import Cutout + +from ppcls.data.preprocess.ops.hide_and_seek import HideAndSeek +from ppcls.data.preprocess.ops.random_erasing import RandomErasing +from ppcls.data.preprocess.ops.grid import GridMask + +from ppcls.data.preprocess.ops.operators import DecodeImage +from ppcls.data.preprocess.ops.operators import ResizeImage +from ppcls.data.preprocess.ops.operators import CropImage +from ppcls.data.preprocess.ops.operators import RandCropImage +from ppcls.data.preprocess.ops.operators import RandFlipImage +from ppcls.data.preprocess.ops.operators import NormalizeImage +from ppcls.data.preprocess.ops.operators import ToCHWImage +from ppcls.data.preprocess.ops.operators import AugMix + +from ppcls.data.preprocess.batch_ops.batch_operators import MixupOperator, CutmixOperator, OpSampler, FmixOperator + +import numpy as np +from PIL import Image + + +def transform(data, ops=[]): + """ transform """ + for op in ops: + data = op(data) + return data + + +class AutoAugment(RawImageNetPolicy): + """ ImageNetPolicy wrapper to auto fit different img types """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + def __call__(self, img): + if not isinstance(img, Image.Image): + img = np.ascontiguousarray(img) + img = Image.fromarray(img) + + img = super().__call__(img) + + if isinstance(img, Image.Image): + img = np.asarray(img) + + return img + + +class RandAugment(RawRandAugment): + """ RandAugment wrapper to auto fit different img types """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + def __call__(self, img): + if not isinstance(img, Image.Image): + img = np.ascontiguousarray(img) + img = Image.fromarray(img) + + img = super().__call__(img) + + if isinstance(img, Image.Image): + img = np.asarray(img) + + return img + + +class TimmAutoAugment(RawTimmAutoAugment): + """ TimmAutoAugment wrapper to auto fit different img tyeps. """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + def __call__(self, img): + if not isinstance(img, Image.Image): + img = np.ascontiguousarray(img) + img = Image.fromarray(img) + + img = super().__call__(img) + + if isinstance(img, Image.Image): + img = np.asarray(img) + + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/batch_ops/__init__.py b/exprements/PaddleClas/ppcls/data/preprocess/batch_ops/__init__.py new file mode 100644 index 00000000..8b137891 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/batch_ops/__init__.py @@ -0,0 +1 @@ + diff --git a/exprements/PaddleClas/ppcls/data/preprocess/batch_ops/batch_operators.py b/exprements/PaddleClas/ppcls/data/preprocess/batch_ops/batch_operators.py new file mode 100644 index 00000000..6f0abb86 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/batch_ops/batch_operators.py @@ -0,0 +1,231 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +from __future__ import unicode_literals +import random + +import numpy as np + +from ppcls.utils import logger +from ppcls.data.preprocess.ops.fmix import sample_mask + + +class BatchOperator(object): + """ BatchOperator """ + + def __init__(self, *args, **kwargs): + pass + + def _unpack(self, batch): + """ _unpack """ + assert isinstance(batch, list), \ + 'batch should be a list filled with tuples (img, label)' + bs = len(batch) + assert bs > 0, 'size of the batch data should > 0' + #imgs, labels = list(zip(*batch)) + imgs = [] + labels = [] + for item in batch: + imgs.append(item[0]) + labels.append(item[1]) + return np.array(imgs), np.array(labels), bs + + def _one_hot(self, targets): + return np.eye(self.class_num, dtype="float32")[targets] + + def _mix_target(self, targets0, targets1, lam): + one_hots0 = self._one_hot(targets0) + one_hots1 = self._one_hot(targets1) + return one_hots0 * lam + one_hots1 * (1 - lam) + + def __call__(self, batch): + return batch + + +class MixupOperator(BatchOperator): + """ Mixup operator + reference: https://arxiv.org/abs/1710.09412 + + """ + + def __init__(self, class_num, alpha: float=1.): + """Build Mixup operator + + Args: + alpha (float, optional): The parameter alpha of mixup. Defaults to 1.. + + Raises: + Exception: The value of parameter is illegal. + """ + if alpha <= 0: + raise Exception( + f"Parameter \"alpha\" of Mixup should be greater than 0. \"alpha\": {alpha}." + ) + if not class_num: + msg = "Please set \"Arch.class_num\" in config if use \"MixupOperator\"." + logger.error(Exception(msg)) + raise Exception(msg) + + self._alpha = alpha + self.class_num = class_num + + def __call__(self, batch): + imgs, labels, bs = self._unpack(batch) + idx = np.random.permutation(bs) + lam = np.random.beta(self._alpha, self._alpha) + imgs = lam * imgs + (1 - lam) * imgs[idx] + targets = self._mix_target(labels, labels[idx], lam) + return list(zip(imgs, targets)) + + +class CutmixOperator(BatchOperator): + """ Cutmix operator + reference: https://arxiv.org/abs/1905.04899 + + """ + + def __init__(self, class_num, alpha=0.2): + """Build Cutmix operator + + Args: + alpha (float, optional): The parameter alpha of cutmix. Defaults to 0.2. + + Raises: + Exception: The value of parameter is illegal. + """ + if alpha <= 0: + raise Exception( + f"Parameter \"alpha\" of Cutmix should be greater than 0. \"alpha\": {alpha}." + ) + if not class_num: + msg = "Please set \"Arch.class_num\" in config if use \"CutmixOperator\"." + logger.error(Exception(msg)) + raise Exception(msg) + + self._alpha = alpha + self.class_num = class_num + + def _rand_bbox(self, size, lam): + """ _rand_bbox """ + w = size[2] + h = size[3] + cut_rat = np.sqrt(1. - lam) + cut_w = int(w * cut_rat) + cut_h = int(h * cut_rat) + + # uniform + cx = np.random.randint(w) + cy = np.random.randint(h) + + bbx1 = np.clip(cx - cut_w // 2, 0, w) + bby1 = np.clip(cy - cut_h // 2, 0, h) + bbx2 = np.clip(cx + cut_w // 2, 0, w) + bby2 = np.clip(cy + cut_h // 2, 0, h) + + return bbx1, bby1, bbx2, bby2 + + def __call__(self, batch): + imgs, labels, bs = self._unpack(batch) + idx = np.random.permutation(bs) + lam = np.random.beta(self._alpha, self._alpha) + + bbx1, bby1, bbx2, bby2 = self._rand_bbox(imgs.shape, lam) + imgs[:, :, bbx1:bbx2, bby1:bby2] = imgs[idx, :, bbx1:bbx2, bby1:bby2] + lam = 1 - (float(bbx2 - bbx1) * (bby2 - bby1) / + (imgs.shape[-2] * imgs.shape[-1])) + targets = self._mix_target(labels, labels[idx], lam) + return list(zip(imgs, targets)) + + +class FmixOperator(BatchOperator): + """ Fmix operator + reference: https://arxiv.org/abs/2002.12047 + + """ + + def __init__(self, + class_num, + alpha=1, + decay_power=3, + max_soft=0., + reformulate=False): + if not class_num: + msg = "Please set \"Arch.class_num\" in config if use \"FmixOperator\"." + logger.error(Exception(msg)) + raise Exception(msg) + + self._alpha = alpha + self._decay_power = decay_power + self._max_soft = max_soft + self._reformulate = reformulate + self.class_num = class_num + + def __call__(self, batch): + imgs, labels, bs = self._unpack(batch) + idx = np.random.permutation(bs) + size = (imgs.shape[2], imgs.shape[3]) + lam, mask = sample_mask(self._alpha, self._decay_power, \ + size, self._max_soft, self._reformulate) + imgs = mask * imgs + (1 - mask) * imgs[idx] + targets = self._mix_target(labels, labels[idx], lam) + return list(zip(imgs, targets)) + + +class OpSampler(object): + """ Sample a operator from """ + + def __init__(self, class_num, **op_dict): + """Build OpSampler + + Raises: + Exception: The parameter \"prob\" of operator(s) are be set error. + """ + if not class_num: + msg = "Please set \"Arch.class_num\" in config if use \"OpSampler\"." + logger.error(Exception(msg)) + raise Exception(msg) + + if len(op_dict) < 1: + msg = f"ConfigWarning: No operator in \"OpSampler\". \"OpSampler\" has been skipped." + logger.warning(msg) + + self.ops = {} + total_prob = 0 + for op_name in op_dict: + param = op_dict[op_name] + if "prob" not in param: + msg = f"ConfigWarning: Parameter \"prob\" should be set when use operator in \"OpSampler\". The operator \"{op_name}\"'s prob has been set \"0\"." + logger.warning(msg) + prob = param.pop("prob", 0) + total_prob += prob + param.update({"class_num": class_num}) + op = eval(op_name)(**param) + self.ops.update({op: prob}) + + if total_prob > 1: + msg = f"ConfigError: The total prob of operators in \"OpSampler\" should be less 1." + logger.error(Exception(msg)) + raise Exception(msg) + + # add "None Op" when total_prob < 1, "None Op" do nothing + self.ops[None] = 1 - total_prob + + def __call__(self, batch): + op = random.choices( + list(self.ops.keys()), weights=list(self.ops.values()), k=1)[0] + # return batch directly when None Op + return op(batch) if op else batch diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/__init__.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/__init__.py new file mode 100644 index 00000000..8b137891 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/__init__.py @@ -0,0 +1 @@ + diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/autoaugment.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/autoaugment.py new file mode 100644 index 00000000..330220a9 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/autoaugment.py @@ -0,0 +1,264 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# This code is based on https://github.com/DeepVoltaire/AutoAugment/blob/master/autoaugment.py + +from PIL import Image, ImageEnhance, ImageOps +import numpy as np +import random + + +class ImageNetPolicy(object): + """ Randomly choose one of the best 24 Sub-policies on ImageNet. + + Example: + >>> policy = ImageNetPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> ImageNetPolicy(), + >>> transforms.ToTensor()]) + """ + + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.4, "posterize", 8, 0.6, "rotate", 9, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "posterize", 7, 0.6, "posterize", 6, fillcolor), + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + SubPolicy(0.4, "equalize", 4, 0.8, "rotate", 8, fillcolor), + SubPolicy(0.6, "solarize", 3, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.8, "posterize", 5, 1.0, "equalize", 2, fillcolor), + SubPolicy(0.2, "rotate", 3, 0.6, "solarize", 8, fillcolor), + SubPolicy(0.6, "equalize", 8, 0.4, "posterize", 6, fillcolor), + SubPolicy(0.8, "rotate", 8, 0.4, "color", 0, fillcolor), + SubPolicy(0.4, "rotate", 9, 0.6, "equalize", 2, fillcolor), + SubPolicy(0.0, "equalize", 7, 0.8, "equalize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + SubPolicy(0.8, "rotate", 8, 1.0, "color", 2, fillcolor), + SubPolicy(0.8, "color", 8, 0.8, "solarize", 7, fillcolor), + SubPolicy(0.4, "sharpness", 7, 0.6, "invert", 8, fillcolor), + SubPolicy(0.6, "shearX", 5, 1.0, "equalize", 9, fillcolor), + SubPolicy(0.4, "color", 0, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor) + ] + + def __call__(self, img, policy_idx=None): + if policy_idx is None or not isinstance(policy_idx, int): + policy_idx = random.randint(0, len(self.policies) - 1) + else: + policy_idx = policy_idx % len(self.policies) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment ImageNet Policy" + + +class CIFAR10Policy(object): + """ Randomly choose one of the best 25 Sub-policies on CIFAR10. + + Example: + >>> policy = CIFAR10Policy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> CIFAR10Policy(), + >>> transforms.ToTensor()]) + """ + + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.1, "invert", 7, 0.2, "contrast", 6, fillcolor), + SubPolicy(0.7, "rotate", 2, 0.3, "translateX", 9, fillcolor), + SubPolicy(0.8, "sharpness", 1, 0.9, "sharpness", 3, fillcolor), + SubPolicy(0.5, "shearY", 8, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.5, "autocontrast", 8, 0.9, "equalize", 2, fillcolor), + SubPolicy(0.2, "shearY", 7, 0.3, "posterize", 7, fillcolor), + SubPolicy(0.4, "color", 3, 0.6, "brightness", 7, fillcolor), + SubPolicy(0.3, "sharpness", 9, 0.7, "brightness", 9, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.5, "equalize", 1, fillcolor), + SubPolicy(0.6, "contrast", 7, 0.6, "sharpness", 5, fillcolor), + SubPolicy(0.7, "color", 7, 0.5, "translateX", 8, fillcolor), + SubPolicy(0.3, "equalize", 7, 0.4, "autocontrast", 8, fillcolor), + SubPolicy(0.4, "translateY", 3, 0.2, "sharpness", 6, fillcolor), + SubPolicy(0.9, "brightness", 6, 0.2, "color", 8, fillcolor), + SubPolicy(0.5, "solarize", 2, 0.0, "invert", 3, fillcolor), + SubPolicy(0.2, "equalize", 0, 0.6, "autocontrast", 0, fillcolor), + SubPolicy(0.2, "equalize", 8, 0.8, "equalize", 4, fillcolor), + SubPolicy(0.9, "color", 9, 0.6, "equalize", 6, fillcolor), + SubPolicy(0.8, "autocontrast", 4, 0.2, "solarize", 8, fillcolor), + SubPolicy(0.1, "brightness", 3, 0.7, "color", 0, fillcolor), + SubPolicy(0.4, "solarize", 5, 0.9, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "translateY", 9, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.9, "autocontrast", 2, 0.8, "solarize", 3, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.1, "invert", 3, fillcolor), + SubPolicy(0.7, "translateY", 9, 0.9, "autocontrast", 1, fillcolor) + ] + + def __call__(self, img, policy_idx=None): + if policy_idx is None or not isinstance(policy_idx, int): + policy_idx = random.randint(0, len(self.policies) - 1) + else: + policy_idx = policy_idx % len(self.policies) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment CIFAR10 Policy" + + +class SVHNPolicy(object): + """ Randomly choose one of the best 25 Sub-policies on SVHN. + + Example: + >>> policy = SVHNPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> SVHNPolicy(), + >>> transforms.ToTensor()]) + """ + + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.9, "shearX", 4, 0.2, "invert", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.7, "invert", 5, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.6, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 3, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "equalize", 1, 0.9, "rotate", 3, fillcolor), + SubPolicy(0.9, "shearX", 4, 0.8, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.4, "invert", 5, fillcolor), + SubPolicy(0.9, "shearY", 5, 0.2, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 6, 0.8, "autocontrast", 1, fillcolor), + SubPolicy(0.6, "equalize", 3, 0.9, "rotate", 3, fillcolor), + SubPolicy(0.9, "shearX", 4, 0.3, "solarize", 3, fillcolor), + SubPolicy(0.8, "shearY", 8, 0.7, "invert", 4, fillcolor), + SubPolicy(0.9, "equalize", 5, 0.6, "translateY", 6, fillcolor), + SubPolicy(0.9, "invert", 4, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.3, "contrast", 3, 0.8, "rotate", 4, fillcolor), + SubPolicy(0.8, "invert", 5, 0.0, "translateY", 2, fillcolor), + SubPolicy(0.7, "shearY", 6, 0.4, "solarize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 0.8, "rotate", 4, fillcolor), + SubPolicy( + 0.3, "shearY", 7, 0.9, "translateX", 3, fillcolor), SubPolicy( + 0.1, "shearX", 6, 0.6, "invert", 5, fillcolor), SubPolicy( + 0.7, "solarize", 2, 0.6, "translateY", 7, + fillcolor), SubPolicy(0.8, "shearY", 4, 0.8, "invert", + 8, fillcolor), SubPolicy( + 0.7, "shearX", 9, 0.8, + "translateY", 3, + fillcolor), SubPolicy( + 0.8, "shearY", 5, 0.7, + "autocontrast", 3, + fillcolor), + SubPolicy(0.7, "shearX", 2, 0.1, "invert", 5, fillcolor) + ] + + def __call__(self, img, policy_idx=None): + if policy_idx is None or not isinstance(policy_idx, int): + policy_idx = random.randint(0, len(self.policies) - 1) + else: + policy_idx = policy_idx % len(self.policies) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment SVHN Policy" + + +class SubPolicy(object): + def __init__(self, + p1, + operation1, + magnitude_idx1, + p2, + operation2, + magnitude_idx2, + fillcolor=(128, 128, 128)): + ranges = { + "shearX": np.linspace(0, 0.3, 10), + "shearY": np.linspace(0, 0.3, 10), + "translateX": np.linspace(0, 150 / 331, 10), + "translateY": np.linspace(0, 150 / 331, 10), + "rotate": np.linspace(0, 30, 10), + "color": np.linspace(0.0, 0.9, 10), + "posterize": np.round(np.linspace(8, 4, 10), 0).astype(np.int), + "solarize": np.linspace(256, 0, 10), + "contrast": np.linspace(0.0, 0.9, 10), + "sharpness": np.linspace(0.0, 0.9, 10), + "brightness": np.linspace(0.0, 0.9, 10), + "autocontrast": [0] * 10, + "equalize": [0] * 10, + "invert": [0] * 10 + } + + # from https://stackoverflow.com/questions/5252170/specify-image-filling-color-when-rotating-in-python-with-pil-and-setting-expand + def rotate_with_fill(img, magnitude): + rot = img.convert("RGBA").rotate(magnitude) + return Image.composite(rot, + Image.new("RGBA", rot.size, (128, ) * 4), + rot).convert(img.mode) + + func = { + "shearX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, magnitude * random.choice([-1, 1]), 0, 0, 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "shearY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, magnitude * random.choice([-1, 1]), 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "translateX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, magnitude * img.size[0] * random.choice([-1, 1]), 0, 1, 0), + fillcolor=fillcolor), + "translateY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, 0, 1, magnitude * img.size[1] * random.choice([-1, 1])), + fillcolor=fillcolor), + "rotate": lambda img, magnitude: rotate_with_fill(img, magnitude), + # "rotate": lambda img, magnitude: img.rotate(magnitude * random.choice([-1, 1])), + "color": lambda img, magnitude: ImageEnhance.Color(img).enhance(1 + magnitude * random.choice([-1, 1])), + "posterize": lambda img, magnitude: ImageOps.posterize(img, magnitude), + "solarize": lambda img, magnitude: ImageOps.solarize(img, magnitude), + "contrast": lambda img, magnitude: ImageEnhance.Contrast(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "sharpness": lambda img, magnitude: ImageEnhance.Sharpness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "brightness": lambda img, magnitude: ImageEnhance.Brightness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "autocontrast": lambda img, magnitude: ImageOps.autocontrast(img), + "equalize": lambda img, magnitude: ImageOps.equalize(img), + "invert": lambda img, magnitude: ImageOps.invert(img) + } + + self.p1 = p1 + self.operation1 = func[operation1] + self.magnitude1 = ranges[operation1][magnitude_idx1] + self.p2 = p2 + self.operation2 = func[operation2] + self.magnitude2 = ranges[operation2][magnitude_idx2] + + def __call__(self, img): + if random.random() < self.p1: + img = self.operation1(img, self.magnitude1) + if random.random() < self.p2: + img = self.operation2(img, self.magnitude2) + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/cutout.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/cutout.py new file mode 100644 index 00000000..b906e145 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/cutout.py @@ -0,0 +1,41 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# This code is based on https://github.com/uoguelph-mlrg/Cutout + +import numpy as np +import random + + +class Cutout(object): + def __init__(self, n_holes=1, length=112): + self.n_holes = n_holes + self.length = length + + def __call__(self, img): + """ cutout_image """ + h, w = img.shape[:2] + mask = np.ones((h, w), np.float32) + + for n in range(self.n_holes): + y = np.random.randint(h) + x = np.random.randint(w) + + y1 = np.clip(y - self.length // 2, 0, h) + y2 = np.clip(y + self.length // 2, 0, h) + x1 = np.clip(x - self.length // 2, 0, w) + x2 = np.clip(x + self.length // 2, 0, w) + + img[y1:y2, x1:x2] = 0 + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/fmix.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/fmix.py new file mode 100644 index 00000000..dc2ef912 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/fmix.py @@ -0,0 +1,217 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +import random + +import numpy as np +from scipy.stats import beta + + +def fftfreqnd(h, w=None, z=None): + """ Get bin values for discrete fourier transform of size (h, w, z) + + :param h: Required, first dimension size + :param w: Optional, second dimension size + :param z: Optional, third dimension size + """ + fz = fx = 0 + fy = np.fft.fftfreq(h) + + if w is not None: + fy = np.expand_dims(fy, -1) + + if w % 2 == 1: + fx = np.fft.fftfreq(w)[:w // 2 + 2] + else: + fx = np.fft.fftfreq(w)[:w // 2 + 1] + + if z is not None: + fy = np.expand_dims(fy, -1) + if z % 2 == 1: + fz = np.fft.fftfreq(z)[:, None] + else: + fz = np.fft.fftfreq(z)[:, None] + + return np.sqrt(fx * fx + fy * fy + fz * fz) + + +def get_spectrum(freqs, decay_power, ch, h, w=0, z=0): + """ Samples a fourier image with given size and frequencies decayed by decay power + + :param freqs: Bin values for the discrete fourier transform + :param decay_power: Decay power for frequency decay prop 1/f**d + :param ch: Number of channels for the resulting mask + :param h: Required, first dimension size + :param w: Optional, second dimension size + :param z: Optional, third dimension size + """ + scale = np.ones(1) / (np.maximum(freqs, np.array([1. / max(w, h, z)])) + **decay_power) + + param_size = [ch] + list(freqs.shape) + [2] + param = np.random.randn(*param_size) + + scale = np.expand_dims(scale, -1)[None, :] + + return scale * param + + +def make_low_freq_image(decay, shape, ch=1): + """ Sample a low frequency image from fourier space + + :param decay_power: Decay power for frequency decay prop 1/f**d + :param shape: Shape of desired mask, list up to 3 dims + :param ch: Number of channels for desired mask + """ + freqs = fftfreqnd(*shape) + spectrum = get_spectrum(freqs, decay, ch, + *shape) #.reshape((1, *shape[:-1], -1)) + spectrum = spectrum[:, 0] + 1j * spectrum[:, 1] + mask = np.real(np.fft.irfftn(spectrum, shape)) + + if len(shape) == 1: + mask = mask[:1, :shape[0]] + if len(shape) == 2: + mask = mask[:1, :shape[0], :shape[1]] + if len(shape) == 3: + mask = mask[:1, :shape[0], :shape[1], :shape[2]] + + mask = mask + mask = (mask - mask.min()) + mask = mask / mask.max() + return mask + + +def sample_lam(alpha, reformulate=False): + """ Sample a lambda from symmetric beta distribution with given alpha + + :param alpha: Alpha value for beta distribution + :param reformulate: If True, uses the reformulation of [1]. + """ + if reformulate: + lam = beta.rvs(alpha + 1, alpha) + else: + lam = beta.rvs(alpha, alpha) + + return lam + + +def binarise_mask(mask, lam, in_shape, max_soft=0.0): + """ Binarises a given low frequency image such that it has mean lambda. + + :param mask: Low frequency image, usually the result of `make_low_freq_image` + :param lam: Mean value of final mask + :param in_shape: Shape of inputs + :param max_soft: Softening value between 0 and 0.5 which smooths hard edges in the mask. + :return: + """ + idx = mask.reshape(-1).argsort()[::-1] + mask = mask.reshape(-1) + num = math.ceil(lam * mask.size) if random.random() > 0.5 else math.floor( + lam * mask.size) + + eff_soft = max_soft + if max_soft > lam or max_soft > (1 - lam): + eff_soft = min(lam, 1 - lam) + + soft = int(mask.size * eff_soft) + num_low = int(num - soft) + num_high = int(num + soft) + + mask[idx[:num_high]] = 1 + mask[idx[num_low:]] = 0 + mask[idx[num_low:num_high]] = np.linspace(1, 0, (num_high - num_low)) + + mask = mask.reshape((1, 1, in_shape[0], in_shape[1])) + return mask + + +def sample_mask(alpha, decay_power, shape, max_soft=0.0, reformulate=False): + """ Samples a mean lambda from beta distribution parametrised by alpha, creates a low frequency image and binarises + it based on this lambda + + :param alpha: Alpha value for beta distribution from which to sample mean of mask + :param decay_power: Decay power for frequency decay prop 1/f**d + :param shape: Shape of desired mask, list up to 3 dims + :param max_soft: Softening value between 0 and 0.5 which smooths hard edges in the mask. + :param reformulate: If True, uses the reformulation of [1]. + """ + if isinstance(shape, int): + shape = (shape, ) + + # Choose lambda + lam = sample_lam(alpha, reformulate) + + # Make mask, get mean / std + mask = make_low_freq_image(decay_power, shape) + mask = binarise_mask(mask, lam, shape, max_soft) + + return float(lam), mask + + +def sample_and_apply(x, + alpha, + decay_power, + shape, + max_soft=0.0, + reformulate=False): + """ + + :param x: Image batch on which to apply fmix of shape [b, c, shape*] + :param alpha: Alpha value for beta distribution from which to sample mean of mask + :param decay_power: Decay power for frequency decay prop 1/f**d + :param shape: Shape of desired mask, list up to 3 dims + :param max_soft: Softening value between 0 and 0.5 which smooths hard edges in the mask. + :param reformulate: If True, uses the reformulation of [1]. + :return: mixed input, permutation indices, lambda value of mix, + """ + lam, mask = sample_mask(alpha, decay_power, shape, max_soft, reformulate) + index = np.random.permutation(x.shape[0]) + + x1, x2 = x * mask, x[index] * (1 - mask) + return x1 + x2, index, lam + + +class FMixBase: + """ FMix augmentation + + Args: + decay_power (float): Decay power for frequency decay prop 1/f**d + alpha (float): Alpha value for beta distribution from which to sample mean of mask + size ([int] | [int, int] | [int, int, int]): Shape of desired mask, list up to 3 dims + max_soft (float): Softening value between 0 and 0.5 which smooths hard edges in the mask. + reformulate (bool): If True, uses the reformulation of [1]. + """ + + def __init__(self, + decay_power=3, + alpha=1, + size=(32, 32), + max_soft=0.0, + reformulate=False): + super().__init__() + self.decay_power = decay_power + self.reformulate = reformulate + self.size = size + self.alpha = alpha + self.max_soft = max_soft + self.index = None + self.lam = None + + def __call__(self, x): + raise NotImplementedError + + def loss(self, *args, **kwargs): + raise NotImplementedError diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/functional.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/functional.py new file mode 100644 index 00000000..9f1369ee --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/functional.py @@ -0,0 +1,138 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# encoding: utf-8 + +import numpy as np +from PIL import Image, ImageOps, ImageEnhance + + + +def int_parameter(level, maxval): + """Helper function to scale `val` between 0 and maxval . + Args: + level: Level of the operation that will be between [0, `PARAMETER_MAX`]. + maxval: Maximum value that the operation can have. This will be scaled to + level/PARAMETER_MAX. + Returns: + An int that results from scaling `maxval` according to `level`. + """ + return int(level * maxval / 10) + + +def float_parameter(level, maxval): + """Helper function to scale `val` between 0 and maxval. + Args: + level: Level of the operation that will be between [0, `PARAMETER_MAX`]. + maxval: Maximum value that the operation can have. This will be scaled to + level/PARAMETER_MAX. + Returns: + A float that results from scaling `maxval` according to `level`. + """ + return float(level) * maxval / 10. + + +def sample_level(n): + return np.random.uniform(low=0.1, high=n) + + +def autocontrast(pil_img, *args): + return ImageOps.autocontrast(pil_img) + + +def equalize(pil_img, *args): + return ImageOps.equalize(pil_img) + + +def posterize(pil_img, level, *args): + level = int_parameter(sample_level(level), 4) + return ImageOps.posterize(pil_img, 4 - level) + + +def rotate(pil_img, level, *args): + degrees = int_parameter(sample_level(level), 30) + if np.random.uniform() > 0.5: + degrees = -degrees + return pil_img.rotate(degrees, resample=Image.BILINEAR) + + +def solarize(pil_img, level, *args): + level = int_parameter(sample_level(level), 256) + return ImageOps.solarize(pil_img, 256 - level) + + +def shear_x(pil_img, level): + level = float_parameter(sample_level(level), 0.3) + if np.random.uniform() > 0.5: + level = -level + return pil_img.transform(pil_img.size, + Image.AFFINE, (1, level, 0, 0, 1, 0), + resample=Image.BILINEAR) + + +def shear_y(pil_img, level): + level = float_parameter(sample_level(level), 0.3) + if np.random.uniform() > 0.5: + level = -level + return pil_img.transform(pil_img.size, + Image.AFFINE, (1, 0, 0, level, 1, 0), + resample=Image.BILINEAR) + + +def translate_x(pil_img, level): + level = int_parameter(sample_level(level), pil_img.size[0] / 3) + if np.random.random() > 0.5: + level = -level + return pil_img.transform(pil_img.size, + Image.AFFINE, (1, 0, level, 0, 1, 0), + resample=Image.BILINEAR) + + +def translate_y(pil_img, level): + level = int_parameter(sample_level(level), pil_img.size[1] / 3) + if np.random.random() > 0.5: + level = -level + return pil_img.transform(pil_img.size, + Image.AFFINE, (1, 0, 0, 0, 1, level), + resample=Image.BILINEAR) + + +# operation that overlaps with ImageNet-C's test set +def color(pil_img, level, *args): + level = float_parameter(sample_level(level), 1.8) + 0.1 + return ImageEnhance.Color(pil_img).enhance(level) + + +# operation that overlaps with ImageNet-C's test set +def contrast(pil_img, level, *args): + level = float_parameter(sample_level(level), 1.8) + 0.1 + return ImageEnhance.Contrast(pil_img).enhance(level) + + +# operation that overlaps with ImageNet-C's test set +def brightness(pil_img, level, *args): + level = float_parameter(sample_level(level), 1.8) + 0.1 + return ImageEnhance.Brightness(pil_img).enhance(level) + + +# operation that overlaps with ImageNet-C's test set +def sharpness(pil_img, level, *args): + level = float_parameter(sample_level(level), 1.8) + 0.1 + return ImageEnhance.Sharpness(pil_img).enhance(level) + + +augmentations = [ + autocontrast, equalize, posterize, rotate, solarize, shear_x, shear_y, + translate_x, translate_y +] diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/grid.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/grid.py new file mode 100644 index 00000000..6f0b2dc8 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/grid.py @@ -0,0 +1,89 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# This code is based on https://github.com/akuxcw/GridMask + +import numpy as np +from PIL import Image +import pdb + +# curr +CURR_EPOCH = 0 +# epoch for the prob to be the upper limit +NUM_EPOCHS = 240 + + +class GridMask(object): + def __init__(self, d1=96, d2=224, rotate=1, ratio=0.5, mode=0, prob=1.): + self.d1 = d1 + self.d2 = d2 + self.rotate = rotate + self.ratio = ratio + self.mode = mode + self.st_prob = prob + self.prob = prob + self.last_prob = -1 + + def set_prob(self): + global CURR_EPOCH + global NUM_EPOCHS + self.prob = self.st_prob * min(1, 1.0 * CURR_EPOCH / NUM_EPOCHS) + + def __call__(self, img): + self.set_prob() + if abs(self.last_prob - self.prob) > 1e-10: + global CURR_EPOCH + global NUM_EPOCHS + print( + "self.prob is updated, self.prob={}, CURR_EPOCH: {}, NUM_EPOCHS: {}". + format(self.prob, CURR_EPOCH, NUM_EPOCHS)) + self.last_prob = self.prob + # print("CURR_EPOCH: {}, NUM_EPOCHS: {}, self.prob is set as: {}".format(CURR_EPOCH, NUM_EPOCHS, self.prob) ) + if np.random.rand() > self.prob: + return img + _, h, w = img.shape + hh = int(1.5 * h) + ww = int(1.5 * w) + d = np.random.randint(self.d1, self.d2) + #d = self.d + self.l = int(d * self.ratio + 0.5) + mask = np.ones((hh, ww), np.float32) + st_h = np.random.randint(d) + st_w = np.random.randint(d) + for i in range(-1, hh // d + 1): + s = d * i + st_h + t = s + self.l + s = max(min(s, hh), 0) + t = max(min(t, hh), 0) + mask[s:t, :] *= 0 + for i in range(-1, ww // d + 1): + s = d * i + st_w + t = s + self.l + s = max(min(s, ww), 0) + t = max(min(t, ww), 0) + mask[:, s:t] *= 0 + r = np.random.randint(self.rotate) + mask = Image.fromarray(np.uint8(mask)) + mask = mask.rotate(r) + mask = np.asarray(mask) + mask = mask[(hh - h) // 2:(hh - h) // 2 + h, (ww - w) // 2:(ww - w) // + 2 + w] + + if self.mode == 1: + mask = 1 - mask + + mask = np.expand_dims(mask, axis=0) + img = (img * mask).astype(img.dtype) + + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/hide_and_seek.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/hide_and_seek.py new file mode 100644 index 00000000..33f25f26 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/hide_and_seek.py @@ -0,0 +1,44 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# This code is based on https://github.com/kkanshul/Hide-and-Seek + +import numpy as np +import random + + +class HideAndSeek(object): + def __init__(self): + # possible grid size, 0 means no hiding + self.grid_sizes = [0, 16, 32, 44, 56] + # hiding probability + self.hide_prob = 0.5 + + def __call__(self, img): + # randomly choose one grid size + grid_size = np.random.choice(self.grid_sizes) + + _, h, w = img.shape + + # hide the patches + if grid_size == 0: + return img + for x in range(0, w, grid_size): + for y in range(0, h, grid_size): + x_end = min(w, x + grid_size) + y_end = min(h, y + grid_size) + if (random.random() <= self.hide_prob): + img[:, x:x_end, y:y_end] = 0 + + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/operators.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/operators.py new file mode 100644 index 00000000..9cdc58b2 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/operators.py @@ -0,0 +1,384 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +from __future__ import unicode_literals + +from functools import partial +import six +import math +import random +import cv2 +import numpy as np +from PIL import Image +from paddle.vision.transforms import ColorJitter as RawColorJitter + +from .autoaugment import ImageNetPolicy +from .functional import augmentations +from ppcls.utils import logger + + +class UnifiedResize(object): + def __init__(self, interpolation=None, backend="cv2"): + _cv2_interp_from_str = { + 'nearest': cv2.INTER_NEAREST, + 'bilinear': cv2.INTER_LINEAR, + 'area': cv2.INTER_AREA, + 'bicubic': cv2.INTER_CUBIC, + 'lanczos': cv2.INTER_LANCZOS4 + } + _pil_interp_from_str = { + 'nearest': Image.NEAREST, + 'bilinear': Image.BILINEAR, + 'bicubic': Image.BICUBIC, + 'box': Image.BOX, + 'lanczos': Image.LANCZOS, + 'hamming': Image.HAMMING + } + + def _pil_resize(src, size, resample): + pil_img = Image.fromarray(src) + pil_img = pil_img.resize(size, resample) + return np.asarray(pil_img) + + if backend.lower() == "cv2": + if isinstance(interpolation, str): + interpolation = _cv2_interp_from_str[interpolation.lower()] + # compatible with opencv < version 4.4.0 + elif interpolation is None: + interpolation = cv2.INTER_LINEAR + self.resize_func = partial(cv2.resize, interpolation=interpolation) + elif backend.lower() == "pil": + if isinstance(interpolation, str): + interpolation = _pil_interp_from_str[interpolation.lower()] + self.resize_func = partial(_pil_resize, resample=interpolation) + else: + logger.warning( + f"The backend of Resize only support \"cv2\" or \"PIL\". \"f{backend}\" is unavailable. Use \"cv2\" instead." + ) + self.resize_func = cv2.resize + + def __call__(self, src, size): + return self.resize_func(src, size) + + +class OperatorParamError(ValueError): + """ OperatorParamError + """ + pass + + +class DecodeImage(object): + """ decode image """ + + def __init__(self, to_rgb=True, to_np=False, channel_first=False): + self.to_rgb = to_rgb + self.to_np = to_np # to numpy + self.channel_first = channel_first # only enabled when to_np is True + + def __call__(self, img): + if six.PY2: + assert type(img) is str and len( + img) > 0, "invalid input 'img' in DecodeImage" + else: + assert type(img) is bytes and len( + img) > 0, "invalid input 'img' in DecodeImage" + data = np.frombuffer(img, dtype='uint8') + img = cv2.imdecode(data, 1) + if self.to_rgb: + assert img.shape[2] == 3, 'invalid shape of image[%s]' % ( + img.shape) + img = img[:, :, ::-1] + + if self.channel_first: + img = img.transpose((2, 0, 1)) + + return img + + +class ResizeImage(object): + """ resize image """ + + def __init__(self, + size=None, + resize_short=None, + interpolation=None, + backend="cv2"): + if resize_short is not None and resize_short > 0: + self.resize_short = resize_short + self.w = None + self.h = None + elif size is not None: + self.resize_short = None + self.w = size if type(size) is int else size[0] + self.h = size if type(size) is int else size[1] + else: + raise OperatorParamError("invalid params for ReisizeImage for '\ + 'both 'size' and 'resize_short' are None") + + self._resize_func = UnifiedResize( + interpolation=interpolation, backend=backend) + + def __call__(self, img): + img_h, img_w = img.shape[:2] + if self.resize_short is not None: + percent = float(self.resize_short) / min(img_w, img_h) + w = int(round(img_w * percent)) + h = int(round(img_h * percent)) + else: + w = self.w + h = self.h + return self._resize_func(img, (w, h)) + + +class CropImage(object): + """ crop image """ + + def __init__(self, size): + if type(size) is int: + self.size = (size, size) + else: + self.size = size # (h, w) + + def __call__(self, img): + w, h = self.size + img_h, img_w = img.shape[:2] + w_start = (img_w - w) // 2 + h_start = (img_h - h) // 2 + + w_end = w_start + w + h_end = h_start + h + return img[h_start:h_end, w_start:w_end, :] + + +class RandCropImage(object): + """ random crop image """ + + def __init__(self, + size, + scale=None, + ratio=None, + interpolation=None, + backend="cv2"): + if type(size) is int: + self.size = (size, size) # (h, w) + else: + self.size = size + + self.scale = [0.08, 1.0] if scale is None else scale + self.ratio = [3. / 4., 4. / 3.] if ratio is None else ratio + + self._resize_func = UnifiedResize( + interpolation=interpolation, backend=backend) + + def __call__(self, img): + size = self.size + scale = self.scale + ratio = self.ratio + + aspect_ratio = math.sqrt(random.uniform(*ratio)) + w = 1. * aspect_ratio + h = 1. / aspect_ratio + + img_h, img_w = img.shape[:2] + + bound = min((float(img_w) / img_h) / (w**2), + (float(img_h) / img_w) / (h**2)) + scale_max = min(scale[1], bound) + scale_min = min(scale[0], bound) + + target_area = img_w * img_h * random.uniform(scale_min, scale_max) + target_size = math.sqrt(target_area) + w = int(target_size * w) + h = int(target_size * h) + + i = random.randint(0, img_w - w) + j = random.randint(0, img_h - h) + + img = img[j:j + h, i:i + w, :] + + return self._resize_func(img, size) + + +class RandFlipImage(object): + """ random flip image + flip_code: + 1: Flipped Horizontally + 0: Flipped Vertically + -1: Flipped Horizontally & Vertically + """ + + def __init__(self, flip_code=1): + assert flip_code in [-1, 0, 1 + ], "flip_code should be a value in [-1, 0, 1]" + self.flip_code = flip_code + + def __call__(self, img): + if random.randint(0, 1) == 1: + return cv2.flip(img, self.flip_code) + else: + return img + + +class AutoAugment(object): + def __init__(self): + self.policy = ImageNetPolicy() + + def __call__(self, img): + from PIL import Image + img = np.ascontiguousarray(img) + img = Image.fromarray(img) + img = self.policy(img) + img = np.asarray(img) + + +class NormalizeImage(object): + """ normalize image such as substract mean, divide std + """ + + def __init__(self, + scale=None, + mean=None, + std=None, + order='chw', + output_fp16=False, + channel_num=3): + if isinstance(scale, str): + scale = eval(scale) + assert channel_num in [ + 3, 4 + ], "channel number of input image should be set to 3 or 4." + self.channel_num = channel_num + self.output_dtype = 'float16' if output_fp16 else 'float32' + self.scale = np.float32(scale if scale is not None else 1.0 / 255.0) + self.order = order + mean = mean if mean is not None else [0.485, 0.456, 0.406] + std = std if std is not None else [0.229, 0.224, 0.225] + + shape = (3, 1, 1) if self.order == 'chw' else (1, 1, 3) + self.mean = np.array(mean).reshape(shape).astype('float32') + self.std = np.array(std).reshape(shape).astype('float32') + + def __call__(self, img): + from PIL import Image + if isinstance(img, Image.Image): + img = np.array(img) + + assert isinstance(img, + np.ndarray), "invalid input 'img' in NormalizeImage" + + img = (img.astype('float32') * self.scale - self.mean) / self.std + + if self.channel_num == 4: + img_h = img.shape[1] if self.order == 'chw' else img.shape[0] + img_w = img.shape[2] if self.order == 'chw' else img.shape[1] + pad_zeros = np.zeros( + (1, img_h, img_w)) if self.order == 'chw' else np.zeros( + (img_h, img_w, 1)) + img = (np.concatenate( + (img, pad_zeros), axis=0) + if self.order == 'chw' else np.concatenate( + (img, pad_zeros), axis=2)) + return img.astype(self.output_dtype) + + +class ToCHWImage(object): + """ convert hwc image to chw image + """ + + def __init__(self): + pass + + def __call__(self, img): + from PIL import Image + if isinstance(img, Image.Image): + img = np.array(img) + + return img.transpose((2, 0, 1)) + + +class AugMix(object): + """ Perform AugMix augmentation and compute mixture. + """ + + def __init__(self, + prob=0.5, + aug_prob_coeff=0.1, + mixture_width=3, + mixture_depth=1, + aug_severity=1): + """ + Args: + prob: Probability of taking augmix + aug_prob_coeff: Probability distribution coefficients. + mixture_width: Number of augmentation chains to mix per augmented example. + mixture_depth: Depth of augmentation chains. -1 denotes stochastic depth in [1, 3]' + aug_severity: Severity of underlying augmentation operators (between 1 to 10). + """ + # fmt: off + self.prob = prob + self.aug_prob_coeff = aug_prob_coeff + self.mixture_width = mixture_width + self.mixture_depth = mixture_depth + self.aug_severity = aug_severity + self.augmentations = augmentations + # fmt: on + + def __call__(self, image): + """Perform AugMix augmentations and compute mixture. + Returns: + mixed: Augmented and mixed image. + """ + if random.random() > self.prob: + # Avoid the warning: the given NumPy array is not writeable + return np.asarray(image).copy() + + ws = np.float32( + np.random.dirichlet([self.aug_prob_coeff] * self.mixture_width)) + m = np.float32( + np.random.beta(self.aug_prob_coeff, self.aug_prob_coeff)) + + # image = Image.fromarray(image) + mix = np.zeros(image.shape) + for i in range(self.mixture_width): + image_aug = image.copy() + image_aug = Image.fromarray(image_aug) + depth = self.mixture_depth if self.mixture_depth > 0 else np.random.randint( + 1, 4) + for _ in range(depth): + op = np.random.choice(self.augmentations) + image_aug = op(image_aug, self.aug_severity) + mix += ws[i] * np.asarray(image_aug) + + mixed = (1 - m) * image + m * mix + return mixed.astype(np.uint8) + + +class ColorJitter(RawColorJitter): + """ColorJitter. + """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + def __call__(self, img): + if not isinstance(img, Image.Image): + img = np.ascontiguousarray(img) + img = Image.fromarray(img) + img = super()._apply_image(img) + if isinstance(img, Image.Image): + img = np.asarray(img) + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/randaugment.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/randaugment.py new file mode 100644 index 00000000..cca59da4 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/randaugment.py @@ -0,0 +1,106 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# This code is based on https://github.com/heartInsert/randaugment + +from PIL import Image, ImageEnhance, ImageOps +import numpy as np +import random + + +class RandAugment(object): + def __init__(self, num_layers=2, magnitude=5, fillcolor=(128, 128, 128)): + self.num_layers = num_layers + self.magnitude = magnitude + self.max_level = 10 + + abso_level = self.magnitude / self.max_level + self.level_map = { + "shearX": 0.3 * abso_level, + "shearY": 0.3 * abso_level, + "translateX": 150.0 / 331 * abso_level, + "translateY": 150.0 / 331 * abso_level, + "rotate": 30 * abso_level, + "color": 0.9 * abso_level, + "posterize": int(4.0 * abso_level), + "solarize": 256.0 * abso_level, + "contrast": 0.9 * abso_level, + "sharpness": 0.9 * abso_level, + "brightness": 0.9 * abso_level, + "autocontrast": 0, + "equalize": 0, + "invert": 0 + } + + # from https://stackoverflow.com/questions/5252170/ + # specify-image-filling-color-when-rotating-in-python-with-pil-and-setting-expand + def rotate_with_fill(img, magnitude): + rot = img.convert("RGBA").rotate(magnitude) + return Image.composite(rot, + Image.new("RGBA", rot.size, (128, ) * 4), + rot).convert(img.mode) + + rnd_ch_op = random.choice + + self.func = { + "shearX": lambda img, magnitude: img.transform( + img.size, + Image.AFFINE, + (1, magnitude * rnd_ch_op([-1, 1]), 0, 0, 1, 0), + Image.BICUBIC, + fillcolor=fillcolor), + "shearY": lambda img, magnitude: img.transform( + img.size, + Image.AFFINE, + (1, 0, 0, magnitude * rnd_ch_op([-1, 1]), 1, 0), + Image.BICUBIC, + fillcolor=fillcolor), + "translateX": lambda img, magnitude: img.transform( + img.size, + Image.AFFINE, + (1, 0, magnitude * img.size[0] * rnd_ch_op([-1, 1]), 0, 1, 0), + fillcolor=fillcolor), + "translateY": lambda img, magnitude: img.transform( + img.size, + Image.AFFINE, + (1, 0, 0, 0, 1, magnitude * img.size[1] * rnd_ch_op([-1, 1])), + fillcolor=fillcolor), + "rotate": lambda img, magnitude: rotate_with_fill(img, magnitude), + "color": lambda img, magnitude: ImageEnhance.Color(img).enhance( + 1 + magnitude * rnd_ch_op([-1, 1])), + "posterize": lambda img, magnitude: + ImageOps.posterize(img, magnitude), + "solarize": lambda img, magnitude: + ImageOps.solarize(img, magnitude), + "contrast": lambda img, magnitude: + ImageEnhance.Contrast(img).enhance( + 1 + magnitude * rnd_ch_op([-1, 1])), + "sharpness": lambda img, magnitude: + ImageEnhance.Sharpness(img).enhance( + 1 + magnitude * rnd_ch_op([-1, 1])), + "brightness": lambda img, magnitude: + ImageEnhance.Brightness(img).enhance( + 1 + magnitude * rnd_ch_op([-1, 1])), + "autocontrast": lambda img, magnitude: + ImageOps.autocontrast(img), + "equalize": lambda img, magnitude: ImageOps.equalize(img), + "invert": lambda img, magnitude: ImageOps.invert(img) + } + + def __call__(self, img): + avaiable_op_names = list(self.level_map.keys()) + for layer_num in range(self.num_layers): + op_name = np.random.choice(avaiable_op_names) + img = self.func[op_name](img, self.level_map[op_name]) + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/random_erasing.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/random_erasing.py new file mode 100644 index 00000000..f234abbb --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/random_erasing.py @@ -0,0 +1,90 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +#This code is adapted from https://github.com/zhunzhong07/Random-Erasing, and refer to Timm. + +from functools import partial + +import math +import random + +import numpy as np + + +class Pixels(object): + def __init__(self, mode="const", mean=[0., 0., 0.]): + self._mode = mode + self._mean = mean + + def __call__(self, h=224, w=224, c=3): + if self._mode == "rand": + return np.random.normal(size=(1, 1, 3)) + elif self._mode == "pixel": + return np.random.normal(size=(h, w, c)) + elif self._mode == "const": + return self._mean + else: + raise Exception( + "Invalid mode in RandomErasing, only support \"const\", \"rand\", \"pixel\"" + ) + + +class RandomErasing(object): + """RandomErasing. + """ + + def __init__(self, + EPSILON=0.5, + sl=0.02, + sh=0.4, + r1=0.3, + mean=[0., 0., 0.], + attempt=100, + use_log_aspect=False, + mode='const'): + self.EPSILON = eval(EPSILON) if isinstance(EPSILON, str) else EPSILON + self.sl = eval(sl) if isinstance(sl, str) else sl + self.sh = eval(sh) if isinstance(sh, str) else sh + r1 = eval(r1) if isinstance(r1, str) else r1 + self.r1 = (math.log(r1), math.log(1 / r1)) if use_log_aspect else ( + r1, 1 / r1) + self.use_log_aspect = use_log_aspect + self.attempt = attempt + self.get_pixels = Pixels(mode, mean) + + def __call__(self, img): + if random.random() > self.EPSILON: + return img + + for _ in range(self.attempt): + area = img.shape[0] * img.shape[1] + + target_area = random.uniform(self.sl, self.sh) * area + aspect_ratio = random.uniform(*self.r1) + if self.use_log_aspect: + aspect_ratio = math.exp(aspect_ratio) + + h = int(round(math.sqrt(target_area * aspect_ratio))) + w = int(round(math.sqrt(target_area / aspect_ratio))) + + if w < img.shape[1] and h < img.shape[0]: + pixels = self.get_pixels(h, w, img.shape[2]) + x1 = random.randint(0, img.shape[0] - h) + y1 = random.randint(0, img.shape[1] - w) + if img.shape[2] == 3: + img[x1:x1 + h, y1:y1 + w, :] = pixels + else: + img[x1:x1 + h, y1:y1 + w, 0] = pixels[0] + return img + return img diff --git a/exprements/PaddleClas/ppcls/data/preprocess/ops/timm_autoaugment.py b/exprements/PaddleClas/ppcls/data/preprocess/ops/timm_autoaugment.py new file mode 100644 index 00000000..dd2994da --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/preprocess/ops/timm_autoaugment.py @@ -0,0 +1,877 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was heavily based on https://github.com/rwightman/pytorch-image-models + +import random +import math +import re +from PIL import Image, ImageOps, ImageEnhance, ImageChops +import PIL +import numpy as np + +IMAGENET_DEFAULT_MEAN = (0.485, 0.456, 0.406) + +_PIL_VER = tuple([int(x) for x in PIL.__version__.split('.')[:2]]) + +_FILL = (128, 128, 128) + +# This signifies the max integer that the controller RNN could predict for the +# augmentation scheme. +_MAX_LEVEL = 10. + +_HPARAMS_DEFAULT = dict( + translate_const=250, + img_mean=_FILL, ) + +_RANDOM_INTERPOLATION = (Image.BILINEAR, Image.BICUBIC) + + +def _pil_interp(method): + if method == 'bicubic': + return Image.BICUBIC + elif method == 'lanczos': + return Image.LANCZOS + elif method == 'hamming': + return Image.HAMMING + else: + # default bilinear, do we want to allow nearest? + return Image.BILINEAR + + +def _interpolation(kwargs): + interpolation = kwargs.pop('resample', Image.BILINEAR) + if isinstance(interpolation, (list, tuple)): + return random.choice(interpolation) + else: + return interpolation + + +def _check_args_tf(kwargs): + if 'fillcolor' in kwargs and _PIL_VER < (5, 0): + kwargs.pop('fillcolor') + kwargs['resample'] = _interpolation(kwargs) + + +def shear_x(img, factor, **kwargs): + _check_args_tf(kwargs) + return img.transform(img.size, Image.AFFINE, (1, factor, 0, 0, 1, 0), + **kwargs) + + +def shear_y(img, factor, **kwargs): + _check_args_tf(kwargs) + return img.transform(img.size, Image.AFFINE, (1, 0, 0, factor, 1, 0), + **kwargs) + + +def translate_x_rel(img, pct, **kwargs): + pixels = pct * img.size[0] + _check_args_tf(kwargs) + return img.transform(img.size, Image.AFFINE, (1, 0, pixels, 0, 1, 0), + **kwargs) + + +def translate_y_rel(img, pct, **kwargs): + pixels = pct * img.size[1] + _check_args_tf(kwargs) + return img.transform(img.size, Image.AFFINE, (1, 0, 0, 0, 1, pixels), + **kwargs) + + +def translate_x_abs(img, pixels, **kwargs): + _check_args_tf(kwargs) + return img.transform(img.size, Image.AFFINE, (1, 0, pixels, 0, 1, 0), + **kwargs) + + +def translate_y_abs(img, pixels, **kwargs): + _check_args_tf(kwargs) + return img.transform(img.size, Image.AFFINE, (1, 0, 0, 0, 1, pixels), + **kwargs) + + +def rotate(img, degrees, **kwargs): + _check_args_tf(kwargs) + if _PIL_VER >= (5, 2): + return img.rotate(degrees, **kwargs) + elif _PIL_VER >= (5, 0): + w, h = img.size + post_trans = (0, 0) + rotn_center = (w / 2.0, h / 2.0) + angle = -math.radians(degrees) + matrix = [ + round(math.cos(angle), 15), + round(math.sin(angle), 15), + 0.0, + round(-math.sin(angle), 15), + round(math.cos(angle), 15), + 0.0, + ] + + def transform(x, y, matrix): + (a, b, c, d, e, f) = matrix + return a * x + b * y + c, d * x + e * y + f + + matrix[2], matrix[5] = transform(-rotn_center[0] - post_trans[0], + -rotn_center[1] - post_trans[1], + matrix) + matrix[2] += rotn_center[0] + matrix[5] += rotn_center[1] + return img.transform(img.size, Image.AFFINE, matrix, **kwargs) + else: + return img.rotate(degrees, resample=kwargs['resample']) + + +def auto_contrast(img, **__): + return ImageOps.autocontrast(img) + + +def invert(img, **__): + return ImageOps.invert(img) + + +def equalize(img, **__): + return ImageOps.equalize(img) + + +def solarize(img, thresh, **__): + return ImageOps.solarize(img, thresh) + + +def solarize_add(img, add, thresh=128, **__): + lut = [] + for i in range(256): + if i < thresh: + lut.append(min(255, i + add)) + else: + lut.append(i) + if img.mode in ("L", "RGB"): + if img.mode == "RGB" and len(lut) == 256: + lut = lut + lut + lut + return img.point(lut) + else: + return img + + +def posterize(img, bits_to_keep, **__): + if bits_to_keep >= 8: + return img + return ImageOps.posterize(img, bits_to_keep) + + +def contrast(img, factor, **__): + return ImageEnhance.Contrast(img).enhance(factor) + + +def color(img, factor, **__): + return ImageEnhance.Color(img).enhance(factor) + + +def brightness(img, factor, **__): + return ImageEnhance.Brightness(img).enhance(factor) + + +def sharpness(img, factor, **__): + return ImageEnhance.Sharpness(img).enhance(factor) + + +def _randomly_negate(v): + """With 50% prob, negate the value""" + return -v if random.random() > 0.5 else v + + +def _rotate_level_to_arg(level, _hparams): + # range [-30, 30] + level = (level / _MAX_LEVEL) * 30. + level = _randomly_negate(level) + return level, + + +def _enhance_level_to_arg(level, _hparams): + # range [0.1, 1.9] + return (level / _MAX_LEVEL) * 1.8 + 0.1, + + +def _enhance_increasing_level_to_arg(level, _hparams): + # the 'no change' level is 1.0, moving away from that towards 0. or 2.0 increases the enhancement blend + # range [0.1, 1.9] + level = (level / _MAX_LEVEL) * .9 + level = 1.0 + _randomly_negate(level) + return level, + + +def _shear_level_to_arg(level, _hparams): + # range [-0.3, 0.3] + level = (level / _MAX_LEVEL) * 0.3 + level = _randomly_negate(level) + return level, + + +def _translate_abs_level_to_arg(level, hparams): + translate_const = hparams['translate_const'] + level = (level / _MAX_LEVEL) * float(translate_const) + level = _randomly_negate(level) + return level, + + +def _translate_rel_level_to_arg(level, hparams): + # default range [-0.45, 0.45] + translate_pct = hparams.get('translate_pct', 0.45) + level = (level / _MAX_LEVEL) * translate_pct + level = _randomly_negate(level) + return level, + + +def _posterize_level_to_arg(level, _hparams): + # As per Tensorflow TPU EfficientNet impl + # range [0, 4], 'keep 0 up to 4 MSB of original image' + # intensity/severity of augmentation decreases with level + return int((level / _MAX_LEVEL) * 4), + + +def _posterize_increasing_level_to_arg(level, hparams): + # As per Tensorflow models research and UDA impl + # range [4, 0], 'keep 4 down to 0 MSB of original image', + # intensity/severity of augmentation increases with level + return 4 - _posterize_level_to_arg(level, hparams)[0], + + +def _posterize_original_level_to_arg(level, _hparams): + # As per original AutoAugment paper description + # range [4, 8], 'keep 4 up to 8 MSB of image' + # intensity/severity of augmentation decreases with level + return int((level / _MAX_LEVEL) * 4) + 4, + + +def _solarize_level_to_arg(level, _hparams): + # range [0, 256] + # intensity/severity of augmentation decreases with level + return int((level / _MAX_LEVEL) * 256), + + +def _solarize_increasing_level_to_arg(level, _hparams): + # range [0, 256] + # intensity/severity of augmentation increases with level + return 256 - _solarize_level_to_arg(level, _hparams)[0], + + +def _solarize_add_level_to_arg(level, _hparams): + # range [0, 110] + return int((level / _MAX_LEVEL) * 110), + + +LEVEL_TO_ARG = { + 'AutoContrast': None, + 'Equalize': None, + 'Invert': None, + 'Rotate': _rotate_level_to_arg, + # There are several variations of the posterize level scaling in various Tensorflow/Google repositories/papers + 'Posterize': _posterize_level_to_arg, + 'PosterizeIncreasing': _posterize_increasing_level_to_arg, + 'PosterizeOriginal': _posterize_original_level_to_arg, + 'Solarize': _solarize_level_to_arg, + 'SolarizeIncreasing': _solarize_increasing_level_to_arg, + 'SolarizeAdd': _solarize_add_level_to_arg, + 'Color': _enhance_level_to_arg, + 'ColorIncreasing': _enhance_increasing_level_to_arg, + 'Contrast': _enhance_level_to_arg, + 'ContrastIncreasing': _enhance_increasing_level_to_arg, + 'Brightness': _enhance_level_to_arg, + 'BrightnessIncreasing': _enhance_increasing_level_to_arg, + 'Sharpness': _enhance_level_to_arg, + 'SharpnessIncreasing': _enhance_increasing_level_to_arg, + 'ShearX': _shear_level_to_arg, + 'ShearY': _shear_level_to_arg, + 'TranslateX': _translate_abs_level_to_arg, + 'TranslateY': _translate_abs_level_to_arg, + 'TranslateXRel': _translate_rel_level_to_arg, + 'TranslateYRel': _translate_rel_level_to_arg, +} + +NAME_TO_OP = { + 'AutoContrast': auto_contrast, + 'Equalize': equalize, + 'Invert': invert, + 'Rotate': rotate, + 'Posterize': posterize, + 'PosterizeIncreasing': posterize, + 'PosterizeOriginal': posterize, + 'Solarize': solarize, + 'SolarizeIncreasing': solarize, + 'SolarizeAdd': solarize_add, + 'Color': color, + 'ColorIncreasing': color, + 'Contrast': contrast, + 'ContrastIncreasing': contrast, + 'Brightness': brightness, + 'BrightnessIncreasing': brightness, + 'Sharpness': sharpness, + 'SharpnessIncreasing': sharpness, + 'ShearX': shear_x, + 'ShearY': shear_y, + 'TranslateX': translate_x_abs, + 'TranslateY': translate_y_abs, + 'TranslateXRel': translate_x_rel, + 'TranslateYRel': translate_y_rel, +} + + +class AugmentOp(object): + def __init__(self, name, prob=0.5, magnitude=10, hparams=None): + hparams = hparams or _HPARAMS_DEFAULT + self.aug_fn = NAME_TO_OP[name] + self.level_fn = LEVEL_TO_ARG[name] + self.prob = prob + self.magnitude = magnitude + self.hparams = hparams.copy() + self.kwargs = dict( + fillcolor=hparams['img_mean'] if 'img_mean' in hparams else _FILL, + resample=hparams['interpolation'] + if 'interpolation' in hparams else _RANDOM_INTERPOLATION, ) + + # If magnitude_std is > 0, we introduce some randomness + # in the usually fixed policy and sample magnitude from a normal distribution + # with mean `magnitude` and std-dev of `magnitude_std`. + # NOTE This is my own hack, being tested, not in papers or reference impls. + self.magnitude_std = self.hparams.get('magnitude_std', 0) + + def __call__(self, img): + if self.prob < 1.0 and random.random() > self.prob: + return img + magnitude = self.magnitude + if self.magnitude_std and self.magnitude_std > 0: + magnitude = random.gauss(magnitude, self.magnitude_std) + magnitude = min(_MAX_LEVEL, max(0, magnitude)) # clip to valid range + level_args = self.level_fn( + magnitude, self.hparams) if self.level_fn is not None else tuple() + return self.aug_fn(img, *level_args, **self.kwargs) + + +def auto_augment_policy_v0(hparams): + # ImageNet v0 policy from TPU EfficientNet impl, cannot find a paper reference. + policy = [ + [('Equalize', 0.8, 1), ('ShearY', 0.8, 4)], + [('Color', 0.4, 9), ('Equalize', 0.6, 3)], + [('Color', 0.4, 1), ('Rotate', 0.6, 8)], + [('Solarize', 0.8, 3), ('Equalize', 0.4, 7)], + [('Solarize', 0.4, 2), ('Solarize', 0.6, 2)], + [('Color', 0.2, 0), ('Equalize', 0.8, 8)], + [('Equalize', 0.4, 8), ('SolarizeAdd', 0.8, 3)], + [('ShearX', 0.2, 9), ('Rotate', 0.6, 8)], + [('Color', 0.6, 1), ('Equalize', 1.0, 2)], + [('Invert', 0.4, 9), ('Rotate', 0.6, 0)], + [('Equalize', 1.0, 9), ('ShearY', 0.6, 3)], + [('Color', 0.4, 7), ('Equalize', 0.6, 0)], + [('Posterize', 0.4, 6), ('AutoContrast', 0.4, 7)], + [('Solarize', 0.6, 8), ('Color', 0.6, 9)], + [('Solarize', 0.2, 4), ('Rotate', 0.8, 9)], + [('Rotate', 1.0, 7), ('TranslateYRel', 0.8, 9)], + [('ShearX', 0.0, 0), ('Solarize', 0.8, 4)], + [('ShearY', 0.8, 0), ('Color', 0.6, 4)], + [('Color', 1.0, 0), ('Rotate', 0.6, 2)], + [('Equalize', 0.8, 4), ('Equalize', 0.0, 8)], + [('Equalize', 1.0, 4), ('AutoContrast', 0.6, 2)], + [('ShearY', 0.4, 7), ('SolarizeAdd', 0.6, 7)], + [('Posterize', 0.8, 2), ('Solarize', 0.6, 10) + ], # This results in black image with Tpu posterize + [('Solarize', 0.6, 8), ('Equalize', 0.6, 1)], + [('Color', 0.8, 6), ('Rotate', 0.4, 5)], + ] + pc = [[AugmentOp(*a, hparams=hparams) for a in sp] for sp in policy] + return pc + + +def auto_augment_policy_v0r(hparams): + # ImageNet v0 policy from TPU EfficientNet impl, with variation of Posterize used + # in Google research implementation (number of bits discarded increases with magnitude) + policy = [ + [('Equalize', 0.8, 1), ('ShearY', 0.8, 4)], + [('Color', 0.4, 9), ('Equalize', 0.6, 3)], + [('Color', 0.4, 1), ('Rotate', 0.6, 8)], + [('Solarize', 0.8, 3), ('Equalize', 0.4, 7)], + [('Solarize', 0.4, 2), ('Solarize', 0.6, 2)], + [('Color', 0.2, 0), ('Equalize', 0.8, 8)], + [('Equalize', 0.4, 8), ('SolarizeAdd', 0.8, 3)], + [('ShearX', 0.2, 9), ('Rotate', 0.6, 8)], + [('Color', 0.6, 1), ('Equalize', 1.0, 2)], + [('Invert', 0.4, 9), ('Rotate', 0.6, 0)], + [('Equalize', 1.0, 9), ('ShearY', 0.6, 3)], + [('Color', 0.4, 7), ('Equalize', 0.6, 0)], + [('PosterizeIncreasing', 0.4, 6), ('AutoContrast', 0.4, 7)], + [('Solarize', 0.6, 8), ('Color', 0.6, 9)], + [('Solarize', 0.2, 4), ('Rotate', 0.8, 9)], + [('Rotate', 1.0, 7), ('TranslateYRel', 0.8, 9)], + [('ShearX', 0.0, 0), ('Solarize', 0.8, 4)], + [('ShearY', 0.8, 0), ('Color', 0.6, 4)], + [('Color', 1.0, 0), ('Rotate', 0.6, 2)], + [('Equalize', 0.8, 4), ('Equalize', 0.0, 8)], + [('Equalize', 1.0, 4), ('AutoContrast', 0.6, 2)], + [('ShearY', 0.4, 7), ('SolarizeAdd', 0.6, 7)], + [('PosterizeIncreasing', 0.8, 2), ('Solarize', 0.6, 10)], + [('Solarize', 0.6, 8), ('Equalize', 0.6, 1)], + [('Color', 0.8, 6), ('Rotate', 0.4, 5)], + ] + pc = [[AugmentOp(*a, hparams=hparams) for a in sp] for sp in policy] + return pc + + +def auto_augment_policy_original(hparams): + # ImageNet policy from https://arxiv.org/abs/1805.09501 + policy = [ + [('PosterizeOriginal', 0.4, 8), ('Rotate', 0.6, 9)], + [('Solarize', 0.6, 5), ('AutoContrast', 0.6, 5)], + [('Equalize', 0.8, 8), ('Equalize', 0.6, 3)], + [('PosterizeOriginal', 0.6, 7), ('PosterizeOriginal', 0.6, 6)], + [('Equalize', 0.4, 7), ('Solarize', 0.2, 4)], + [('Equalize', 0.4, 4), ('Rotate', 0.8, 8)], + [('Solarize', 0.6, 3), ('Equalize', 0.6, 7)], + [('PosterizeOriginal', 0.8, 5), ('Equalize', 1.0, 2)], + [('Rotate', 0.2, 3), ('Solarize', 0.6, 8)], + [('Equalize', 0.6, 8), ('PosterizeOriginal', 0.4, 6)], + [('Rotate', 0.8, 8), ('Color', 0.4, 0)], + [('Rotate', 0.4, 9), ('Equalize', 0.6, 2)], + [('Equalize', 0.0, 7), ('Equalize', 0.8, 8)], + [('Invert', 0.6, 4), ('Equalize', 1.0, 8)], + [('Color', 0.6, 4), ('Contrast', 1.0, 8)], + [('Rotate', 0.8, 8), ('Color', 1.0, 2)], + [('Color', 0.8, 8), ('Solarize', 0.8, 7)], + [('Sharpness', 0.4, 7), ('Invert', 0.6, 8)], + [('ShearX', 0.6, 5), ('Equalize', 1.0, 9)], + [('Color', 0.4, 0), ('Equalize', 0.6, 3)], + [('Equalize', 0.4, 7), ('Solarize', 0.2, 4)], + [('Solarize', 0.6, 5), ('AutoContrast', 0.6, 5)], + [('Invert', 0.6, 4), ('Equalize', 1.0, 8)], + [('Color', 0.6, 4), ('Contrast', 1.0, 8)], + [('Equalize', 0.8, 8), ('Equalize', 0.6, 3)], + ] + pc = [[AugmentOp(*a, hparams=hparams) for a in sp] for sp in policy] + return pc + + +def auto_augment_policy_originalr(hparams): + # ImageNet policy from https://arxiv.org/abs/1805.09501 with research posterize variation + policy = [ + [('PosterizeIncreasing', 0.4, 8), ('Rotate', 0.6, 9)], + [('Solarize', 0.6, 5), ('AutoContrast', 0.6, 5)], + [('Equalize', 0.8, 8), ('Equalize', 0.6, 3)], + [('PosterizeIncreasing', 0.6, 7), ('PosterizeIncreasing', 0.6, 6)], + [('Equalize', 0.4, 7), ('Solarize', 0.2, 4)], + [('Equalize', 0.4, 4), ('Rotate', 0.8, 8)], + [('Solarize', 0.6, 3), ('Equalize', 0.6, 7)], + [('PosterizeIncreasing', 0.8, 5), ('Equalize', 1.0, 2)], + [('Rotate', 0.2, 3), ('Solarize', 0.6, 8)], + [('Equalize', 0.6, 8), ('PosterizeIncreasing', 0.4, 6)], + [('Rotate', 0.8, 8), ('Color', 0.4, 0)], + [('Rotate', 0.4, 9), ('Equalize', 0.6, 2)], + [('Equalize', 0.0, 7), ('Equalize', 0.8, 8)], + [('Invert', 0.6, 4), ('Equalize', 1.0, 8)], + [('Color', 0.6, 4), ('Contrast', 1.0, 8)], + [('Rotate', 0.8, 8), ('Color', 1.0, 2)], + [('Color', 0.8, 8), ('Solarize', 0.8, 7)], + [('Sharpness', 0.4, 7), ('Invert', 0.6, 8)], + [('ShearX', 0.6, 5), ('Equalize', 1.0, 9)], + [('Color', 0.4, 0), ('Equalize', 0.6, 3)], + [('Equalize', 0.4, 7), ('Solarize', 0.2, 4)], + [('Solarize', 0.6, 5), ('AutoContrast', 0.6, 5)], + [('Invert', 0.6, 4), ('Equalize', 1.0, 8)], + [('Color', 0.6, 4), ('Contrast', 1.0, 8)], + [('Equalize', 0.8, 8), ('Equalize', 0.6, 3)], + ] + pc = [[AugmentOp(*a, hparams=hparams) for a in sp] for sp in policy] + return pc + + +def auto_augment_policy(name='v0', hparams=None): + hparams = hparams or _HPARAMS_DEFAULT + if name == 'original': + return auto_augment_policy_original(hparams) + elif name == 'originalr': + return auto_augment_policy_originalr(hparams) + elif name == 'v0': + return auto_augment_policy_v0(hparams) + elif name == 'v0r': + return auto_augment_policy_v0r(hparams) + else: + assert False, 'Unknown AA policy (%s)' % name + + +class AutoAugment(object): + def __init__(self, policy): + self.policy = policy + + def __call__(self, img): + sub_policy = random.choice(self.policy) + for op in sub_policy: + img = op(img) + return img + + +def auto_augment_transform(config_str, hparams): + """ + Create a AutoAugment transform + + :param config_str: String defining configuration of auto augmentation. Consists of multiple sections separated by + dashes ('-'). The first section defines the AutoAugment policy (one of 'v0', 'v0r', 'original', 'originalr'). + The remaining sections, not order sepecific determine + 'mstd' - float std deviation of magnitude noise applied + Ex 'original-mstd0.5' results in AutoAugment with original policy, magnitude_std 0.5 + + :param hparams: Other hparams (kwargs) for the AutoAugmentation scheme + + :return: A callable Transform Op + """ + config = config_str.split('-') + policy_name = config[0] + config = config[1:] + for c in config: + cs = re.split(r'(\d.*)', c) + if len(cs) < 2: + continue + key, val = cs[:2] + if key == 'mstd': + # noise param injected via hparams for now + hparams.setdefault('magnitude_std', float(val)) + else: + assert False, 'Unknown AutoAugment config section' + aa_policy = auto_augment_policy(policy_name, hparams=hparams) + return AutoAugment(aa_policy) + + +_RAND_TRANSFORMS = [ + 'AutoContrast', + 'Equalize', + 'Invert', + 'Rotate', + 'Posterize', + 'Solarize', + 'SolarizeAdd', + 'Color', + 'Contrast', + 'Brightness', + 'Sharpness', + 'ShearX', + 'ShearY', + 'TranslateXRel', + 'TranslateYRel', + #'Cutout' # NOTE I've implement this as random erasing separately +] + +_RAND_INCREASING_TRANSFORMS = [ + 'AutoContrast', + 'Equalize', + 'Invert', + 'Rotate', + 'PosterizeIncreasing', + 'SolarizeIncreasing', + 'SolarizeAdd', + 'ColorIncreasing', + 'ContrastIncreasing', + 'BrightnessIncreasing', + 'SharpnessIncreasing', + 'ShearX', + 'ShearY', + 'TranslateXRel', + 'TranslateYRel', + #'Cutout' # NOTE I've implement this as random erasing separately +] + +# These experimental weights are based loosely on the relative improvements mentioned in paper. +# They may not result in increased performance, but could likely be tuned to so. +_RAND_CHOICE_WEIGHTS_0 = { + 'Rotate': 0.3, + 'ShearX': 0.2, + 'ShearY': 0.2, + 'TranslateXRel': 0.1, + 'TranslateYRel': 0.1, + 'Color': .025, + 'Sharpness': 0.025, + 'AutoContrast': 0.025, + 'Solarize': .005, + 'SolarizeAdd': .005, + 'Contrast': .005, + 'Brightness': .005, + 'Equalize': .005, + 'Posterize': 0, + 'Invert': 0, +} + + +def _select_rand_weights(weight_idx=0, transforms=None): + transforms = transforms or _RAND_TRANSFORMS + assert weight_idx == 0 # only one set of weights currently + rand_weights = _RAND_CHOICE_WEIGHTS_0 + probs = [rand_weights[k] for k in transforms] + probs /= np.sum(probs) + return probs + + +def rand_augment_ops(magnitude=10, hparams=None, transforms=None): + hparams = hparams or _HPARAMS_DEFAULT + transforms = transforms or _RAND_TRANSFORMS + return [ + AugmentOp( + name, prob=0.5, magnitude=magnitude, hparams=hparams) + for name in transforms + ] + + +class RandAugment(object): + def __init__(self, ops, num_layers=2, choice_weights=None): + self.ops = ops + self.num_layers = num_layers + self.choice_weights = choice_weights + + def __call__(self, img): + # no replacement when using weighted choice + ops = np.random.choice( + self.ops, + self.num_layers, + replace=self.choice_weights is None, + p=self.choice_weights) + for op in ops: + img = op(img) + return img + + +def rand_augment_transform(config_str, hparams): + """ + Create a RandAugment transform + + :param config_str: String defining configuration of random augmentation. Consists of multiple sections separated by + dashes ('-'). The first section defines the specific variant of rand augment (currently only 'rand'). The remaining + sections, not order sepecific determine + 'm' - integer magnitude of rand augment + 'n' - integer num layers (number of transform ops selected per image) + 'w' - integer probabiliy weight index (index of a set of weights to influence choice of op) + 'mstd' - float std deviation of magnitude noise applied + 'inc' - integer (bool), use augmentations that increase in severity with magnitude (default: 0) + Ex 'rand-m9-n3-mstd0.5' results in RandAugment with magnitude 9, num_layers 3, magnitude_std 0.5 + 'rand-mstd1-w0' results in magnitude_std 1.0, weights 0, default magnitude of 10 and num_layers 2 + + :param hparams: Other hparams (kwargs) for the RandAugmentation scheme + + :return: A callable Transform Op + """ + magnitude = _MAX_LEVEL # default to _MAX_LEVEL for magnitude (currently 10) + num_layers = 2 # default to 2 ops per image + weight_idx = None # default to no probability weights for op choice + transforms = _RAND_TRANSFORMS + config = config_str.split('-') + assert config[0] == 'rand' + config = config[1:] + for c in config: + cs = re.split(r'(\d.*)', c) + if len(cs) < 2: + continue + key, val = cs[:2] + if key == 'mstd': + # noise param injected via hparams for now + hparams.setdefault('magnitude_std', float(val)) + elif key == 'inc': + if bool(val): + transforms = _RAND_INCREASING_TRANSFORMS + elif key == 'm': + magnitude = int(val) + elif key == 'n': + num_layers = int(val) + elif key == 'w': + weight_idx = int(val) + else: + assert False, 'Unknown RandAugment config section' + ra_ops = rand_augment_ops( + magnitude=magnitude, hparams=hparams, transforms=transforms) + choice_weights = None if weight_idx is None else _select_rand_weights( + weight_idx) + return RandAugment(ra_ops, num_layers, choice_weights=choice_weights) + + +_AUGMIX_TRANSFORMS = [ + 'AutoContrast', + 'ColorIncreasing', # not in paper + 'ContrastIncreasing', # not in paper + 'BrightnessIncreasing', # not in paper + 'SharpnessIncreasing', # not in paper + 'Equalize', + 'Rotate', + 'PosterizeIncreasing', + 'SolarizeIncreasing', + 'ShearX', + 'ShearY', + 'TranslateXRel', + 'TranslateYRel', +] + + +def augmix_ops(magnitude=10, hparams=None, transforms=None): + hparams = hparams or _HPARAMS_DEFAULT + transforms = transforms or _AUGMIX_TRANSFORMS + return [ + AugmentOp( + name, prob=1.0, magnitude=magnitude, hparams=hparams) + for name in transforms + ] + + +class AugMixAugment(object): + """ AugMix Transform + Adapted and improved from impl here: https://github.com/google-research/augmix/blob/master/imagenet.py + From paper: 'AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty - + https://arxiv.org/abs/1912.02781 + """ + + def __init__(self, ops, alpha=1., width=3, depth=-1, blended=False): + self.ops = ops + self.alpha = alpha + self.width = width + self.depth = depth + self.blended = blended # blended mode is faster but not well tested + + def _calc_blended_weights(self, ws, m): + ws = ws * m + cump = 1. + rws = [] + for w in ws[::-1]: + alpha = w / cump + cump *= (1 - alpha) + rws.append(alpha) + return np.array(rws[::-1], dtype=np.float32) + + def _apply_blended(self, img, mixing_weights, m): + # This is my first crack and implementing a slightly faster mixed augmentation. Instead + # of accumulating the mix for each chain in a Numpy array and then blending with original, + # it recomputes the blending coefficients and applies one PIL image blend per chain. + # TODO the results appear in the right ballpark but they differ by more than rounding. + img_orig = img.copy() + ws = self._calc_blended_weights(mixing_weights, m) + for w in ws: + depth = self.depth if self.depth > 0 else np.random.randint(1, 4) + ops = np.random.choice(self.ops, depth, replace=True) + img_aug = img_orig # no ops are in-place, deep copy not necessary + for op in ops: + img_aug = op(img_aug) + img = Image.blend(img, img_aug, w) + return img + + def _apply_basic(self, img, mixing_weights, m): + # This is a literal adaptation of the paper/official implementation without normalizations and + # PIL <-> Numpy conversions between every op. It is still quite CPU compute heavy compared to the + # typical augmentation transforms, could use a GPU / Kornia implementation. + img_shape = img.size[0], img.size[1], len(img.getbands()) + mixed = np.zeros(img_shape, dtype=np.float32) + for mw in mixing_weights: + depth = self.depth if self.depth > 0 else np.random.randint(1, 4) + ops = np.random.choice(self.ops, depth, replace=True) + img_aug = img # no ops are in-place, deep copy not necessary + for op in ops: + img_aug = op(img_aug) + mixed += mw * np.asarray(img_aug, dtype=np.float32) + np.clip(mixed, 0, 255., out=mixed) + mixed = Image.fromarray(mixed.astype(np.uint8)) + return Image.blend(img, mixed, m) + + def __call__(self, img): + mixing_weights = np.float32( + np.random.dirichlet([self.alpha] * self.width)) + m = np.float32(np.random.beta(self.alpha, self.alpha)) + if self.blended: + mixed = self._apply_blended(img, mixing_weights, m) + else: + mixed = self._apply_basic(img, mixing_weights, m) + return mixed + + +def augment_and_mix_transform(config_str, hparams): + """ Create AugMix transform + + :param config_str: String defining configuration of random augmentation. Consists of multiple sections separated by + dashes ('-'). The first section defines the specific variant of rand augment (currently only 'rand'). The remaining + sections, not order sepecific determine + 'm' - integer magnitude (severity) of augmentation mix (default: 3) + 'w' - integer width of augmentation chain (default: 3) + 'd' - integer depth of augmentation chain (-1 is random [1, 3], default: -1) + 'b' - integer (bool), blend each branch of chain into end result without a final blend, less CPU (default: 0) + 'mstd' - float std deviation of magnitude noise applied (default: 0) + Ex 'augmix-m5-w4-d2' results in AugMix with severity 5, chain width 4, chain depth 2 + + :param hparams: Other hparams (kwargs) for the Augmentation transforms + + :return: A callable Transform Op + """ + magnitude = 3 + width = 3 + depth = -1 + alpha = 1. + blended = False + config = config_str.split('-') + assert config[0] == 'augmix' + config = config[1:] + for c in config: + cs = re.split(r'(\d.*)', c) + if len(cs) < 2: + continue + key, val = cs[:2] + if key == 'mstd': + # noise param injected via hparams for now + hparams.setdefault('magnitude_std', float(val)) + elif key == 'm': + magnitude = int(val) + elif key == 'w': + width = int(val) + elif key == 'd': + depth = int(val) + elif key == 'a': + alpha = float(val) + elif key == 'b': + blended = bool(val) + else: + assert False, 'Unknown AugMix config section' + ops = augmix_ops(magnitude=magnitude, hparams=hparams) + return AugMixAugment( + ops, alpha=alpha, width=width, depth=depth, blended=blended) + + +class RawTimmAutoAugment(object): + """TimmAutoAugment API for PaddleClas.""" + + def __init__(self, + config_str="rand-m9-mstd0.5-inc1", + interpolation="bicubic", + img_size=224, + mean=IMAGENET_DEFAULT_MEAN): + if isinstance(img_size, (tuple, list)): + img_size_min = min(img_size) + else: + img_size_min = img_size + + aa_params = dict( + translate_const=int(img_size_min * 0.45), + img_mean=tuple([min(255, round(255 * x)) for x in mean]), ) + if interpolation and interpolation != 'random': + aa_params['interpolation'] = _pil_interp(interpolation) + if config_str.startswith('rand'): + self.augment_func = rand_augment_transform(config_str, aa_params) + elif config_str.startswith('augmix'): + aa_params['translate_pct'] = 0.3 + self.augment_func = augment_and_mix_transform(config_str, + aa_params) + elif config_str.startswith('auto'): + self.augment_func = auto_augment_transform(config_str, aa_params) + else: + raise Exception( + "ConfigError: The TimmAutoAugment Op only support RandAugment, AutoAugment, AugMix, and the config_str only starts with \"rand\", \"augmix\", \"auto\"." + ) + + def __call__(self, img): + return self.augment_func(img) diff --git a/exprements/PaddleClas/ppcls/data/utils/__init__.py b/exprements/PaddleClas/ppcls/data/utils/__init__.py new file mode 100644 index 00000000..61d5aa21 --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/utils/__init__.py @@ -0,0 +1,13 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. \ No newline at end of file diff --git a/exprements/PaddleClas/ppcls/data/utils/get_image_list.py b/exprements/PaddleClas/ppcls/data/utils/get_image_list.py new file mode 100644 index 00000000..6f10935a --- /dev/null +++ b/exprements/PaddleClas/ppcls/data/utils/get_image_list.py @@ -0,0 +1,49 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import argparse +import base64 +import numpy as np + + +def get_image_list(img_file): + imgs_lists = [] + if img_file is None or not os.path.exists(img_file): + raise Exception("not found any img file in {}".format(img_file)) + + img_end = ['jpg', 'png', 'jpeg', 'JPEG', 'JPG', 'bmp'] + if os.path.isfile(img_file) and img_file.split('.')[-1] in img_end: + imgs_lists.append(img_file) + elif os.path.isdir(img_file): + for single_file in os.listdir(img_file): + if single_file.split('.')[-1] in img_end: + imgs_lists.append(os.path.join(img_file, single_file)) + if len(imgs_lists) == 0: + raise Exception("not found any img file in {}".format(img_file)) + imgs_lists = sorted(imgs_lists) + return imgs_lists + + +def get_image_list_from_label_file(image_path, label_file_path): + imgs_lists = [] + gt_labels = [] + with open(label_file_path, "r") as fin: + lines = fin.readlines() + for line in lines: + image_name, label = line.strip("\n").split() + label = int(label) + imgs_lists.append(os.path.join(image_path, image_name)) + gt_labels.append(int(label)) + return imgs_lists, gt_labels diff --git a/exprements/PaddleClas/ppcls/engine/__init__.py b/exprements/PaddleClas/ppcls/engine/__init__.py new file mode 100644 index 00000000..e69de29b diff --git a/exprements/PaddleClas/ppcls/engine/engine.py b/exprements/PaddleClas/ppcls/engine/engine.py new file mode 100644 index 00000000..7f04221c --- /dev/null +++ b/exprements/PaddleClas/ppcls/engine/engine.py @@ -0,0 +1,463 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import os +import platform +import paddle +import paddle.distributed as dist +from visualdl import LogWriter +from paddle import nn +import numpy as np +import random + +from ppcls.utils.check import check_gpu +from ppcls.utils.misc import AverageMeter +from ppcls.utils import logger +from ppcls.utils.logger import init_logger +from ppcls.utils.config import print_config +from ppcls.data import build_dataloader +from ppcls.arch import build_model, RecModel, DistillationModel, TheseusLayer +from ppcls.arch import apply_to_static +from ppcls.loss import build_loss +from ppcls.metric import build_metrics +from ppcls.optimizer import build_optimizer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url +from ppcls.utils.save_load import init_model +from ppcls.utils import save_load + +from ppcls.data.utils.get_image_list import get_image_list +from ppcls.data.postprocess import build_postprocess +from ppcls.data import create_operators +from ppcls.engine.train import train_epoch +from ppcls.engine import evaluation +from ppcls.arch.gears.identity_head import IdentityHead + + +class Engine(object): + def __init__(self, config, mode="train"): + assert mode in ["train", "eval", "infer", "export"] + self.mode = mode + self.config = config + self.eval_mode = self.config["Global"].get("eval_mode", + "classification") + if "Head" in self.config["Arch"] or self.config["Arch"].get("is_rec", + False): + self.is_rec = True + else: + self.is_rec = False + + # set seed + seed = self.config["Global"].get("seed", False) + if seed or seed == 0: + assert isinstance(seed, int), "The 'seed' must be a integer!" + paddle.seed(seed) + np.random.seed(seed) + random.seed(seed) + + # init logger + self.output_dir = self.config['Global']['output_dir'] + log_file = os.path.join(self.output_dir, self.config["Arch"]["name"], + f"{mode}.log") + init_logger(name='root', log_file=log_file) + print_config(config) + + # init train_func and eval_func + assert self.eval_mode in ["classification", "retrieval"], logger.error( + "Invalid eval mode: {}".format(self.eval_mode)) + self.train_epoch_func = train_epoch + self.eval_func = getattr(evaluation, self.eval_mode + "_eval") + + self.use_dali = self.config['Global'].get("use_dali", False) + + # for visualdl + self.vdl_writer = None + if self.config['Global'][ + 'use_visualdl'] and mode == "train" and dist.get_rank() == 0: + vdl_writer_path = os.path.join(self.output_dir, "vdl") + if not os.path.exists(vdl_writer_path): + os.makedirs(vdl_writer_path) + self.vdl_writer = LogWriter(logdir=vdl_writer_path) + + # set device + assert self.config["Global"]["device"] in ["cpu", "gpu", "xpu", "npu"] + self.device = paddle.set_device(self.config["Global"]["device"]) + logger.info('train with paddle {} and device {}'.format( + paddle.__version__, self.device)) + + # AMP training + self.amp = True if "AMP" in self.config and self.mode == "train" else False + if self.amp and self.config["AMP"] is not None: + self.scale_loss = self.config["AMP"].get("scale_loss", 1.0) + self.use_dynamic_loss_scaling = self.config["AMP"].get( + "use_dynamic_loss_scaling", False) + else: + self.scale_loss = 1.0 + self.use_dynamic_loss_scaling = False + if self.amp: + AMP_RELATED_FLAGS_SETTING = { + 'FLAGS_cudnn_batchnorm_spatial_persistent': 1, + 'FLAGS_max_inplace_grad_add': 8, + } + paddle.fluid.set_flags(AMP_RELATED_FLAGS_SETTING) + + if "class_num" in config["Global"]: + global_class_num = config["Global"]["class_num"] + if "class_num" not in config["Arch"]: + config["Arch"]["class_num"] = global_class_num + msg = f"The Global.class_num will be deprecated. Please use Arch.class_num instead. Arch.class_num has been set to {global_class_num}." + else: + msg = "The Global.class_num will be deprecated. Please use Arch.class_num instead. The Global.class_num has been ignored." + logger.warning(msg) + #TODO(gaotingquan): support rec + class_num = config["Arch"].get("class_num", None) + self.config["DataLoader"].update({"class_num": class_num}) + # build dataloader + if self.mode == 'train': + self.train_dataloader = build_dataloader( + self.config["DataLoader"], "Train", self.device, self.use_dali) + if self.mode == "eval" or (self.mode == "train" and + self.config["Global"]["eval_during_train"]): + if self.eval_mode == "classification": + self.eval_dataloader = build_dataloader( + self.config["DataLoader"], "Eval", self.device, + self.use_dali) + elif self.eval_mode == "retrieval": + self.gallery_query_dataloader = None + if len(self.config["DataLoader"]["Eval"].keys()) == 1: + key = list(self.config["DataLoader"]["Eval"].keys())[0] + self.gallery_query_dataloader = build_dataloader( + self.config["DataLoader"]["Eval"], key, self.device, + self.use_dali) + else: + self.gallery_dataloader = build_dataloader( + self.config["DataLoader"]["Eval"], "Gallery", + self.device, self.use_dali) + self.query_dataloader = build_dataloader( + self.config["DataLoader"]["Eval"], "Query", + self.device, self.use_dali) + + # build loss + if self.mode == "train": + loss_info = self.config["Loss"]["Train"] + self.train_loss_func = build_loss(loss_info) + if self.mode == "eval" or (self.mode == "train" and + self.config["Global"]["eval_during_train"]): + loss_config = self.config.get("Loss", None) + if loss_config is not None: + loss_config = loss_config.get("Eval") + if loss_config is not None: + self.eval_loss_func = build_loss(loss_config) + else: + self.eval_loss_func = None + else: + self.eval_loss_func = None + + # build metric + if self.mode == 'train': + metric_config = self.config.get("Metric") + if metric_config is not None: + metric_config = metric_config.get("Train") + if metric_config is not None: + if hasattr(self.train_dataloader, "collate_fn"): + for m_idx, m in enumerate(metric_config): + if "TopkAcc" in m: + msg = f"'TopkAcc' metric can not be used when setting 'batch_transform_ops' in config. The 'TopkAcc' metric has been removed." + logger.warning(msg) + break + metric_config.pop(m_idx) + self.train_metric_func = build_metrics(metric_config) + else: + self.train_metric_func = None + else: + self.train_metric_func = None + + if self.mode == "eval" or (self.mode == "train" and + self.config["Global"]["eval_during_train"]): + metric_config = self.config.get("Metric") + if self.eval_mode == "classification": + if metric_config is not None: + metric_config = metric_config.get("Eval") + if metric_config is not None: + self.eval_metric_func = build_metrics(metric_config) + elif self.eval_mode == "retrieval": + if metric_config is None: + metric_config = [{"name": "Recallk", "topk": (1, 5)}] + else: + metric_config = metric_config["Eval"] + self.eval_metric_func = build_metrics(metric_config) + else: + self.eval_metric_func = None + + # build model + self.model = build_model(self.config) + # set @to_static for benchmark, skip this by default. + apply_to_static(self.config, self.model) + + # load_pretrain + if self.config["Global"]["pretrained_model"] is not None: + if self.config["Global"]["pretrained_model"].startswith("http"): + load_dygraph_pretrain_from_url( + self.model, self.config["Global"]["pretrained_model"]) + else: + load_dygraph_pretrain( + self.model, self.config["Global"]["pretrained_model"]) + + # build optimizer + if self.mode == 'train': + self.optimizer, self.lr_sch = build_optimizer( + self.config["Optimizer"], self.config["Global"]["epochs"], + len(self.train_dataloader), [self.model]) + + # for amp training + if self.amp: + self.scaler = paddle.amp.GradScaler( + init_loss_scaling=self.scale_loss, + use_dynamic_loss_scaling=self.use_dynamic_loss_scaling) + amp_level = self.config['AMP'].get("level", "O1") + if amp_level not in ["O1", "O2"]: + msg = "[Parameter Error]: The optimize level of AMP only support 'O1' and 'O2'. The level has been set 'O1'." + logger.warning(msg) + self.config['AMP']["level"] = "O1" + amp_level = "O1" + self.model, self.optimizer = paddle.amp.decorate( + models=self.model, + optimizers=self.optimizer, + level=amp_level, + save_dtype='float32') + + # for distributed + world_size = dist.get_world_size() + self.config["Global"]["distributed"] = world_size != 1 + if world_size != 4 and self.mode == "train": + msg = f"The training strategy in config files provided by PaddleClas is based on 4 gpus. But the number of gpus is {world_size} in current training. Please modify the stategy (learning rate, batch size and so on) if use config files in PaddleClas to train." + logger.warning(msg) + if self.config["Global"]["distributed"]: + dist.init_parallel_env() + self.model = paddle.DataParallel(self.model) + + # build postprocess for infer + if self.mode == 'infer': + self.preprocess_func = create_operators(self.config["Infer"][ + "transforms"]) + self.postprocess_func = build_postprocess(self.config["Infer"][ + "PostProcess"]) + + def train(self): + assert self.mode == "train" + print_batch_step = self.config['Global']['print_batch_step'] + save_interval = self.config["Global"]["save_interval"] + best_metric = { + "metric": 0.0, + "epoch": 0, + } + # key: + # val: metrics list word + self.output_info = dict() + self.time_info = { + "batch_cost": AverageMeter( + "batch_cost", '.5f', postfix=" s,"), + "reader_cost": AverageMeter( + "reader_cost", ".5f", postfix=" s,"), + } + # global iter counter + self.global_step = 0 + + if self.config["Global"]["checkpoints"] is not None: + metric_info = init_model(self.config["Global"], self.model, + self.optimizer) + if metric_info is not None: + best_metric.update(metric_info) + + self.max_iter = len(self.train_dataloader) - 1 if platform.system( + ) == "Windows" else len(self.train_dataloader) + for epoch_id in range(best_metric["epoch"] + 1, + self.config["Global"]["epochs"] + 1): + acc = 0.0 + # for one epoch train + self.train_epoch_func(self, epoch_id, print_batch_step) + + if self.use_dali: + self.train_dataloader.reset() + metric_msg = ", ".join([ + "{}: {:.5f}".format(key, self.output_info[key].avg) + for key in self.output_info + ]) + logger.info("[Train][Epoch {}/{}][Avg]{}".format( + epoch_id, self.config["Global"]["epochs"], metric_msg)) + self.output_info.clear() + + # eval model and save model if possible + if self.config["Global"][ + "eval_during_train"] and epoch_id % self.config["Global"][ + "eval_interval"] == 0: + acc = self.eval(epoch_id) + if acc > best_metric["metric"]: + best_metric["metric"] = acc + best_metric["epoch"] = epoch_id + save_load.save_model( + self.model, + self.optimizer, + best_metric, + self.output_dir, + model_name=self.config["Arch"]["name"], + prefix="best_model") + logger.info("[Eval][Epoch {}][best metric: {}]".format( + epoch_id, best_metric["metric"])) + logger.scaler( + name="eval_acc", + value=acc, + step=epoch_id, + writer=self.vdl_writer) + + self.model.train() + + # save model + if epoch_id % save_interval == 0: + save_load.save_model( + self.model, + self.optimizer, {"metric": acc, + "epoch": epoch_id}, + self.output_dir, + model_name=self.config["Arch"]["name"], + prefix="epoch_{}".format(epoch_id)) + # save the latest model + save_load.save_model( + self.model, + self.optimizer, {"metric": acc, + "epoch": epoch_id}, + self.output_dir, + model_name=self.config["Arch"]["name"], + prefix="latest") + + if self.vdl_writer is not None: + self.vdl_writer.close() + + @paddle.no_grad() + def eval(self, epoch_id=0): + assert self.mode in ["train", "eval"] + self.model.eval() + eval_result = self.eval_func(self, epoch_id) + self.model.train() + return eval_result + + @paddle.no_grad() + def infer(self): + assert self.mode == "infer" and self.eval_mode == "classification" + total_trainer = dist.get_world_size() + local_rank = dist.get_rank() + image_list = get_image_list(self.config["Infer"]["infer_imgs"]) + # data split + image_list = image_list[local_rank::total_trainer] + + batch_size = self.config["Infer"]["batch_size"] + self.model.eval() + batch_data = [] + image_file_list = [] + for idx, image_file in enumerate(image_list): + with open(image_file, 'rb') as f: + x = f.read() + for process in self.preprocess_func: + x = process(x) + batch_data.append(x) + image_file_list.append(image_file) + if len(batch_data) >= batch_size or idx == len(image_list) - 1: + batch_tensor = paddle.to_tensor(batch_data) + out = self.model(batch_tensor) + if isinstance(out, list): + out = out[0] + if isinstance(out, dict) and "logits" in out: + out = out["logits"] + if isinstance(out, dict) and "output" in out: + out = out["output"] + result = self.postprocess_func(out, image_file_list) + print(result) + batch_data.clear() + image_file_list.clear() + + def export(self): + assert self.mode == "export" + use_multilabel = self.config["Global"].get("use_multilabel", False) + model = ExportModel(self.config["Arch"], self.model, use_multilabel) + if self.config["Global"]["pretrained_model"] is not None: + load_dygraph_pretrain(model.base_model, + self.config["Global"]["pretrained_model"]) + + model.eval() + save_path = os.path.join(self.config["Global"]["save_inference_dir"], + "inference") + if model.quanter: + model.quanter.save_quantized_model( + model.base_model, + save_path, + input_spec=[ + paddle.static.InputSpec( + shape=[None] + self.config["Global"]["image_shape"], + dtype='float32') + ]) + else: + model = paddle.jit.to_static( + model, + input_spec=[ + paddle.static.InputSpec( + shape=[None] + self.config["Global"]["image_shape"], + dtype='float32') + ]) + paddle.jit.save(model, save_path) + + +class ExportModel(TheseusLayer): + """ + ExportModel: add softmax onto the model + """ + + def __init__(self, config, model, use_multilabel): + super().__init__() + self.base_model = model + # we should choose a final model to export + if isinstance(self.base_model, DistillationModel): + self.infer_model_name = config["infer_model_name"] + else: + self.infer_model_name = None + + self.infer_output_key = config.get("infer_output_key", None) + if self.infer_output_key == "features" and isinstance(self.base_model, + RecModel): + self.base_model.head = IdentityHead() + if use_multilabel: + self.out_act = nn.Sigmoid() + else: + if config.get("infer_add_softmax", True): + self.out_act = nn.Softmax(axis=-1) + else: + self.out_act = None + + def eval(self): + self.training = False + for layer in self.sublayers(): + layer.training = False + layer.eval() + + def forward(self, x): + x = self.base_model(x) + if isinstance(x, list): + x = x[0] + if self.infer_model_name is not None: + x = x[self.infer_model_name] + if self.infer_output_key is not None: + x = x[self.infer_output_key] + if self.out_act is not None: + x = self.out_act(x) + return x diff --git a/exprements/PaddleClas/ppcls/engine/evaluation/__init__.py b/exprements/PaddleClas/ppcls/engine/evaluation/__init__.py new file mode 100644 index 00000000..e0cd7788 --- /dev/null +++ b/exprements/PaddleClas/ppcls/engine/evaluation/__init__.py @@ -0,0 +1,16 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from ppcls.engine.evaluation.classification import classification_eval +from ppcls.engine.evaluation.retrieval import retrieval_eval diff --git a/exprements/PaddleClas/ppcls/engine/evaluation/classification.py b/exprements/PaddleClas/ppcls/engine/evaluation/classification.py new file mode 100644 index 00000000..d7b5c476 --- /dev/null +++ b/exprements/PaddleClas/ppcls/engine/evaluation/classification.py @@ -0,0 +1,169 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import time +import platform +import paddle + +from ppcls.utils.misc import AverageMeter +from ppcls.utils import logger + + +def classification_eval(engine, epoch_id=0): + output_info = dict() + time_info = { + "batch_cost": AverageMeter( + "batch_cost", '.5f', postfix=" s,"), + "reader_cost": AverageMeter( + "reader_cost", ".5f", postfix=" s,"), + } + print_batch_step = engine.config["Global"]["print_batch_step"] + + metric_key = None + tic = time.time() + accum_samples = 0 + total_samples = len( + engine.eval_dataloader. + dataset) if not engine.use_dali else engine.eval_dataloader.size + max_iter = len(engine.eval_dataloader) - 1 if platform.system( + ) == "Windows" else len(engine.eval_dataloader) + for iter_id, batch in enumerate(engine.eval_dataloader): + if iter_id >= max_iter: + break + if iter_id == 5: + for key in time_info: + time_info[key].reset() + if engine.use_dali: + batch = [ + paddle.to_tensor(batch[0]['data']), + paddle.to_tensor(batch[0]['label']) + ] + time_info["reader_cost"].update(time.time() - tic) + batch_size = batch[0].shape[0] + batch[0] = paddle.to_tensor(batch[0]).astype("float32") + if not engine.config["Global"].get("use_multilabel", False): + batch[1] = batch[1].reshape([-1, 1]).astype("int64") + + # image input + if engine.amp: + amp_level = engine.config['AMP'].get("level", "O1").upper() + with paddle.amp.auto_cast( + custom_black_list={ + "flatten_contiguous_range", "greater_than" + }, + level=amp_level): + out = engine.model(batch[0]) + # calc loss + if engine.eval_loss_func is not None: + loss_dict = engine.eval_loss_func(out, batch[1]) + for key in loss_dict: + if key not in output_info: + output_info[key] = AverageMeter(key, '7.5f') + output_info[key].update(loss_dict[key].numpy()[0], + batch_size) + else: + out = engine.model(batch[0]) + # calc loss + if engine.eval_loss_func is not None: + loss_dict = engine.eval_loss_func(out, batch[1]) + for key in loss_dict: + if key not in output_info: + output_info[key] = AverageMeter(key, '7.5f') + output_info[key].update(loss_dict[key].numpy()[0], + batch_size) + + # just for DistributedBatchSampler issue: repeat sampling + current_samples = batch_size * paddle.distributed.get_world_size() + accum_samples += current_samples + + # calc metric + if engine.eval_metric_func is not None: + if paddle.distributed.get_world_size() > 1: + label_list = [] + paddle.distributed.all_gather(label_list, batch[1]) + labels = paddle.concat(label_list, 0) + + if isinstance(out, dict): + if "Student" in out: + out = out["Student"] + elif "logits" in out: + out = out["logits"] + else: + msg = "Error: Wrong key in out!" + raise Exception(msg) + if isinstance(out, list): + pred = [] + for x in out: + pred_list = [] + paddle.distributed.all_gather(pred_list, x) + pred_x = paddle.concat(pred_list, 0) + pred.append(pred_x) + else: + pred_list = [] + paddle.distributed.all_gather(pred_list, out) + pred = paddle.concat(pred_list, 0) + + if accum_samples > total_samples and not engine.use_dali: + pred = pred[:total_samples + current_samples - + accum_samples] + labels = labels[:total_samples + current_samples - + accum_samples] + current_samples = total_samples + current_samples - accum_samples + metric_dict = engine.eval_metric_func(pred, labels) + else: + metric_dict = engine.eval_metric_func(out, batch[1]) + + for key in metric_dict: + if metric_key is None: + metric_key = key + if key not in output_info: + output_info[key] = AverageMeter(key, '7.5f') + + output_info[key].update(metric_dict[key].numpy()[0], + current_samples) + + time_info["batch_cost"].update(time.time() - tic) + + if iter_id % print_batch_step == 0: + time_msg = "s, ".join([ + "{}: {:.5f}".format(key, time_info[key].avg) + for key in time_info + ]) + + ips_msg = "ips: {:.5f} images/sec".format( + batch_size / time_info["batch_cost"].avg) + + metric_msg = ", ".join([ + "{}: {:.5f}".format(key, output_info[key].val) + for key in output_info + ]) + logger.info("[Eval][Epoch {}][Iter: {}/{}]{}, {}, {}".format( + epoch_id, iter_id, + len(engine.eval_dataloader), metric_msg, time_msg, ips_msg)) + + tic = time.time() + if engine.use_dali: + engine.eval_dataloader.reset() + metric_msg = ", ".join([ + "{}: {:.5f}".format(key, output_info[key].avg) for key in output_info + ]) + logger.info("[Eval][Epoch {}][Avg]{}".format(epoch_id, metric_msg)) + + # do not try to save best eval.model + if engine.eval_metric_func is None: + return -1 + # return 1st metric in the dict + return output_info[metric_key].avg diff --git a/exprements/PaddleClas/ppcls/engine/evaluation/retrieval.py b/exprements/PaddleClas/ppcls/engine/evaluation/retrieval.py new file mode 100644 index 00000000..8471a42c --- /dev/null +++ b/exprements/PaddleClas/ppcls/engine/evaluation/retrieval.py @@ -0,0 +1,171 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import platform +import paddle +from ppcls.utils import logger + + +def retrieval_eval(engine, epoch_id=0): + engine.model.eval() + # step1. build gallery + if engine.gallery_query_dataloader is not None: + gallery_feas, gallery_img_id, gallery_unique_id = cal_feature( + engine, name='gallery_query') + query_feas, query_img_id, query_query_id = gallery_feas, gallery_img_id, gallery_unique_id + else: + gallery_feas, gallery_img_id, gallery_unique_id = cal_feature( + engine, name='gallery') + query_feas, query_img_id, query_query_id = cal_feature( + engine, name='query') + + # step2. do evaluation + sim_block_size = engine.config["Global"].get("sim_block_size", 64) + sections = [sim_block_size] * (len(query_feas) // sim_block_size) + if len(query_feas) % sim_block_size: + sections.append(len(query_feas) % sim_block_size) + fea_blocks = paddle.split(query_feas, num_or_sections=sections) + if query_query_id is not None: + query_id_blocks = paddle.split( + query_query_id, num_or_sections=sections) + image_id_blocks = paddle.split(query_img_id, num_or_sections=sections) + metric_key = None + + if engine.eval_loss_func is None: + metric_dict = {metric_key: 0.} + else: + metric_dict = dict() + for block_idx, block_fea in enumerate(fea_blocks): + similarity_matrix = paddle.matmul( + block_fea, gallery_feas, transpose_y=True) + if query_query_id is not None: + query_id_block = query_id_blocks[block_idx] + query_id_mask = (query_id_block != gallery_unique_id.t()) + + image_id_block = image_id_blocks[block_idx] + image_id_mask = (image_id_block != gallery_img_id.t()) + + keep_mask = paddle.logical_or(query_id_mask, image_id_mask) + similarity_matrix = similarity_matrix * keep_mask.astype( + "float32") + else: + keep_mask = None + + metric_tmp = engine.eval_metric_func(similarity_matrix, + image_id_blocks[block_idx], + gallery_img_id, keep_mask) + + for key in metric_tmp: + if key not in metric_dict: + metric_dict[key] = metric_tmp[key] * block_fea.shape[ + 0] / len(query_feas) + else: + metric_dict[key] += metric_tmp[key] * block_fea.shape[ + 0] / len(query_feas) + + metric_info_list = [] + for key in metric_dict: + if metric_key is None: + metric_key = key + metric_info_list.append("{}: {:.5f}".format(key, metric_dict[key])) + metric_msg = ", ".join(metric_info_list) + logger.info("[Eval][Epoch {}][Avg]{}".format(epoch_id, metric_msg)) + + return metric_dict[metric_key] + + +def cal_feature(engine, name='gallery'): + all_feas = None + all_image_id = None + all_unique_id = None + has_unique_id = False + + if name == 'gallery': + dataloader = engine.gallery_dataloader + elif name == 'query': + dataloader = engine.query_dataloader + elif name == 'gallery_query': + dataloader = engine.gallery_query_dataloader + else: + raise RuntimeError("Only support gallery or query dataset") + + max_iter = len(dataloader) - 1 if platform.system() == "Windows" else len( + dataloader) + for idx, batch in enumerate(dataloader): # load is very time-consuming + if idx >= max_iter: + break + if idx % engine.config["Global"]["print_batch_step"] == 0: + logger.info( + f"{name} feature calculation process: [{idx}/{len(dataloader)}]" + ) + if engine.use_dali: + batch = [ + paddle.to_tensor(batch[0]['data']), + paddle.to_tensor(batch[0]['label']) + ] + batch = [paddle.to_tensor(x) for x in batch] + batch[1] = batch[1].reshape([-1, 1]).astype("int64") + if len(batch) == 3: + has_unique_id = True + batch[2] = batch[2].reshape([-1, 1]).astype("int64") + out = engine.model(batch[0], batch[1]) + if "Student" in out: + out = out["Student"] + batch_feas = out["features"] + + # do norm + if engine.config["Global"].get("feature_normalize", True): + feas_norm = paddle.sqrt( + paddle.sum(paddle.square(batch_feas), axis=1, keepdim=True)) + batch_feas = paddle.divide(batch_feas, feas_norm) + + # do binarize + if engine.config["Global"].get("feature_binarize") == "round": + batch_feas = paddle.round(batch_feas).astype("float32") * 2.0 - 1.0 + + if engine.config["Global"].get("feature_binarize") == "sign": + batch_feas = paddle.sign(batch_feas).astype("float32") + + if all_feas is None: + all_feas = batch_feas + if has_unique_id: + all_unique_id = batch[2] + all_image_id = batch[1] + else: + all_feas = paddle.concat([all_feas, batch_feas]) + all_image_id = paddle.concat([all_image_id, batch[1]]) + if has_unique_id: + all_unique_id = paddle.concat([all_unique_id, batch[2]]) + + if engine.use_dali: + dataloader.reset() + + if paddle.distributed.get_world_size() > 1: + feat_list = [] + img_id_list = [] + unique_id_list = [] + paddle.distributed.all_gather(feat_list, all_feas) + paddle.distributed.all_gather(img_id_list, all_image_id) + all_feas = paddle.concat(feat_list, axis=0) + all_image_id = paddle.concat(img_id_list, axis=0) + if has_unique_id: + paddle.distributed.all_gather(unique_id_list, all_unique_id) + all_unique_id = paddle.concat(unique_id_list, axis=0) + + logger.info("Build {} done, all feat shape: {}, begin to eval..".format( + name, all_feas.shape)) + return all_feas, all_image_id, all_unique_id diff --git a/exprements/PaddleClas/ppcls/engine/train/__init__.py b/exprements/PaddleClas/ppcls/engine/train/__init__.py new file mode 100644 index 00000000..800d3a41 --- /dev/null +++ b/exprements/PaddleClas/ppcls/engine/train/__init__.py @@ -0,0 +1,14 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from ppcls.engine.train.train import train_epoch diff --git a/exprements/PaddleClas/ppcls/engine/train/train.py b/exprements/PaddleClas/ppcls/engine/train/train.py new file mode 100644 index 00000000..3b02bac8 --- /dev/null +++ b/exprements/PaddleClas/ppcls/engine/train/train.py @@ -0,0 +1,83 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import, division, print_function + +import time +import paddle +from ppcls.engine.train.utils import update_loss, update_metric, log_info +from ppcls.utils import profiler + + +def train_epoch(engine, epoch_id, print_batch_step): + tic = time.time() + v_current = [int(i) for i in paddle.__version__.split(".")] + for iter_id, batch in enumerate(engine.train_dataloader): + if iter_id >= engine.max_iter: + break + profiler.add_profiler_step(engine.config["profiler_options"]) + if iter_id == 5: + for key in engine.time_info: + engine.time_info[key].reset() + engine.time_info["reader_cost"].update(time.time() - tic) + if engine.use_dali: + batch = [ + paddle.to_tensor(batch[0]['data']), + paddle.to_tensor(batch[0]['label']) + ] + batch_size = batch[0].shape[0] + if not engine.config["Global"].get("use_multilabel", False): + batch[1] = batch[1].reshape([batch_size, -1]) + engine.global_step += 1 + + # image input + if engine.amp: + amp_level = engine.config['AMP'].get("level", "O1").upper() + with paddle.amp.auto_cast( + custom_black_list={ + "flatten_contiguous_range", "greater_than" + }, + level=amp_level): + out = forward(engine, batch) + loss_dict = engine.train_loss_func(out, batch[1]) + else: + out = forward(engine, batch) + loss_dict = engine.train_loss_func(out, batch[1]) + + # step opt and lr + if engine.amp: + scaled = engine.scaler.scale(loss_dict["loss"]) + scaled.backward() + engine.scaler.minimize(engine.optimizer, scaled) + else: + loss_dict["loss"].backward() + engine.optimizer.step() + engine.optimizer.clear_grad() + engine.lr_sch.step() + + # below code just for logging + # update metric_for_logger + update_metric(engine, out, batch, batch_size) + # update_loss_for_logger + update_loss(engine, loss_dict, batch_size) + engine.time_info["batch_cost"].update(time.time() - tic) + if iter_id % print_batch_step == 0: + log_info(engine, batch_size, epoch_id, iter_id) + tic = time.time() + + +def forward(engine, batch): + if not engine.is_rec: + return engine.model(batch[0]) + else: + return engine.model(batch[0], batch[1]) diff --git a/exprements/PaddleClas/ppcls/engine/train/utils.py b/exprements/PaddleClas/ppcls/engine/train/utils.py new file mode 100644 index 00000000..92eb35d7 --- /dev/null +++ b/exprements/PaddleClas/ppcls/engine/train/utils.py @@ -0,0 +1,72 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import, division, print_function + +import datetime +from ppcls.utils import logger +from ppcls.utils.misc import AverageMeter + + +def update_metric(trainer, out, batch, batch_size): + # calc metric + if trainer.train_metric_func is not None: + metric_dict = trainer.train_metric_func(out, batch[-1]) + for key in metric_dict: + if key not in trainer.output_info: + trainer.output_info[key] = AverageMeter(key, '7.5f') + trainer.output_info[key].update(metric_dict[key].numpy()[0], + batch_size) + + +def update_loss(trainer, loss_dict, batch_size): + # update_output_info + for key in loss_dict: + if key not in trainer.output_info: + trainer.output_info[key] = AverageMeter(key, '7.5f') + trainer.output_info[key].update(loss_dict[key].numpy()[0], batch_size) + + +def log_info(trainer, batch_size, epoch_id, iter_id): + lr_msg = "lr: {:.5f}".format(trainer.lr_sch.get_lr()) + metric_msg = ", ".join([ + "{}: {:.5f}".format(key, trainer.output_info[key].avg) + for key in trainer.output_info + ]) + time_msg = "s, ".join([ + "{}: {:.5f}".format(key, trainer.time_info[key].avg) + for key in trainer.time_info + ]) + + ips_msg = "ips: {:.5f} images/sec".format( + batch_size / trainer.time_info["batch_cost"].avg) + eta_sec = ((trainer.config["Global"]["epochs"] - epoch_id + 1 + ) * len(trainer.train_dataloader) - iter_id + ) * trainer.time_info["batch_cost"].avg + eta_msg = "eta: {:s}".format(str(datetime.timedelta(seconds=int(eta_sec)))) + logger.info("[Train][Epoch {}/{}][Iter: {}/{}]{}, {}, {}, {}, {}".format( + epoch_id, trainer.config["Global"]["epochs"], iter_id, + len(trainer.train_dataloader), lr_msg, metric_msg, time_msg, ips_msg, + eta_msg)) + + logger.scaler( + name="lr", + value=trainer.lr_sch.get_lr(), + step=trainer.global_step, + writer=trainer.vdl_writer) + for key in trainer.output_info: + logger.scaler( + name="train_{}".format(key), + value=trainer.output_info[key].avg, + step=trainer.global_step, + writer=trainer.vdl_writer) diff --git a/exprements/PaddleClas/ppcls/loss/__init__.py b/exprements/PaddleClas/ppcls/loss/__init__.py new file mode 100644 index 00000000..d15dab9d --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/__init__.py @@ -0,0 +1,67 @@ +import copy + +import paddle +import paddle.nn as nn +from ppcls.utils import logger + +from .celoss import CELoss, MixCELoss +from .googlenetloss import GoogLeNetLoss +from .centerloss import CenterLoss +from .emlloss import EmlLoss +from .msmloss import MSMLoss +from .npairsloss import NpairsLoss +from .trihardloss import TriHardLoss +from .triplet import TripletLoss, TripletLossV2 +from .supconloss import SupConLoss +from .pairwisecosface import PairwiseCosface +from .dmlloss import DMLLoss +from .distanceloss import DistanceLoss + +from .distillationloss import DistillationCELoss +from .distillationloss import DistillationGTCELoss +from .distillationloss import DistillationDMLLoss +from .distillationloss import DistillationDistanceLoss +from .distillationloss import DistillationRKDLoss +from .multilabelloss import MultiLabelLoss + +from .deephashloss import DSHSDLoss, LCDSHLoss + + +class CombinedLoss(nn.Layer): + def __init__(self, config_list): + super().__init__() + self.loss_func = [] + self.loss_weight = [] + assert isinstance(config_list, list), ( + 'operator config should be a list') + for config in config_list: + assert isinstance(config, + dict) and len(config) == 1, "yaml format error" + name = list(config)[0] + param = config[name] + assert "weight" in param, "weight must be in param, but param just contains {}".format( + param.keys()) + self.loss_weight.append(param.pop("weight")) + self.loss_func.append(eval(name)(**param)) + + def __call__(self, input, batch): + loss_dict = {} + # just for accelerate classification traing speed + if len(self.loss_func) == 1: + loss = self.loss_func[0](input, batch) + loss_dict.update(loss) + loss_dict["loss"] = list(loss.values())[0] + else: + for idx, loss_func in enumerate(self.loss_func): + loss = loss_func(input, batch) + weight = self.loss_weight[idx] + loss = {key: loss[key] * weight for key in loss} + loss_dict.update(loss) + loss_dict["loss"] = paddle.add_n(list(loss_dict.values())) + return loss_dict + + +def build_loss(config): + module_class = CombinedLoss(copy.deepcopy(config)) + logger.debug("build loss {} success.".format(module_class)) + return module_class diff --git a/exprements/PaddleClas/ppcls/loss/celoss.py b/exprements/PaddleClas/ppcls/loss/celoss.py new file mode 100644 index 00000000..a7892617 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/celoss.py @@ -0,0 +1,67 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import warnings + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + +from ppcls.utils import logger + + +class CELoss(nn.Layer): + """ + Cross entropy loss + """ + + def __init__(self, epsilon=None): + super().__init__() + if epsilon is not None and (epsilon <= 0 or epsilon >= 1): + epsilon = None + self.epsilon = epsilon + + def _labelsmoothing(self, target, class_num): + if len(target.shape) == 1 or target.shape[-1] != class_num: + one_hot_target = F.one_hot(target, class_num) + else: + one_hot_target = target + soft_target = F.label_smooth(one_hot_target, epsilon=self.epsilon) + soft_target = paddle.reshape(soft_target, shape=[-1, class_num]) + return soft_target + + def forward(self, x, label): + if isinstance(x, dict): + x = x["logits"] + if self.epsilon is not None: + class_num = x.shape[-1] + label = self._labelsmoothing(label, class_num) + x = -F.log_softmax(x, axis=-1) + loss = paddle.sum(x * label, axis=-1) + else: + if label.shape[-1] == x.shape[-1]: + label = F.softmax(label, axis=-1) + soft_label = True + else: + soft_label = False + loss = F.cross_entropy(x, label=label, soft_label=soft_label) + loss = loss.mean() + return {"CELoss": loss} + + +class MixCELoss(object): + def __init__(self, *args, **kwargs): + msg = "\"MixCELos\" is deprecated, please use \"CELoss\" instead." + logger.error(DeprecationWarning(msg)) + raise DeprecationWarning(msg) diff --git a/exprements/PaddleClas/ppcls/loss/centerloss.py b/exprements/PaddleClas/ppcls/loss/centerloss.py new file mode 100644 index 00000000..d85b3f2a --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/centerloss.py @@ -0,0 +1,54 @@ +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class CenterLoss(nn.Layer): + def __init__(self, num_classes=5013, feat_dim=2048): + super(CenterLoss, self).__init__() + self.num_classes = num_classes + self.feat_dim = feat_dim + self.centers = paddle.randn( + shape=[self.num_classes, self.feat_dim]).astype( + "float64") #random center + + def __call__(self, input, target): + """ + inputs: network output: {"features: xxx", "logits": xxxx} + target: image label + """ + feats = input["features"] + labels = target + batch_size = feats.shape[0] + + #calc feat * feat + dist1 = paddle.sum(paddle.square(feats), axis=1, keepdim=True) + dist1 = paddle.expand(dist1, [batch_size, self.num_classes]) + + #dist2 of centers + dist2 = paddle.sum(paddle.square(self.centers), axis=1, + keepdim=True) #num_classes + dist2 = paddle.expand(dist2, + [self.num_classes, batch_size]).astype("float64") + dist2 = paddle.transpose(dist2, [1, 0]) + + #first x * x + y * y + distmat = paddle.add(dist1, dist2) + tmp = paddle.matmul(feats, paddle.transpose(self.centers, [1, 0])) + distmat = distmat - 2.0 * tmp + + #generate the mask + classes = paddle.arange(self.num_classes).astype("int64") + labels = paddle.expand( + paddle.unsqueeze(labels, 1), (batch_size, self.num_classes)) + mask = paddle.equal( + paddle.expand(classes, [batch_size, self.num_classes]), + labels).astype("float64") #get mask + + dist = paddle.multiply(distmat, mask) + loss = paddle.sum(paddle.clip(dist, min=1e-12, max=1e+12)) / batch_size + + return {'CenterLoss': loss} diff --git a/exprements/PaddleClas/ppcls/loss/comfunc.py b/exprements/PaddleClas/ppcls/loss/comfunc.py new file mode 100644 index 00000000..277bdd6b --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/comfunc.py @@ -0,0 +1,45 @@ +# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np + + +def rerange_index(batch_size, samples_each_class): + tmp = np.arange(0, batch_size * batch_size) + tmp = tmp.reshape(-1, batch_size) + rerange_index = [] + + for i in range(batch_size): + step = i // samples_each_class + start = step * samples_each_class + end = (step + 1) * samples_each_class + + pos_idx = [] + neg_idx = [] + for j, k in enumerate(tmp[i]): + if j >= start and j < end: + if j == i: + pos_idx.insert(0, k) + else: + pos_idx.append(k) + else: + neg_idx.append(k) + rerange_index += (pos_idx + neg_idx) + + rerange_index = np.array(rerange_index).astype(np.int32) + return rerange_index diff --git a/exprements/PaddleClas/ppcls/loss/deephashloss.py b/exprements/PaddleClas/ppcls/loss/deephashloss.py new file mode 100644 index 00000000..c9a58dc7 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/deephashloss.py @@ -0,0 +1,92 @@ +#copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +#Unless required by applicable law or agreed to in writing, software +#distributed under the License is distributed on an "AS IS" BASIS, +#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +#See the License for the specific language governing permissions and +#limitations under the License. + +import paddle +import paddle.nn as nn + +class DSHSDLoss(nn.Layer): + """ + # DSHSD(IEEE ACCESS 2019) + # paper [Deep Supervised Hashing Based on Stable Distribution](https://ieeexplore.ieee.org/document/8648432/) + # [DSHSD] epoch:70, bit:48, dataset:cifar10-1, MAP:0.809, Best MAP: 0.809 + # [DSHSD] epoch:250, bit:48, dataset:nuswide_21, MAP:0.809, Best MAP: 0.815 + # [DSHSD] epoch:135, bit:48, dataset:imagenet, MAP:0.647, Best MAP: 0.647 + """ + def __init__(self, alpha, multi_label=False): + super(DSHSDLoss, self).__init__() + self.alpha = alpha + self.multi_label = multi_label + + def forward(self, input, label): + feature = input["features"] + logits = input["logits"] + + dist = paddle.sum(paddle.square( + (paddle.unsqueeze(feature, 1) - paddle.unsqueeze(feature, 0))), + axis=2) + + # label to ont-hot + label = paddle.flatten(label) + n_class = logits.shape[1] + label = paddle.nn.functional.one_hot(label, n_class).astype("float32") + + s = (paddle.matmul( + label, label, transpose_y=True) == 0).astype("float32") + margin = 2 * feature.shape[1] + Ld = (1 - s) / 2 * dist + s / 2 * (margin - dist).clip(min=0) + Ld = Ld.mean() + + if self.multi_label: + # multiple labels classification loss + Lc = (logits - label * logits + ( + (1 + (-logits).exp()).log())).sum(axis=1).mean() + else: + # single labels classification loss + Lc = (-paddle.nn.functional.softmax(logits).log() * label).sum( + axis=1).mean() + + return {"dshsdloss": Lc + Ld * self.alpha} + + +class LCDSHLoss(nn.Layer): + """ + # paper [Locality-Constrained Deep Supervised Hashing for Image Retrieval](https://www.ijcai.org/Proceedings/2017/0499.pdf) + # [LCDSH] epoch:145, bit:48, dataset:cifar10-1, MAP:0.798, Best MAP: 0.798 + # [LCDSH] epoch:183, bit:48, dataset:nuswide_21, MAP:0.833, Best MAP: 0.834 + """ + def __init__(self, n_class, _lambda): + super(LCDSHLoss, self).__init__() + self._lambda = _lambda + self.n_class = n_class + + def forward(self, input, label): + feature = input["features"] + + # label to ont-hot + label = paddle.flatten(label) + label = paddle.nn.functional.one_hot(label, self.n_class).astype("float32") + + s = 2 * (paddle.matmul(label, label, transpose_y=True) > 0).astype("float32") - 1 + inner_product = paddle.matmul(feature, feature, transpose_y=True) * 0.5 + + inner_product = inner_product.clip(min=-50, max=50) + L1 = paddle.log(1 + paddle.exp(-s * inner_product)).mean() + + b = feature.sign() + inner_product_ = paddle.matmul(b, b, transpose_y=True) * 0.5 + sigmoid = paddle.nn.Sigmoid() + L2 = (sigmoid(inner_product) - sigmoid(inner_product_)).pow(2).mean() + + return {"lcdshloss": L1 + self._lambda * L2} + diff --git a/exprements/PaddleClas/ppcls/loss/distanceloss.py b/exprements/PaddleClas/ppcls/loss/distanceloss.py new file mode 100644 index 00000000..0a09f0cb --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/distanceloss.py @@ -0,0 +1,43 @@ +#copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +#Unless required by applicable law or agreed to in writing, software +#distributed under the License is distributed on an "AS IS" BASIS, +#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +#See the License for the specific language governing permissions and +#limitations under the License. + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + +from paddle.nn import L1Loss +from paddle.nn import MSELoss as L2Loss +from paddle.nn import SmoothL1Loss + + +class DistanceLoss(nn.Layer): + """ + DistanceLoss: + mode: loss mode + """ + + def __init__(self, mode="l2", **kargs): + super().__init__() + assert mode in ["l1", "l2", "smooth_l1"] + if mode == "l1": + self.loss_func = nn.L1Loss(**kargs) + elif mode == "l2": + self.loss_func = nn.MSELoss(**kargs) + elif mode == "smooth_l1": + self.loss_func = nn.SmoothL1Loss(**kargs) + self.mode = mode + + def forward(self, x, y): + loss = self.loss_func(x, y) + return {"loss_{}".format(self.mode): loss} diff --git a/exprements/PaddleClas/ppcls/loss/distillationloss.py b/exprements/PaddleClas/ppcls/loss/distillationloss.py new file mode 100644 index 00000000..0340234b --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/distillationloss.py @@ -0,0 +1,174 @@ +#copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +#Unless required by applicable law or agreed to in writing, software +#distributed under the License is distributed on an "AS IS" BASIS, +#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +#See the License for the specific language governing permissions and +#limitations under the License. + +import paddle +import paddle.nn as nn + +from .celoss import CELoss +from .dmlloss import DMLLoss +from .distanceloss import DistanceLoss +from .rkdloss import RKdAngle, RkdDistance + + +class DistillationCELoss(CELoss): + """ + DistillationCELoss + """ + + def __init__(self, + model_name_pairs=[], + epsilon=None, + key=None, + name="loss_ce"): + super().__init__(epsilon=epsilon) + assert isinstance(model_name_pairs, list) + self.key = key + self.model_name_pairs = model_name_pairs + self.name = name + + def forward(self, predicts, batch): + loss_dict = dict() + for idx, pair in enumerate(self.model_name_pairs): + out1 = predicts[pair[0]] + out2 = predicts[pair[1]] + if self.key is not None: + out1 = out1[self.key] + out2 = out2[self.key] + loss = super().forward(out1, out2) + for key in loss: + loss_dict["{}_{}_{}".format(key, pair[0], pair[1])] = loss[key] + return loss_dict + + +class DistillationGTCELoss(CELoss): + """ + DistillationGTCELoss + """ + + def __init__(self, + model_names=[], + epsilon=None, + key=None, + name="loss_gt_ce"): + super().__init__(epsilon=epsilon) + assert isinstance(model_names, list) + self.key = key + self.model_names = model_names + self.name = name + + def forward(self, predicts, batch): + loss_dict = dict() + for name in self.model_names: + out = predicts[name] + if self.key is not None: + out = out[self.key] + loss = super().forward(out, batch) + for key in loss: + loss_dict["{}_{}".format(key, name)] = loss[key] + return loss_dict + + +class DistillationDMLLoss(DMLLoss): + """ + """ + + def __init__(self, + model_name_pairs=[], + act="softmax", + key=None, + name="loss_dml"): + super().__init__(act=act) + assert isinstance(model_name_pairs, list) + self.key = key + self.model_name_pairs = model_name_pairs + self.name = name + + def forward(self, predicts, batch): + loss_dict = dict() + for idx, pair in enumerate(self.model_name_pairs): + out1 = predicts[pair[0]] + out2 = predicts[pair[1]] + if self.key is not None: + out1 = out1[self.key] + out2 = out2[self.key] + loss = super().forward(out1, out2) + if isinstance(loss, dict): + for key in loss: + loss_dict["{}_{}_{}_{}".format(key, pair[0], pair[1], + idx)] = loss[key] + else: + loss_dict["{}_{}".format(self.name, idx)] = loss + return loss_dict + + +class DistillationDistanceLoss(DistanceLoss): + """ + """ + + def __init__(self, + mode="l2", + model_name_pairs=[], + key=None, + name="loss_", + **kargs): + super().__init__(mode=mode, **kargs) + assert isinstance(model_name_pairs, list) + self.key = key + self.model_name_pairs = model_name_pairs + self.name = name + mode + + def forward(self, predicts, batch): + loss_dict = dict() + for idx, pair in enumerate(self.model_name_pairs): + out1 = predicts[pair[0]] + out2 = predicts[pair[1]] + if self.key is not None: + out1 = out1[self.key] + out2 = out2[self.key] + loss = super().forward(out1, out2) + for key in loss: + loss_dict["{}_{}_{}".format(self.name, key, idx)] = loss[key] + return loss_dict + + +class DistillationRKDLoss(nn.Layer): + def __init__(self, + target_size=None, + model_name_pairs=(["Student", "Teacher"], ), + student_keepkeys=[], + teacher_keepkeys=[]): + super().__init__() + self.student_keepkeys = student_keepkeys + self.teacher_keepkeys = teacher_keepkeys + self.model_name_pairs = model_name_pairs + assert len(self.student_keepkeys) == len(self.teacher_keepkeys) + + self.rkd_angle_loss = RKdAngle(target_size=target_size) + self.rkd_dist_loss = RkdDistance(target_size=target_size) + + def __call__(self, predicts, batch): + loss_dict = {} + for m1, m2 in self.model_name_pairs: + for idx, ( + student_name, teacher_name + ) in enumerate(zip(self.student_keepkeys, self.teacher_keepkeys)): + student_out = predicts[m1][student_name] + teacher_out = predicts[m2][teacher_name] + + loss_dict[f"loss_angle_{idx}_{m1}_{m2}"] = self.rkd_angle_loss( + student_out, teacher_out) + loss_dict[f"loss_dist_{idx}_{m1}_{m2}"] = self.rkd_dist_loss( + student_out, teacher_out) + + return loss_dict diff --git a/exprements/PaddleClas/ppcls/loss/dmlloss.py b/exprements/PaddleClas/ppcls/loss/dmlloss.py new file mode 100644 index 00000000..48bf6c02 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/dmlloss.py @@ -0,0 +1,50 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class DMLLoss(nn.Layer): + """ + DMLLoss + """ + + def __init__(self, act="softmax", eps=1e-12): + super().__init__() + if act is not None: + assert act in ["softmax", "sigmoid"] + if act == "softmax": + self.act = nn.Softmax(axis=-1) + elif act == "sigmoid": + self.act = nn.Sigmoid() + else: + self.act = None + self.eps = eps + + def _kldiv(self, x, target): + class_num = x.shape[-1] + cost = target * paddle.log( + (target + self.eps) / (x + self.eps)) * class_num + return cost + + def forward(self, x, target): + if self.act is not None: + x = self.act(x) + target = self.act(target) + loss = self._kldiv(x, target) + self._kldiv(target, x) + loss = loss / 2 + loss = paddle.mean(loss) + return {"DMLLoss": loss} diff --git a/exprements/PaddleClas/ppcls/loss/emlloss.py b/exprements/PaddleClas/ppcls/loss/emlloss.py new file mode 100644 index 00000000..97357038 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/emlloss.py @@ -0,0 +1,97 @@ +# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import math +import paddle +import numpy as np +from .comfunc import rerange_index + + +class EmlLoss(paddle.nn.Layer): + def __init__(self, batch_size=40, samples_each_class=2): + super(EmlLoss, self).__init__() + assert (batch_size % samples_each_class == 0) + self.samples_each_class = samples_each_class + self.batch_size = batch_size + self.rerange_index = rerange_index(batch_size, samples_each_class) + self.thresh = 20.0 + self.beta = 100000 + + def surrogate_function(self, beta, theta, bias): + x = theta * paddle.exp(bias) + output = paddle.log(1 + beta * x) / math.log(1 + beta) + return output + + def surrogate_function_approximate(self, beta, theta, bias): + output = ( + paddle.log(theta) + bias + math.log(beta)) / math.log(1 + beta) + return output + + def surrogate_function_stable(self, beta, theta, target, thresh): + max_gap = paddle.to_tensor(thresh, dtype='float32') + max_gap.stop_gradient = True + + target_max = paddle.maximum(target, max_gap) + target_min = paddle.minimum(target, max_gap) + + loss1 = self.surrogate_function(beta, theta, target_min) + loss2 = self.surrogate_function_approximate(beta, theta, target_max) + bias = self.surrogate_function(beta, theta, max_gap) + loss = loss1 + loss2 - bias + return loss + + def forward(self, input, target=None): + features = input["features"] + samples_each_class = self.samples_each_class + batch_size = self.batch_size + rerange_index = self.rerange_index + + #calc distance + diffs = paddle.unsqueeze( + features, axis=1) - paddle.unsqueeze( + features, axis=0) + similary_matrix = paddle.sum(paddle.square(diffs), axis=-1) + + tmp = paddle.reshape(similary_matrix, shape=[-1, 1]) + rerange_index = paddle.to_tensor(rerange_index) + tmp = paddle.gather(tmp, index=rerange_index) + similary_matrix = paddle.reshape(tmp, shape=[-1, batch_size]) + + ignore, pos, neg = paddle.split( + similary_matrix, + num_or_sections=[ + 1, samples_each_class - 1, batch_size - samples_each_class + ], + axis=1) + ignore.stop_gradient = True + + pos_max = paddle.max(pos, axis=1, keepdim=True) + pos = paddle.exp(pos - pos_max) + pos_mean = paddle.mean(pos, axis=1, keepdim=True) + + neg_min = paddle.min(neg, axis=1, keepdim=True) + neg = paddle.exp(neg_min - neg) + neg_mean = paddle.mean(neg, axis=1, keepdim=True) + + bias = pos_max - neg_min + theta = paddle.multiply(neg_mean, pos_mean) + + loss = self.surrogate_function_stable(self.beta, theta, bias, + self.thresh) + loss = paddle.mean(loss) + return {"emlloss": loss} diff --git a/exprements/PaddleClas/ppcls/loss/googlenetloss.py b/exprements/PaddleClas/ppcls/loss/googlenetloss.py new file mode 100644 index 00000000..c580aa61 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/googlenetloss.py @@ -0,0 +1,41 @@ +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class GoogLeNetLoss(nn.Layer): + """ + Cross entropy loss used after googlenet + """ + def __init__(self, epsilon=None): + super().__init__() + assert (epsilon is None or epsilon <= 0 or epsilon >= 1), "googlenet is not support label_smooth" + + + def forward(self, inputs, label): + input0, input1, input2 = inputs + if isinstance(input0, dict): + input0 = input0["logits"] + if isinstance(input1, dict): + input1 = input1["logits"] + if isinstance(input2, dict): + input2 = input2["logits"] + + loss0 = F.cross_entropy(input0, label=label, soft_label=False) + loss1 = F.cross_entropy(input1, label=label, soft_label=False) + loss2 = F.cross_entropy(input2, label=label, soft_label=False) + loss = loss0 + 0.3 * loss1 + 0.3 * loss2 + loss = loss.mean() + return {"GooleNetLoss": loss} diff --git a/exprements/PaddleClas/ppcls/loss/msmloss.py b/exprements/PaddleClas/ppcls/loss/msmloss.py new file mode 100644 index 00000000..3aa0dd8b --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/msmloss.py @@ -0,0 +1,78 @@ +# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import paddle +from .comfunc import rerange_index + + +class MSMLoss(paddle.nn.Layer): + """ + MSMLoss Loss, based on triplet loss. USE P * K samples. + the batch size is fixed. Batch_size = P * K; but the K may vary between batches. + same label gather together + + supported_metrics = [ + 'euclidean', + 'sqeuclidean', + 'cityblock', + ] + only consider samples_each_class = 2 + """ + + def __init__(self, batch_size=120, samples_each_class=2, margin=0.1): + super(MSMLoss, self).__init__() + self.margin = margin + self.samples_each_class = samples_each_class + self.batch_size = batch_size + self.rerange_index = rerange_index(batch_size, samples_each_class) + + def forward(self, input, target=None): + #normalization + features = input["features"] + features = self._nomalize(features) + samples_each_class = self.samples_each_class + rerange_index = paddle.to_tensor(self.rerange_index) + + #calc sm + diffs = paddle.unsqueeze( + features, axis=1) - paddle.unsqueeze( + features, axis=0) + similary_matrix = paddle.sum(paddle.square(diffs), axis=-1) + + #rerange + tmp = paddle.reshape(similary_matrix, shape=[-1, 1]) + tmp = paddle.gather(tmp, index=rerange_index) + similary_matrix = paddle.reshape(tmp, shape=[-1, self.batch_size]) + + #split + ignore, pos, neg = paddle.split( + similary_matrix, + num_or_sections=[1, samples_each_class - 1, -1], + axis=1) + ignore.stop_gradient = True + + hard_pos = paddle.max(pos) + hard_neg = paddle.min(neg) + + loss = hard_pos + self.margin - hard_neg + loss = paddle.nn.ReLU()(loss) + return {"msmloss": loss} + + def _nomalize(self, input): + input_norm = paddle.sqrt( + paddle.sum(paddle.square(input), axis=1, keepdim=True)) + return paddle.divide(input, input_norm) diff --git a/exprements/PaddleClas/ppcls/loss/multilabelloss.py b/exprements/PaddleClas/ppcls/loss/multilabelloss.py new file mode 100644 index 00000000..d30d5b8d --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/multilabelloss.py @@ -0,0 +1,43 @@ +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class MultiLabelLoss(nn.Layer): + """ + Multi-label loss + """ + + def __init__(self, epsilon=None): + super().__init__() + if epsilon is not None and (epsilon <= 0 or epsilon >= 1): + epsilon = None + self.epsilon = epsilon + + def _labelsmoothing(self, target, class_num): + if target.ndim == 1 or target.shape[-1] != class_num: + one_hot_target = F.one_hot(target, class_num) + else: + one_hot_target = target + soft_target = F.label_smooth(one_hot_target, epsilon=self.epsilon) + soft_target = paddle.reshape(soft_target, shape=[-1, class_num]) + return soft_target + + def _binary_crossentropy(self, input, target, class_num): + if self.epsilon is not None: + target = self._labelsmoothing(target, class_num) + cost = F.binary_cross_entropy_with_logits( + logit=input, label=target) + else: + cost = F.binary_cross_entropy_with_logits( + logit=input, label=target) + + return cost + + def forward(self, x, target): + if isinstance(x, dict): + x = x["logits"] + class_num = x.shape[-1] + loss = self._binary_crossentropy(x, target, class_num) + loss = loss.mean() + return {"MultiLabelLoss": loss} diff --git a/exprements/PaddleClas/ppcls/loss/npairsloss.py b/exprements/PaddleClas/ppcls/loss/npairsloss.py new file mode 100644 index 00000000..d4b359e8 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/npairsloss.py @@ -0,0 +1,38 @@ +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import paddle + + +class NpairsLoss(paddle.nn.Layer): + def __init__(self, reg_lambda=0.01): + super(NpairsLoss, self).__init__() + self.reg_lambda = reg_lambda + + def forward(self, input, target=None): + """ + anchor and positive(should include label) + """ + features = input["features"] + reg_lambda = self.reg_lambda + batch_size = features.shape[0] + fea_dim = features.shape[1] + num_class = batch_size // 2 + + #reshape + out_feas = paddle.reshape(features, shape=[-1, 2, fea_dim]) + anc_feas, pos_feas = paddle.split(out_feas, num_or_sections=2, axis=1) + anc_feas = paddle.squeeze(anc_feas, axis=1) + pos_feas = paddle.squeeze(pos_feas, axis=1) + + #get simi matrix + similarity_matrix = paddle.matmul( + anc_feas, pos_feas, transpose_y=True) #get similarity matrix + sparse_labels = paddle.arange(0, num_class, dtype='int64') + xentloss = paddle.nn.CrossEntropyLoss()( + similarity_matrix, sparse_labels) #by default: mean + + #l2 norm + reg = paddle.mean(paddle.sum(paddle.square(features), axis=1)) + l2loss = 0.5 * reg_lambda * reg + return {"npairsloss": xentloss + l2loss} diff --git a/exprements/PaddleClas/ppcls/loss/pairwisecosface.py b/exprements/PaddleClas/ppcls/loss/pairwisecosface.py new file mode 100644 index 00000000..beb80686 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/pairwisecosface.py @@ -0,0 +1,55 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import math +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class PairwiseCosface(nn.Layer): + def __init__(self, margin, gamma): + super(PairwiseCosface, self).__init__() + self.margin = margin + self.gamma = gamma + + def forward(self, embedding, targets): + if isinstance(embedding, dict): + embedding = embedding['features'] + # Normalize embedding features + embedding = F.normalize(embedding, axis=1) + dist_mat = paddle.matmul(embedding, embedding, transpose_y=True) + + N = dist_mat.shape[0] + is_pos = targets.reshape([N,1]).expand([N,N]).equal(paddle.t(targets.reshape([N,1]).expand([N,N]))).astype('float') + is_neg = targets.reshape([N,1]).expand([N,N]).not_equal(paddle.t(targets.reshape([N,1]).expand([N,N]))).astype('float') + + # Mask scores related to itself + is_pos = is_pos - paddle.eye(N, N) + + s_p = dist_mat * is_pos + s_n = dist_mat * is_neg + + logit_p = -self.gamma * s_p + (-99999999.) * (1 - is_pos) + logit_n = self.gamma * (s_n + self.margin) + (-99999999.) * (1 - is_neg) + + loss = F.softplus(paddle.logsumexp(logit_p, axis=1) + paddle.logsumexp(logit_n, axis=1)).mean() + + return {"PairwiseCosface": loss} + + diff --git a/exprements/PaddleClas/ppcls/loss/rkdloss.py b/exprements/PaddleClas/ppcls/loss/rkdloss.py new file mode 100644 index 00000000..e6ffea27 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/rkdloss.py @@ -0,0 +1,97 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +def pdist(e, squared=False, eps=1e-12): + e_square = e.pow(2).sum(axis=1) + prod = paddle.mm(e, e.t()) + res = (e_square.unsqueeze(1) + e_square.unsqueeze(0) - 2 * prod).clip( + min=eps) + + if not squared: + res = res.sqrt() + return res + + +class RKdAngle(nn.Layer): + # reference: https://github.com/lenscloth/RKD/blob/master/metric/loss.py + def __init__(self, target_size=None): + super().__init__() + if target_size is not None: + self.avgpool = paddle.nn.AdaptiveAvgPool2D(target_size) + else: + self.avgpool = None + + def forward(self, student, teacher): + # GAP to reduce memory + if self.avgpool is not None: + # NxC1xH1xW1 -> NxC1x1x1 + student = self.avgpool(student) + # NxC2xH2xW2 -> NxC2x1x1 + teacher = self.avgpool(teacher) + + # reshape for feature map distillation + bs = student.shape[0] + student = student.reshape([bs, -1]) + teacher = teacher.reshape([bs, -1]) + + td = (teacher.unsqueeze(0) - teacher.unsqueeze(1)) + norm_td = F.normalize(td, p=2, axis=2) + t_angle = paddle.bmm(norm_td, norm_td.transpose([0, 2, 1])).reshape( + [-1, 1]) + + sd = (student.unsqueeze(0) - student.unsqueeze(1)) + norm_sd = F.normalize(sd, p=2, axis=2) + s_angle = paddle.bmm(norm_sd, norm_sd.transpose([0, 2, 1])).reshape( + [-1, 1]) + loss = F.smooth_l1_loss(s_angle, t_angle, reduction='mean') + return loss + + +class RkdDistance(nn.Layer): + # reference: https://github.com/lenscloth/RKD/blob/master/metric/loss.py + def __init__(self, eps=1e-12, target_size=1): + super().__init__() + self.eps = eps + if target_size is not None: + self.avgpool = paddle.nn.AdaptiveAvgPool2D(target_size) + else: + self.avgpool = None + + def forward(self, student, teacher): + # GAP to reduce memory + if self.avgpool is not None: + # NxC1xH1xW1 -> NxC1x1x1 + student = self.avgpool(student) + # NxC2xH2xW2 -> NxC2x1x1 + teacher = self.avgpool(teacher) + + bs = student.shape[0] + student = student.reshape([bs, -1]) + teacher = teacher.reshape([bs, -1]) + + t_d = pdist(teacher, squared=False) + mean_td = t_d.mean() + t_d = t_d / (mean_td + self.eps) + + d = pdist(student, squared=False) + mean_d = d.mean() + d = d / (mean_d + self.eps) + + loss = F.smooth_l1_loss(d, t_d, reduction="mean") + return loss diff --git a/exprements/PaddleClas/ppcls/loss/supconloss.py b/exprements/PaddleClas/ppcls/loss/supconloss.py new file mode 100644 index 00000000..3dd33bc1 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/supconloss.py @@ -0,0 +1,108 @@ +import paddle +from paddle import nn + + +class SupConLoss(nn.Layer): + """Supervised Contrastive Learning: https://arxiv.org/pdf/2004.11362.pdf. + It also supports the unsupervised contrastive loss in SimCLR""" + + def __init__(self, + views=16, + temperature=0.07, + contrast_mode='all', + base_temperature=0.07, + normalize_feature=True): + super(SupConLoss, self).__init__() + self.temperature = paddle.to_tensor(temperature) + self.contrast_mode = contrast_mode + self.base_temperature = paddle.to_tensor(base_temperature) + self.num_ids = None + self.views = views + self.normalize_feature = normalize_feature + + def forward(self, features, labels, mask=None): + """Compute loss for model. If both `labels` and `mask` are None, + it degenerates to SimCLR unsupervised loss: + https://arxiv.org/pdf/2002.05709.pdf + Args: + features: hidden vector of shape [bsz, n_views, ...]. + labels: ground truth of shape [bsz]. + mask: contrastive mask of shape [bsz, bsz], mask_{i,j}=1 if sample j + has the same class as sample i. Can be asymmetric. + Returns: + A loss scalar. + """ + features = features["features"] + if self.num_ids is None: + self.num_ids = int(features.shape[0] / self.views) + + if self.normalize_feature: + features = 1. * features / (paddle.expand_as( + paddle.norm( + features, p=2, axis=-1, keepdim=True), features) + 1e-12) + features = features.reshape([self.num_ids, self.views, -1]) + labels = labels.reshape([self.num_ids, self.views])[:, 0] + + if len(features.shape) < 3: + raise ValueError('`features` needs to be [bsz, n_views, ...],' + 'at least 3 dimensions are required') + if len(features.shape) > 3: + features = features.reshape( + [features.shape[0], features.shape[1], -1]) + + batch_size = features.shape[0] + if labels is not None and mask is not None: + raise ValueError('Cannot define both `labels` and `mask`') + elif labels is None and mask is None: + mask = paddle.eye(batch_size, dtype='float32') + elif labels is not None: + labels = labels.reshape([-1, 1]) + if labels.shape[0] != batch_size: + raise ValueError( + 'Num of labels does not match num of features') + mask = paddle.cast( + paddle.equal(labels, paddle.t(labels)), 'float32') + else: + mask = paddle.cast(mask, 'float32') + + contrast_count = features.shape[1] + contrast_feature = paddle.concat( + paddle.unbind( + features, axis=1), axis=0) + if self.contrast_mode == 'one': + anchor_feature = features[:, 0] + anchor_count = 1 + elif self.contrast_mode == 'all': + anchor_feature = contrast_feature + anchor_count = contrast_count + else: + raise ValueError('Unknown mode: {}'.format(self.contrast_mode)) + + # compute logits + anchor_dot_contrast = paddle.divide( + paddle.matmul(anchor_feature, paddle.t(contrast_feature)), + self.temperature) + # for numerical stability + logits_max = paddle.max(anchor_dot_contrast, axis=1, keepdim=True) + logits = anchor_dot_contrast - logits_max.detach() + + # tile mask + mask = paddle.tile(mask, [anchor_count, contrast_count]) + + logits_mask = 1 - paddle.eye(batch_size * anchor_count) + mask = mask * logits_mask + + # compute log_prob + exp_logits = paddle.exp(logits) * logits_mask + log_prob = logits - paddle.log( + paddle.sum(exp_logits, axis=1, keepdim=True)) + + # compute mean of log-likelihood over positive + mean_log_prob_pos = paddle.sum((mask * log_prob), + axis=1) / paddle.sum(mask, axis=1) + + # loss + loss = -(self.temperature / self.base_temperature) * mean_log_prob_pos + loss = paddle.mean(loss.reshape([anchor_count, batch_size])) + + return {"SupConLoss": loss} diff --git a/exprements/PaddleClas/ppcls/loss/trihardloss.py b/exprements/PaddleClas/ppcls/loss/trihardloss.py new file mode 100644 index 00000000..132c604d --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/trihardloss.py @@ -0,0 +1,82 @@ +# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import paddle +from .comfunc import rerange_index + + +class TriHardLoss(paddle.nn.Layer): + """ + TriHard Loss, based on triplet loss. USE P * K samples. + the batch size is fixed. Batch_size = P * K; but the K may vary between batches. + same label gather together + + supported_metrics = [ + 'euclidean', + 'sqeuclidean', + 'cityblock', + ] + only consider samples_each_class = 2 + """ + + def __init__(self, batch_size=120, samples_each_class=2, margin=0.1): + super(TriHardLoss, self).__init__() + self.margin = margin + self.samples_each_class = samples_each_class + self.batch_size = batch_size + self.rerange_index = rerange_index(batch_size, samples_each_class) + + def forward(self, input, target=None): + features = input["features"] + assert (self.batch_size == features.shape[0]) + + #normalization + features = self._nomalize(features) + samples_each_class = self.samples_each_class + rerange_index = paddle.to_tensor(self.rerange_index) + + #calc sm + diffs = paddle.unsqueeze( + features, axis=1) - paddle.unsqueeze( + features, axis=0) + similary_matrix = paddle.sum(paddle.square(diffs), axis=-1) + + #rerange + tmp = paddle.reshape(similary_matrix, shape=[-1, 1]) + tmp = paddle.gather(tmp, index=rerange_index) + similary_matrix = paddle.reshape(tmp, shape=[-1, self.batch_size]) + + #split + ignore, pos, neg = paddle.split( + similary_matrix, + num_or_sections=[1, samples_each_class - 1, -1], + axis=1) + + ignore.stop_gradient = True + hard_pos = paddle.max(pos, axis=1) + hard_neg = paddle.min(neg, axis=1) + + loss = hard_pos + self.margin - hard_neg + loss = paddle.nn.ReLU()(loss) + loss = paddle.mean(loss) + return {"trihardloss": loss} + + def _nomalize(self, input): + input_norm = paddle.sqrt( + paddle.sum(paddle.square(input), axis=1, keepdim=True)) + return paddle.divide(input, input_norm) diff --git a/exprements/PaddleClas/ppcls/loss/triplet.py b/exprements/PaddleClas/ppcls/loss/triplet.py new file mode 100644 index 00000000..d1c7eec9 --- /dev/null +++ b/exprements/PaddleClas/ppcls/loss/triplet.py @@ -0,0 +1,137 @@ +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import paddle +import paddle.nn as nn + + +class TripletLossV2(nn.Layer): + """Triplet loss with hard positive/negative mining. + Args: + margin (float): margin for triplet. + """ + + def __init__(self, margin=0.5, normalize_feature=True): + super(TripletLossV2, self).__init__() + self.margin = margin + self.ranking_loss = paddle.nn.loss.MarginRankingLoss(margin=margin) + self.normalize_feature = normalize_feature + + def forward(self, input, target): + """ + Args: + inputs: feature matrix with shape (batch_size, feat_dim) + target: ground truth labels with shape (num_classes) + """ + inputs = input["features"] + + if self.normalize_feature: + inputs = 1. * inputs / (paddle.expand_as( + paddle.norm( + inputs, p=2, axis=-1, keepdim=True), inputs) + 1e-12) + + bs = inputs.shape[0] + + # compute distance + dist = paddle.pow(inputs, 2).sum(axis=1, keepdim=True).expand([bs, bs]) + dist = dist + dist.t() + dist = paddle.addmm( + input=dist, x=inputs, y=inputs.t(), alpha=-2.0, beta=1.0) + dist = paddle.clip(dist, min=1e-12).sqrt() + + # hard negative mining + is_pos = paddle.expand(target, ( + bs, bs)).equal(paddle.expand(target, (bs, bs)).t()) + is_neg = paddle.expand(target, ( + bs, bs)).not_equal(paddle.expand(target, (bs, bs)).t()) + + # `dist_ap` means distance(anchor, positive) + ## both `dist_ap` and `relative_p_inds` with shape [N, 1] + ''' + dist_ap, relative_p_inds = paddle.max( + paddle.reshape(dist[is_pos], (bs, -1)), axis=1, keepdim=True) + # `dist_an` means distance(anchor, negative) + # both `dist_an` and `relative_n_inds` with shape [N, 1] + dist_an, relative_n_inds = paddle.min( + paddle.reshape(dist[is_neg], (bs, -1)), axis=1, keepdim=True) + ''' + dist_ap = paddle.max(paddle.reshape( + paddle.masked_select(dist, is_pos), (bs, -1)), + axis=1, + keepdim=True) + # `dist_an` means distance(anchor, negative) + # both `dist_an` and `relative_n_inds` with shape [N, 1] + dist_an = paddle.min(paddle.reshape( + paddle.masked_select(dist, is_neg), (bs, -1)), + axis=1, + keepdim=True) + # shape [N] + dist_ap = paddle.squeeze(dist_ap, axis=1) + dist_an = paddle.squeeze(dist_an, axis=1) + + # Compute ranking hinge loss + y = paddle.ones_like(dist_an) + loss = self.ranking_loss(dist_an, dist_ap, y) + return {"TripletLossV2": loss} + + +class TripletLoss(nn.Layer): + """Triplet loss with hard positive/negative mining. + Reference: + Hermans et al. In Defense of the Triplet Loss for Person Re-Identification. arXiv:1703.07737. + Code imported from https://github.com/Cysu/open-reid/blob/master/reid/loss/triplet.py. + Args: + margin (float): margin for triplet. + """ + + def __init__(self, margin=1.0): + super(TripletLoss, self).__init__() + self.margin = margin + self.ranking_loss = paddle.nn.loss.MarginRankingLoss(margin=margin) + + def forward(self, input, target): + """ + Args: + inputs: feature matrix with shape (batch_size, feat_dim) + target: ground truth labels with shape (num_classes) + """ + inputs = input["features"] + + bs = inputs.shape[0] + # Compute pairwise distance, replace by the official when merged + dist = paddle.pow(inputs, 2).sum(axis=1, keepdim=True).expand([bs, bs]) + dist = dist + dist.t() + dist = paddle.addmm( + input=dist, x=inputs, y=inputs.t(), alpha=-2.0, beta=1.0) + dist = paddle.clip(dist, min=1e-12).sqrt() + + mask = paddle.equal( + target.expand([bs, bs]), target.expand([bs, bs]).t()) + mask_numpy_idx = mask.numpy() + dist_ap, dist_an = [], [] + for i in range(bs): + # dist_ap_i = paddle.to_tensor(dist[i].numpy()[mask_numpy_idx[i]].max(),dtype='float64').unsqueeze(0) + # dist_ap_i.stop_gradient = False + # dist_ap.append(dist_ap_i) + dist_ap.append( + max([ + dist[i][j] if mask_numpy_idx[i][j] == True else float( + "-inf") for j in range(bs) + ]).unsqueeze(0)) + # dist_an_i = paddle.to_tensor(dist[i].numpy()[mask_numpy_idx[i] == False].min(), dtype='float64').unsqueeze(0) + # dist_an_i.stop_gradient = False + # dist_an.append(dist_an_i) + dist_an.append( + min([ + dist[i][k] if mask_numpy_idx[i][k] == False else float( + "inf") for k in range(bs) + ]).unsqueeze(0)) + + dist_ap = paddle.concat(dist_ap, axis=0) + dist_an = paddle.concat(dist_an, axis=0) + + # Compute ranking hinge loss + y = paddle.ones_like(dist_an) + loss = self.ranking_loss(dist_an, dist_ap, y) + return {"TripletLoss": loss} diff --git a/exprements/PaddleClas/ppcls/metric/__init__.py b/exprements/PaddleClas/ppcls/metric/__init__.py new file mode 100644 index 00000000..94721235 --- /dev/null +++ b/exprements/PaddleClas/ppcls/metric/__init__.py @@ -0,0 +1,51 @@ +#copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +#Unless required by applicable law or agreed to in writing, software +#distributed under the License is distributed on an "AS IS" BASIS, +#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +#See the License for the specific language governing permissions and +#limitations under the License. + +from paddle import nn +import copy +from collections import OrderedDict + +from .metrics import TopkAcc, mAP, mINP, Recallk, Precisionk +from .metrics import DistillationTopkAcc +from .metrics import GoogLeNetTopkAcc +from .metrics import HammingDistance, AccuracyScore + + +class CombinedMetrics(nn.Layer): + def __init__(self, config_list): + super().__init__() + self.metric_func_list = [] + assert isinstance(config_list, list), ( + 'operator config should be a list') + for config in config_list: + assert isinstance(config, + dict) and len(config) == 1, "yaml format error" + metric_name = list(config)[0] + metric_params = config[metric_name] + if metric_params is not None: + self.metric_func_list.append( + eval(metric_name)(**metric_params)) + else: + self.metric_func_list.append(eval(metric_name)()) + + def __call__(self, *args, **kwargs): + metric_dict = OrderedDict() + for idx, metric_func in enumerate(self.metric_func_list): + metric_dict.update(metric_func(*args, **kwargs)) + return metric_dict + + +def build_metrics(config): + metrics_list = CombinedMetrics(copy.deepcopy(config)) + return metrics_list diff --git a/exprements/PaddleClas/ppcls/metric/metrics.py b/exprements/PaddleClas/ppcls/metric/metrics.py new file mode 100644 index 00000000..7c6407e7 --- /dev/null +++ b/exprements/PaddleClas/ppcls/metric/metrics.py @@ -0,0 +1,309 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import numpy as np +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + +from sklearn.metrics import hamming_loss +from sklearn.metrics import accuracy_score as accuracy_metric +from sklearn.metrics import multilabel_confusion_matrix +from sklearn.preprocessing import binarize + + +class TopkAcc(nn.Layer): + def __init__(self, topk=(1, 5)): + super().__init__() + assert isinstance(topk, (int, list, tuple)) + if isinstance(topk, int): + topk = [topk] + self.topk = topk + + def forward(self, x, label): + if isinstance(x, dict): + x = x["logits"] + + metric_dict = dict() + for k in self.topk: + metric_dict["top{}".format(k)] = paddle.metric.accuracy( + x, label, k=k) + return metric_dict + + +class mAP(nn.Layer): + def __init__(self): + super().__init__() + + def forward(self, similarities_matrix, query_img_id, gallery_img_id, + keep_mask): + metric_dict = dict() + + choosen_indices = paddle.argsort( + similarities_matrix, axis=1, descending=True) + gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0]) + gallery_labels_transpose = paddle.broadcast_to( + gallery_labels_transpose, + shape=[ + choosen_indices.shape[0], gallery_labels_transpose.shape[1] + ]) + choosen_label = paddle.index_sample(gallery_labels_transpose, + choosen_indices) + equal_flag = paddle.equal(choosen_label, query_img_id) + if keep_mask is not None: + keep_mask = paddle.index_sample( + keep_mask.astype('float32'), choosen_indices) + equal_flag = paddle.logical_and(equal_flag, + keep_mask.astype('bool')) + equal_flag = paddle.cast(equal_flag, 'float32') + + num_rel = paddle.sum(equal_flag, axis=1) + num_rel = paddle.greater_than(num_rel, paddle.to_tensor(0.)) + num_rel_index = paddle.nonzero(num_rel.astype("int")) + num_rel_index = paddle.reshape(num_rel_index, [num_rel_index.shape[0]]) + equal_flag = paddle.index_select(equal_flag, num_rel_index, axis=0) + + acc_sum = paddle.cumsum(equal_flag, axis=1) + div = paddle.arange(acc_sum.shape[1]).astype("float32") + 1 + precision = paddle.divide(acc_sum, div) + + #calc map + precision_mask = paddle.multiply(equal_flag, precision) + ap = paddle.sum(precision_mask, axis=1) / paddle.sum(equal_flag, + axis=1) + metric_dict["mAP"] = paddle.mean(ap).numpy()[0] + return metric_dict + + +class mINP(nn.Layer): + def __init__(self): + super().__init__() + + def forward(self, similarities_matrix, query_img_id, gallery_img_id, + keep_mask): + metric_dict = dict() + + choosen_indices = paddle.argsort( + similarities_matrix, axis=1, descending=True) + gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0]) + gallery_labels_transpose = paddle.broadcast_to( + gallery_labels_transpose, + shape=[ + choosen_indices.shape[0], gallery_labels_transpose.shape[1] + ]) + choosen_label = paddle.index_sample(gallery_labels_transpose, + choosen_indices) + equal_flag = paddle.equal(choosen_label, query_img_id) + if keep_mask is not None: + keep_mask = paddle.index_sample( + keep_mask.astype('float32'), choosen_indices) + equal_flag = paddle.logical_and(equal_flag, + keep_mask.astype('bool')) + equal_flag = paddle.cast(equal_flag, 'float32') + + num_rel = paddle.sum(equal_flag, axis=1) + num_rel = paddle.greater_than(num_rel, paddle.to_tensor(0.)) + num_rel_index = paddle.nonzero(num_rel.astype("int")) + num_rel_index = paddle.reshape(num_rel_index, [num_rel_index.shape[0]]) + equal_flag = paddle.index_select(equal_flag, num_rel_index, axis=0) + + #do accumulative sum + div = paddle.arange(equal_flag.shape[1]).astype("float32") + 2 + minus = paddle.divide(equal_flag, div) + auxilary = paddle.subtract(equal_flag, minus) + hard_index = paddle.argmax(auxilary, axis=1).astype("float32") + all_INP = paddle.divide(paddle.sum(equal_flag, axis=1), hard_index) + mINP = paddle.mean(all_INP) + metric_dict["mINP"] = mINP.numpy()[0] + return metric_dict + + +class Recallk(nn.Layer): + def __init__(self, topk=(1, 5)): + super().__init__() + assert isinstance(topk, (int, list, tuple)) + if isinstance(topk, int): + topk = [topk] + self.topk = topk + + def forward(self, similarities_matrix, query_img_id, gallery_img_id, + keep_mask): + metric_dict = dict() + + #get cmc + choosen_indices = paddle.argsort( + similarities_matrix, axis=1, descending=True) + gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0]) + gallery_labels_transpose = paddle.broadcast_to( + gallery_labels_transpose, + shape=[ + choosen_indices.shape[0], gallery_labels_transpose.shape[1] + ]) + choosen_label = paddle.index_sample(gallery_labels_transpose, + choosen_indices) + equal_flag = paddle.equal(choosen_label, query_img_id) + if keep_mask is not None: + keep_mask = paddle.index_sample( + keep_mask.astype('float32'), choosen_indices) + equal_flag = paddle.logical_and(equal_flag, + keep_mask.astype('bool')) + equal_flag = paddle.cast(equal_flag, 'float32') + real_query_num = paddle.sum(equal_flag, axis=1) + real_query_num = paddle.sum( + paddle.greater_than(real_query_num, paddle.to_tensor(0.)).astype( + "float32")) + + acc_sum = paddle.cumsum(equal_flag, axis=1) + mask = paddle.greater_than(acc_sum, + paddle.to_tensor(0.)).astype("float32") + all_cmc = (paddle.sum(mask, axis=0) / real_query_num).numpy() + + for k in self.topk: + metric_dict["recall{}".format(k)] = all_cmc[k - 1] + return metric_dict + + +class Precisionk(nn.Layer): + def __init__(self, topk=(1, 5)): + super().__init__() + assert isinstance(topk, (int, list, tuple)) + if isinstance(topk, int): + topk = [topk] + self.topk = topk + + def forward(self, similarities_matrix, query_img_id, gallery_img_id, + keep_mask): + metric_dict = dict() + + #get cmc + choosen_indices = paddle.argsort( + similarities_matrix, axis=1, descending=True) + gallery_labels_transpose = paddle.transpose(gallery_img_id, [1, 0]) + gallery_labels_transpose = paddle.broadcast_to( + gallery_labels_transpose, + shape=[ + choosen_indices.shape[0], gallery_labels_transpose.shape[1] + ]) + choosen_label = paddle.index_sample(gallery_labels_transpose, + choosen_indices) + equal_flag = paddle.equal(choosen_label, query_img_id) + if keep_mask is not None: + keep_mask = paddle.index_sample( + keep_mask.astype('float32'), choosen_indices) + equal_flag = paddle.logical_and(equal_flag, + keep_mask.astype('bool')) + equal_flag = paddle.cast(equal_flag, 'float32') + + Ns = paddle.arange(gallery_img_id.shape[0]) + 1 + equal_flag_cumsum = paddle.cumsum(equal_flag, axis=1) + Precision_at_k = (paddle.mean(equal_flag_cumsum, axis=0) / Ns).numpy() + + for k in self.topk: + metric_dict["precision@{}".format(k)] = Precision_at_k[k - 1] + + return metric_dict + + +class DistillationTopkAcc(TopkAcc): + def __init__(self, model_key, feature_key=None, topk=(1, 5)): + super().__init__(topk=topk) + self.model_key = model_key + self.feature_key = feature_key + + def forward(self, x, label): + if isinstance(x, dict): + x = x[self.model_key] + if self.feature_key is not None: + x = x[self.feature_key] + return super().forward(x, label) + + +class GoogLeNetTopkAcc(TopkAcc): + def __init__(self, topk=(1, 5)): + super().__init__() + assert isinstance(topk, (int, list, tuple)) + if isinstance(topk, int): + topk = [topk] + self.topk = topk + + def forward(self, x, label): + return super().forward(x[0], label) + + +class MutiLabelMetric(object): + def __init__(self): + pass + + def _multi_hot_encode(self, logits, threshold=0.5): + return binarize(logits, threshold=threshold) + + def __call__(self, output): + output = F.sigmoid(output) + preds = self._multi_hot_encode(logits=output.numpy(), threshold=0.5) + return preds + + +class HammingDistance(MutiLabelMetric): + """ + Soft metric based label for multilabel classification + Returns: + The smaller the return value is, the better model is. + """ + + def __init__(self): + super().__init__() + + def __call__(self, output, target): + preds = super().__call__(output) + metric_dict = dict() + metric_dict["HammingDistance"] = paddle.to_tensor( + hamming_loss(target, preds)) + return metric_dict + + +class AccuracyScore(MutiLabelMetric): + """ + Hard metric for multilabel classification + Args: + base: ["sample", "label"], default="sample" + if "sample", return metric score based sample, + if "label", return metric score based label. + Returns: + accuracy: + """ + + def __init__(self, base="label"): + super().__init__() + assert base in ["sample", "label" + ], 'must be one of ["sample", "label"]' + self.base = base + + def __call__(self, output, target): + preds = super().__call__(output) + metric_dict = dict() + if self.base == "sample": + accuracy = accuracy_metric(target, preds) + elif self.base == "label": + mcm = multilabel_confusion_matrix(target, preds) + tns = mcm[:, 0, 0] + fns = mcm[:, 1, 0] + tps = mcm[:, 1, 1] + fps = mcm[:, 0, 1] + accuracy = (sum(tps) + sum(tns)) / ( + sum(tps) + sum(tns) + sum(fns) + sum(fps)) + precision = sum(tps) / (sum(tps) + sum(fps)) + recall = sum(tps) / (sum(tps) + sum(fns)) + F1 = 2 * (accuracy * recall) / (accuracy + recall) + metric_dict["AccuracyScore"] = paddle.to_tensor(accuracy) + return metric_dict diff --git a/exprements/PaddleClas/ppcls/optimizer/__init__.py b/exprements/PaddleClas/ppcls/optimizer/__init__.py new file mode 100644 index 00000000..61db39f8 --- /dev/null +++ b/exprements/PaddleClas/ppcls/optimizer/__init__.py @@ -0,0 +1,72 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import copy +import paddle + +from ppcls.utils import logger + +from . import optimizer + +__all__ = ['build_optimizer'] + + +def build_lr_scheduler(lr_config, epochs, step_each_epoch): + from . import learning_rate + lr_config.update({'epochs': epochs, 'step_each_epoch': step_each_epoch}) + if 'name' in lr_config: + lr_name = lr_config.pop('name') + lr = getattr(learning_rate, lr_name)(**lr_config) + if isinstance(lr, paddle.optimizer.lr.LRScheduler): + return lr + else: + return lr() + else: + lr = lr_config['learning_rate'] + return lr + + +# model_list is None in static graph +def build_optimizer(config, epochs, step_each_epoch, model_list=None): + config = copy.deepcopy(config) + # step1 build lr + lr = build_lr_scheduler(config.pop('lr'), epochs, step_each_epoch) + logger.debug("build lr ({}) success..".format(lr)) + # step2 build regularization + if 'regularizer' in config and config['regularizer'] is not None: + if 'weight_decay' in config: + logger.warning( + "ConfigError: Only one of regularizer and weight_decay can be set in Optimizer Config. \"weight_decay\" has been ignored." + ) + reg_config = config.pop('regularizer') + reg_name = reg_config.pop('name') + 'Decay' + reg = getattr(paddle.regularizer, reg_name)(**reg_config) + config["weight_decay"] = reg + logger.debug("build regularizer ({}) success..".format(reg)) + # step3 build optimizer + optim_name = config.pop('name') + if 'clip_norm' in config: + clip_norm = config.pop('clip_norm') + grad_clip = paddle.nn.ClipGradByNorm(clip_norm=clip_norm) + else: + grad_clip = None + optim = getattr(optimizer, optim_name)(learning_rate=lr, + grad_clip=grad_clip, + **config)(model_list=model_list) + logger.debug("build optimizer ({}) success..".format(optim)) + return optim, lr diff --git a/exprements/PaddleClas/ppcls/optimizer/learning_rate.py b/exprements/PaddleClas/ppcls/optimizer/learning_rate.py new file mode 100644 index 00000000..b59387dd --- /dev/null +++ b/exprements/PaddleClas/ppcls/optimizer/learning_rate.py @@ -0,0 +1,326 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import (absolute_import, division, print_function, + unicode_literals) + +from paddle.optimizer import lr +from paddle.optimizer.lr import LRScheduler + +from ppcls.utils import logger + + +class Linear(object): + """ + Linear learning rate decay + Args: + lr (float): The initial learning rate. It is a python float number. + epochs(int): The decay step size. It determines the decay cycle. + end_lr(float, optional): The minimum final learning rate. Default: 0.0001. + power(float, optional): Power of polynomial. Default: 1.0. + warmup_epoch(int): The epoch numbers for LinearWarmup. Default: 0. + warmup_start_lr(float): Initial learning rate of warm up. Default: 0.0. + last_epoch (int, optional): The index of last epoch. Can be set to restart training. Default: -1, means initial learning rate. + """ + + def __init__(self, + learning_rate, + epochs, + step_each_epoch, + end_lr=0.0, + power=1.0, + warmup_epoch=0, + warmup_start_lr=0.0, + last_epoch=-1, + **kwargs): + super().__init__() + if warmup_epoch >= epochs: + msg = f"When using warm up, the value of \"Global.epochs\" must be greater than value of \"Optimizer.lr.warmup_epoch\". The value of \"Optimizer.lr.warmup_epoch\" has been set to {epochs}." + logger.warning(msg) + warmup_epoch = epochs + self.learning_rate = learning_rate + self.steps = (epochs - warmup_epoch) * step_each_epoch + self.end_lr = end_lr + self.power = power + self.last_epoch = last_epoch + self.warmup_steps = round(warmup_epoch * step_each_epoch) + self.warmup_start_lr = warmup_start_lr + + def __call__(self): + learning_rate = lr.PolynomialDecay( + learning_rate=self.learning_rate, + decay_steps=self.steps, + end_lr=self.end_lr, + power=self.power, + last_epoch=self. + last_epoch) if self.steps > 0 else self.learning_rate + if self.warmup_steps > 0: + learning_rate = lr.LinearWarmup( + learning_rate=learning_rate, + warmup_steps=self.warmup_steps, + start_lr=self.warmup_start_lr, + end_lr=self.learning_rate, + last_epoch=self.last_epoch) + return learning_rate + + +class Cosine(object): + """ + Cosine learning rate decay + lr = 0.05 * (math.cos(epoch * (math.pi / epochs)) + 1) + Args: + lr(float): initial learning rate + step_each_epoch(int): steps each epoch + epochs(int): total training epochs + eta_min(float): Minimum learning rate. Default: 0.0. + warmup_epoch(int): The epoch numbers for LinearWarmup. Default: 0. + warmup_start_lr(float): Initial learning rate of warm up. Default: 0.0. + last_epoch (int, optional): The index of last epoch. Can be set to restart training. Default: -1, means initial learning rate. + """ + + def __init__(self, + learning_rate, + step_each_epoch, + epochs, + eta_min=0.0, + warmup_epoch=0, + warmup_start_lr=0.0, + last_epoch=-1, + **kwargs): + super().__init__() + if warmup_epoch >= epochs: + msg = f"When using warm up, the value of \"Global.epochs\" must be greater than value of \"Optimizer.lr.warmup_epoch\". The value of \"Optimizer.lr.warmup_epoch\" has been set to {epochs}." + logger.warning(msg) + warmup_epoch = epochs + self.learning_rate = learning_rate + self.T_max = (epochs - warmup_epoch) * step_each_epoch + self.eta_min = eta_min + self.last_epoch = last_epoch + self.warmup_steps = round(warmup_epoch * step_each_epoch) + self.warmup_start_lr = warmup_start_lr + + def __call__(self): + learning_rate = lr.CosineAnnealingDecay( + learning_rate=self.learning_rate, + T_max=self.T_max, + eta_min=self.eta_min, + last_epoch=self. + last_epoch) if self.T_max > 0 else self.learning_rate + if self.warmup_steps > 0: + learning_rate = lr.LinearWarmup( + learning_rate=learning_rate, + warmup_steps=self.warmup_steps, + start_lr=self.warmup_start_lr, + end_lr=self.learning_rate, + last_epoch=self.last_epoch) + return learning_rate + + +class Step(object): + """ + Piecewise learning rate decay + Args: + step_each_epoch(int): steps each epoch + learning_rate (float): The initial learning rate. It is a python float number. + step_size (int): the interval to update. + gamma (float, optional): The Ratio that the learning rate will be reduced. ``new_lr = origin_lr * gamma`` . + It should be less than 1.0. Default: 0.1. + warmup_epoch(int): The epoch numbers for LinearWarmup. Default: 0. + warmup_start_lr(float): Initial learning rate of warm up. Default: 0.0. + last_epoch (int, optional): The index of last epoch. Can be set to restart training. Default: -1, means initial learning rate. + """ + + def __init__(self, + learning_rate, + step_size, + step_each_epoch, + epochs, + gamma, + warmup_epoch=0, + warmup_start_lr=0.0, + last_epoch=-1, + **kwargs): + super().__init__() + if warmup_epoch >= epochs: + msg = f"When using warm up, the value of \"Global.epochs\" must be greater than value of \"Optimizer.lr.warmup_epoch\". The value of \"Optimizer.lr.warmup_epoch\" has been set to {epochs}." + logger.warning(msg) + warmup_epoch = epochs + self.step_size = step_each_epoch * step_size + self.learning_rate = learning_rate + self.gamma = gamma + self.last_epoch = last_epoch + self.warmup_steps = round(warmup_epoch * step_each_epoch) + self.warmup_start_lr = warmup_start_lr + + def __call__(self): + learning_rate = lr.StepDecay( + learning_rate=self.learning_rate, + step_size=self.step_size, + gamma=self.gamma, + last_epoch=self.last_epoch) + if self.warmup_steps > 0: + learning_rate = lr.LinearWarmup( + learning_rate=learning_rate, + warmup_steps=self.warmup_steps, + start_lr=self.warmup_start_lr, + end_lr=self.learning_rate, + last_epoch=self.last_epoch) + return learning_rate + + +class Piecewise(object): + """ + Piecewise learning rate decay + Args: + boundaries(list): A list of steps numbers. The type of element in the list is python int. + values(list): A list of learning rate values that will be picked during different epoch boundaries. + The type of element in the list is python float. + warmup_epoch(int): The epoch numbers for LinearWarmup. Default: 0. + warmup_start_lr(float): Initial learning rate of warm up. Default: 0.0. + last_epoch (int, optional): The index of last epoch. Can be set to restart training. Default: -1, means initial learning rate. + """ + + def __init__(self, + step_each_epoch, + decay_epochs, + values, + epochs, + warmup_epoch=0, + warmup_start_lr=0.0, + last_epoch=-1, + **kwargs): + super().__init__() + if warmup_epoch >= epochs: + msg = f"When using warm up, the value of \"Global.epochs\" must be greater than value of \"Optimizer.lr.warmup_epoch\". The value of \"Optimizer.lr.warmup_epoch\" has been set to {epochs}." + logger.warning(msg) + warmup_epoch = epochs + self.boundaries = [step_each_epoch * e for e in decay_epochs] + self.values = values + self.last_epoch = last_epoch + self.warmup_steps = round(warmup_epoch * step_each_epoch) + self.warmup_start_lr = warmup_start_lr + + def __call__(self): + learning_rate = lr.PiecewiseDecay( + boundaries=self.boundaries, + values=self.values, + last_epoch=self.last_epoch) + if self.warmup_steps > 0: + learning_rate = lr.LinearWarmup( + learning_rate=learning_rate, + warmup_steps=self.warmup_steps, + start_lr=self.warmup_start_lr, + end_lr=self.values[0], + last_epoch=self.last_epoch) + return learning_rate + + +class MultiStepDecay(LRScheduler): + """ + Update the learning rate by ``gamma`` once ``epoch`` reaches one of the milestones. + The algorithm can be described as the code below. + .. code-block:: text + learning_rate = 0.5 + milestones = [30, 50] + gamma = 0.1 + if epoch < 30: + learning_rate = 0.5 + elif epoch < 50: + learning_rate = 0.05 + else: + learning_rate = 0.005 + Args: + learning_rate (float): The initial learning rate. It is a python float number. + milestones (tuple|list): List or tuple of each boundaries. Must be increasing. + gamma (float, optional): The Ratio that the learning rate will be reduced. ``new_lr = origin_lr * gamma`` . + It should be less than 1.0. Default: 0.1. + last_epoch (int, optional): The index of last epoch. Can be set to restart training. Default: -1, means initial learning rate. + verbose (bool, optional): If ``True``, prints a message to stdout for each update. Default: ``False`` . + + Returns: + ``MultiStepDecay`` instance to schedule learning rate. + Examples: + + .. code-block:: python + import paddle + import numpy as np + # train on default dynamic graph mode + linear = paddle.nn.Linear(10, 10) + scheduler = paddle.optimizer.lr.MultiStepDecay(learning_rate=0.5, milestones=[2, 4, 6], gamma=0.8, verbose=True) + sgd = paddle.optimizer.SGD(learning_rate=scheduler, parameters=linear.parameters()) + for epoch in range(20): + for batch_id in range(5): + x = paddle.uniform([10, 10]) + out = linear(x) + loss = paddle.mean(out) + loss.backward() + sgd.step() + sgd.clear_gradients() + scheduler.step() # If you update learning rate each step + # scheduler.step() # If you update learning rate each epoch + # train on static graph mode + paddle.enable_static() + main_prog = paddle.static.Program() + start_prog = paddle.static.Program() + with paddle.static.program_guard(main_prog, start_prog): + x = paddle.static.data(name='x', shape=[None, 4, 5]) + y = paddle.static.data(name='y', shape=[None, 4, 5]) + z = paddle.static.nn.fc(x, 100) + loss = paddle.mean(z) + scheduler = paddle.optimizer.lr.MultiStepDecay(learning_rate=0.5, milestones=[2, 4, 6], gamma=0.8, verbose=True) + sgd = paddle.optimizer.SGD(learning_rate=scheduler) + sgd.minimize(loss) + exe = paddle.static.Executor() + exe.run(start_prog) + for epoch in range(20): + for batch_id in range(5): + out = exe.run( + main_prog, + feed={ + 'x': np.random.randn(3, 4, 5).astype('float32'), + 'y': np.random.randn(3, 4, 5).astype('float32') + }, + fetch_list=loss.name) + scheduler.step() # If you update learning rate each step + # scheduler.step() # If you update learning rate each epoch + """ + + def __init__(self, + learning_rate, + milestones, + epochs, + step_each_epoch, + gamma=0.1, + last_epoch=-1, + verbose=False): + if not isinstance(milestones, (tuple, list)): + raise TypeError( + "The type of 'milestones' in 'MultiStepDecay' must be 'tuple, list', but received %s." + % type(milestones)) + if not all([ + milestones[i] < milestones[i + 1] + for i in range(len(milestones) - 1) + ]): + raise ValueError('The elements of milestones must be incremented') + if gamma >= 1.0: + raise ValueError('gamma should be < 1.0.') + self.milestones = [x * step_each_epoch for x in milestones] + self.gamma = gamma + super().__init__(learning_rate, last_epoch, verbose) + + def get_lr(self): + for i in range(len(self.milestones)): + if self.last_epoch < self.milestones[i]: + return self.base_lr * (self.gamma**i) + return self.base_lr * (self.gamma**len(self.milestones)) diff --git a/exprements/PaddleClas/ppcls/optimizer/optimizer.py b/exprements/PaddleClas/ppcls/optimizer/optimizer.py new file mode 100644 index 00000000..4422ea70 --- /dev/null +++ b/exprements/PaddleClas/ppcls/optimizer/optimizer.py @@ -0,0 +1,217 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +from paddle import optimizer as optim +import paddle + +from ppcls.utils import logger + + +class Momentum(object): + """ + Simple Momentum optimizer with velocity state. + Args: + learning_rate (float|Variable) - The learning rate used to update parameters. + Can be a float value or a Variable with one float value as data element. + momentum (float) - Momentum factor. + regularization (WeightDecayRegularizer, optional) - The strategy of regularization. + """ + + def __init__(self, + learning_rate, + momentum, + weight_decay=None, + grad_clip=None, + multi_precision=True): + super().__init__() + self.learning_rate = learning_rate + self.momentum = momentum + self.weight_decay = weight_decay + self.grad_clip = grad_clip + self.multi_precision = multi_precision + + def __call__(self, model_list): + # model_list is None in static graph + parameters = sum([m.parameters() for m in model_list], + []) if model_list else None + opt = optim.Momentum( + learning_rate=self.learning_rate, + momentum=self.momentum, + weight_decay=self.weight_decay, + grad_clip=self.grad_clip, + multi_precision=self.multi_precision, + parameters=parameters) + if hasattr(opt, '_use_multi_tensor'): + opt = optim.Momentum( + learning_rate=self.learning_rate, + momentum=self.momentum, + weight_decay=self.weight_decay, + grad_clip=self.grad_clip, + multi_precision=self.multi_precision, + parameters=parameters, + use_multi_tensor=True) + return opt + + +class Adam(object): + def __init__(self, + learning_rate=0.001, + beta1=0.9, + beta2=0.999, + epsilon=1e-08, + parameter_list=None, + weight_decay=None, + grad_clip=None, + name=None, + lazy_mode=False, + multi_precision=False): + self.learning_rate = learning_rate + self.beta1 = beta1 + self.beta2 = beta2 + self.epsilon = epsilon + self.parameter_list = parameter_list + self.learning_rate = learning_rate + self.weight_decay = weight_decay + self.grad_clip = grad_clip + self.name = name + self.lazy_mode = lazy_mode + self.multi_precision = multi_precision + + def __call__(self, model_list): + # model_list is None in static graph + parameters = sum([m.parameters() for m in model_list], + []) if model_list else None + opt = optim.Adam( + learning_rate=self.learning_rate, + beta1=self.beta1, + beta2=self.beta2, + epsilon=self.epsilon, + weight_decay=self.weight_decay, + grad_clip=self.grad_clip, + name=self.name, + lazy_mode=self.lazy_mode, + multi_precision=self.multi_precision, + parameters=parameters) + return opt + + +class RMSProp(object): + """ + Root Mean Squared Propagation (RMSProp) is an unpublished, adaptive learning rate method. + Args: + learning_rate (float|Variable) - The learning rate used to update parameters. + Can be a float value or a Variable with one float value as data element. + momentum (float) - Momentum factor. + rho (float) - rho value in equation. + epsilon (float) - avoid division by zero, default is 1e-6. + regularization (WeightDecayRegularizer, optional) - The strategy of regularization. + """ + + def __init__(self, + learning_rate, + momentum=0.0, + rho=0.95, + epsilon=1e-6, + weight_decay=None, + grad_clip=None, + multi_precision=False): + super().__init__() + self.learning_rate = learning_rate + self.momentum = momentum + self.rho = rho + self.epsilon = epsilon + self.weight_decay = weight_decay + self.grad_clip = grad_clip + + def __call__(self, model_list): + # model_list is None in static graph + parameters = sum([m.parameters() for m in model_list], + []) if model_list else None + opt = optim.RMSProp( + learning_rate=self.learning_rate, + momentum=self.momentum, + rho=self.rho, + epsilon=self.epsilon, + weight_decay=self.weight_decay, + grad_clip=self.grad_clip, + parameters=parameters) + return opt + + +class AdamW(object): + def __init__(self, + learning_rate=0.001, + beta1=0.9, + beta2=0.999, + epsilon=1e-8, + weight_decay=None, + multi_precision=False, + grad_clip=None, + no_weight_decay_name=None, + one_dim_param_no_weight_decay=False, + **args): + super().__init__() + self.learning_rate = learning_rate + self.beta1 = beta1 + self.beta2 = beta2 + self.epsilon = epsilon + self.grad_clip = grad_clip + self.weight_decay = weight_decay + self.multi_precision = multi_precision + self.no_weight_decay_name_list = no_weight_decay_name.split( + ) if no_weight_decay_name else [] + self.one_dim_param_no_weight_decay = one_dim_param_no_weight_decay + + def __call__(self, model_list): + # model_list is None in static graph + parameters = sum([m.parameters() for m in model_list], + []) if model_list else None + + # TODO(gaotingquan): model_list is None when in static graph, "no_weight_decay" not work. + if model_list is None: + if self.one_dim_param_no_weight_decay or len( + self.no_weight_decay_name_list) != 0: + msg = "\"AdamW\" does not support setting \"no_weight_decay\" in static graph. Please use dynamic graph." + logger.error(Exception(msg)) + raise Exception(msg) + + self.no_weight_decay_param_name_list = [ + p.name for model in model_list for n, p in model.named_parameters() + if any(nd in n for nd in self.no_weight_decay_name_list) + ] if model_list else [] + + if self.one_dim_param_no_weight_decay: + self.no_weight_decay_param_name_list += [ + p.name for model in model_list + for n, p in model.named_parameters() if len(p.shape) == 1 + ] if model_list else [] + + opt = optim.AdamW( + learning_rate=self.learning_rate, + beta1=self.beta1, + beta2=self.beta2, + epsilon=self.epsilon, + parameters=parameters, + weight_decay=self.weight_decay, + multi_precision=self.multi_precision, + grad_clip=self.grad_clip, + apply_decay_param_fun=self._apply_decay_param_fun) + return opt + + def _apply_decay_param_fun(self, name): + return name not in self.no_weight_decay_param_name_list diff --git a/exprements/PaddleClas/ppcls/static/program.py b/exprements/PaddleClas/ppcls/static/program.py new file mode 100644 index 00000000..b3534a2c --- /dev/null +++ b/exprements/PaddleClas/ppcls/static/program.py @@ -0,0 +1,449 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import os +import time +import numpy as np + +from collections import OrderedDict + +import paddle +import paddle.nn.functional as F + +from paddle.distributed import fleet +from paddle.distributed.fleet import DistributedStrategy + +# from ppcls.optimizer import OptimizerBuilder +# from ppcls.optimizer.learning_rate import LearningRateBuilder + +from ppcls.arch import build_model +from ppcls.loss import build_loss +from ppcls.metric import build_metrics +from ppcls.optimizer import build_optimizer +from ppcls.optimizer import build_lr_scheduler + +from ppcls.utils.misc import AverageMeter +from ppcls.utils import logger, profiler + + +def create_feeds(image_shape, use_mix=False, class_num=None, dtype="float32"): + """ + Create feeds as model input + + Args: + image_shape(list[int]): model input shape, such as [3, 224, 224] + use_mix(bool): whether to use mix(include mixup, cutmix, fmix) + class_num(int): the class number of network, required if use_mix + + Returns: + feeds(dict): dict of model input variables + """ + feeds = OrderedDict() + feeds['data'] = paddle.static.data( + name="data", shape=[None] + image_shape, dtype=dtype) + + if use_mix: + if class_num is None: + msg = "When use MixUp, CutMix and so on, you must set class_num." + logger.error(msg) + raise Exception(msg) + feeds['target'] = paddle.static.data( + name="target", shape=[None, class_num], dtype="float32") + else: + feeds['label'] = paddle.static.data( + name="label", shape=[None, 1], dtype="int64") + + return feeds + + +def create_fetchs(out, + feeds, + architecture, + topk=5, + epsilon=None, + class_num=None, + use_mix=False, + config=None, + mode="Train"): + """ + Create fetchs as model outputs(included loss and measures), + will call create_loss and create_metric(if use_mix). + Args: + out(variable): model output variable + feeds(dict): dict of model input variables. + If use mix_up, it will not include label. + architecture(dict): architecture information, + name(such as ResNet50) is needed + topk(int): usually top5 + epsilon(float): parameter for label smoothing, 0.0 <= epsilon <= 1.0 + class_num(int): the class number of network, required if use_mix + use_mix(bool): whether to use mix(include mixup, cutmix, fmix) + config(dict): model config + + Returns: + fetchs(dict): dict of model outputs(included loss and measures) + """ + fetchs = OrderedDict() + # build loss + if use_mix: + if class_num is None: + msg = "When use MixUp, CutMix and so on, you must set class_num." + logger.error(msg) + raise Exception(msg) + target = paddle.reshape(feeds['target'], [-1, class_num]) + else: + target = paddle.reshape(feeds['label'], [-1, 1]) + + loss_func = build_loss(config["Loss"][mode]) + loss_dict = loss_func(out, target) + + loss_out = loss_dict["loss"] + fetchs['loss'] = (loss_out, AverageMeter('loss', '7.4f', need_avg=True)) + + # build metric + if not use_mix: + metric_func = build_metrics(config["Metric"][mode]) + + metric_dict = metric_func(out, target) + + for key in metric_dict: + if mode != "Train" and paddle.distributed.get_world_size() > 1: + paddle.distributed.all_reduce( + metric_dict[key], op=paddle.distributed.ReduceOp.SUM) + metric_dict[key] = metric_dict[ + key] / paddle.distributed.get_world_size() + + fetchs[key] = (metric_dict[key], AverageMeter( + key, '7.4f', need_avg=True)) + + return fetchs + + +def create_optimizer(config, step_each_epoch): + # create learning_rate instance + optimizer, lr_sch = build_optimizer( + config["Optimizer"], config["Global"]["epochs"], step_each_epoch) + return optimizer, lr_sch + + +def create_strategy(config): + """ + Create build strategy and exec strategy. + + Args: + config(dict): config + + Returns: + build_strategy: build strategy + exec_strategy: exec strategy + """ + build_strategy = paddle.static.BuildStrategy() + exec_strategy = paddle.static.ExecutionStrategy() + + exec_strategy.num_threads = 1 + exec_strategy.num_iteration_per_drop_scope = ( + 10000 + if 'AMP' in config and config.AMP.get("level", "O1") == "O2" else 10) + + fuse_op = True if 'AMP' in config else False + + fuse_bn_act_ops = config.get('fuse_bn_act_ops', fuse_op) + fuse_elewise_add_act_ops = config.get('fuse_elewise_add_act_ops', fuse_op) + fuse_bn_add_act_ops = config.get('fuse_bn_add_act_ops', fuse_op) + enable_addto = config.get('enable_addto', fuse_op) + + build_strategy.fuse_bn_act_ops = fuse_bn_act_ops + build_strategy.fuse_elewise_add_act_ops = fuse_elewise_add_act_ops + build_strategy.fuse_bn_add_act_ops = fuse_bn_add_act_ops + build_strategy.enable_addto = enable_addto + + return build_strategy, exec_strategy + + +def dist_optimizer(config, optimizer): + """ + Create a distributed optimizer based on a normal optimizer + + Args: + config(dict): + optimizer(): a normal optimizer + + Returns: + optimizer: a distributed optimizer + """ + build_strategy, exec_strategy = create_strategy(config) + + dist_strategy = DistributedStrategy() + dist_strategy.execution_strategy = exec_strategy + dist_strategy.build_strategy = build_strategy + + dist_strategy.nccl_comm_num = 1 + dist_strategy.fuse_all_reduce_ops = True + dist_strategy.fuse_grad_size_in_MB = 16 + optimizer = fleet.distributed_optimizer(optimizer, strategy=dist_strategy) + + return optimizer + + +def mixed_precision_optimizer(config, optimizer): + if 'AMP' in config: + amp_cfg = config.AMP if config.AMP else dict() + scale_loss = amp_cfg.get('scale_loss', 1.0) + use_dynamic_loss_scaling = amp_cfg.get('use_dynamic_loss_scaling', + False) + use_pure_fp16 = amp_cfg.get("level", "O1") == "O2" + optimizer = paddle.static.amp.decorate( + optimizer, + init_loss_scaling=scale_loss, + use_dynamic_loss_scaling=use_dynamic_loss_scaling, + use_pure_fp16=use_pure_fp16, + use_fp16_guard=True) + + return optimizer + + +def build(config, + main_prog, + startup_prog, + class_num=None, + step_each_epoch=100, + is_train=True, + is_distributed=True): + """ + Build a program using a model and an optimizer + 1. create feeds + 2. create a dataloader + 3. create a model + 4. create fetchs + 5. create an optimizer + + Args: + config(dict): config + main_prog(): main program + startup_prog(): startup program + class_num(int): the class number of network, required if use_mix + is_train(bool): train or eval + is_distributed(bool): whether to use distributed training method + + Returns: + dataloader(): a bridge between the model and the data + fetchs(dict): dict of model outputs(included loss and measures) + """ + with paddle.static.program_guard(main_prog, startup_prog): + with paddle.utils.unique_name.guard(): + mode = "Train" if is_train else "Eval" + use_mix = "batch_transform_ops" in config["DataLoader"][mode][ + "dataset"] + feeds = create_feeds( + config["Global"]["image_shape"], + use_mix, + class_num=class_num, + dtype="float32") + + # build model + # data_format should be assigned in arch-dict + input_image_channel = config["Global"]["image_shape"][ + 0] # default as [3, 224, 224] + model = build_model(config) + out = model(feeds["data"]) + # end of build model + + fetchs = create_fetchs( + out, + feeds, + config["Arch"], + epsilon=config.get('ls_epsilon'), + class_num=class_num, + use_mix=use_mix, + config=config, + mode=mode) + lr_scheduler = None + optimizer = None + if is_train: + optimizer, lr_scheduler = build_optimizer( + config["Optimizer"], config["Global"]["epochs"], + step_each_epoch) + optimizer = mixed_precision_optimizer(config, optimizer) + if is_distributed: + optimizer = dist_optimizer(config, optimizer) + optimizer.minimize(fetchs['loss'][0]) + return fetchs, lr_scheduler, feeds, optimizer + + +def compile(config, program, loss_name=None, share_prog=None): + """ + Compile the program + + Args: + config(dict): config + program(): the program which is wrapped by + loss_name(str): loss name + share_prog(): the shared program, used for evaluation during training + + Returns: + compiled_program(): a compiled program + """ + build_strategy, exec_strategy = create_strategy(config) + + compiled_program = paddle.static.CompiledProgram( + program).with_data_parallel( + share_vars_from=share_prog, + loss_name=loss_name, + build_strategy=build_strategy, + exec_strategy=exec_strategy) + + return compiled_program + + +total_step = 0 + + +def run(dataloader, + exe, + program, + feeds, + fetchs, + epoch=0, + mode='train', + config=None, + vdl_writer=None, + lr_scheduler=None, + profiler_options=None): + """ + Feed data to the model and fetch the measures and loss + + Args: + dataloader(paddle io dataloader): + exe(): + program(): + fetchs(dict): dict of measures and the loss + epoch(int): epoch of training or evaluation + model(str): log only + + Returns: + """ + fetch_list = [f[0] for f in fetchs.values()] + metric_dict = OrderedDict([("lr", AverageMeter( + 'lr', 'f', postfix=",", need_avg=False))]) + + for k in fetchs: + metric_dict[k] = fetchs[k][1] + + metric_dict["batch_time"] = AverageMeter( + 'batch_cost', '.5f', postfix=" s,") + metric_dict["reader_time"] = AverageMeter( + 'reader_cost', '.5f', postfix=" s,") + + for m in metric_dict.values(): + m.reset() + + use_dali = config["Global"].get('use_dali', False) + tic = time.time() + + if not use_dali: + dataloader = dataloader() + + idx = 0 + batch_size = None + while True: + # The DALI maybe raise RuntimeError for some particular images, such as ImageNet1k/n04418357_26036.JPEG + try: + batch = next(dataloader) + except StopIteration: + break + except RuntimeError: + logger.warning( + "Except RuntimeError when reading data from dataloader, try to read once again..." + ) + continue + idx += 1 + # ignore the warmup iters + if idx == 5: + metric_dict["batch_time"].reset() + metric_dict["reader_time"].reset() + + metric_dict['reader_time'].update(time.time() - tic) + + profiler.add_profiler_step(profiler_options) + + if use_dali: + batch_size = batch[0]["data"].shape()[0] + feed_dict = batch[0] + else: + batch_size = batch[0].shape()[0] + feed_dict = { + key.name: batch[idx] + for idx, key in enumerate(feeds.values()) + } + + metrics = exe.run(program=program, + feed=feed_dict, + fetch_list=fetch_list) + + for name, m in zip(fetchs.keys(), metrics): + metric_dict[name].update(np.mean(m), batch_size) + metric_dict["batch_time"].update(time.time() - tic) + if mode == "train": + metric_dict['lr'].update(lr_scheduler.get_lr()) + + fetchs_str = ' '.join([ + str(metric_dict[key].mean) + if "time" in key else str(metric_dict[key].value) + for key in metric_dict + ]) + ips_info = " ips: {:.5f} images/sec.".format( + batch_size / metric_dict["batch_time"].avg) + fetchs_str += ips_info + + if lr_scheduler is not None: + lr_scheduler.step() + + if vdl_writer: + global total_step + logger.scaler('loss', metrics[0][0], total_step, vdl_writer) + total_step += 1 + if mode == 'eval': + if idx % config.get('print_interval', 10) == 0: + logger.info("{:s} step:{:<4d} {:s}".format(mode, idx, + fetchs_str)) + else: + epoch_str = "epoch:{:<3d}".format(epoch) + step_str = "{:s} step:{:<4d}".format(mode, idx) + + if idx % config.get('print_interval', 10) == 0: + logger.info("{:s} {:s} {:s}".format(epoch_str, step_str, + fetchs_str)) + + tic = time.time() + + end_str = ' '.join([str(m.mean) for m in metric_dict.values()] + + [metric_dict["batch_time"].total]) + ips_info = "ips: {:.5f} images/sec.".format(batch_size / + metric_dict["batch_time"].avg) + if mode == 'eval': + logger.info("END {:s} {:s} {:s}".format(mode, end_str, ips_info)) + else: + end_epoch_str = "END epoch:{:<3d}".format(epoch) + logger.info("{:s} {:s} {:s} {:s}".format(end_epoch_str, mode, end_str, + ips_info)) + if use_dali: + dataloader.reset() + + # return top1_acc in order to save the best model + if mode == 'eval': + return fetchs["top1"][1].avg diff --git a/exprements/PaddleClas/ppcls/static/run_dali.sh b/exprements/PaddleClas/ppcls/static/run_dali.sh new file mode 100644 index 00000000..5bf0ef4c --- /dev/null +++ b/exprements/PaddleClas/ppcls/static/run_dali.sh @@ -0,0 +1,8 @@ +#!/usr/bin/env bash + +export CUDA_VISIBLE_DEVICES="0,1,2,3" + +python3.7 -m paddle.distributed.launch \ + --gpus="0,1,2,3" \ + ppcls/static/train.py \ + -c ./ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml diff --git a/exprements/PaddleClas/ppcls/static/save_load.py b/exprements/PaddleClas/ppcls/static/save_load.py new file mode 100644 index 00000000..13badfdd --- /dev/null +++ b/exprements/PaddleClas/ppcls/static/save_load.py @@ -0,0 +1,139 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import errno +import os +import re +import shutil +import tempfile + +import paddle + +from ppcls.utils import logger + +__all__ = ['init_model', 'save_model'] + + +def _mkdir_if_not_exist(path): + """ + mkdir if not exists, ignore the exception when multiprocess mkdir together + """ + if not os.path.exists(path): + try: + os.makedirs(path) + except OSError as e: + if e.errno == errno.EEXIST and os.path.isdir(path): + logger.warning( + 'be happy if some process has already created {}'.format( + path)) + else: + raise OSError('Failed to mkdir {}'.format(path)) + + +def _load_state(path): + if os.path.exists(path + '.pdopt'): + # XXX another hack to ignore the optimizer state + tmp = tempfile.mkdtemp() + dst = os.path.join(tmp, os.path.basename(os.path.normpath(path))) + shutil.copy(path + '.pdparams', dst + '.pdparams') + state = paddle.static.load_program_state(dst) + shutil.rmtree(tmp) + else: + state = paddle.static.load_program_state(path) + return state + + +def load_params(exe, prog, path, ignore_params=None): + """ + Load model from the given path. + Args: + exe (fluid.Executor): The fluid.Executor object. + prog (fluid.Program): load weight to which Program object. + path (string): URL string or loca model path. + ignore_params (list): ignore variable to load when finetuning. + It can be specified by finetune_exclude_pretrained_params + and the usage can refer to the document + docs/advanced_tutorials/TRANSFER_LEARNING.md + """ + if not (os.path.isdir(path) or os.path.exists(path + '.pdparams')): + raise ValueError("Model pretrain path {} does not " + "exists.".format(path)) + + logger.info("Loading parameters from {}...".format(path)) + + ignore_set = set() + state = _load_state(path) + + # ignore the parameter which mismatch the shape + # between the model and pretrain weight. + all_var_shape = {} + for block in prog.blocks: + for param in block.all_parameters(): + all_var_shape[param.name] = param.shape + ignore_set.update([ + name for name, shape in all_var_shape.items() + if name in state and shape != state[name].shape + ]) + + if ignore_params: + all_var_names = [var.name for var in prog.list_vars()] + ignore_list = filter( + lambda var: any([re.match(name, var) for name in ignore_params]), + all_var_names) + ignore_set.update(list(ignore_list)) + + if len(ignore_set) > 0: + for k in ignore_set: + if k in state: + logger.warning( + 'variable {} is already excluded automatically'.format(k)) + del state[k] + + paddle.static.set_program_state(prog, state) + + +def init_model(config, program, exe): + """ + load model from checkpoint or pretrained_model + """ + checkpoints = config.get('checkpoints') + if checkpoints: + paddle.static.load(program, checkpoints, exe) + logger.info("Finish initing model from {}".format(checkpoints)) + return + + pretrained_model = config.get('pretrained_model') + if pretrained_model: + if not isinstance(pretrained_model, list): + pretrained_model = [pretrained_model] + for pretrain in pretrained_model: + load_params(exe, program, pretrain) + logger.info("Finish initing model from {}".format(pretrained_model)) + + +def save_model(program, model_path, epoch_id, prefix='ppcls'): + """ + save model to the target path + """ + if paddle.distributed.get_rank() != 0: + return + model_path = os.path.join(model_path, str(epoch_id)) + _mkdir_if_not_exist(model_path) + model_prefix = os.path.join(model_path, prefix) + paddle.static.save(program, model_prefix) + logger.info("Already save model in {}".format(model_path)) diff --git a/exprements/PaddleClas/ppcls/static/train.py b/exprements/PaddleClas/ppcls/static/train.py new file mode 100644 index 00000000..9c03598b --- /dev/null +++ b/exprements/PaddleClas/ppcls/static/train.py @@ -0,0 +1,209 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import argparse +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(__dir__) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../../'))) + +import paddle +from paddle.distributed import fleet +from visualdl import LogWriter + +from ppcls.data import build_dataloader +from ppcls.utils.config import get_config, print_config +from ppcls.utils import logger +from ppcls.utils.logger import init_logger +from ppcls.static.save_load import init_model, save_model +from ppcls.static import program + + +def parse_args(): + parser = argparse.ArgumentParser("PaddleClas train script") + parser.add_argument( + '-c', + '--config', + type=str, + default='configs/ResNet/ResNet50.yaml', + help='config file path') + parser.add_argument( + '-p', + '--profiler_options', + type=str, + default=None, + help='The option of profiler, which should be in format \"key1=value1;key2=value2;key3=value3\".' + ) + parser.add_argument( + '-o', + '--override', + action='append', + default=[], + help='config options to be overridden') + args = parser.parse_args() + return args + + +def main(args): + """ + all the config of training paradigm should be in config["Global"] + """ + config = get_config(args.config, overrides=args.override, show=False) + global_config = config["Global"] + + mode = "train" + + log_file = os.path.join(global_config['output_dir'], + config["Arch"]["name"], f"{mode}.log") + init_logger(name='root', log_file=log_file) + print_config(config) + + if global_config.get("is_distributed", True): + fleet.init(is_collective=True) + # assign the device + use_gpu = global_config.get("use_gpu", True) + # amp related config + if 'AMP' in config: + AMP_RELATED_FLAGS_SETTING = { + 'FLAGS_cudnn_exhaustive_search': 1, + 'FLAGS_conv_workspace_size_limit': 1500, + 'FLAGS_cudnn_batchnorm_spatial_persistent': 1, + 'FLAGS_max_inplace_grad_add': 8, + } + os.environ['FLAGS_cudnn_batchnorm_spatial_persistent'] = '1' + paddle.fluid.set_flags(AMP_RELATED_FLAGS_SETTING) + + use_xpu = global_config.get("use_xpu", False) + use_npu = global_config.get("use_npu", False) + assert ( + use_gpu and use_xpu and use_npu + ) is not True, "gpu, xpu and npu can not be true in the same time in static mode!" + + if use_gpu: + device = paddle.set_device('gpu') + elif use_xpu: + device = paddle.set_device('xpu') + elif use_npu: + device = paddle.set_device('npu') + else: + device = paddle.set_device('cpu') + + # visualDL + vdl_writer = None + if global_config["use_visualdl"]: + vdl_dir = os.path.join(global_config["output_dir"], "vdl") + vdl_writer = LogWriter(vdl_dir) + + # build dataloader + eval_dataloader = None + use_dali = global_config.get('use_dali', False) + + class_num = config["Arch"].get("class_num", None) + config["DataLoader"].update({"class_num": class_num}) + train_dataloader = build_dataloader( + config["DataLoader"], "Train", device=device, use_dali=use_dali) + if global_config["eval_during_train"]: + eval_dataloader = build_dataloader( + config["DataLoader"], "Eval", device=device, use_dali=use_dali) + + step_each_epoch = len(train_dataloader) + + # startup_prog is used to do some parameter init work, + # and train prog is used to hold the network + startup_prog = paddle.static.Program() + train_prog = paddle.static.Program() + + best_top1_acc = 0.0 # best top1 acc record + + train_fetchs, lr_scheduler, train_feeds, optimizer = program.build( + config, + train_prog, + startup_prog, + class_num, + step_each_epoch=step_each_epoch, + is_train=True, + is_distributed=global_config.get("is_distributed", True)) + + if global_config["eval_during_train"]: + eval_prog = paddle.static.Program() + eval_fetchs, _, eval_feeds, _ = program.build( + config, + eval_prog, + startup_prog, + is_train=False, + is_distributed=global_config.get("is_distributed", True)) + # clone to prune some content which is irrelevant in eval_prog + eval_prog = eval_prog.clone(for_test=True) + + # create the "Executor" with the statement of which device + exe = paddle.static.Executor(device) + # Parameter initialization + exe.run(startup_prog) + # load pretrained models or checkpoints + init_model(global_config, train_prog, exe) + + if 'AMP' in config and config.AMP.get("level", "O1") == "O2": + optimizer.amp_init( + device, + scope=paddle.static.global_scope(), + test_program=eval_prog + if global_config["eval_during_train"] else None) + + if not global_config.get("is_distributed", True): + compiled_train_prog = program.compile( + config, train_prog, loss_name=train_fetchs["loss"][0].name) + else: + compiled_train_prog = train_prog + + if eval_dataloader is not None: + compiled_eval_prog = program.compile(config, eval_prog) + + for epoch_id in range(global_config["epochs"]): + # 1. train with train dataset + program.run(train_dataloader, exe, compiled_train_prog, train_feeds, + train_fetchs, epoch_id, 'train', config, vdl_writer, + lr_scheduler, args.profiler_options) + # 2. evaate with eval dataset + if global_config["eval_during_train"] and epoch_id % global_config[ + "eval_interval"] == 0: + top1_acc = program.run(eval_dataloader, exe, compiled_eval_prog, + eval_feeds, eval_fetchs, epoch_id, "eval", + config) + if top1_acc > best_top1_acc: + best_top1_acc = top1_acc + message = "The best top1 acc {:.5f}, in epoch: {:d}".format( + best_top1_acc, epoch_id) + logger.info(message) + if epoch_id % global_config["save_interval"] == 0: + + model_path = os.path.join(global_config["output_dir"], + config["Arch"]["name"]) + save_model(train_prog, model_path, "best_model") + + # 3. save the persistable model + if epoch_id % global_config["save_interval"] == 0: + model_path = os.path.join(global_config["output_dir"], + config["Arch"]["name"]) + save_model(train_prog, model_path, epoch_id) + + +if __name__ == '__main__': + paddle.enable_static() + args = parse_args() + main(args) diff --git a/exprements/PaddleClas/ppcls/utils/__init__.py b/exprements/PaddleClas/ppcls/utils/__init__.py new file mode 100644 index 00000000..632cc788 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/__init__.py @@ -0,0 +1,27 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from . import logger +from . import misc +from . import model_zoo +from . import metrics + +from .save_load import init_model, save_model +from .config import get_config +from .misc import AverageMeter +from .metrics import multi_hot_encode +from .metrics import hamming_distance +from .metrics import accuracy_score +from .metrics import precision_recall_fscore +from .metrics import mean_average_precision diff --git a/exprements/PaddleClas/ppcls/utils/check.py b/exprements/PaddleClas/ppcls/utils/check.py new file mode 100644 index 00000000..bc703081 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/check.py @@ -0,0 +1,151 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import os +import sys + +import paddle +from paddle import is_compiled_with_cuda + +from ppcls.arch import get_architectures +from ppcls.arch import similar_architectures +from ppcls.arch import get_blacklist_model_in_static_mode +from ppcls.utils import logger + + +def check_version(): + """ + Log error and exit when the installed version of paddlepaddle is + not satisfied. + """ + err = "PaddlePaddle version 1.8.0 or higher is required, " \ + "or a suitable develop version is satisfied as well. \n" \ + "Please make sure the version is good with your code." + try: + pass + # paddle.utils.require_version('0.0.0') + except Exception: + logger.error(err) + sys.exit(1) + + +def check_gpu(): + """ + Log error and exit when using paddlepaddle cpu version. + """ + err = "You are using paddlepaddle cpu version! Please try to " \ + "install paddlepaddle-gpu to run model on GPU." + + try: + assert is_compiled_with_cuda() + except AssertionError: + logger.error(err) + sys.exit(1) + + +def check_architecture(architecture): + """ + check architecture and recommend similar architectures + """ + assert isinstance(architecture, dict), \ + ("the type of architecture({}) should be dict". format(architecture)) + assert "name" in architecture, \ + ("name must be in the architecture keys, just contains: {}". format( + architecture.keys())) + + similar_names = similar_architectures(architecture["name"], + get_architectures()) + model_list = ', '.join(similar_names) + err = "Architecture [{}] is not exist! Maybe you want: [{}]" \ + "".format(architecture["name"], model_list) + try: + assert architecture["name"] in similar_names + except AssertionError: + logger.error(err) + sys.exit(1) + + +def check_model_with_running_mode(architecture): + """ + check whether the model is consistent with the operating mode + """ + # some model are not supported in the static mode + blacklist = get_blacklist_model_in_static_mode() + if not paddle.in_dynamic_mode() and architecture["name"] in blacklist: + logger.error("Model: {} is not supported in the staic mode.".format( + architecture["name"])) + sys.exit(1) + return + + +def check_mix(architecture, use_mix=False): + """ + check mix parameter + """ + err = "Cannot use mix processing in GoogLeNet, " \ + "please set use_mix = False." + try: + if architecture["name"] == "GoogLeNet": + assert use_mix is not True + except AssertionError: + logger.error(err) + sys.exit(1) + + +def check_classes_num(classes_num): + """ + check classes_num + """ + err = "classes_num({}) should be a positive integer" \ + "and larger than 1".format(classes_num) + try: + assert isinstance(classes_num, int) + assert classes_num > 1 + except AssertionError: + logger.error(err) + sys.exit(1) + + +def check_data_dir(path): + """ + check cata_dir + """ + err = "Data path is not exist, please given a right path" \ + "".format(path) + try: + assert os.isdir(path) + except AssertionError: + logger.error(err) + sys.exit(1) + + +def check_function_params(config, key): + """ + check specify config + """ + k_config = config.get(key) + assert k_config is not None, \ + ('{} is required in config'.format(key)) + + assert k_config.get('function'), \ + ('function is required {} config'.format(key)) + params = k_config.get('params') + assert params is not None, \ + ('params is required in {} config'.format(key)) + assert isinstance(params, dict), \ + ('the params in {} config should be a dict'.format(key)) diff --git a/exprements/PaddleClas/ppcls/utils/config.py b/exprements/PaddleClas/ppcls/utils/config.py new file mode 100644 index 00000000..e3277c48 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/config.py @@ -0,0 +1,210 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import copy +import argparse +import yaml +from ppcls.utils import logger +from ppcls.utils import check +__all__ = ['get_config'] + + +class AttrDict(dict): + def __getattr__(self, key): + return self[key] + + def __setattr__(self, key, value): + if key in self.__dict__: + self.__dict__[key] = value + else: + self[key] = value + + def __deepcopy__(self, content): + return copy.deepcopy(dict(self)) + + +def create_attr_dict(yaml_config): + from ast import literal_eval + for key, value in yaml_config.items(): + if type(value) is dict: + yaml_config[key] = value = AttrDict(value) + if isinstance(value, str): + try: + value = literal_eval(value) + except BaseException: + pass + if isinstance(value, AttrDict): + create_attr_dict(yaml_config[key]) + else: + yaml_config[key] = value + + +def parse_config(cfg_file): + """Load a config file into AttrDict""" + with open(cfg_file, 'r') as fopen: + yaml_config = AttrDict(yaml.load(fopen, Loader=yaml.SafeLoader)) + create_attr_dict(yaml_config) + return yaml_config + + +def print_dict(d, delimiter=0): + """ + Recursively visualize a dict and + indenting acrrording by the relationship of keys. + """ + placeholder = "-" * 60 + for k, v in sorted(d.items()): + if isinstance(v, dict): + logger.info("{}{} : ".format(delimiter * " ", k)) + print_dict(v, delimiter + 4) + elif isinstance(v, list) and len(v) >= 1 and isinstance(v[0], dict): + logger.info("{}{} : ".format(delimiter * " ", k)) + for value in v: + print_dict(value, delimiter + 4) + else: + logger.info("{}{} : {}".format(delimiter * " ", k, v)) + if k.isupper(): + logger.info(placeholder) + + +def print_config(config): + """ + visualize configs + Arguments: + config: configs + """ + logger.advertise() + print_dict(config) + + +def check_config(config): + """ + Check config + """ + check.check_version() + use_gpu = config.get('use_gpu', True) + if use_gpu: + check.check_gpu() + architecture = config.get('ARCHITECTURE') + #check.check_architecture(architecture) + use_mix = config.get('use_mix', False) + check.check_mix(architecture, use_mix) + classes_num = config.get('classes_num') + check.check_classes_num(classes_num) + mode = config.get('mode', 'train') + if mode.lower() == 'train': + check.check_function_params(config, 'LEARNING_RATE') + check.check_function_params(config, 'OPTIMIZER') + + +def override(dl, ks, v): + """ + Recursively replace dict of list + Args: + dl(dict or list): dict or list to be replaced + ks(list): list of keys + v(str): value to be replaced + """ + + def str2num(v): + try: + return eval(v) + except Exception: + return v + + assert isinstance(dl, (list, dict)), ("{} should be a list or a dict") + assert len(ks) > 0, ('lenght of keys should larger than 0') + if isinstance(dl, list): + k = str2num(ks[0]) + if len(ks) == 1: + assert k < len(dl), ('index({}) out of range({})'.format(k, dl)) + dl[k] = str2num(v) + else: + override(dl[k], ks[1:], v) + else: + if len(ks) == 1: + # assert ks[0] in dl, ('{} is not exist in {}'.format(ks[0], dl)) + if not ks[0] in dl: + print('A new filed ({}) detected!'.format(ks[0], dl)) + dl[ks[0]] = str2num(v) + else: + override(dl[ks[0]], ks[1:], v) + + +def override_config(config, options=None): + """ + Recursively override the config + Args: + config(dict): dict to be replaced + options(list): list of pairs(key0.key1.idx.key2=value) + such as: [ + 'topk=2', + 'VALID.transforms.1.ResizeImage.resize_short=300' + ] + Returns: + config(dict): replaced config + """ + if options is not None: + for opt in options: + assert isinstance(opt, str), ( + "option({}) should be a str".format(opt)) + assert "=" in opt, ( + "option({}) should contain a =" + "to distinguish between key and value".format(opt)) + pair = opt.split('=') + assert len(pair) == 2, ("there can be only a = in the option") + key, value = pair + keys = key.split('.') + override(config, keys, value) + return config + + +def get_config(fname, overrides=None, show=False): + """ + Read config from file + """ + assert os.path.exists(fname), ( + 'config file({}) is not exist'.format(fname)) + config = parse_config(fname) + override_config(config, overrides) + if show: + print_config(config) + # check_config(config) + return config + + +def parse_args(): + parser = argparse.ArgumentParser("generic-image-rec train script") + parser.add_argument( + '-c', + '--config', + type=str, + default='configs/config.yaml', + help='config file path') + parser.add_argument( + '-o', + '--override', + action='append', + default=[], + help='config options to be overridden') + parser.add_argument( + '-p', + '--profiler_options', + type=str, + default=None, + help='The option of profiler, which should be in format \"key1=value1;key2=value2;key3=value3\".' + ) + args = parser.parse_args() + return args diff --git a/exprements/PaddleClas/ppcls/utils/download.py b/exprements/PaddleClas/ppcls/utils/download.py new file mode 100644 index 00000000..9c457504 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/download.py @@ -0,0 +1,319 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import os +import sys +import os.path as osp +import shutil +import requests +import hashlib +import tarfile +import zipfile +import time +from collections import OrderedDict +from tqdm import tqdm + +from ppcls.utils import logger + +__all__ = ['get_weights_path_from_url'] + +WEIGHTS_HOME = osp.expanduser("~/.paddleclas/weights") + +DOWNLOAD_RETRY_LIMIT = 3 + + +def is_url(path): + """ + Whether path is URL. + Args: + path (string): URL string or not. + """ + return path.startswith('http://') or path.startswith('https://') + + +def get_weights_path_from_url(url, md5sum=None): + """Get weights path from WEIGHT_HOME, if not exists, + download it from url. + + Args: + url (str): download url + md5sum (str): md5 sum of download package + + Returns: + str: a local path to save downloaded weights. + + Examples: + .. code-block:: python + + from paddle.utils.download import get_weights_path_from_url + + resnet18_pretrained_weight_url = 'https://paddle-hapi.bj.bcebos.com/models/resnet18.pdparams' + local_weight_path = get_weights_path_from_url(resnet18_pretrained_weight_url) + + """ + path = get_path_from_url(url, WEIGHTS_HOME, md5sum) + return path + + +def _map_path(url, root_dir): + # parse path after download under root_dir + fname = osp.split(url)[-1] + fpath = fname + return osp.join(root_dir, fpath) + + +def _get_unique_endpoints(trainer_endpoints): + # Sorting is to avoid different environmental variables for each card + trainer_endpoints.sort() + ips = set() + unique_endpoints = set() + for endpoint in trainer_endpoints: + ip = endpoint.split(":")[0] + if ip in ips: + continue + ips.add(ip) + unique_endpoints.add(endpoint) + logger.info("unique_endpoints {}".format(unique_endpoints)) + return unique_endpoints + + +def get_path_from_url(url, + root_dir, + md5sum=None, + check_exist=True, + decompress=True): + """ Download from given url to root_dir. + if file or directory specified by url is exists under + root_dir, return the path directly, otherwise download + from url and decompress it, return the path. + + Args: + url (str): download url + root_dir (str): root dir for downloading, it should be + WEIGHTS_HOME or DATASET_HOME + md5sum (str): md5 sum of download package + + Returns: + str: a local path to save downloaded models & weights & datasets. + """ + + from paddle.fluid.dygraph.parallel import ParallelEnv + + assert is_url(url), "downloading from {} not a url".format(url) + # parse path after download to decompress under root_dir + fullpath = _map_path(url, root_dir) + # Mainly used to solve the problem of downloading data from different + # machines in the case of multiple machines. Different ips will download + # data, and the same ip will only download data once. + unique_endpoints = _get_unique_endpoints(ParallelEnv() + .trainer_endpoints[:]) + if osp.exists(fullpath) and check_exist and _md5check(fullpath, md5sum): + logger.info("Found {}".format(fullpath)) + else: + if ParallelEnv().current_endpoint in unique_endpoints: + fullpath = _download(url, root_dir, md5sum) + else: + while not os.path.exists(fullpath): + time.sleep(1) + + if ParallelEnv().current_endpoint in unique_endpoints: + if decompress and (tarfile.is_tarfile(fullpath) or + zipfile.is_zipfile(fullpath)): + fullpath = _decompress(fullpath) + + return fullpath + + +def _download(url, path, md5sum=None): + """ + Download from url, save to path. + + url (str): download url + path (str): download to given path + """ + if not osp.exists(path): + os.makedirs(path) + + fname = osp.split(url)[-1] + fullname = osp.join(path, fname) + retry_cnt = 0 + + while not (osp.exists(fullname) and _md5check(fullname, md5sum)): + if retry_cnt < DOWNLOAD_RETRY_LIMIT: + retry_cnt += 1 + else: + raise RuntimeError("Download from {} failed. " + "Retry limit reached".format(url)) + + logger.info("Downloading {} from {}".format(fname, url)) + + try: + req = requests.get(url, stream=True) + except Exception as e: # requests.exceptions.ConnectionError + logger.info( + "Downloading {} from {} failed {} times with exception {}". + format(fname, url, retry_cnt + 1, str(e))) + time.sleep(1) + continue + + if req.status_code != 200: + raise RuntimeError("Downloading from {} failed with code " + "{}!".format(url, req.status_code)) + + # For protecting download interupted, download to + # tmp_fullname firstly, move tmp_fullname to fullname + # after download finished + tmp_fullname = fullname + "_tmp" + total_size = req.headers.get('content-length') + with open(tmp_fullname, 'wb') as f: + if total_size: + with tqdm(total=(int(total_size) + 1023) // 1024) as pbar: + for chunk in req.iter_content(chunk_size=1024): + f.write(chunk) + pbar.update(1) + else: + for chunk in req.iter_content(chunk_size=1024): + if chunk: + f.write(chunk) + shutil.move(tmp_fullname, fullname) + + return fullname + + +def _md5check(fullname, md5sum=None): + if md5sum is None: + return True + + logger.info("File {} md5 checking...".format(fullname)) + md5 = hashlib.md5() + with open(fullname, 'rb') as f: + for chunk in iter(lambda: f.read(4096), b""): + md5.update(chunk) + calc_md5sum = md5.hexdigest() + + if calc_md5sum != md5sum: + logger.info("File {} md5 check failed, {}(calc) != " + "{}(base)".format(fullname, calc_md5sum, md5sum)) + return False + return True + + +def _decompress(fname): + """ + Decompress for zip and tar file + """ + logger.info("Decompressing {}...".format(fname)) + + # For protecting decompressing interupted, + # decompress to fpath_tmp directory firstly, if decompress + # successed, move decompress files to fpath and delete + # fpath_tmp and remove download compress file. + + if tarfile.is_tarfile(fname): + uncompressed_path = _uncompress_file_tar(fname) + elif zipfile.is_zipfile(fname): + uncompressed_path = _uncompress_file_zip(fname) + else: + raise TypeError("Unsupport compress file type {}".format(fname)) + + return uncompressed_path + + +def _uncompress_file_zip(filepath): + files = zipfile.ZipFile(filepath, 'r') + file_list = files.namelist() + + file_dir = os.path.dirname(filepath) + + if _is_a_single_file(file_list): + rootpath = file_list[0] + uncompressed_path = os.path.join(file_dir, rootpath) + + for item in file_list: + files.extract(item, file_dir) + + elif _is_a_single_dir(file_list): + rootpath = os.path.splitext(file_list[0])[0].split(os.sep)[-1] + uncompressed_path = os.path.join(file_dir, rootpath) + + for item in file_list: + files.extract(item, file_dir) + + else: + rootpath = os.path.splitext(filepath)[0].split(os.sep)[-1] + uncompressed_path = os.path.join(file_dir, rootpath) + if not os.path.exists(uncompressed_path): + os.makedirs(uncompressed_path) + for item in file_list: + files.extract(item, os.path.join(file_dir, rootpath)) + + files.close() + + return uncompressed_path + + +def _uncompress_file_tar(filepath, mode="r:*"): + files = tarfile.open(filepath, mode) + file_list = files.getnames() + + file_dir = os.path.dirname(filepath) + + if _is_a_single_file(file_list): + rootpath = file_list[0] + uncompressed_path = os.path.join(file_dir, rootpath) + for item in file_list: + files.extract(item, file_dir) + elif _is_a_single_dir(file_list): + rootpath = os.path.splitext(file_list[0])[0].split(os.sep)[-1] + uncompressed_path = os.path.join(file_dir, rootpath) + for item in file_list: + files.extract(item, file_dir) + else: + rootpath = os.path.splitext(filepath)[0].split(os.sep)[-1] + uncompressed_path = os.path.join(file_dir, rootpath) + if not os.path.exists(uncompressed_path): + os.makedirs(uncompressed_path) + + for item in file_list: + files.extract(item, os.path.join(file_dir, rootpath)) + + files.close() + + return uncompressed_path + + +def _is_a_single_file(file_list): + if len(file_list) == 1 and file_list[0].find(os.sep) < -1: + return True + return False + + +def _is_a_single_dir(file_list): + new_file_list = [] + for file_path in file_list: + if '/' in file_path: + file_path = file_path.replace('/', os.sep) + elif '\\' in file_path: + file_path = file_path.replace('\\', os.sep) + new_file_list.append(file_path) + + file_name = new_file_list[0].split(os.sep)[0] + for i in range(1, len(new_file_list)): + if file_name != new_file_list[i].split(os.sep)[0]: + return False + return True diff --git a/exprements/PaddleClas/ppcls/utils/ema.py b/exprements/PaddleClas/ppcls/utils/ema.py new file mode 100644 index 00000000..b54cdb1b --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/ema.py @@ -0,0 +1,63 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +import numpy as np + + +class ExponentialMovingAverage(): + """ + Exponential Moving Average + Code was heavily based on https://github.com/Wanger-SJTU/SegToolbox.Pytorch/blob/master/lib/utils/ema.py + """ + + def __init__(self, model, decay, thres_steps=True): + self._model = model + self._decay = decay + self._thres_steps = thres_steps + self._shadow = {} + self._backup = {} + + def register(self): + self._update_step = 0 + for name, param in self._model.named_parameters(): + if param.stop_gradient is False: + self._shadow[name] = param.numpy().copy() + + def update(self): + decay = min(self._decay, (1 + self._update_step) / ( + 10 + self._update_step)) if self._thres_steps else self._decay + for name, param in self._model.named_parameters(): + if param.stop_gradient is False: + assert name in self._shadow + new_val = np.array(param.numpy().copy()) + old_val = np.array(self._shadow[name]) + new_average = decay * old_val + (1 - decay) * new_val + self._shadow[name] = new_average + self._update_step += 1 + return decay + + def apply(self): + for name, param in self._model.named_parameters(): + if param.stop_gradient is False: + assert name in self._shadow + self._backup[name] = np.array(param.numpy().copy()) + param.set_value(np.array(self._shadow[name])) + + def restore(self): + for name, param in self._model.named_parameters(): + if param.stop_gradient is False: + assert name in self._backup + param.set_value(self._backup[name]) + self._backup = {} diff --git a/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/fm_vis.py b/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/fm_vis.py new file mode 100644 index 00000000..a5368b10 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/fm_vis.py @@ -0,0 +1,97 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import numpy as np +import cv2 +import utils +import argparse +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(__dir__) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../../..'))) + +import paddle +from paddle.distributed import ParallelEnv + +from resnet import ResNet50 +from ppcls.utils.save_load import load_dygraph_pretrain + + +def parse_args(): + def str2bool(v): + return v.lower() in ("true", "t", "1") + + parser = argparse.ArgumentParser() + parser.add_argument("-i", "--image_file", required=True, type=str) + parser.add_argument("-c", "--channel_num", type=int) + parser.add_argument("-p", "--pretrained_model", type=str) + parser.add_argument("--show", type=str2bool, default=False) + parser.add_argument("--interpolation", type=int, default=1) + parser.add_argument("--save_path", type=str, default=None) + parser.add_argument("--use_gpu", type=str2bool, default=True) + + return parser.parse_args() + + +def create_operators(interpolation=1): + size = 224 + img_mean = [0.485, 0.456, 0.406] + img_std = [0.229, 0.224, 0.225] + img_scale = 1.0 / 255.0 + + resize_op = utils.ResizeImage( + resize_short=256, interpolation=interpolation) + crop_op = utils.CropImage(size=(size, size)) + normalize_op = utils.NormalizeImage( + scale=img_scale, mean=img_mean, std=img_std) + totensor_op = utils.ToTensor() + + return [resize_op, crop_op, normalize_op, totensor_op] + + +def preprocess(data, ops): + for op in ops: + data = op(data) + return data + + +def main(): + args = parse_args() + operators = create_operators(args.interpolation) + # assign the place + place = 'gpu:{}'.format(ParallelEnv().dev_id) if args.use_gpu else 'cpu' + place = paddle.set_device(place) + + net = ResNet50() + load_dygraph_pretrain(net, args.pretrained_model) + + img = cv2.imread(args.image_file, cv2.IMREAD_COLOR) + data = preprocess(img, operators) + data = np.expand_dims(data, axis=0) + data = paddle.to_tensor(data) + net.eval() + _, fm = net(data) + assert args.channel_num >= 0 and args.channel_num <= fm.shape[ + 1], "the channel is out of the range, should be in {} but got {}".format( + [0, fm.shape[1]], args.channel_num) + + fm = (np.squeeze(fm[0][args.channel_num].numpy()) * 255).astype(np.uint8) + fm = cv2.resize(fm, (img.shape[1], img.shape[0])) + if args.save_path is not None: + print("the feature map is saved in path: {}".format(args.save_path)) + cv2.imwrite(args.save_path, fm) + + +if __name__ == "__main__": + main() diff --git a/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/resnet.py b/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/resnet.py new file mode 100644 index 00000000..b7588141 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/resnet.py @@ -0,0 +1,535 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import numpy as np +import paddle +from paddle import ParamAttr +import paddle.nn as nn +from paddle.nn import Conv2D, BatchNorm, Linear +from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D +from paddle.nn.initializer import Uniform +import math + +from ppcls.arch.backbone.base.theseus_layer import TheseusLayer +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "ResNet18": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_pretrained.pdparams", + "ResNet18_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_vd_pretrained.pdparams", + "ResNet34": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_pretrained.pdparams", + "ResNet34_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_pretrained.pdparams", + "ResNet50": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_pretrained.pdparams", + "ResNet50_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_pretrained.pdparams", + "ResNet101": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_pretrained.pdparams", + "ResNet101_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_vd_pretrained.pdparams", + "ResNet152": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_pretrained.pdparams", + "ResNet152_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_vd_pretrained.pdparams", + "ResNet200_vd": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet200_vd_pretrained.pdparams", +} + +__all__ = MODEL_URLS.keys() +''' +ResNet config: dict. + key: depth of ResNet. + values: config's dict of specific model. + keys: + block_type: Two different blocks in ResNet, BasicBlock and BottleneckBlock are optional. + block_depth: The number of blocks in different stages in ResNet. + num_channels: The number of channels to enter the next stage. +''' +NET_CONFIG = { + "18": { + "block_type": "BasicBlock", + "block_depth": [2, 2, 2, 2], + "num_channels": [64, 64, 128, 256] + }, + "34": { + "block_type": "BasicBlock", + "block_depth": [3, 4, 6, 3], + "num_channels": [64, 64, 128, 256] + }, + "50": { + "block_type": "BottleneckBlock", + "block_depth": [3, 4, 6, 3], + "num_channels": [64, 256, 512, 1024] + }, + "101": { + "block_type": "BottleneckBlock", + "block_depth": [3, 4, 23, 3], + "num_channels": [64, 256, 512, 1024] + }, + "152": { + "block_type": "BottleneckBlock", + "block_depth": [3, 8, 36, 3], + "num_channels": [64, 256, 512, 1024] + }, + "200": { + "block_type": "BottleneckBlock", + "block_depth": [3, 12, 48, 3], + "num_channels": [64, 256, 512, 1024] + }, +} + + +class ConvBNLayer(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + filter_size, + stride=1, + groups=1, + is_vd_mode=False, + act=None, + lr_mult=1.0, + data_format="NCHW"): + super().__init__() + self.is_vd_mode = is_vd_mode + self.act = act + self.avg_pool = AvgPool2D( + kernel_size=2, stride=2, padding=0, ceil_mode=True) + self.conv = Conv2D( + in_channels=num_channels, + out_channels=num_filters, + kernel_size=filter_size, + stride=stride, + padding=(filter_size - 1) // 2, + groups=groups, + weight_attr=ParamAttr(learning_rate=lr_mult), + bias_attr=False, + data_format=data_format) + self.bn = BatchNorm( + num_filters, + param_attr=ParamAttr(learning_rate=lr_mult), + bias_attr=ParamAttr(learning_rate=lr_mult), + data_layout=data_format) + self.relu = nn.ReLU() + + def forward(self, x): + if self.is_vd_mode: + x = self.avg_pool(x) + x = self.conv(x) + x = self.bn(x) + if self.act: + x = self.relu(x) + return x + + +class BottleneckBlock(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + if_first=False, + lr_mult=1.0, + data_format="NCHW"): + super().__init__() + + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + act="relu", + lr_mult=lr_mult, + data_format=data_format) + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + stride=stride, + act="relu", + lr_mult=lr_mult, + data_format=data_format) + self.conv2 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters * 4, + filter_size=1, + act=None, + lr_mult=lr_mult, + data_format=data_format) + + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters * 4, + filter_size=1, + stride=stride if if_first else 1, + is_vd_mode=False if if_first else True, + lr_mult=lr_mult, + data_format=data_format) + self.relu = nn.ReLU() + self.shortcut = shortcut + + def forward(self, x): + identity = x + x = self.conv0(x) + x = self.conv1(x) + x = self.conv2(x) + + if self.shortcut: + short = identity + else: + short = self.short(identity) + x = paddle.add(x=x, y=short) + x = self.relu(x) + return x + + +class BasicBlock(TheseusLayer): + def __init__(self, + num_channels, + num_filters, + stride, + shortcut=True, + if_first=False, + lr_mult=1.0, + data_format="NCHW"): + super().__init__() + + self.stride = stride + self.conv0 = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=3, + stride=stride, + act="relu", + lr_mult=lr_mult, + data_format=data_format) + self.conv1 = ConvBNLayer( + num_channels=num_filters, + num_filters=num_filters, + filter_size=3, + act=None, + lr_mult=lr_mult, + data_format=data_format) + if not shortcut: + self.short = ConvBNLayer( + num_channels=num_channels, + num_filters=num_filters, + filter_size=1, + stride=stride if if_first else 1, + is_vd_mode=False if if_first else True, + lr_mult=lr_mult, + data_format=data_format) + self.shortcut = shortcut + self.relu = nn.ReLU() + + def forward(self, x): + identity = x + x = self.conv0(x) + x = self.conv1(x) + if self.shortcut: + short = identity + else: + short = self.short(identity) + x = paddle.add(x=x, y=short) + x = self.relu(x) + return x + + +class ResNet(TheseusLayer): + """ + ResNet + Args: + config: dict. config of ResNet. + version: str="vb". Different version of ResNet, version vd can perform better. + class_num: int=1000. The number of classes. + lr_mult_list: list. Control the learning rate of different stages. + Returns: + model: nn.Layer. Specific ResNet model depends on args. + """ + + def __init__(self, + config, + version="vb", + class_num=1000, + lr_mult_list=[1.0, 1.0, 1.0, 1.0, 1.0], + data_format="NCHW", + input_image_channel=3, + return_patterns=None): + super().__init__() + + self.cfg = config + self.lr_mult_list = lr_mult_list + self.is_vd_mode = version == "vd" + self.class_num = class_num + self.num_filters = [64, 128, 256, 512] + self.block_depth = self.cfg["block_depth"] + self.block_type = self.cfg["block_type"] + self.num_channels = self.cfg["num_channels"] + self.channels_mult = 1 if self.num_channels[-1] == 256 else 4 + + assert isinstance(self.lr_mult_list, ( + list, tuple + )), "lr_mult_list should be in (list, tuple) but got {}".format( + type(self.lr_mult_list)) + assert len(self.lr_mult_list + ) == 5, "lr_mult_list length should be 5 but got {}".format( + len(self.lr_mult_list)) + + self.stem_cfg = { + #num_channels, num_filters, filter_size, stride + "vb": [[input_image_channel, 64, 7, 2]], + "vd": + [[input_image_channel, 32, 3, 2], [32, 32, 3, 1], [32, 64, 3, 1]] + } + + self.stem = nn.Sequential(* [ + ConvBNLayer( + num_channels=in_c, + num_filters=out_c, + filter_size=k, + stride=s, + act="relu", + lr_mult=self.lr_mult_list[0], + data_format=data_format) + for in_c, out_c, k, s in self.stem_cfg[version] + ]) + + self.max_pool = MaxPool2D( + kernel_size=3, stride=2, padding=1, data_format=data_format) + block_list = [] + for block_idx in range(len(self.block_depth)): + shortcut = False + for i in range(self.block_depth[block_idx]): + block_list.append(globals()[self.block_type]( + num_channels=self.num_channels[block_idx] if i == 0 else + self.num_filters[block_idx] * self.channels_mult, + num_filters=self.num_filters[block_idx], + stride=2 if i == 0 and block_idx != 0 else 1, + shortcut=shortcut, + if_first=block_idx == i == 0 if version == "vd" else True, + lr_mult=self.lr_mult_list[block_idx + 1], + data_format=data_format)) + shortcut = True + self.blocks = nn.Sequential(*block_list) + + self.avg_pool = AdaptiveAvgPool2D(1, data_format=data_format) + self.flatten = nn.Flatten() + self.avg_pool_channels = self.num_channels[-1] * 2 + stdv = 1.0 / math.sqrt(self.avg_pool_channels * 1.0) + self.fc = Linear( + self.avg_pool_channels, + self.class_num, + weight_attr=ParamAttr(initializer=Uniform(-stdv, stdv))) + + self.data_format = data_format + if return_patterns is not None: + self.update_res(return_patterns) + self.register_forward_post_hook(self._return_dict_hook) + + def forward(self, x): + with paddle.static.amp.fp16_guard(): + if self.data_format == "NHWC": + x = paddle.transpose(x, [0, 2, 3, 1]) + x.stop_gradient = True + x = self.stem(x) + fm = x + x = self.max_pool(x) + x = self.blocks(x) + x = self.avg_pool(x) + x = self.flatten(x) + x = self.fc(x) + return x, fm + + +def _load_pretrained(pretrained, model, model_url, use_ssld): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def ResNet18(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet18 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet18` model depends on args. + """ + model = ResNet(config=NET_CONFIG["18"], version="vb", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet18"], use_ssld) + return model + + +def ResNet18_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet18_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet18_vd` model depends on args. + """ + model = ResNet(config=NET_CONFIG["18"], version="vd", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet18_vd"], use_ssld) + return model + + +def ResNet34(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet34 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet34` model depends on args. + """ + model = ResNet(config=NET_CONFIG["34"], version="vb", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet34"], use_ssld) + return model + + +def ResNet34_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet34_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet34_vd` model depends on args. + """ + model = ResNet(config=NET_CONFIG["34"], version="vd", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet34_vd"], use_ssld) + return model + + +def ResNet50(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet50 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet50` model depends on args. + """ + model = ResNet(config=NET_CONFIG["50"], version="vb", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet50"], use_ssld) + return model + + +def ResNet50_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet50_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet50_vd` model depends on args. + """ + model = ResNet(config=NET_CONFIG["50"], version="vd", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet50_vd"], use_ssld) + return model + + +def ResNet101(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet101 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet101` model depends on args. + """ + model = ResNet(config=NET_CONFIG["101"], version="vb", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet101"], use_ssld) + return model + + +def ResNet101_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet101_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet101_vd` model depends on args. + """ + model = ResNet(config=NET_CONFIG["101"], version="vd", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet101_vd"], use_ssld) + return model + + +def ResNet152(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet152 + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet152` model depends on args. + """ + model = ResNet(config=NET_CONFIG["152"], version="vb", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet152"], use_ssld) + return model + + +def ResNet152_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet152_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet152_vd` model depends on args. + """ + model = ResNet(config=NET_CONFIG["152"], version="vd", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet152_vd"], use_ssld) + return model + + +def ResNet200_vd(pretrained=False, use_ssld=False, **kwargs): + """ + ResNet200_vd + Args: + pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise. + If str, means the path of the pretrained model. + use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True. + Returns: + model: nn.Layer. Specific `ResNet200_vd` model depends on args. + """ + model = ResNet(config=NET_CONFIG["200"], version="vd", **kwargs) + _load_pretrained(pretrained, model, MODEL_URLS["ResNet200_vd"], use_ssld) + return model diff --git a/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/utils.py b/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/utils.py new file mode 100644 index 00000000..7c701493 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/feature_maps_visualization/utils.py @@ -0,0 +1,85 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import cv2 +import numpy as np + + +class DecodeImage(object): + def __init__(self, to_rgb=True): + self.to_rgb = to_rgb + + def __call__(self, img): + data = np.frombuffer(img, dtype='uint8') + img = cv2.imdecode(data, 1) + if self.to_rgb: + assert img.shape[2] == 3, 'invalid shape of image[%s]' % ( + img.shape) + img = img[:, :, ::-1] + + return img + + +class ResizeImage(object): + def __init__(self, resize_short=None, interpolation=1): + self.resize_short = resize_short + self.interpolation = interpolation + + def __call__(self, img): + img_h, img_w = img.shape[:2] + percent = float(self.resize_short) / min(img_w, img_h) + w = int(round(img_w * percent)) + h = int(round(img_h * percent)) + return cv2.resize(img, (w, h), interpolation=self.interpolation) + + +class CropImage(object): + def __init__(self, size): + if type(size) is int: + self.size = (size, size) + else: + self.size = size + + def __call__(self, img): + w, h = self.size + img_h, img_w = img.shape[:2] + w_start = (img_w - w) // 2 + h_start = (img_h - h) // 2 + + w_end = w_start + w + h_end = h_start + h + return img[h_start:h_end, w_start:w_end, :] + + +class NormalizeImage(object): + def __init__(self, scale=None, mean=None, std=None): + self.scale = np.float32(scale if scale is not None else 1.0 / 255.0) + mean = mean if mean is not None else [0.485, 0.456, 0.406] + std = std if std is not None else [0.229, 0.224, 0.225] + + shape = (1, 1, 3) + self.mean = np.array(mean).reshape(shape).astype('float32') + self.std = np.array(std).reshape(shape).astype('float32') + + def __call__(self, img): + return (img.astype('float32') * self.scale - self.mean) / self.std + + +class ToTensor(object): + def __init__(self): + pass + + def __call__(self, img): + img = img.transpose((2, 0, 1)) + return img diff --git a/exprements/PaddleClas/ppcls/utils/gallery2fc.py b/exprements/PaddleClas/ppcls/utils/gallery2fc.py new file mode 100644 index 00000000..67b08529 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/gallery2fc.py @@ -0,0 +1,119 @@ +# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import paddle +import cv2 + +from ppcls.arch import build_model +from ppcls.utils.config import parse_config, parse_args +from ppcls.utils.save_load import load_dygraph_pretrain +from ppcls.utils.logger import init_logger +from ppcls.data import create_operators +from ppcls.arch.slim import quantize_model + + +class GalleryLayer(paddle.nn.Layer): + def __init__(self, configs): + super().__init__() + self.configs = configs + embedding_size = self.configs["Arch"]["Head"]["embedding_size"] + self.batch_size = self.configs["IndexProcess"]["batch_size"] + self.image_shape = self.configs["Global"]["image_shape"].copy() + self.image_shape.insert(0, self.batch_size) + + image_root = self.configs["IndexProcess"]["image_root"] + data_file = self.configs["IndexProcess"]["data_file"] + delimiter = self.configs["IndexProcess"]["delimiter"] + self.gallery_images = [] + gallery_docs = [] + gallery_labels = [] + + with open(data_file, 'r', encoding='utf-8') as f: + lines = f.readlines() + for ori_line in lines: + line = ori_line.strip().split(delimiter) + text_num = len(line) + assert text_num >= 2, f"line({ori_line}) must be splitted into at least 2 parts, but got {text_num}" + image_file = os.path.join(image_root, line[0]) + + self.gallery_images.append(image_file) + gallery_docs.append(ori_line.strip()) + gallery_labels.append(line[1].strip()) + self.gallery_layer = paddle.nn.Linear(embedding_size, len(self.gallery_images), bias_attr=False) + self.gallery_layer.skip_quant = True + output_label_str = "" + for i, label_i in enumerate(gallery_labels): + output_label_str += "{} {}\n".format(i, label_i) + output_path = configs["Global"]["save_inference_dir"] + "_label.txt" + with open(output_path, "w") as f: + f.write(output_label_str) + + def forward(self, x, label=None): + x = paddle.nn.functional.normalize(x) + x = self.gallery_layer(x) + return x + + def build_gallery_layer(self, feature_extractor): + transform_configs = self.configs["IndexProcess"]["transform_ops"] + preprocess_ops = create_operators(transform_configs) + embedding_size = self.configs["Arch"]["Head"]["embedding_size"] + batch_index = 0 + input_tensor = paddle.zeros(self.image_shape) + gallery_feature = paddle.zeros((len(self.gallery_images), embedding_size)) + for i, image_path in enumerate(self.gallery_images): + image = cv2.imread(image_path)[:, :, ::-1] + for op in preprocess_ops: + image = op(image) + input_tensor[batch_index] = image + batch_index += 1 + if batch_index == self.batch_size or i == len(self.gallery_images) - 1: + batch_feature = feature_extractor(input_tensor)["features"] + for j in range(batch_index): + feature = batch_feature[j] + norm_feature = paddle.nn.functional.normalize(feature, axis=0) + gallery_feature[i - batch_index + j + 1] = norm_feature + self.gallery_layer.set_state_dict({"_layer.weight": gallery_feature.T}) + + +def export_fuse_model(configs): + slim_config = configs["Slim"].copy() + configs["Slim"] = None + fuse_model = build_model(configs) + fuse_model.head = GalleryLayer(configs) + configs["Slim"] = slim_config + quantize_model(configs, fuse_model) + load_dygraph_pretrain(fuse_model, configs["Global"]["pretrained_model"]) + fuse_model.eval() + fuse_model.head.build_gallery_layer(fuse_model) + save_path = configs["Global"]["save_inference_dir"] + fuse_model.quanter.save_quantized_model( + fuse_model, + save_path, + input_spec=[ + paddle.static.InputSpec( + shape=[None] + configs["Global"]["image_shape"], + dtype='float32') + ]) + + +def main(): + args = parse_args() + configs = parse_config(args.config) + init_logger(name='gallery2fc') + export_fuse_model(configs) + + +if __name__ == '__main__': + main() diff --git a/exprements/PaddleClas/ppcls/utils/imagenet1k_label_list.txt b/exprements/PaddleClas/ppcls/utils/imagenet1k_label_list.txt new file mode 100644 index 00000000..376e1802 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/imagenet1k_label_list.txt @@ -0,0 +1,1000 @@ +0 tench, Tinca tinca +1 goldfish, Carassius auratus +2 great white shark, white shark, man-eater, man-eating shark, Carcharodon carcharias +3 tiger shark, Galeocerdo cuvieri +4 hammerhead, hammerhead shark +5 electric ray, crampfish, numbfish, torpedo +6 stingray +7 cock +8 hen +9 ostrich, Struthio camelus +10 brambling, Fringilla montifringilla +11 goldfinch, Carduelis carduelis +12 house finch, linnet, Carpodacus mexicanus +13 junco, snowbird +14 indigo bunting, indigo finch, indigo bird, Passerina cyanea +15 robin, American robin, Turdus migratorius +16 bulbul +17 jay +18 magpie +19 chickadee +20 water ouzel, dipper +21 kite +22 bald eagle, American eagle, Haliaeetus leucocephalus +23 vulture +24 great grey owl, great gray owl, Strix nebulosa +25 European fire salamander, Salamandra salamandra +26 common newt, Triturus vulgaris +27 eft +28 spotted salamander, Ambystoma maculatum +29 axolotl, mud puppy, Ambystoma mexicanum +30 bullfrog, Rana catesbeiana +31 tree frog, tree-frog +32 tailed frog, bell toad, ribbed toad, tailed toad, Ascaphus trui +33 loggerhead, loggerhead turtle, Caretta caretta +34 leatherback turtle, leatherback, leathery turtle, Dermochelys coriacea +35 mud turtle +36 terrapin +37 box turtle, box tortoise +38 banded gecko +39 common iguana, iguana, Iguana iguana +40 American chameleon, anole, Anolis carolinensis +41 whiptail, whiptail lizard +42 agama +43 frilled lizard, Chlamydosaurus kingi +44 alligator lizard +45 Gila monster, Heloderma suspectum +46 green lizard, Lacerta viridis +47 African chameleon, Chamaeleo chamaeleon +48 Komodo dragon, Komodo lizard, dragon lizard, giant lizard, Varanus komodoensis +49 African crocodile, Nile crocodile, Crocodylus niloticus +50 American alligator, Alligator mississipiensis +51 triceratops +52 thunder snake, worm snake, Carphophis amoenus +53 ringneck snake, ring-necked snake, ring snake +54 hognose snake, puff adder, sand viper +55 green snake, grass snake +56 king snake, kingsnake +57 garter snake, grass snake +58 water snake +59 vine snake +60 night snake, Hypsiglena torquata +61 boa constrictor, Constrictor constrictor +62 rock python, rock snake, Python sebae +63 Indian cobra, Naja naja +64 green mamba +65 sea snake +66 horned viper, cerastes, sand viper, horned asp, Cerastes cornutus +67 diamondback, diamondback rattlesnake, Crotalus adamanteus +68 sidewinder, horned rattlesnake, Crotalus cerastes +69 trilobite +70 harvestman, daddy longlegs, Phalangium opilio +71 scorpion +72 black and gold garden spider, Argiope aurantia +73 barn spider, Araneus cavaticus +74 garden spider, Aranea diademata +75 black widow, Latrodectus mactans +76 tarantula +77 wolf spider, hunting spider +78 tick +79 centipede +80 black grouse +81 ptarmigan +82 ruffed grouse, partridge, Bonasa umbellus +83 prairie chicken, prairie grouse, prairie fowl +84 peacock +85 quail +86 partridge +87 African grey, African gray, Psittacus erithacus +88 macaw +89 sulphur-crested cockatoo, Kakatoe galerita, Cacatua galerita +90 lorikeet +91 coucal +92 bee eater +93 hornbill +94 hummingbird +95 jacamar +96 toucan +97 drake +98 red-breasted merganser, Mergus serrator +99 goose +100 black swan, Cygnus atratus +101 tusker +102 echidna, spiny anteater, anteater +103 platypus, duckbill, duckbilled platypus, duck-billed platypus, Ornithorhynchus anatinus +104 wallaby, brush kangaroo +105 koala, koala bear, kangaroo bear, native bear, Phascolarctos cinereus +106 wombat +107 jellyfish +108 sea anemone, anemone +109 brain coral +110 flatworm, platyhelminth +111 nematode, nematode worm, roundworm +112 conch +113 snail +114 slug +115 sea slug, nudibranch +116 chiton, coat-of-mail shell, sea cradle, polyplacophore +117 chambered nautilus, pearly nautilus, nautilus +118 Dungeness crab, Cancer magister +119 rock crab, Cancer irroratus +120 fiddler crab +121 king crab, Alaska crab, Alaskan king crab, Alaska king crab, Paralithodes camtschatica +122 American lobster, Northern lobster, Maine lobster, Homarus americanus +123 spiny lobster, langouste, rock lobster, crawfish, crayfish, sea crawfish +124 crayfish, crawfish, crawdad, crawdaddy +125 hermit crab +126 isopod +127 white stork, Ciconia ciconia +128 black stork, Ciconia nigra +129 spoonbill +130 flamingo +131 little blue heron, Egretta caerulea +132 American egret, great white heron, Egretta albus +133 bittern +134 crane +135 limpkin, Aramus pictus +136 European gallinule, Porphyrio porphyrio +137 American coot, marsh hen, mud hen, water hen, Fulica americana +138 bustard +139 ruddy turnstone, Arenaria interpres +140 red-backed sandpiper, dunlin, Erolia alpina +141 redshank, Tringa totanus +142 dowitcher +143 oystercatcher, oyster catcher +144 pelican +145 king penguin, Aptenodytes patagonica +146 albatross, mollymawk +147 grey whale, gray whale, devilfish, Eschrichtius gibbosus, Eschrichtius robustus +148 killer whale, killer, orca, grampus, sea wolf, Orcinus orca +149 dugong, Dugong dugon +150 sea lion +151 Chihuahua +152 Japanese spaniel +153 Maltese dog, Maltese terrier, Maltese +154 Pekinese, Pekingese, Peke +155 Shih-Tzu +156 Blenheim spaniel +157 papillon +158 toy terrier +159 Rhodesian ridgeback +160 Afghan hound, Afghan +161 basset, basset hound +162 beagle +163 bloodhound, sleuthhound +164 bluetick +165 black-and-tan coonhound +166 Walker hound, Walker foxhound +167 English foxhound +168 redbone +169 borzoi, Russian wolfhound +170 Irish wolfhound +171 Italian greyhound +172 whippet +173 Ibizan hound, Ibizan Podenco +174 Norwegian elkhound, elkhound +175 otterhound, otter hound +176 Saluki, gazelle hound +177 Scottish deerhound, deerhound +178 Weimaraner +179 Staffordshire bullterrier, Staffordshire bull terrier +180 American Staffordshire terrier, Staffordshire terrier, American pit bull terrier, pit bull terrier +181 Bedlington terrier +182 Border terrier +183 Kerry blue terrier +184 Irish terrier +185 Norfolk terrier +186 Norwich terrier +187 Yorkshire terrier +188 wire-haired fox terrier +189 Lakeland terrier +190 Sealyham terrier, Sealyham +191 Airedale, Airedale terrier +192 cairn, cairn terrier +193 Australian terrier +194 Dandie Dinmont, Dandie Dinmont terrier +195 Boston bull, Boston terrier +196 miniature schnauzer +197 giant schnauzer +198 standard schnauzer +199 Scotch terrier, Scottish terrier, Scottie +200 Tibetan terrier, chrysanthemum dog +201 silky terrier, Sydney silky +202 soft-coated wheaten terrier +203 West Highland white terrier +204 Lhasa, Lhasa apso +205 flat-coated retriever +206 curly-coated retriever +207 golden retriever +208 Labrador retriever +209 Chesapeake Bay retriever +210 German short-haired pointer +211 vizsla, Hungarian pointer +212 English setter +213 Irish setter, red setter +214 Gordon setter +215 Brittany spaniel +216 clumber, clumber spaniel +217 English springer, English springer spaniel +218 Welsh springer spaniel +219 cocker spaniel, English cocker spaniel, cocker +220 Sussex spaniel +221 Irish water spaniel +222 kuvasz +223 schipperke +224 groenendael +225 malinois +226 briard +227 kelpie +228 komondor +229 Old English sheepdog, bobtail +230 Shetland sheepdog, Shetland sheep dog, Shetland +231 collie +232 Border collie +233 Bouvier des Flandres, Bouviers des Flandres +234 Rottweiler +235 German shepherd, German shepherd dog, German police dog, alsatian +236 Doberman, Doberman pinscher +237 miniature pinscher +238 Greater Swiss Mountain dog +239 Bernese mountain dog +240 Appenzeller +241 EntleBucher +242 boxer +243 bull mastiff +244 Tibetan mastiff +245 French bulldog +246 Great Dane +247 Saint Bernard, St Bernard +248 Eskimo dog, husky +249 malamute, malemute, Alaskan malamute +250 Siberian husky +251 dalmatian, coach dog, carriage dog +252 affenpinscher, monkey pinscher, monkey dog +253 basenji +254 pug, pug-dog +255 Leonberg +256 Newfoundland, Newfoundland dog +257 Great Pyrenees +258 Samoyed, Samoyede +259 Pomeranian +260 chow, chow chow +261 keeshond +262 Brabancon griffon +263 Pembroke, Pembroke Welsh corgi +264 Cardigan, Cardigan Welsh corgi +265 toy poodle +266 miniature poodle +267 standard poodle +268 Mexican hairless +269 timber wolf, grey wolf, gray wolf, Canis lupus +270 white wolf, Arctic wolf, Canis lupus tundrarum +271 red wolf, maned wolf, Canis rufus, Canis niger +272 coyote, prairie wolf, brush wolf, Canis latrans +273 dingo, warrigal, warragal, Canis dingo +274 dhole, Cuon alpinus +275 African hunting dog, hyena dog, Cape hunting dog, Lycaon pictus +276 hyena, hyaena +277 red fox, Vulpes vulpes +278 kit fox, Vulpes macrotis +279 Arctic fox, white fox, Alopex lagopus +280 grey fox, gray fox, Urocyon cinereoargenteus +281 tabby, tabby cat +282 tiger cat +283 Persian cat +284 Siamese cat, Siamese +285 Egyptian cat +286 cougar, puma, catamount, mountain lion, painter, panther, Felis concolor +287 lynx, catamount +288 leopard, Panthera pardus +289 snow leopard, ounce, Panthera uncia +290 jaguar, panther, Panthera onca, Felis onca +291 lion, king of beasts, Panthera leo +292 tiger, Panthera tigris +293 cheetah, chetah, Acinonyx jubatus +294 brown bear, bruin, Ursus arctos +295 American black bear, black bear, Ursus americanus, Euarctos americanus +296 ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus +297 sloth bear, Melursus ursinus, Ursus ursinus +298 mongoose +299 meerkat, mierkat +300 tiger beetle +301 ladybug, ladybeetle, lady beetle, ladybird, ladybird beetle +302 ground beetle, carabid beetle +303 long-horned beetle, longicorn, longicorn beetle +304 leaf beetle, chrysomelid +305 dung beetle +306 rhinoceros beetle +307 weevil +308 fly +309 bee +310 ant, emmet, pismire +311 grasshopper, hopper +312 cricket +313 walking stick, walkingstick, stick insect +314 cockroach, roach +315 mantis, mantid +316 cicada, cicala +317 leafhopper +318 lacewing, lacewing fly +319 dragonfly, darning needle, devil's darning needle, sewing needle, snake feeder, snake doctor, mosquito hawk, skeeter hawk +320 damselfly +321 admiral +322 ringlet, ringlet butterfly +323 monarch, monarch butterfly, milkweed butterfly, Danaus plexippus +324 cabbage butterfly +325 sulphur butterfly, sulfur butterfly +326 lycaenid, lycaenid butterfly +327 starfish, sea star +328 sea urchin +329 sea cucumber, holothurian +330 wood rabbit, cottontail, cottontail rabbit +331 hare +332 Angora, Angora rabbit +333 hamster +334 porcupine, hedgehog +335 fox squirrel, eastern fox squirrel, Sciurus niger +336 marmot +337 beaver +338 guinea pig, Cavia cobaya +339 sorrel +340 zebra +341 hog, pig, grunter, squealer, Sus scrofa +342 wild boar, boar, Sus scrofa +343 warthog +344 hippopotamus, hippo, river horse, Hippopotamus amphibius +345 ox +346 water buffalo, water ox, Asiatic buffalo, Bubalus bubalis +347 bison +348 ram, tup +349 bighorn, bighorn sheep, cimarron, Rocky Mountain bighorn, Rocky Mountain sheep, Ovis canadensis +350 ibex, Capra ibex +351 hartebeest +352 impala, Aepyceros melampus +353 gazelle +354 Arabian camel, dromedary, Camelus dromedarius +355 llama +356 weasel +357 mink +358 polecat, fitch, foulmart, foumart, Mustela putorius +359 black-footed ferret, ferret, Mustela nigripes +360 otter +361 skunk, polecat, wood pussy +362 badger +363 armadillo +364 three-toed sloth, ai, Bradypus tridactylus +365 orangutan, orang, orangutang, Pongo pygmaeus +366 gorilla, Gorilla gorilla +367 chimpanzee, chimp, Pan troglodytes +368 gibbon, Hylobates lar +369 siamang, Hylobates syndactylus, Symphalangus syndactylus +370 guenon, guenon monkey +371 patas, hussar monkey, Erythrocebus patas +372 baboon +373 macaque +374 langur +375 colobus, colobus monkey +376 proboscis monkey, Nasalis larvatus +377 marmoset +378 capuchin, ringtail, Cebus capucinus +379 howler monkey, howler +380 titi, titi monkey +381 spider monkey, Ateles geoffroyi +382 squirrel monkey, Saimiri sciureus +383 Madagascar cat, ring-tailed lemur, Lemur catta +384 indri, indris, Indri indri, Indri brevicaudatus +385 Indian elephant, Elephas maximus +386 African elephant, Loxodonta africana +387 lesser panda, red panda, panda, bear cat, cat bear, Ailurus fulgens +388 giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca +389 barracouta, snoek +390 eel +391 coho, cohoe, coho salmon, blue jack, silver salmon, Oncorhynchus kisutch +392 rock beauty, Holocanthus tricolor +393 anemone fish +394 sturgeon +395 gar, garfish, garpike, billfish, Lepisosteus osseus +396 lionfish +397 puffer, pufferfish, blowfish, globefish +398 abacus +399 abaya +400 academic gown, academic robe, judge's robe +401 accordion, piano accordion, squeeze box +402 acoustic guitar +403 aircraft carrier, carrier, flattop, attack aircraft carrier +404 airliner +405 airship, dirigible +406 altar +407 ambulance +408 amphibian, amphibious vehicle +409 analog clock +410 apiary, bee house +411 apron +412 ashcan, trash can, garbage can, wastebin, ash bin, ash-bin, ashbin, dustbin, trash barrel, trash bin +413 assault rifle, assault gun +414 backpack, back pack, knapsack, packsack, rucksack, haversack +415 bakery, bakeshop, bakehouse +416 balance beam, beam +417 balloon +418 ballpoint, ballpoint pen, ballpen, Biro +419 Band Aid +420 banjo +421 bannister, banister, balustrade, balusters, handrail +422 barbell +423 barber chair +424 barbershop +425 barn +426 barometer +427 barrel, cask +428 barrow, garden cart, lawn cart, wheelbarrow +429 baseball +430 basketball +431 bassinet +432 bassoon +433 bathing cap, swimming cap +434 bath towel +435 bathtub, bathing tub, bath, tub +436 beach wagon, station wagon, wagon, estate car, beach waggon, station waggon, waggon +437 beacon, lighthouse, beacon light, pharos +438 beaker +439 bearskin, busby, shako +440 beer bottle +441 beer glass +442 bell cote, bell cot +443 bib +444 bicycle-built-for-two, tandem bicycle, tandem +445 bikini, two-piece +446 binder, ring-binder +447 binoculars, field glasses, opera glasses +448 birdhouse +449 boathouse +450 bobsled, bobsleigh, bob +451 bolo tie, bolo, bola tie, bola +452 bonnet, poke bonnet +453 bookcase +454 bookshop, bookstore, bookstall +455 bottlecap +456 bow +457 bow tie, bow-tie, bowtie +458 brass, memorial tablet, plaque +459 brassiere, bra, bandeau +460 breakwater, groin, groyne, mole, bulwark, seawall, jetty +461 breastplate, aegis, egis +462 broom +463 bucket, pail +464 buckle +465 bulletproof vest +466 bullet train, bullet +467 butcher shop, meat market +468 cab, hack, taxi, taxicab +469 caldron, cauldron +470 candle, taper, wax light +471 cannon +472 canoe +473 can opener, tin opener +474 cardigan +475 car mirror +476 carousel, carrousel, merry-go-round, roundabout, whirligig +477 carpenter's kit, tool kit +478 carton +479 car wheel +480 cash machine, cash dispenser, automated teller machine, automatic teller machine, automated teller, automatic teller, ATM +481 cassette +482 cassette player +483 castle +484 catamaran +485 CD player +486 cello, violoncello +487 cellular telephone, cellular phone, cellphone, cell, mobile phone +488 chain +489 chainlink fence +490 chain mail, ring mail, mail, chain armor, chain armour, ring armor, ring armour +491 chain saw, chainsaw +492 chest +493 chiffonier, commode +494 chime, bell, gong +495 china cabinet, china closet +496 Christmas stocking +497 church, church building +498 cinema, movie theater, movie theatre, movie house, picture palace +499 cleaver, meat cleaver, chopper +500 cliff dwelling +501 cloak +502 clog, geta, patten, sabot +503 cocktail shaker +504 coffee mug +505 coffeepot +506 coil, spiral, volute, whorl, helix +507 combination lock +508 computer keyboard, keypad +509 confectionery, confectionary, candy store +510 container ship, containership, container vessel +511 convertible +512 corkscrew, bottle screw +513 cornet, horn, trumpet, trump +514 cowboy boot +515 cowboy hat, ten-gallon hat +516 cradle +517 crane +518 crash helmet +519 crate +520 crib, cot +521 Crock Pot +522 croquet ball +523 crutch +524 cuirass +525 dam, dike, dyke +526 desk +527 desktop computer +528 dial telephone, dial phone +529 diaper, nappy, napkin +530 digital clock +531 digital watch +532 dining table, board +533 dishrag, dishcloth +534 dishwasher, dish washer, dishwashing machine +535 disk brake, disc brake +536 dock, dockage, docking facility +537 dogsled, dog sled, dog sleigh +538 dome +539 doormat, welcome mat +540 drilling platform, offshore rig +541 drum, membranophone, tympan +542 drumstick +543 dumbbell +544 Dutch oven +545 electric fan, blower +546 electric guitar +547 electric locomotive +548 entertainment center +549 envelope +550 espresso maker +551 face powder +552 feather boa, boa +553 file, file cabinet, filing cabinet +554 fireboat +555 fire engine, fire truck +556 fire screen, fireguard +557 flagpole, flagstaff +558 flute, transverse flute +559 folding chair +560 football helmet +561 forklift +562 fountain +563 fountain pen +564 four-poster +565 freight car +566 French horn, horn +567 frying pan, frypan, skillet +568 fur coat +569 garbage truck, dustcart +570 gasmask, respirator, gas helmet +571 gas pump, gasoline pump, petrol pump, island dispenser +572 goblet +573 go-kart +574 golf ball +575 golfcart, golf cart +576 gondola +577 gong, tam-tam +578 gown +579 grand piano, grand +580 greenhouse, nursery, glasshouse +581 grille, radiator grille +582 grocery store, grocery, food market, market +583 guillotine +584 hair slide +585 hair spray +586 half track +587 hammer +588 hamper +589 hand blower, blow dryer, blow drier, hair dryer, hair drier +590 hand-held computer, hand-held microcomputer +591 handkerchief, hankie, hanky, hankey +592 hard disc, hard disk, fixed disk +593 harmonica, mouth organ, harp, mouth harp +594 harp +595 harvester, reaper +596 hatchet +597 holster +598 home theater, home theatre +599 honeycomb +600 hook, claw +601 hoopskirt, crinoline +602 horizontal bar, high bar +603 horse cart, horse-cart +604 hourglass +605 iPod +606 iron, smoothing iron +607 jack-o'-lantern +608 jean, blue jean, denim +609 jeep, landrover +610 jersey, T-shirt, tee shirt +611 jigsaw puzzle +612 jinrikisha, ricksha, rickshaw +613 joystick +614 kimono +615 knee pad +616 knot +617 lab coat, laboratory coat +618 ladle +619 lampshade, lamp shade +620 laptop, laptop computer +621 lawn mower, mower +622 lens cap, lens cover +623 letter opener, paper knife, paperknife +624 library +625 lifeboat +626 lighter, light, igniter, ignitor +627 limousine, limo +628 liner, ocean liner +629 lipstick, lip rouge +630 Loafer +631 lotion +632 loudspeaker, speaker, speaker unit, loudspeaker system, speaker system +633 loupe, jeweler's loupe +634 lumbermill, sawmill +635 magnetic compass +636 mailbag, postbag +637 mailbox, letter box +638 maillot +639 maillot, tank suit +640 manhole cover +641 maraca +642 marimba, xylophone +643 mask +644 matchstick +645 maypole +646 maze, labyrinth +647 measuring cup +648 medicine chest, medicine cabinet +649 megalith, megalithic structure +650 microphone, mike +651 microwave, microwave oven +652 military uniform +653 milk can +654 minibus +655 miniskirt, mini +656 minivan +657 missile +658 mitten +659 mixing bowl +660 mobile home, manufactured home +661 Model T +662 modem +663 monastery +664 monitor +665 moped +666 mortar +667 mortarboard +668 mosque +669 mosquito net +670 motor scooter, scooter +671 mountain bike, all-terrain bike, off-roader +672 mountain tent +673 mouse, computer mouse +674 mousetrap +675 moving van +676 muzzle +677 nail +678 neck brace +679 necklace +680 nipple +681 notebook, notebook computer +682 obelisk +683 oboe, hautboy, hautbois +684 ocarina, sweet potato +685 odometer, hodometer, mileometer, milometer +686 oil filter +687 organ, pipe organ +688 oscilloscope, scope, cathode-ray oscilloscope, CRO +689 overskirt +690 oxcart +691 oxygen mask +692 packet +693 paddle, boat paddle +694 paddlewheel, paddle wheel +695 padlock +696 paintbrush +697 pajama, pyjama, pj's, jammies +698 palace +699 panpipe, pandean pipe, syrinx +700 paper towel +701 parachute, chute +702 parallel bars, bars +703 park bench +704 parking meter +705 passenger car, coach, carriage +706 patio, terrace +707 pay-phone, pay-station +708 pedestal, plinth, footstall +709 pencil box, pencil case +710 pencil sharpener +711 perfume, essence +712 Petri dish +713 photocopier +714 pick, plectrum, plectron +715 pickelhaube +716 picket fence, paling +717 pickup, pickup truck +718 pier +719 piggy bank, penny bank +720 pill bottle +721 pillow +722 ping-pong ball +723 pinwheel +724 pirate, pirate ship +725 pitcher, ewer +726 plane, carpenter's plane, woodworking plane +727 planetarium +728 plastic bag +729 plate rack +730 plow, plough +731 plunger, plumber's helper +732 Polaroid camera, Polaroid Land camera +733 pole +734 police van, police wagon, paddy wagon, patrol wagon, wagon, black Maria +735 poncho +736 pool table, billiard table, snooker table +737 pop bottle, soda bottle +738 pot, flowerpot +739 potter's wheel +740 power drill +741 prayer rug, prayer mat +742 printer +743 prison, prison house +744 projectile, missile +745 projector +746 puck, hockey puck +747 punching bag, punch bag, punching ball, punchball +748 purse +749 quill, quill pen +750 quilt, comforter, comfort, puff +751 racer, race car, racing car +752 racket, racquet +753 radiator +754 radio, wireless +755 radio telescope, radio reflector +756 rain barrel +757 recreational vehicle, RV, R.V. +758 reel +759 reflex camera +760 refrigerator, icebox +761 remote control, remote +762 restaurant, eating house, eating place, eatery +763 revolver, six-gun, six-shooter +764 rifle +765 rocking chair, rocker +766 rotisserie +767 rubber eraser, rubber, pencil eraser +768 rugby ball +769 rule, ruler +770 running shoe +771 safe +772 safety pin +773 saltshaker, salt shaker +774 sandal +775 sarong +776 sax, saxophone +777 scabbard +778 scale, weighing machine +779 school bus +780 schooner +781 scoreboard +782 screen, CRT screen +783 screw +784 screwdriver +785 seat belt, seatbelt +786 sewing machine +787 shield, buckler +788 shoe shop, shoe-shop, shoe store +789 shoji +790 shopping basket +791 shopping cart +792 shovel +793 shower cap +794 shower curtain +795 ski +796 ski mask +797 sleeping bag +798 slide rule, slipstick +799 sliding door +800 slot, one-armed bandit +801 snorkel +802 snowmobile +803 snowplow, snowplough +804 soap dispenser +805 soccer ball +806 sock +807 solar dish, solar collector, solar furnace +808 sombrero +809 soup bowl +810 space bar +811 space heater +812 space shuttle +813 spatula +814 speedboat +815 spider web, spider's web +816 spindle +817 sports car, sport car +818 spotlight, spot +819 stage +820 steam locomotive +821 steel arch bridge +822 steel drum +823 stethoscope +824 stole +825 stone wall +826 stopwatch, stop watch +827 stove +828 strainer +829 streetcar, tram, tramcar, trolley, trolley car +830 stretcher +831 studio couch, day bed +832 stupa, tope +833 submarine, pigboat, sub, U-boat +834 suit, suit of clothes +835 sundial +836 sunglass +837 sunglasses, dark glasses, shades +838 sunscreen, sunblock, sun blocker +839 suspension bridge +840 swab, swob, mop +841 sweatshirt +842 swimming trunks, bathing trunks +843 swing +844 switch, electric switch, electrical switch +845 syringe +846 table lamp +847 tank, army tank, armored combat vehicle, armoured combat vehicle +848 tape player +849 teapot +850 teddy, teddy bear +851 television, television system +852 tennis ball +853 thatch, thatched roof +854 theater curtain, theatre curtain +855 thimble +856 thresher, thrasher, threshing machine +857 throne +858 tile roof +859 toaster +860 tobacco shop, tobacconist shop, tobacconist +861 toilet seat +862 torch +863 totem pole +864 tow truck, tow car, wrecker +865 toyshop +866 tractor +867 trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +868 tray +869 trench coat +870 tricycle, trike, velocipede +871 trimaran +872 tripod +873 triumphal arch +874 trolleybus, trolley coach, trackless trolley +875 trombone +876 tub, vat +877 turnstile +878 typewriter keyboard +879 umbrella +880 unicycle, monocycle +881 upright, upright piano +882 vacuum, vacuum cleaner +883 vase +884 vault +885 velvet +886 vending machine +887 vestment +888 viaduct +889 violin, fiddle +890 volleyball +891 waffle iron +892 wall clock +893 wallet, billfold, notecase, pocketbook +894 wardrobe, closet, press +895 warplane, military plane +896 washbasin, handbasin, washbowl, lavabo, wash-hand basin +897 washer, automatic washer, washing machine +898 water bottle +899 water jug +900 water tower +901 whiskey jug +902 whistle +903 wig +904 window screen +905 window shade +906 Windsor tie +907 wine bottle +908 wing +909 wok +910 wooden spoon +911 wool, woolen, woollen +912 worm fence, snake fence, snake-rail fence, Virginia fence +913 wreck +914 yawl +915 yurt +916 web site, website, internet site, site +917 comic book +918 crossword puzzle, crossword +919 street sign +920 traffic light, traffic signal, stoplight +921 book jacket, dust cover, dust jacket, dust wrapper +922 menu +923 plate +924 guacamole +925 consomme +926 hot pot, hotpot +927 trifle +928 ice cream, icecream +929 ice lolly, lolly, lollipop, popsicle +930 French loaf +931 bagel, beigel +932 pretzel +933 cheeseburger +934 hotdog, hot dog, red hot +935 mashed potato +936 head cabbage +937 broccoli +938 cauliflower +939 zucchini, courgette +940 spaghetti squash +941 acorn squash +942 butternut squash +943 cucumber, cuke +944 artichoke, globe artichoke +945 bell pepper +946 cardoon +947 mushroom +948 Granny Smith +949 strawberry +950 orange +951 lemon +952 fig +953 pineapple, ananas +954 banana +955 jackfruit, jak, jack +956 custard apple +957 pomegranate +958 hay +959 carbonara +960 chocolate sauce, chocolate syrup +961 dough +962 meat loaf, meatloaf +963 pizza, pizza pie +964 potpie +965 burrito +966 red wine +967 espresso +968 cup +969 eggnog +970 alp +971 bubble +972 cliff, drop, drop-off +973 coral reef +974 geyser +975 lakeside, lakeshore +976 promontory, headland, head, foreland +977 sandbar, sand bar +978 seashore, coast, seacoast, sea-coast +979 valley, vale +980 volcano +981 ballplayer, baseball player +982 groom, bridegroom +983 scuba diver +984 rapeseed +985 daisy +986 yellow lady's slipper, yellow lady-slipper, Cypripedium calceolus, Cypripedium parviflorum +987 corn +988 acorn +989 hip, rose hip, rosehip +990 buckeye, horse chestnut, conker +991 coral fungus +992 agaric +993 gyromitra +994 stinkhorn, carrion fungus +995 earthstar +996 hen-of-the-woods, hen of the woods, Polyporus frondosus, Grifola frondosa +997 bolete +998 ear, spike, capitulum +999 toilet tissue, toilet paper, bathroom tissue diff --git a/exprements/PaddleClas/ppcls/utils/logger.py b/exprements/PaddleClas/ppcls/utils/logger.py new file mode 100644 index 00000000..d4faaa9b --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/logger.py @@ -0,0 +1,137 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import sys + +import logging +import datetime +import paddle.distributed as dist + +_logger = None + + +def init_logger(name='root', log_file=None, log_level=logging.INFO): + """Initialize and get a logger by name. + If the logger has not been initialized, this method will initialize the + logger by adding one or two handlers, otherwise the initialized logger will + be directly returned. During initialization, a StreamHandler will always be + added. If `log_file` is specified a FileHandler will also be added. + Args: + name (str): Logger name. + log_file (str | None): The log filename. If specified, a FileHandler + will be added to the logger. + log_level (int): The logger level. Note that only the process of + rank 0 is affected, and other processes will set the level to + "Error" thus be silent most of the time. + Returns: + logging.Logger: The expected logger. + """ + global _logger + assert _logger is None, "logger should not be initialized twice or more." + _logger = logging.getLogger(name) + + formatter = logging.Formatter( + '[%(asctime)s] %(name)s %(levelname)s: %(message)s', + datefmt="%Y/%m/%d %H:%M:%S") + + stream_handler = logging.StreamHandler(stream=sys.stdout) + stream_handler.setFormatter(formatter) + _logger.addHandler(stream_handler) + if log_file is not None and dist.get_rank() == 0: + log_file_folder = os.path.split(log_file)[0] + os.makedirs(log_file_folder, exist_ok=True) + file_handler = logging.FileHandler(log_file, 'a') + file_handler.setFormatter(formatter) + _logger.addHandler(file_handler) + if dist.get_rank() == 0: + _logger.setLevel(log_level) + else: + _logger.setLevel(logging.ERROR) + + +def log_at_trainer0(log): + """ + logs will print multi-times when calling Fleet API. + Only display single log and ignore the others. + """ + + def wrapper(fmt, *args): + if dist.get_rank() == 0: + log(fmt, *args) + + return wrapper + + +@log_at_trainer0 +def info(fmt, *args): + _logger.info(fmt, *args) + + +@log_at_trainer0 +def debug(fmt, *args): + _logger.debug(fmt, *args) + + +@log_at_trainer0 +def warning(fmt, *args): + _logger.warning(fmt, *args) + + +@log_at_trainer0 +def error(fmt, *args): + _logger.error(fmt, *args) + + +def scaler(name, value, step, writer): + """ + This function will draw a scalar curve generated by the visualdl. + Usage: Install visualdl: pip3 install visualdl==2.0.0b4 + and then: + visualdl --logdir ./scalar --host 0.0.0.0 --port 8830 + to preview loss corve in real time. + """ + if writer is None: + return + writer.add_scalar(tag=name, step=step, value=value) + + +def advertise(): + """ + Show the advertising message like the following: + + =========================================================== + == PaddleClas is powered by PaddlePaddle ! == + =========================================================== + == == + == For more info please go to the following website. == + == == + == https://github.com/PaddlePaddle/PaddleClas == + =========================================================== + + """ + copyright = "PaddleClas is powered by PaddlePaddle !" + ad = "For more info please go to the following website." + website = "https://github.com/PaddlePaddle/PaddleClas" + AD_LEN = 6 + len(max([copyright, ad, website], key=len)) + + info("\n{0}\n{1}\n{2}\n{3}\n{4}\n{5}\n{6}\n{7}\n".format( + "=" * (AD_LEN + 4), + "=={}==".format(copyright.center(AD_LEN)), + "=" * (AD_LEN + 4), + "=={}==".format(' ' * AD_LEN), + "=={}==".format(ad.center(AD_LEN)), + "=={}==".format(' ' * AD_LEN), + "=={}==".format(website.center(AD_LEN)), + "=" * (AD_LEN + 4), )) diff --git a/exprements/PaddleClas/ppcls/utils/metrics.py b/exprements/PaddleClas/ppcls/utils/metrics.py new file mode 100644 index 00000000..b0db68a7 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/metrics.py @@ -0,0 +1,107 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +from sklearn.metrics import hamming_loss +from sklearn.metrics import accuracy_score as accuracy_metric +from sklearn.metrics import multilabel_confusion_matrix +from sklearn.metrics import precision_recall_fscore_support +from sklearn.metrics import average_precision_score +from sklearn.preprocessing import binarize + +import numpy as np + +__all__ = ["multi_hot_encode", "hamming_distance", "accuracy_score", "precision_recall_fscore", "mean_average_precision"] + + +def multi_hot_encode(logits, threshold=0.5): + """ + Encode logits to multi-hot by elementwise for multilabel + """ + + return binarize(logits, threshold=threshold) + + +def hamming_distance(output, target): + """ + Soft metric based label for multilabel classification + Returns: + The smaller the return value is, the better model is. + """ + + return hamming_loss(target, output) + + +def accuracy_score(output, target, base="sample"): + """ + Hard metric for multilabel classification + Args: + output: + target: + base: ["sample", "label"], default="sample" + if "sample", return metric score based sample, + if "label", return metric score based label. + Returns: + accuracy: + """ + + assert base in ["sample", "label"], 'must be one of ["sample", "label"]' + + if base == "sample": + accuracy = accuracy_metric(target, output) + elif base == "label": + mcm = multilabel_confusion_matrix(target, output) + tns = mcm[:, 0, 0] + fns = mcm[:, 1, 0] + tps = mcm[:, 1, 1] + fps = mcm[:, 0, 1] + + accuracy = (sum(tps) + sum(tns)) / (sum(tps) + sum(tns) + sum(fns) + sum(fps)) + + return accuracy + + +def precision_recall_fscore(output, target): + """ + Metric based label for multilabel classification + Returns: + precisions: + recalls: + fscores: + """ + + precisions, recalls, fscores, _ = precision_recall_fscore_support(target, output) + + return precisions, recalls, fscores + + +def mean_average_precision(logits, target): + """ + Calculate average precision + Args: + logits: probability from network before sigmoid or softmax + target: ground truth, 0 or 1 + """ + if not (isinstance(logits, np.ndarray) and isinstance(target, np.ndarray)): + raise TypeError("logits and target should be np.ndarray.") + + aps = [] + for i in range(target.shape[1]): + ap = average_precision_score(target[:, i], logits[:, i]) + aps.append(ap) + + return np.mean(aps) diff --git a/exprements/PaddleClas/ppcls/utils/misc.py b/exprements/PaddleClas/ppcls/utils/misc.py new file mode 100644 index 00000000..08ab7b6f --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/misc.py @@ -0,0 +1,63 @@ +# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +__all__ = ['AverageMeter'] + + +class AverageMeter(object): + """ + Computes and stores the average and current value + Code was based on https://github.com/pytorch/examples/blob/master/imagenet/main.py + """ + + def __init__(self, name='', fmt='f', postfix="", need_avg=True): + self.name = name + self.fmt = fmt + self.postfix = postfix + self.need_avg = need_avg + self.reset() + + def reset(self): + """ reset """ + self.val = 0 + self.avg = 0 + self.sum = 0 + self.count = 0 + + def update(self, val, n=1): + """ update """ + self.val = val + self.sum += val * n + self.count += n + self.avg = self.sum / self.count + + @property + def total(self): + return '{self.name}_sum: {self.sum:{self.fmt}}{self.postfix}'.format( + self=self) + + @property + def total_minute(self): + return '{self.name} {s:{self.fmt}}{self.postfix} min'.format( + s=self.sum / 60, self=self) + + @property + def mean(self): + return '{self.name}: {self.avg:{self.fmt}}{self.postfix}'.format( + self=self) if self.need_avg else '' + + @property + def value(self): + return '{self.name}: {self.val:{self.fmt}}{self.postfix}'.format( + self=self) diff --git a/exprements/PaddleClas/ppcls/utils/model_zoo.py b/exprements/PaddleClas/ppcls/utils/model_zoo.py new file mode 100644 index 00000000..fc527f6a --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/model_zoo.py @@ -0,0 +1,213 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import os +import requests +import shutil +import tarfile +import tqdm +import zipfile + +from ppcls.arch import similar_architectures +from ppcls.utils import logger + +__all__ = ['get'] + +DOWNLOAD_RETRY_LIMIT = 3 + + +class UrlError(Exception): + """ UrlError + """ + + def __init__(self, url='', code=''): + message = "Downloading from {} failed with code {}!".format(url, code) + super(UrlError, self).__init__(message) + + +class ModelNameError(Exception): + """ ModelNameError + """ + + def __init__(self, message=''): + super(ModelNameError, self).__init__(message) + + +class RetryError(Exception): + """ RetryError + """ + + def __init__(self, url='', times=''): + message = "Download from {} failed. Retry({}) limit reached".format( + url, times) + super(RetryError, self).__init__(message) + + +def _get_url(architecture, postfix="pdparams"): + prefix = "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/" + fname = architecture + "_pretrained." + postfix + return prefix + fname + + +def _move_and_merge_tree(src, dst): + """ + Move src directory to dst, if dst is already exists, + merge src to dst + """ + if not os.path.exists(dst): + shutil.move(src, dst) + elif os.path.isfile(src): + shutil.move(src, dst) + else: + for fp in os.listdir(src): + src_fp = os.path.join(src, fp) + dst_fp = os.path.join(dst, fp) + if os.path.isdir(src_fp): + if os.path.isdir(dst_fp): + _move_and_merge_tree(src_fp, dst_fp) + else: + shutil.move(src_fp, dst_fp) + elif os.path.isfile(src_fp) and \ + not os.path.isfile(dst_fp): + shutil.move(src_fp, dst_fp) + + +def _download(url, path): + """ + Download from url, save to path. + url (str): download url + path (str): download to given path + """ + if not os.path.exists(path): + os.makedirs(path) + + fname = os.path.split(url)[-1] + fullname = os.path.join(path, fname) + retry_cnt = 0 + + while not os.path.exists(fullname): + if retry_cnt < DOWNLOAD_RETRY_LIMIT: + retry_cnt += 1 + else: + raise RetryError(url, DOWNLOAD_RETRY_LIMIT) + + logger.info("Downloading {} from {}".format(fname, url)) + + req = requests.get(url, stream=True) + if req.status_code != 200: + raise UrlError(url, req.status_code) + + # For protecting download interupted, download to + # tmp_fullname firstly, move tmp_fullname to fullname + # after download finished + tmp_fullname = fullname + "_tmp" + total_size = req.headers.get('content-length') + with open(tmp_fullname, 'wb') as f: + if total_size: + for chunk in tqdm.tqdm( + req.iter_content(chunk_size=1024), + total=(int(total_size) + 1023) // 1024, + unit='KB'): + f.write(chunk) + else: + for chunk in req.iter_content(chunk_size=1024): + if chunk: + f.write(chunk) + shutil.move(tmp_fullname, fullname) + + return fullname + + +def _decompress(fname): + """ + Decompress for zip and tar file + """ + logger.info("Decompressing {}...".format(fname)) + + # For protecting decompressing interupted, + # decompress to fpath_tmp directory firstly, if decompress + # successed, move decompress files to fpath and delete + # fpath_tmp and remove download compress file. + fpath = os.path.split(fname)[0] + fpath_tmp = os.path.join(fpath, 'tmp') + if os.path.isdir(fpath_tmp): + shutil.rmtree(fpath_tmp) + os.makedirs(fpath_tmp) + + if fname.find('tar') >= 0: + with tarfile.open(fname) as tf: + tf.extractall(path=fpath_tmp) + elif fname.find('zip') >= 0: + with zipfile.ZipFile(fname) as zf: + zf.extractall(path=fpath_tmp) + else: + raise TypeError("Unsupport compress file type {}".format(fname)) + + fs = os.listdir(fpath_tmp) + assert len( + fs + ) == 1, "There should just be 1 pretrained path in an archive file but got {}.".format( + len(fs)) + + f = fs[0] + src_dir = os.path.join(fpath_tmp, f) + dst_dir = os.path.join(fpath, f) + _move_and_merge_tree(src_dir, dst_dir) + + shutil.rmtree(fpath_tmp) + os.remove(fname) + + return f + + +def _get_pretrained(): + with open('./ppcls/utils/pretrained.list') as flist: + pretrained = [line.strip() for line in flist] + return pretrained + + +def _check_pretrained_name(architecture): + assert isinstance(architecture, str), \ + ("the type of architecture({}) should be str". format(architecture)) + pretrained = _get_pretrained() + similar_names = similar_architectures(architecture, pretrained) + model_list = ', '.join(similar_names) + err = "{} is not exist! Maybe you want: [{}]" \ + "".format(architecture, model_list) + if architecture not in similar_names: + raise ModelNameError(err) + + +def list_models(): + pretrained = _get_pretrained() + msg = "All avialable pretrained models are as follows: {}".format( + pretrained) + logger.info(msg) + return + + +def get(architecture, path, decompress=False, postfix="pdparams"): + """ + Get the pretrained model. + """ + _check_pretrained_name(architecture) + url = _get_url(architecture, postfix=postfix) + fname = _download(url, path) + if postfix == "tar" and decompress: + _decompress(fname) + logger.info("download {} finished ".format(fname)) diff --git a/exprements/PaddleClas/ppcls/utils/pretrained.list b/exprements/PaddleClas/ppcls/utils/pretrained.list new file mode 100644 index 00000000..36d70f5a --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/pretrained.list @@ -0,0 +1,121 @@ +ResNet18 +ResNet34 +ResNet50 +ResNet101 +ResNet152 +ResNet50_vc +ResNet18_vd +ResNet34_vd +ResNet50_vd +ResNet50_vd_v2 +ResNet101_vd +ResNet152_vd +ResNet200_vd +ResNet50_vd_ssld +ResNet50_vd_ssld_v2 +Fix_ResNet50_vd_ssld_v2 +ResNet101_vd_ssld +MobileNetV3_large_x0_35 +MobileNetV3_large_x0_5 +MobileNetV3_large_x0_75 +MobileNetV3_large_x1_0 +MobileNetV3_large_x1_25 +MobileNetV3_small_x0_35 +MobileNetV3_small_x0_5 +MobileNetV3_small_x0_75 +MobileNetV3_small_x1_0 +MobileNetV3_small_x1_25 +MobileNetV3_large_x1_0_ssld +MobileNetV3_large_x1_0_ssld_int8 +MobileNetV3_small_x1_0_ssld +MobileNetV2_x0_25 +MobileNetV2_x0_5 +MobileNetV2_x0_75 +MobileNetV2 +MobileNetV2_x1_5 +MobileNetV2_x2_0 +MobileNetV2_ssld +MobileNetV1_x0_25 +MobileNetV1_x0_5 +MobileNetV1_x0_75 +MobileNetV1 +MobileNetV1_ssld +ShuffleNetV2_x0_25 +ShuffleNetV2_x0_33 +ShuffleNetV2_x0_5 +ShuffleNetV2 +ShuffleNetV2_x1_5 +ShuffleNetV2_x2_0 +ShuffleNetV2_swish +ResNeXt50_32x4d +ResNeXt50_64x4d +ResNeXt101_32x4d +ResNeXt101_64x4d +ResNeXt152_32x4d +ResNeXt152_64x4d +ResNeXt50_vd_32x4d +ResNeXt50_vd_64x4d +ResNeXt101_vd_32x4d +ResNeXt101_vd_64x4d +ResNeXt152_vd_32x4d +ResNeXt152_vd_64x4d +SE_ResNet18_vd +SE_ResNet34_vd +SE_ResNet50_vd +SE_ResNeXt50_32x4d +SE_ResNeXt101_32x4d +SE_ResNeXt50_vd_32x4d +SENet154_vd +Res2Net50_26w_4s +Res2Net50_vd_26w_4s +Res2Net50_14w_8s +Res2Net101_vd_26w_4s +Res2Net200_vd_26w_4s +GoogLeNet +InceptionV4 +Xception41 +Xception41_deeplab +Xception65 +Xception65_deeplab +Xception71 +HRNet_W18_C +HRNet_W30_C +HRNet_W32_C +HRNet_W40_C +HRNet_W44_C +HRNet_W48_C +HRNet_W64_C +DPN68 +DPN92 +DPN98 +DPN107 +DPN131 +DenseNet121 +DenseNet161 +DenseNet169 +DenseNet201 +DenseNet264 +EfficientNetB0_small +EfficientNetB0 +EfficientNetB1 +EfficientNetB2 +EfficientNetB3 +EfficientNetB4 +EfficientNetB5 +EfficientNetB6 +EfficientNetB7 +ResNeXt101_32x8d_wsl +ResNeXt101_32x16d_wsl +ResNeXt101_32x32d_wsl +ResNeXt101_32x48d_wsl +Fix_ResNeXt101_32x48d_wsl +AlexNet +SqueezeNet1_0 +SqueezeNet1_1 +VGG11 +VGG13 +VGG16 +VGG19 +DarkNet53_ImageNet1k +ResNet50_ACNet_deploy +CSPResNet50_leaky diff --git a/exprements/PaddleClas/ppcls/utils/profiler.py b/exprements/PaddleClas/ppcls/utils/profiler.py new file mode 100644 index 00000000..7cf945a2 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/profiler.py @@ -0,0 +1,111 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import sys +import paddle + +# A global variable to record the number of calling times for profiler +# functions. It is used to specify the tracing range of training steps. +_profiler_step_id = 0 + +# A global variable to avoid parsing from string every time. +_profiler_options = None + + +class ProfilerOptions(object): + ''' + Use a string to initialize a ProfilerOptions. + The string should be in the format: "key1=value1;key2=value;key3=value3". + For example: + "profile_path=model.profile" + "batch_range=[50, 60]; profile_path=model.profile" + "batch_range=[50, 60]; tracer_option=OpDetail; profile_path=model.profile" + + ProfilerOptions supports following key-value pair: + batch_range - a integer list, e.g. [100, 110]. + state - a string, the optional values are 'CPU', 'GPU' or 'All'. + sorted_key - a string, the optional values are 'calls', 'total', + 'max', 'min' or 'ave. + tracer_option - a string, the optional values are 'Default', 'OpDetail', + 'AllOpDetail'. + profile_path - a string, the path to save the serialized profile data, + which can be used to generate a timeline. + exit_on_finished - a boolean. + ''' + + def __init__(self, options_str): + assert isinstance(options_str, str) + + self._options = { + 'batch_range': [10, 20], + 'state': 'All', + 'sorted_key': 'total', + 'tracer_option': 'Default', + 'profile_path': '/tmp/profile', + 'exit_on_finished': True + } + self._parse_from_string(options_str) + + def _parse_from_string(self, options_str): + for kv in options_str.replace(' ', '').split(';'): + key, value = kv.split('=') + if key == 'batch_range': + value_list = value.replace('[', '').replace(']', '').split(',') + value_list = list(map(int, value_list)) + if len(value_list) >= 2 and value_list[0] >= 0 and value_list[ + 1] > value_list[0]: + self._options[key] = value_list + elif key == 'exit_on_finished': + self._options[key] = value.lower() in ("yes", "true", "t", "1") + elif key in [ + 'state', 'sorted_key', 'tracer_option', 'profile_path' + ]: + self._options[key] = value + + def __getitem__(self, name): + if self._options.get(name, None) is None: + raise ValueError( + "ProfilerOptions does not have an option named %s." % name) + return self._options[name] + + +def add_profiler_step(options_str=None): + ''' + Enable the operator-level timing using PaddlePaddle's profiler. + The profiler uses a independent variable to count the profiler steps. + One call of this function is treated as a profiler step. + + Args: + profiler_options - a string to initialize the ProfilerOptions. + Default is None, and the profiler is disabled. + ''' + if options_str is None: + return + + global _profiler_step_id + global _profiler_options + + if _profiler_options is None: + _profiler_options = ProfilerOptions(options_str) + + if _profiler_step_id == _profiler_options['batch_range'][0]: + paddle.utils.profiler.start_profiler( + _profiler_options['state'], _profiler_options['tracer_option']) + elif _profiler_step_id == _profiler_options['batch_range'][1]: + paddle.utils.profiler.stop_profiler(_profiler_options['sorted_key'], + _profiler_options['profile_path']) + if _profiler_options['exit_on_finished']: + sys.exit(0) + + _profiler_step_id += 1 diff --git a/exprements/PaddleClas/ppcls/utils/save_load.py b/exprements/PaddleClas/ppcls/utils/save_load.py new file mode 100644 index 00000000..625a2848 --- /dev/null +++ b/exprements/PaddleClas/ppcls/utils/save_load.py @@ -0,0 +1,136 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import errno +import os +import re +import shutil +import tempfile + +import paddle +from ppcls.utils import logger +from .download import get_weights_path_from_url + +__all__ = ['init_model', 'save_model', 'load_dygraph_pretrain'] + + +def _mkdir_if_not_exist(path): + """ + mkdir if not exists, ignore the exception when multiprocess mkdir together + """ + if not os.path.exists(path): + try: + os.makedirs(path) + except OSError as e: + if e.errno == errno.EEXIST and os.path.isdir(path): + logger.warning( + 'be happy if some process has already created {}'.format( + path)) + else: + raise OSError('Failed to mkdir {}'.format(path)) + + +def load_dygraph_pretrain(model, path=None): + if not (os.path.isdir(path) or os.path.exists(path + '.pdparams')): + raise ValueError("Model pretrain path {} does not " + "exists.".format(path)) + param_state_dict = paddle.load(path + ".pdparams") + model.set_dict(param_state_dict) + return + + +def load_dygraph_pretrain_from_url(model, pretrained_url, use_ssld=False): + if use_ssld: + pretrained_url = pretrained_url.replace("_pretrained", + "_ssld_pretrained") + local_weight_path = get_weights_path_from_url(pretrained_url).replace( + ".pdparams", "") + load_dygraph_pretrain(model, path=local_weight_path) + return + + +def load_distillation_model(model, pretrained_model): + logger.info("In distillation mode, teacher model will be " + "loaded firstly before student model.") + + if not isinstance(pretrained_model, list): + pretrained_model = [pretrained_model] + + teacher = model.teacher if hasattr(model, + "teacher") else model._layers.teacher + student = model.student if hasattr(model, + "student") else model._layers.student + load_dygraph_pretrain(teacher, path=pretrained_model[0]) + logger.info("Finish initing teacher model from {}".format( + pretrained_model)) + # load student model + if len(pretrained_model) >= 2: + load_dygraph_pretrain(student, path=pretrained_model[1]) + logger.info("Finish initing student model from {}".format( + pretrained_model)) + + +def init_model(config, net, optimizer=None): + """ + load model from checkpoint or pretrained_model + """ + checkpoints = config.get('checkpoints') + if checkpoints and optimizer is not None: + assert os.path.exists(checkpoints + ".pdparams"), \ + "Given dir {}.pdparams not exist.".format(checkpoints) + assert os.path.exists(checkpoints + ".pdopt"), \ + "Given dir {}.pdopt not exist.".format(checkpoints) + para_dict = paddle.load(checkpoints + ".pdparams") + opti_dict = paddle.load(checkpoints + ".pdopt") + metric_dict = paddle.load(checkpoints + ".pdstates") + net.set_dict(para_dict) + optimizer.set_state_dict(opti_dict) + logger.info("Finish load checkpoints from {}".format(checkpoints)) + return metric_dict + + pretrained_model = config.get('pretrained_model') + use_distillation = config.get('use_distillation', False) + if pretrained_model: + if use_distillation: + load_distillation_model(net, pretrained_model) + else: # common load + load_dygraph_pretrain(net, path=pretrained_model) + logger.info( + logger.coloring("Finish load pretrained model from {}".format( + pretrained_model), "HEADER")) + + +def save_model(net, + optimizer, + metric_info, + model_path, + model_name="", + prefix='ppcls'): + """ + save model to the target path + """ + if paddle.distributed.get_rank() != 0: + return + model_path = os.path.join(model_path, model_name) + _mkdir_if_not_exist(model_path) + model_path = os.path.join(model_path, prefix) + + paddle.save(net.state_dict(), model_path + ".pdparams") + paddle.save(optimizer.state_dict(), model_path + ".pdopt") + paddle.save(metric_info, model_path + ".pdstates") + logger.info("Already save model in {}".format(model_path)) diff --git a/exprements/PaddleClas/requirements.txt b/exprements/PaddleClas/requirements.txt new file mode 100644 index 00000000..79f548c2 --- /dev/null +++ b/exprements/PaddleClas/requirements.txt @@ -0,0 +1,11 @@ +prettytable +ujson +opencv-python==4.4.0.46 +pillow +tqdm +PyYAML +visualdl >= 2.2.0 +scipy +scikit-learn==0.23.2 +gast==0.3.3 +faiss-cpu==1.7.1.post2 diff --git a/exprements/PaddleClas/setup.py b/exprements/PaddleClas/setup.py new file mode 100644 index 00000000..a17e77d4 --- /dev/null +++ b/exprements/PaddleClas/setup.py @@ -0,0 +1,58 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from io import open +from setuptools import setup + +with open('requirements.txt', encoding="utf-8-sig") as f: + requirements = f.readlines() + + +def readme(): + with open('docs/en/whl_en.md', encoding="utf-8-sig") as f: + README = f.read() + return README + + +setup( + name='paddleclas', + packages=['paddleclas'], + package_dir={'paddleclas': ''}, + include_package_data=True, + entry_points={ + "console_scripts": ["paddleclas= paddleclas.paddleclas:main"] + }, + version='0.0.0', + install_requires=requirements, + license='Apache License 2.0', + description='Awesome Image Classification toolkits based on PaddlePaddle ', + long_description=readme(), + long_description_content_type='text/markdown', + url='https://github.com/PaddlePaddle/PaddleClas', + download_url='https://github.com/PaddlePaddle/PaddleClas.git', + keywords=[ + 'A treasure chest for image classification powered by PaddlePaddle.' + ], + classifiers=[ + 'Intended Audience :: Developers', + 'Operating System :: OS Independent', + 'Natural Language :: Chinese (Simplified)', + 'Programming Language :: Python :: 3', + 'Programming Language :: Python :: 3.2', + 'Programming Language :: Python :: 3.3', + 'Programming Language :: Python :: 3.4', + 'Programming Language :: Python :: 3.5', + 'Programming Language :: Python :: 3.6', + 'Programming Language :: Python :: 3.7', 'Topic :: Utilities' + ], ) diff --git a/exprements/PaddleClas/test_tipc/README.md b/exprements/PaddleClas/test_tipc/README.md new file mode 100644 index 00000000..4869f6e1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/README.md @@ -0,0 +1,108 @@ + +# 飞桨训推一体认证 + +## 1. 简介 + +飞桨除了基本的模型训练和预测，还提供了支持多端多平台的高性能推理部署工具。本文档提供了PaddleClas中所有模型的飞桨训推一体认证 (Training and Inference Pipeline Certification(TIPC)) 信息和测试工具，方便用户查阅每种模型的训练推理部署打通情况，并可以进行一键测试。 + +
+ +
+ +## 2. 汇总信息 + +打通情况汇总如下，已填写的部分表示可以使用本工具进行一键测试，未填写的表示正在支持中。 + +**字段说明：** +- 基础训练预测：包括模型训练、Paddle Inference Python预测。 +- 更多训练方式：包括多机多卡、混合精度。 +- 模型压缩：包括裁剪、离线/在线量化、蒸馏。 +- 其他预测部署：包括Paddle Inference C++预测、Paddle Serving部署、Paddle-Lite部署等。 + +更详细的mkldnn、Tensorrt等预测加速相关功能的支持情况可以查看各测试工具的[更多教程](#more)。 +| 算法论文 | 模型名称 | 模型类型 | 基础
训练预测 | 更多
训练方式 | 模型压缩 | 其他预测部署 | +| :--- | :--- | :----: | :--------: | :---- | :---- | :---- | +| ResNet |ResNet50_vd | 分类 | 支持 | 多机多卡
混合精度 | FPGM裁剪
PACT量化| | +| MobileNetV3 |MobileNetV3_large_x1_0 | 分类 | 支持 | 多机多卡
混合精度 | FPGM裁剪
PACT量化| | +| PPLCNet |PPLCNet_x2_5 | 分类 | 支持 | 多机多卡
混合精度 | - | | + +## 3. 测试工具简介 +### 目录介绍 +``` +./test_tipc/ +├── common_func.sh #test_*.sh会调用到的公共函数 +├── config # 配置文件目录 +│ ├── MobileNetV3 # MobileNetV3系列模型测试配置文件目录 +│ │ ├── MobileNetV3_large_x1_0_train_infer_python.txt #基础训练预测配置文件 +│ │ ├── MobileNetV3_large_x1_0_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt #多机多卡训练预测配置文件 +│ │ └── MobileNetV3_large_x1_0_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt #混合精度训练预测配置文件 +│ └── ResNet # ResNet系列模型测试配置文件目录 +│ ├── ResNet50_vd_train_infer_python.txt #基础训练预测配置文件 +│ ├── ResNet50_vd_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt #多机多卡训练预测配置文件 +│ └── ResNet50_vd_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt #混合精度训练预测配置文件 +| ...... +├── docs +│ ├── guide.png +│ └── test.png +├── prepare.sh # 完成test_*.sh运行所需要的数据和模型下载 +├── README.md # 使用文档 +├── results # 预先保存的预测结果，用于和实际预测结果进行精读比对 +└── test_train_inference_python.sh # 测试python训练预测的主程序 +``` + +### 测试流程概述 +使用本工具，可以测试不同功能的支持情况，以及预测结果是否对齐，测试流程如下： +
+ +
+ +1. 运行`prepare.sh`准备测试所需数据和模型； +2. 运行要测试的功能对应的测试脚本`test_*.sh`，产出log，由log可以看到不同配置是否运行成功； +3. 用`compare_results.py`对比log中的预测结果和预存在results目录下的结果，判断预测精度是否符合预期（在误差范围内）。 + +测试单项功能仅需两行命令，**如需测试不同模型/功能，替换配置文件即可**，命令格式如下： + +```shell +# 功能：准备数据 +# 格式：bash + 运行脚本 + 参数1: 配置文件选择 + 参数2: 模式选择 +bash test_tipc/prepare.sh configs/[model_name]/[params_file_name] [Mode] + +# 功能：运行测试 +# 格式：bash + 运行脚本 + 参数1: 配置文件选择 + 参数2: 模式选择 +bash test_tipc/test_train_inference_python.sh configs/[model_name]/[params_file_name] [Mode] + +``` + +例如，测试基本训练预测功能的`lite_train_lite_infer`模式，运行： + +```shell +# 准备数据 +bash test_tipc/prepare.sh ./test_tipc/configs/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt 'lite_train_lite_infer' +# 运行测试 +bash test_tipc/test_train_inference_python.sh ./test_tipc/configs/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt 'lite_train_lite_infer' + +``` + +关于本示例命令的更多信息可查看[基础训练预测使用文档](docs/test_train_inference_python.md)。 + +### 配置文件命名规范 + +在`configs`目录下，**按模型系列划分为子目录**，子目录中存放所有该模型系列测试需要用到的配置文件，如`MobileNetV3`文件夹下存放了所有`MobileNetV3`系列模型的配置文件。配置文件的命名遵循如下规范： + +1. 基础训练预测配置简单命名为：`ModelName_train_infer_python.txt`，表示**Linux环境下单机、不使用混合精度训练+python预测**，其完整命名对应`ModelName_train_linux_gpu_normal_normal_infer_python_linux_gpu_cpu.txt`，由于本配置文件使用频率较高，这里进行了名称简化。其中`ModelName`指具体模型名称 +2. 其他带训练配置命名格式为：`ModelName_train_训练硬件环境(linux_gpu/linux_dcu/…)_是否多机(fleet/normal)_是否混合精度(amp/normal)_预测模式(infer/lite/serving/js)_语言(cpp/python/java)_预测硬件环境(ModelName_linux_gpu/mac/jetson/opencl_arm_gpu/...).txt`。如，linux gpu下多机多卡+混合精度链条测试对应配置 `ModelName_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt`，linux dcu下基础训练预测对应配置 `ModelName_train_linux_dcu_normal_normal_infer_python_linux_dcu.txt`。 +3. 仅预测的配置（如serving、lite等）命名格式：`ModelName_model_训练硬件环境(ModelName_linux_gpu/linux_dcu/…)_是否多机(fleet/normal)_是否混合精度(amp/normal)_(infer/lite/serving/js)_语言(cpp/python/java)_预测硬件环境(ModelName_linux_gpu/mac/jetson/opencl_arm_gpu/...).txt`，即，与2相比，仅第二个字段从train换为model，测试时模型直接下载获取，这里的“训练硬件环境”表示所测试的模型是在哪种环境下训练得到的。 + +**根据上述命名规范，可以直接从子目录名称和配置文件名找到需要测试的场景和功能对应的配置文件。** + + + +## 4 开始测试 + +各功能测试中涉及混合精度、裁剪、量化等训练相关，及mkldnn、Tensorrt等多种预测相关参数配置，请点击下方相应链接了解更多细节和使用教程： + +- [test_train_inference_python 使用](docs/test_train_inference_python.md)：测试基于Python的模型训练、评估、推理等基本功能，包括裁剪、量化、蒸馏。 +- [test_inference_cpp 使用](docs/test_inference_cpp.md) ：测试基于C++的模型推理。 +- [test_serving 使用](docs/test_serving.md) ：测试基于Paddle Serving的服务化部署功能。 +- [test_lite_arm_cpu_cpp 使用](docs/test_lite_arm_cpu_cpp.md): 测试基于Paddle-Lite的ARM CPU端c++预测部署功能. +- [test_paddle2onnx 使用](docs/test_paddle2onnx.md)：测试Paddle2ONNX的模型转化功能，并验证正确性。 diff --git a/exprements/PaddleClas/test_tipc/common_func.sh b/exprements/PaddleClas/test_tipc/common_func.sh new file mode 100644 index 00000000..3f0fa66b --- /dev/null +++ b/exprements/PaddleClas/test_tipc/common_func.sh @@ -0,0 +1,65 @@ +#!/bin/bash + +function func_parser_key(){ + strs=$1 + IFS=":" + array=(${strs}) + tmp=${array[0]} + echo ${tmp} +} + +function func_parser_value(){ + strs=$1 + IFS=":" + array=(${strs}) + tmp=${array[1]} + echo ${tmp} +} + +function func_set_params(){ + key=$1 + value=$2 + if [ ${key}x = "null"x ];then + echo " " + elif [[ ${value} = "null" ]] || [[ ${value} = " " ]] || [ ${#value} -le 0 ];then + echo " " + else + echo "${key}=${value}" + fi +} + +function func_parser_params(){ + strs=$1 + IFS=":" + array=(${strs}) + key=${array[0]} + tmp=${array[1]} + IFS="|" + res="" + for _params in ${tmp[*]}; do + IFS="=" + array=(${_params}) + mode=${array[0]} + value=${array[1]} + if [[ ${mode} = ${MODE} ]]; then + IFS="|" + #echo $(func_set_params "${mode}" "${value}") + echo $value + break + fi + IFS="|" + done + echo ${res} +} + +function status_check(){ + last_status=$1 # the exit code + run_command=$2 + run_log=$3 + if [ $last_status -eq 0 ]; then + echo -e "\033[33m Run successfully with command - ${run_command}! \033[0m" | tee -a ${run_log} + else + echo -e "\033[33m Run failed with command - ${run_command}! \033[0m" | tee -a ${run_log} + fi +} + diff --git a/exprements/PaddleClas/test_tipc/compare_results.py b/exprements/PaddleClas/test_tipc/compare_results.py new file mode 100644 index 00000000..e28410ed --- /dev/null +++ b/exprements/PaddleClas/test_tipc/compare_results.py @@ -0,0 +1,140 @@ +import numpy as np +import os +import subprocess +import json +import argparse +import glob + + +def init_args(): + parser = argparse.ArgumentParser() + # params for testing assert allclose + parser.add_argument("--atol", type=float, default=1e-3) + parser.add_argument("--rtol", type=float, default=1e-3) + parser.add_argument("--gt_file", type=str, default="") + parser.add_argument("--log_file", type=str, default="") + parser.add_argument("--precision", type=str, default="fp32") + return parser + + +def parse_args(): + parser = init_args() + return parser.parse_args() + + +def run_shell_command(cmd): + p = subprocess.Popen( + cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True) + out, err = p.communicate() + + if p.returncode == 0: + return out.decode('utf-8') + else: + return None + + +def parser_results_from_log_by_name(log_path, names_list): + if not os.path.exists(log_path): + raise ValueError("The log file {} does not exists!".format(log_path)) + + if names_list is None or len(names_list) < 1: + return [] + + parser_results = {} + for name in names_list: + cmd = "grep {} {}".format(name, log_path) + outs = run_shell_command(cmd) + outs = outs.split("\n")[0] + result = outs.split("{}".format(name))[-1] + try: + result = json.loads(result) + except: + result = np.array([int(r) for r in result.split()]).reshape(-1, 4) + parser_results[name] = result + return parser_results + + +def load_gt_from_file(gt_file): + if not os.path.exists(gt_file): + raise ValueError("The log file {} does not exists!".format(gt_file)) + with open(gt_file, 'r') as f: + data = f.readlines() + f.close() + parser_gt = {} + for line in data: + image_name, result = line.strip("\n").split("\t") + image_name = image_name.split('/')[-1] + try: + result = json.loads(result) + except: + result = np.array([int(r) for r in result.split()]).reshape(-1, 4) + parser_gt[image_name] = result + return parser_gt + + +def load_gt_from_txts(gt_file): + gt_list = glob.glob(gt_file) + gt_collection = {} + for gt_f in gt_list: + gt_dict = load_gt_from_file(gt_f) + basename = os.path.basename(gt_f) + if "fp32" in basename: + gt_collection["fp32"] = [gt_dict, gt_f] + elif "fp16" in basename: + gt_collection["fp16"] = [gt_dict, gt_f] + elif "int8" in basename: + gt_collection["int8"] = [gt_dict, gt_f] + else: + continue + return gt_collection + + +def collect_predict_from_logs(log_path, key_list): + log_list = glob.glob(log_path) + pred_collection = {} + for log_f in log_list: + pred_dict = parser_results_from_log_by_name(log_f, key_list) + key = os.path.basename(log_f) + pred_collection[key] = pred_dict + + return pred_collection + + +def testing_assert_allclose(dict_x, dict_y, atol=1e-7, rtol=1e-7): + for k in dict_x: + np.testing.assert_allclose( + np.array(dict_x[k]), np.array(dict_y[k]), atol=atol, rtol=rtol) + + +if __name__ == "__main__": + # Usage: + # python3.7 tests/compare_results.py --gt_file=./tests/results/*.txt --log_file=./tests/output/infer_*.log + + args = parse_args() + + gt_collection = load_gt_from_txts(args.gt_file) + key_list = gt_collection["fp32"][0].keys() + + pred_collection = collect_predict_from_logs(args.log_file, key_list) + for filename in pred_collection.keys(): + if "fp32" in filename: + gt_dict, gt_filename = gt_collection["fp32"] + elif "fp16" in filename: + gt_dict, gt_filename = gt_collection["fp16"] + elif "int8" in filename: + gt_dict, gt_filename = gt_collection["int8"] + else: + continue + pred_dict = pred_collection[filename] + + try: + testing_assert_allclose( + gt_dict, pred_dict, atol=args.atol, rtol=args.rtol) + print( + "Assert allclose passed! The results of {} and {} are consistent!". + format(filename, gt_filename)) + except Exception as E: + print(E) + raise ValueError( + "The results of {} and the results of {} are inconsistent!". + format(filename, gt_filename)) diff --git a/exprements/PaddleClas/test_tipc/config/AlexNet/AlexNet_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/AlexNet/AlexNet_train_infer_python.txt new file mode 100644 index 00000000..7662f5d9 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/AlexNet/AlexNet_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:AlexNet +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/AlexNet/AlexNet.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/AlexNet/AlexNet.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/AlexNet/AlexNet.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/AlexNet_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/CSPNet/CSPDarkNet53_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/CSPNet/CSPDarkNet53_train_infer_python.txt new file mode 100644 index 00000000..cd5ec06d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/CSPNet/CSPDarkNet53_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:CSPDarkNet53 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/CSPNet/CSPDarkNet53.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSPNet/CSPDarkNet53.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSPNet/CSPDarkNet53.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSPDarkNet53_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=288 -o PreProcess.transform_ops.1.CropImage.size=256 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA102_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA102_train_infer_python.txt new file mode 100644 index 00000000..0a85e2a3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA102_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA102 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA102.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA102.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA102.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA102x2_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA102x2_train_infer_python.txt new file mode 100644 index 00000000..fff0ee1c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA102x2_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA102x2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA102x2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA102x2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA102x2.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x2_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA102x_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA102x_train_infer_python.txt new file mode 100644 index 00000000..db7af68c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA102x_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA102x +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA102x.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA102x.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA102x.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA169_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA169_train_infer_python.txt new file mode 100644 index 00000000..0abc0e21 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA169_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA169 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA169.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA169.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA169.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA169_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA34_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA34_train_infer_python.txt new file mode 100644 index 00000000..fa064056 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA34_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA34 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA34.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA34.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA34.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA34_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA46_c_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA46_c_train_infer_python.txt new file mode 100644 index 00000000..85034529 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA46_c_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA46_c +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA46_c.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA46_c.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA46_c.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA46_c_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA46x_c_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA46x_c_train_infer_python.txt new file mode 100644 index 00000000..ee6c58ad --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA46x_c_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA46x_c +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA46x_c.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA46x_c.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA46x_c.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA46x_c_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA60_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA60_train_infer_python.txt new file mode 100644 index 00000000..e14fcabe --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA60_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA60 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA60.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA60.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA60.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA60x_c_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA60x_c_train_infer_python.txt new file mode 100644 index 00000000..cd45206c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA60x_c_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA60x_c +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA60x_c.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA60x_c.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA60x_c.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_c_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DLA/DLA60x_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DLA/DLA60x_train_infer_python.txt new file mode 100644 index 00000000..51b7e77c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DLA/DLA60x_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA60x +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA60x.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA60x.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA60x.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DPN/DPN107_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DPN/DPN107_train_infer_python.txt new file mode 100644 index 00000000..646bce35 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DPN/DPN107_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN107 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN107.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN107.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN107.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN107_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DPN/DPN131_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DPN/DPN131_train_infer_python.txt new file mode 100644 index 00000000..1aaea6a4 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DPN/DPN131_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN131 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN131.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN131.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN131.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN131_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DPN/DPN68_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DPN/DPN68_train_infer_python.txt new file mode 100644 index 00000000..d26e5e3f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DPN/DPN68_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN68 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN68.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN68.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN68.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN68_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DPN/DPN92_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DPN/DPN92_train_infer_python.txt new file mode 100644 index 00000000..3d0d0778 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DPN/DPN92_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN92 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN92.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN92.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN92.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN92_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DPN/DPN98_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DPN/DPN98_train_infer_python.txt new file mode 100644 index 00000000..e4cee43c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DPN/DPN98_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN98 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN98.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN98.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN98.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN98_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DarkNet/DarkNet53_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DarkNet/DarkNet53_train_infer_python.txt new file mode 100644 index 00000000..b10918ea --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DarkNet/DarkNet53_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DarkNet53 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DarkNet53_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=292 -o PreProcess.transform_ops.1.CropImage.size=256 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_base_patch16_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_base_patch16_224_train_infer_python.txt new file mode 100644 index 00000000..d10fc201 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_base_patch16_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_base_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_base_patch16_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_base_patch16_384_train_infer_python.txt new file mode 100644 index 00000000..c8fc60e0 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_base_patch16_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_base_patch16_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_small_patch16_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_small_patch16_224_train_infer_python.txt new file mode 100644 index 00000000..83ea4c31 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_small_patch16_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_small_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_small_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_infer_python.txt new file mode 100644 index 00000000..cb185023 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_tiny_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet121_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet121_train_infer_python.txt new file mode 100644 index 00000000..9f89e4cb --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet121_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet121 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet121.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet121.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet121.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet121_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet161_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet161_train_infer_python.txt new file mode 100644 index 00000000..379cc29e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet161_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet161 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet161.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet161.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet161.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet161_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet169_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet169_train_infer_python.txt new file mode 100644 index 00000000..c2e2bbb5 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet169_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet169 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet169.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet169.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet169.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet169_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet201_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet201_train_infer_python.txt new file mode 100644 index 00000000..08dedc4d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet201_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet201 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet201.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet201.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet201.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet201_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet264_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet264_train_infer_python.txt new file mode 100644 index 00000000..146ad68c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/DenseNet/DenseNet264_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet264 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet264.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet264.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet264.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet264_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_25_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_25_train_infer_python.txt new file mode 100644 index 00000000..82ead104 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_25_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_5_train_infer_python.txt new file mode 100644 index 00000000..38b376f3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_75_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_75_train_infer_python.txt new file mode 100644 index 00000000..dc634980 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x0_75_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x1_0_train_infer_python.txt new file mode 100644 index 00000000..c5a4a5af --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ESNet/ESNet_x1_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB0_train_infer_python.txt new file mode 100644 index 00000000..23b3ef1c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB1_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB1_train_infer_python.txt new file mode 100644 index 00000000..1c431454 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB1_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB1.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB1_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=272 -o PreProcess.transform_ops.1.CropImage.size=240 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB2_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB2_train_infer_python.txt new file mode 100644 index 00000000..0ac16c3d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB2_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB2.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB2_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=292 -o PreProcess.transform_ops.1.CropImage.size=260 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB3_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB3_train_infer_python.txt new file mode 100644 index 00000000..0d0ca3db --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB3_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=332 -o PreProcess.transform_ops.1.CropImage.size=300 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB4_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB4_train_infer_python.txt new file mode 100644 index 00000000..92d4cb81 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB4_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB4 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB4.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB4.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB4.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB4_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=412 -o PreProcess.transform_ops.1.CropImage.size=380 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB5_train_infer_python.txt new file mode 100644 index 00000000..02c93574 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=488 -o PreProcess.transform_ops.1.CropImage.size=456 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB6_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB6_train_infer_python.txt new file mode 100644 index 00000000..392c0482 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB6_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB6 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB6.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB6.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB6.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB6_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=560 -o PreProcess.transform_ops.1.CropImage.size=528 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB7_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB7_train_infer_python.txt new file mode 100644 index 00000000..7265243e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/EfficientNet/EfficientNetB7_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB7 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB7.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB7.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB7.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB7_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=632 -o PreProcess.transform_ops.1.CropImage.size=600 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_custom_sampler.txt b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_custom_sampler.txt new file mode 100644 index 00000000..c1dbc896 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_custom_sampler.txt @@ -0,0 +1,27 @@ +===========================train_params=========================== +model_name:GeneralRecognition_PPLCNet_x2_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml -o DataLoader.Train.sampler.name="PKSampler" -o DataLoader.Train.sampler.sample_per_id=2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:null +null:null +## + diff --git a/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_infer_python.txt new file mode 100644 index 00000000..40e6675a --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GeneralRecognition_PPLCNet_x2_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/pretrain/general_PPLCNet_x2_5_pretrained_v1.0.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_rec.py -c configs/inference_rec.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.rec_inference_model_dir:../inference +-o Global.infer_imgs:../dataset/Aliproduct/demo_test/ +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_multicard_eval.txt b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_multicard_eval.txt new file mode 100644 index 00000000..165cfa9f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_multicard_eval.txt @@ -0,0 +1,26 @@ +===========================train_params=========================== +model_name:GeneralRecognition_PPLCNet_x2_5 +python:python3.7 +gpu_list:0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:null +null:null +## diff --git a/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_no_eval.txt b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_no_eval.txt new file mode 100644 index 00000000..1e167518 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_no_eval.txt @@ -0,0 +1,26 @@ +===========================train_params=========================== +model_name:GeneralRecognition_PPLCNet_x2_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o Global.eval_during_train=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:null +null:null +## diff --git a/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x0_5_train_infer_python.txt new file mode 100644 index 00000000..7df8dee4 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x0_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GhostNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x1_0_train_infer_python.txt new file mode 100644 index 00000000..7d3184b4 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x1_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GhostNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x1_3_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x1_3_train_infer_python.txt new file mode 100644 index 00000000..e186eea4 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/GhostNet/GhostNet_x1_3_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GhostNet_x1_3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W18_C_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W18_C_train_infer_python.txt new file mode 100644 index 00000000..3883580f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W18_C_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W18_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W18_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W30_C_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W30_C_train_infer_python.txt new file mode 100644 index 00000000..cba02656 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W30_C_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W30_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W30_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W30_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W30_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W30_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W32_C_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W32_C_train_infer_python.txt new file mode 100644 index 00000000..197a76ba --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W32_C_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W32_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W32_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W32_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W32_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W32_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W40_C_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W40_C_train_infer_python.txt new file mode 100644 index 00000000..1deb8917 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W40_C_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W40_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W40_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W40_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W40_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W40_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W44_C_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W44_C_train_infer_python.txt new file mode 100644 index 00000000..10fab3d3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W44_C_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W44_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W44_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W44_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W44_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W44_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W48_C_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W48_C_train_infer_python.txt new file mode 100644 index 00000000..6de8e26a --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W48_C_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W48_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W48_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W48_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W48_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W48_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W64_C_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W64_C_train_infer_python.txt new file mode 100644 index 00000000..c99c5e6c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HRNet/HRNet_W64_C_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W64_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W64_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W64_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W64_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W64_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet39_ds_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet39_ds_train_infer_python.txt new file mode 100644 index 00000000..5cda671b --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet39_ds_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet39_ds +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet39_ds.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet39_ds.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet39_ds.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet39_ds_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet68_ds_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet68_ds_train_infer_python.txt new file mode 100644 index 00000000..bfe601ee --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet68_ds_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet68_ds +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68_ds.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68_ds.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68_ds.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_ds_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet68_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet68_train_infer_python.txt new file mode 100644 index 00000000..b61ea632 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet68_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet68 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet85_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet85_train_infer_python.txt new file mode 100644 index 00000000..02237ba7 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/HarDNet/HarDNet85_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet85 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet85.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet85.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet85.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet85_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Inception/GoogLeNet_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Inception/GoogLeNet_train_infer_python.txt new file mode 100644 index 00000000..ebbb86dd --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Inception/GoogLeNet_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GoogLeNet +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Inception/GoogLeNet.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Inception/GoogLeNet.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Inception/GoogLeNet.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GoogLeNet_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Inception/InceptionV3_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Inception/InceptionV3_train_infer_python.txt new file mode 100644 index 00000000..6a7960c6 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Inception/InceptionV3_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:InceptionV3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Inception/InceptionV3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Inception/InceptionV3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Inception/InceptionV3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Inception/InceptionV4_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Inception/InceptionV4_train_infer_python.txt new file mode 100644 index 00000000..1e1d7fd5 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Inception/InceptionV4_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:InceptionV4 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Inception/InceptionV4.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Inception/InceptionV4.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Inception/InceptionV4.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV4_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_128S_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_128S_train_infer_python.txt new file mode 100644 index 00000000..772049b8 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_128S_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_128S +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_128_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_128_train_infer_python.txt new file mode 100644 index 00000000..8c8699ed --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_128_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_128 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_128.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_128.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_128.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_192_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_192_train_infer_python.txt new file mode 100644 index 00000000..c2ac115f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_192_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_192 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_192.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_192.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_192.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_192_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_256_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_256_train_infer_python.txt new file mode 100644 index 00000000..f58880c8 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_256_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_256 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_256.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_256.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_256.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_256_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_384_train_infer_python.txt new file mode 100644 index 00000000..0ae0d40f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/LeViT/LeViT_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_L_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_L_train_infer_python.txt new file mode 100644 index 00000000..46fcbc1d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_L_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MixNet_L +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MixNet/MixNet_L.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MixNet/MixNet_L.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MixNet/MixNet_L.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_L_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_M_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_M_train_infer_python.txt new file mode 100644 index 00000000..87be78a5 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_M_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MixNet_M +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MixNet/MixNet_M.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MixNet/MixNet_M.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MixNet/MixNet_M.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_M_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_S_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_S_train_infer_python.txt new file mode 100644 index 00000000..42e5df7f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MixNet/MixNet_S_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MixNet_S +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MixNet/MixNet_S.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MixNet/MixNet_S.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MixNet/MixNet_S.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_S_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_train_infer_python.txt new file mode 100644 index 00000000..65fee152 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_infer_python.txt new file mode 100644 index 00000000..a9f63dfa --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_infer_python.txt new file mode 100644 index 00000000..379cbd66 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_infer_python.txt new file mode 100644 index 00000000..fef81fa3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt new file mode 100644 index 00000000..fb0c8a24 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_infer_python.txt new file mode 100644 index 00000000..9183e01e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_infer_python.txt new file mode 100644 index 00000000..bef4e1d9 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_infer_python.txt new file mode 100644 index 00000000..a7887316 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_infer_python.txt new file mode 100644 index 00000000..d5f9bf41 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_infer_python.txt new file mode 100644 index 00000000..f4455704 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_infer_python.txt new file mode 100644 index 00000000..2797a2d2 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_infer_python.txt new file mode 100644 index 00000000..6f9165ca --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_infer_python.txt new file mode 100644 index 00000000..da486711 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt new file mode 100644 index 00000000..25ecbd5f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt new file mode 100644 index 00000000..c7c956f3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_0 +python:python3.7 +gpu_list:xx.xx.xx.xx,yy.yy.yy.yy;0,1 +-o Global.device:gpu +-o Global.auto_cast:null|amp +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.save_inference_dir=./inference +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/MobileNetV3_large_x1_0_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt new file mode 100644 index 00000000..f251e9c9 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:amp +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.save_inference_dir=./inference +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/MobileNetV3_large_x1_0_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_infer_python.txt new file mode 100644 index 00000000..4b62f541 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_infer_python.txt new file mode 100644 index 00000000..2e40f899 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_infer_python.txt new file mode 100644 index 00000000..e8a2b0e1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_infer_python.txt new file mode 100644 index 00000000..65a86713 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_infer_python.txt new file mode 100644 index 00000000..6d2334b4 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_infer_python.txt new file mode 100644 index 00000000..6a730b7e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x1_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PP-ShiTu/PPShiTu_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/PP-ShiTu/PPShiTu_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt new file mode 100644 index 00000000..b86aa9b1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PP-ShiTu/PPShiTu_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt @@ -0,0 +1,19 @@ +===========================cpp_infer_params=========================== +model_name:PPShiTu +cpp_infer_type:shitu +feature_inference_model_dir:./feature_inference/ +det_inference_model_dir:./det_inference +cls_inference_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/general_PPLCNet_x2_5_lite_v1.0_infer.tar +det_inference_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer.tar +infer_quant:False +inference_cmd:./deploy/cpp_shitu/build/pp_shitu -c inference_drink.yaml +use_gpu:True|False +enable_mkldnn:True|False +cpu_threads:1|6 +batch_size:1 +use_tensorrt:False|True +precision:fp32|fp16 +data_dir:./dataset/drink_dataset_v1.0 +benchmark:True +generate_yaml_cmd:python3 test_tipc/generate_cpp_yaml.py +transform_index_cmd:python3 deploy/cpp_shitu/tools/transform_id_map.py -c inference_drink.yaml diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_infer_python.txt new file mode 100644 index 00000000..66afc24d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_infer_python.txt new file mode 100644 index 00000000..9aa5362f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_infer_python.txt new file mode 100644 index 00000000..277ef813 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_infer_python.txt new file mode 100644 index 00000000..14b5ec34 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_infer_python.txt new file mode 100644 index 00000000..97bec1f6 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_infer_python.txt new file mode 100644 index 00000000..6faaddb1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_infer_python.txt new file mode 100644 index 00000000..36145d52 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_infer_python.txt new file mode 100644 index 00000000..f369e740 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x2_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt new file mode 100644 index 00000000..66375655 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x2_5 +python:python3.7 +gpu_list:xx.xx.xx.xx,yy.yy.yy.yy;0,1 +-o Global.device:gpu +-o Global.auto_cast:null|amp +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt new file mode 100644 index 00000000..74c6d046 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x2_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:amp +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B0.txt b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B0.txt new file mode 100644 index 00000000..1e0a3800 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B0.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:PVT_V2_B0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B0.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B1.txt b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B1.txt new file mode 100644 index 00000000..489f8c80 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B1.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:PVT_V2_B1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B1.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B2.txt b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B2.txt new file mode 100644 index 00000000..8d274cd2 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B2.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:PVT_V2_B2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B2_Linear.txt b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B2_Linear.txt new file mode 100644 index 00000000..2fd530c9 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B2_Linear.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:PVT_V2_B2_Linear +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2_Linear.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B3.txt b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B3.txt new file mode 100644 index 00000000..f375fcf2 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B3.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:PVT_V2_B3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B3.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B4.txt b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B4.txt new file mode 100644 index 00000000..f0c7084f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B4.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:PVT_V2_B4 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B4.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B5.txt b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B5.txt new file mode 100644 index 00000000..9383851c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/PVTV2/PVT_V2_B5.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:PVT_V2_B5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B5.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_0_train_infer_python.txt new file mode 100644 index 00000000..352ff5ec --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_3_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_3_train_infer_python.txt new file mode 100644 index 00000000..4c76badb --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_3_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_1_3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_5_train_infer_python.txt new file mode 100644 index 00000000..b302911e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_1_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_2_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_2_0_train_infer_python.txt new file mode 100644 index 00000000..c52d563c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_2_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_3_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_3_0_train_infer_python.txt new file mode 100644 index 00000000..a2047c44 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ReXNet/ReXNet_3_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_3_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_3_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_3_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_3_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_3_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/RedNet/RedNet101_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet101_train_infer_python.txt new file mode 100644 index 00000000..6d7d3b9d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet101_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet101 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet101.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet101.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet101.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet101_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/RedNet/RedNet152_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet152_train_infer_python.txt new file mode 100644 index 00000000..82e8189d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet152_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet152 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet152.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet152.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet152.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet152_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/RedNet/RedNet26_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet26_train_infer_python.txt new file mode 100644 index 00000000..53e5ce00 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet26_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet26 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet26.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet26.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet26.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet26_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/RedNet/RedNet38_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet38_train_infer_python.txt new file mode 100644 index 00000000..3ec24716 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet38_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet38 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet38.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet38.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet38.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet38_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/RedNet/RedNet50_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet50_train_infer_python.txt new file mode 100644 index 00000000..bb36200a --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/RedNet/RedNet50_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet50 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet50.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet50.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet50.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet50_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_infer_python.txt new file mode 100644 index 00000000..f5542adc --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net101_vd_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net101_vd_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net101_vd_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net101_vd_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net101_vd_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_infer_python.txt new file mode 100644 index 00000000..40d7f410 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net200_vd_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net200_vd_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net200_vd_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net200_vd_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net200_vd_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_infer_python.txt new file mode 100644 index 00000000..c6300bc9 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net50_14w_8s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_14w_8s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_14w_8s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_14w_8s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_14w_8s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_infer_python.txt new file mode 100644 index 00000000..74915258 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net50_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_infer_python.txt new file mode 100644 index 00000000..cbdc2563 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net50_vd_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_vd_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_vd_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_vd_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_infer_python.txt new file mode 100644 index 00000000..66ce66a5 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeSt50_fast_1s1x64d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50_fast_1s1x64d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50_fast_1s1x64d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50_fast_1s1x64d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_fast_1s1x64d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeSt/ResNeSt50_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeSt/ResNeSt50_train_infer_python.txt new file mode 100644 index 00000000..2d1bfb2a --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeSt/ResNeSt50_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeSt50 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_infer_python.txt new file mode 100644 index 00000000..eb21af69 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_infer_python.txt new file mode 100644 index 00000000..8e4ac811 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_infer_python.txt new file mode 100644 index 00000000..cf9211f6 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_infer_python.txt new file mode 100644 index 00000000..eae4d0e4 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_vd_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_infer_python.txt new file mode 100644 index 00000000..eb21af69 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_infer_python.txt new file mode 100644 index 00000000..c8a85f6f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt152_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_infer_python.txt new file mode 100644 index 00000000..4abaa874 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt152_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_infer_python.txt new file mode 100644 index 00000000..8d8a4ea1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt152_vd_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_infer_python.txt new file mode 100644 index 00000000..4cb942f1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_infer_python.txt new file mode 100644 index 00000000..0cad348c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_infer_python.txt new file mode 100644 index 00000000..fc745601 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_infer_python.txt new file mode 100644 index 00000000..b2817dba --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_vd_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet101_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet101_train_infer_python.txt new file mode 100644 index 00000000..f10760d7 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet101_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet101 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet101.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet101.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet101.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet101_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet101_vd_train_infer_python.txt new file mode 100644 index 00000000..7027d30c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet101_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet101_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet101_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet101_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet101_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet152_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet152_train_infer_python.txt new file mode 100644 index 00000000..1d4cc80d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet152_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet152 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet152.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet152.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet152.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet152_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet152_vd_train_infer_python.txt new file mode 100644 index 00000000..40913286 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet152_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet152_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet152_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet152_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet152_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet18_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet18_train_infer_python.txt new file mode 100644 index 00000000..06d55049 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet18_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet18 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet18.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet18.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet18.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet18_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet18_vd_train_infer_python.txt new file mode 100644 index 00000000..d3d2bcf1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet18_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet18_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet18_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet18_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet18_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet200_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet200_vd_train_infer_python.txt new file mode 100644 index 00000000..e742e87a --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet200_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet200_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet200_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet200_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet200_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet200_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet34_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet34_train_infer_python.txt new file mode 100644 index 00000000..4a520b55 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet34_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet34 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet34.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet34.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet34.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet34_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet34_vd_train_infer_python.txt new file mode 100644 index 00000000..b956c83d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet34_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet34_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet34_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet34_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet34_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_train_infer_python.txt new file mode 100644 index 00000000..6775f952 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt new file mode 100644 index 00000000..51c73f13 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt @@ -0,0 +1,18 @@ +===========================cpp_infer_params=========================== +model_name:ResNet50_vd +cpp_infer_type:cls +cls_inference_model_dir:./cls_inference/ +det_inference_model_dir: +cls_inference_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/ResNet50_vd_inference.tar +det_inference_url: +infer_quant:False +inference_cmd:./deploy/cpp/build/clas_system -c inference_cls.yaml +use_gpu:True|False +enable_mkldnn:True|False +cpu_threads:1|6 +batch_size:1 +use_tensorrt:False|True +precision:fp32|fp16 +image_dir:./dataset/ILSVRC2012/val +benchmark:True +generate_yaml_cmd:python3 test_tipc/generate_cpp_yaml.py diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_paddle2onnx_python_linux_cpu.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_paddle2onnx_python_linux_cpu.txt new file mode 100644 index 00000000..163bb484 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_paddle2onnx_python_linux_cpu.txt @@ -0,0 +1,14 @@ +===========================paddle2onnx_params=========================== +model_name:ResNet50_vd +python:python3.7 +2onnx: paddle2onnx +--model_dir:./deploy/models/ResNet50_vd_infer/ +--model_filename:inference.pdmodel +--params_filename:inference.pdiparams +--save_file:./deploy/models/ResNet50_vd_infer/inference.onnx +--opset_version:10 +--enable_onnx_checker:True +inference: python/predict_cls.py -c configs/inference_cls.yaml +Global.use_onnx:True +Global.inference_model_dir:models/ResNet50_vd_infer/ +Global.use_gpu:False diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt new file mode 100644 index 00000000..d16ecd3f --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt new file mode 100644 index 00000000..2cb69154 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50_vd +python:python3.7 +gpu_list:xx.xx.xx.xx,yy.yy.yy.yy;0,1 +-o Global.device:gpu +-o Global.auto_cast:null|amp +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./inference +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/ResNet50_vd_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt new file mode 100644 index 00000000..22c0f8db --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:amp +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./inference +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/ResNet50_vd_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_use_dali.txt b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_use_dali.txt new file mode 100644 index 00000000..9d4bf7f3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ResNet/ResNet50_vd_train_linux_gpu_use_dali.txt @@ -0,0 +1,22 @@ +===========================train_params=========================== +model_name:ResNet50_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null|amp +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o Global.use_dali=True +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## diff --git a/exprements/PaddleClas/test_tipc/config/SENet/SENet154_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SENet/SENet154_vd_train_infer_python.txt new file mode 100644 index 00000000..29d29547 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SENet/SENet154_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SENet154_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SENet154_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SENet154_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SENet154_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SENet154_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_infer_python.txt new file mode 100644 index 00000000..023dceba --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNeXt101_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt101_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_infer_python.txt new file mode 100644 index 00000000..e6c6db79 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNeXt50_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_infer_python.txt new file mode 100644 index 00000000..c2704f99 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNeXt50_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet18_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet18_vd_train_infer_python.txt new file mode 100644 index 00000000..095b1b98 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet18_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNet18_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNet18_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNet18_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNet18_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet18_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet34_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet34_vd_train_infer_python.txt new file mode 100644 index 00000000..260bc637 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet34_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNet34_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNet34_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNet34_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNet34_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet34_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet50_vd_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet50_vd_train_infer_python.txt new file mode 100644 index 00000000..b486f3c6 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SENet/SE_ResNet50_vd_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNet50_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNet50_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet50_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_infer_python.txt new file mode 100644 index 00000000..616e269c --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_swish +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_swish.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_swish.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_swish.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_infer_python.txt new file mode 100644 index 00000000..df208125 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_infer_python.txt new file mode 100644 index 00000000..91df1ac5 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_33 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_infer_python.txt new file mode 100644 index 00000000..e0ee7532 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt new file mode 100644 index 00000000..a0f354c0 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_infer_python.txt new file mode 100644 index 00000000..f9ab9c15 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_infer_python.txt new file mode 100644 index 00000000..eea768a3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_infer_python.txt new file mode 100644 index 00000000..49c59a9e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SqueezeNet1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_infer_python.txt new file mode 100644 index 00000000..1b957af7 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SqueezeNet1_1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_1.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_1_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_infer_python.txt new file mode 100644 index 00000000..e7161c8d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_base_patch4_window12_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_infer_python.txt new file mode 100644 index 00000000..2307086a --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_base_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_infer_python.txt new file mode 100644 index 00000000..f1a1873d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_large_patch4_window12_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_infer_python.txt new file mode 100644 index 00000000..359899e1 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_large_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_infer_python.txt new file mode 100644 index 00000000..7f9edb81 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_small_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_small_patch4_window7_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt new file mode 100644 index 00000000..80b76ed6 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_tiny_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_tiny_patch4_window7_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/TNT/TNT_small_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/TNT/TNT_small_train_infer_python.txt new file mode 100644 index 00000000..20d42648 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/TNT/TNT_small_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:TNT_small +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/TNT/TNT_small.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/TNT/TNT_small.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/TNT/TNT_small.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/TNT_small_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_base_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_base_train_infer_python.txt new file mode 100644 index 00000000..a08a7326 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_base_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:alt_gvt_base +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_base_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_large_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_large_train_infer_python.txt new file mode 100644 index 00000000..d75f8a14 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_large_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:alt_gvt_large +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_large_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_small_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_small_train_infer_python.txt new file mode 100644 index 00000000..d0686e5d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Twins/alt_gvt_small_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:alt_gvt_small +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_small_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_base_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_base_train_infer_python.txt new file mode 100644 index 00000000..76486b81 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_base_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:pcpvt_base +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Twins/pcpvt_base.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/pcpvt_base.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/pcpvt_base.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_base_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_large_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_large_train_infer_python.txt new file mode 100644 index 00000000..928c0b30 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_large_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:pcpvt_large +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Twins/pcpvt_large.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/pcpvt_large.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/pcpvt_large.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_large_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_small_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_small_train_infer_python.txt new file mode 100644 index 00000000..9ce13dfb --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Twins/pcpvt_small_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:pcpvt_small +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Twins/pcpvt_small.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/pcpvt_small.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/pcpvt_small.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_small_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VGG/VGG11_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VGG/VGG11_train_infer_python.txt new file mode 100644 index 00000000..4d5f7b10 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VGG/VGG11_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG11 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG11.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG11.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG11.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG11_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VGG/VGG13_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VGG/VGG13_train_infer_python.txt new file mode 100644 index 00000000..737a56bb --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VGG/VGG13_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG13 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG13.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG13.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG13.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG13_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VGG/VGG16_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VGG/VGG16_train_infer_python.txt new file mode 100644 index 00000000..b8bf46f0 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VGG/VGG16_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG16 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG16.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG16.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG16.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG16_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VGG/VGG19_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VGG/VGG19_train_infer_python.txt new file mode 100644 index 00000000..3f5a8f3e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VGG/VGG19_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG19 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG19.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG19.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG19.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG19_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_infer_python.txt new file mode 100644 index 00000000..d9f6e2f9 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_base_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_infer_python.txt new file mode 100644 index 00000000..ed88b515 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_base_patch16_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_infer_python.txt new file mode 100644 index 00000000..9c3abcdc --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_base_patch32_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch32_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch32_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch32_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch32_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_infer_python.txt new file mode 100644 index 00000000..ec77af6d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_large_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_infer_python.txt new file mode 100644 index 00000000..e05c6df9 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_large_patch16_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_infer_python.txt new file mode 100644 index 00000000..43b313f5 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_large_patch32_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch32_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch32_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch32_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch32_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_infer_python.txt new file mode 100644 index 00000000..6fc2cd03 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_small_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_small_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_small_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_small_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_small_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Xception/Xception41_deeplab_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Xception/Xception41_deeplab_train_infer_python.txt new file mode 100644 index 00000000..b9c16215 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Xception/Xception41_deeplab_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception41_deeplab +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception41_deeplab.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception41_deeplab.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception41_deeplab.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_deeplab_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Xception/Xception41_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Xception/Xception41_train_infer_python.txt new file mode 100644 index 00000000..7d7fbf10 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Xception/Xception41_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception41 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception41.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception41.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception41.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Xception/Xception65_deeplab_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Xception/Xception65_deeplab_train_infer_python.txt new file mode 100644 index 00000000..adf12cb8 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Xception/Xception65_deeplab_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception65_deeplab +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception65_deeplab.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception65_deeplab.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception65_deeplab.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_deeplab_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Xception/Xception65_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Xception/Xception65_train_infer_python.txt new file mode 100644 index 00000000..463e20df --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Xception/Xception65_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception65 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception65.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception65.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception65.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/config/Xception/Xception71_train_infer_python.txt b/exprements/PaddleClas/test_tipc/config/Xception/Xception71_train_infer_python.txt new file mode 100644 index 00000000..52ea37d0 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/config/Xception/Xception71_train_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception71 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception71.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception71.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception71.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception71_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/exprements/PaddleClas/test_tipc/docs/compare_cpp_right.png b/exprements/PaddleClas/test_tipc/docs/compare_cpp_right.png new file mode 100644 index 00000000..f9d0ba8e Binary files /dev/null and b/exprements/PaddleClas/test_tipc/docs/compare_cpp_right.png differ diff --git a/exprements/PaddleClas/test_tipc/docs/compare_cpp_wrong.png b/exprements/PaddleClas/test_tipc/docs/compare_cpp_wrong.png new file mode 100644 index 00000000..621d446b Binary files /dev/null and b/exprements/PaddleClas/test_tipc/docs/compare_cpp_wrong.png differ diff --git a/exprements/PaddleClas/test_tipc/docs/compare_right.png b/exprements/PaddleClas/test_tipc/docs/compare_right.png new file mode 100644 index 00000000..3d74ef1c Binary files /dev/null and b/exprements/PaddleClas/test_tipc/docs/compare_right.png differ diff --git a/exprements/PaddleClas/test_tipc/docs/compare_wrong.png b/exprements/PaddleClas/test_tipc/docs/compare_wrong.png new file mode 100644 index 00000000..26ad576d Binary files /dev/null and b/exprements/PaddleClas/test_tipc/docs/compare_wrong.png differ diff --git a/exprements/PaddleClas/test_tipc/docs/guide.png b/exprements/PaddleClas/test_tipc/docs/guide.png new file mode 100644 index 00000000..319ac819 Binary files /dev/null and b/exprements/PaddleClas/test_tipc/docs/guide.png differ diff --git a/exprements/PaddleClas/test_tipc/docs/install.md b/exprements/PaddleClas/test_tipc/docs/install.md new file mode 100644 index 00000000..35e2866b --- /dev/null +++ b/exprements/PaddleClas/test_tipc/docs/install.md @@ -0,0 +1,121 @@ +## 1. 环境准备 + +本教程适用于test_tipc目录下基础功能测试的运行环境搭建。 + +推荐环境： +- CUDA 10.1/10.2 +- CUDNN 7.6/cudnn8.1 +- TensorRT 6.1.0.5 / 7.1 / 7.2 + +环境配置可以选择docker镜像安装，或者在本地环境Python搭建环境。推荐使用docker镜像安装，避免不必要的环境配置。 + +## 2. Docker 镜像安装 + +推荐docker镜像安装，按照如下命令创建镜像，当前目录映射到镜像中的`/paddle`目录下 +``` +nvidia-docker run --name paddle -it -v $PWD:/paddle paddlepaddle/paddle:latest-dev-cuda10.1-cudnn7-gcc82 /bin/bash +cd /paddle + +# 安装带TRT的paddle +pip3.7 install https://paddle-wheel.bj.bcebos.com/with-trt/2.1.3/linux-gpu-cuda10.1-cudnn7-mkl-gcc8.2-trt6-avx/paddlepaddle_gpu-2.1.3.post101-cp37-cp37m-linux_x86_64.whl +``` + +## 3 Python 环境构建 + +非docker环境下，环境配置比较灵活，推荐环境组合配置： +- CUDA10.1 + CUDNN7.6 + TensorRT 6 +- CUDA10.2 + CUDNN8.1 + TensorRT 7 +- CUDA11.1 + CUDNN8.1 + TensorRT 7 + +下面以 CUDA10.2 + CUDNN8.1 + TensorRT 7 配置为例，介绍环境配置的流程。 + +### 3.1 安装CUDNN + +如果当前环境满足CUDNN版本的要求，可以跳过此步骤。 + +以CUDNN8.1 安装安装为例，安装步骤如下，首先下载CUDNN，从[Nvidia官网](https://developer.nvidia.com/rdp/cudnn-archive)下载CUDNN8.1版本，下载符合当前系统版本的三个deb文件，分别是： +- cuDNN Runtime Library ，如：libcudnn8_8.1.0.77-1+cuda10.2_amd64.deb +- cuDNN Developer Library ，如：libcudnn8-dev_8.1.0.77-1+cuda10.2_amd64.deb +- cuDNN Code Samples，如：libcudnn8-samples_8.1.0.77-1+cuda10.2_amd64.deb + +deb安装可以参考[官方文档](https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html#installlinux-deb)，安装方式如下 +``` +# x.x.x表示下载的版本号 +# $HOME为工作目录 +sudo dpkg -i libcudnn8_x.x.x-1+cudax.x_arm64.deb +sudo dpkg -i libcudnn8-dev_8.x.x.x-1+cudax.x_arm64.deb +sudo dpkg -i libcudnn8-samples_8.x.x.x-1+cudax.x_arm64.deb + +# 验证是否正确安装 +cp -r /usr/src/cudnn_samples_v8/ $HOME +cd $HOME/cudnn_samples_v8/mnistCUDNN + +# 编译 +make clean && make +./mnistCUDNN +``` +如果运行mnistCUDNN完后提示运行成功，则表示安装成功。如果运行后出现freeimage相关的报错，需要按照提示安装freeimage库: +``` +sudo apt-get install libfreeimage-dev +sudo apt-get install libfreeimage +``` + +### 3.2 安装TensorRT + +首先，从[Nvidia官网TensorRT板块](https://developer.nvidia.com/tensorrt-getting-started)下载TensorRT，这里选择7.1.3.4版本的TensorRT，注意选择适合自己系统版本和CUDA版本的TensorRT，另外建议下载TAR package的安装包。 + +以Ubuntu16.04+CUDA10.2为例，下载并解压后可以参考[官方文档](https://docs.nvidia.com/deeplearning/tensorrt/archives/tensorrt-713/install-guide/index.html#installing-tar)的安装步骤，按照如下步骤安装: +``` +# 以下安装命令中 '${version}' 为下载的TensorRT版本，如7.1.3.4 +# 设置环境变量， 为解压后的TensorRT的lib目录 +export LD_LIBRARY_PATH=$LD_LIBRARY_PATH: + +# 安装TensorRT +cd TensorRT-${version}/python +pip3.7 install tensorrt-*-cp3x-none-linux_x86_64.whl + +# 安装graphsurgeon +cd TensorRT-${version}/graphsurgeon +``` + + +### 3.3 安装PaddlePaddle + +下载支持TensorRT版本的Paddle安装包，注意安装包的TensorRT版本需要与本地TensorRT一致，下载[链接](https://paddleinference.paddlepaddle.org.cn/user_guides/download_lib.html#python) +选择下载 linux-cuda10.2-trt7-gcc8.2 Python3.7版本的Paddle： +``` +# 从下载链接中可以看到是paddle2.1.1-cuda10.2-cudnn8.1版本 +wget https://paddle-wheel.bj.bcebos.com/with-trt/2.1.1-gpu-cuda10.2-cudnn8.1-mkl-gcc8.2/paddlepaddle_gpu-2.1.1-cp37-cp37m-linux_x86_64.whl +pip3.7 install -U paddlepaddle_gpu-2.1.1-cp37-cp37m-linux_x86_64.whl +``` + +## 4. 安装PaddleClas依赖 +``` +# 安装AutoLog +git clone https://github.com/LDOUBLEV/AutoLog +cd AutoLog +pip3.7 install -r requirements.txt +python3.7 setup.py bdist_wheel +pip3.7 install ./dist/auto_log-1.0.0-py3-none-any.whl + +# 下载Clas代码 +cd ../ +git clone https://github.com/PaddlePaddle/PaddleClas + +``` + +安装PaddleClas依赖： +``` +cd PaddleClas +pip3.7 install -r requirements.txt +``` + +## FAQ : +Q. You are using Paddle compiled with TensorRT, but TensorRT dynamic library is not found. Ignore this if TensorRT is not needed. + +A. 问题一般是当前安装paddle版本带TRT，但是本地环境找不到TensorRT的预测库，需要下载TensorRT库，解压后设置环境变量LD_LIBRARY_PATH; +如： +``` +export LD_LIBRARY_PATH=/usr/local/python3.7.0/lib:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/paddle/package/TensorRT-6.0.1.5/lib +``` +或者问题是下载的TensorRT版本和当前paddle中编译的TRT版本不匹配，需要下载版本相符的TensorRT重新安装。 diff --git a/exprements/PaddleClas/test_tipc/docs/test.png b/exprements/PaddleClas/test_tipc/docs/test.png new file mode 100644 index 00000000..f99f23d7 Binary files /dev/null and b/exprements/PaddleClas/test_tipc/docs/test.png differ diff --git a/exprements/PaddleClas/test_tipc/docs/test_inference_cpp.md b/exprements/PaddleClas/test_tipc/docs/test_inference_cpp.md new file mode 100644 index 00000000..eabf8774 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/docs/test_inference_cpp.md @@ -0,0 +1,86 @@ +# C++预测功能测试 + +C++预测功能测试的主程序为`test_inference_cpp.sh`，可以测试基于C++预测库的模型推理功能。 + +## 1. 测试结论汇总 + +基于训练是否使用量化，进行本测试的模型可以分为`正常模型`和`量化模型`，这两类模型对应的C++预测功能汇总如下： + +| 模型类型 |device | batchsize | tensorrt | mkldnn | cpu多线程 | +| ---- | ---- | ---- | :----: | :----: | :----: | +| 正常模型 | GPU | 1/6 | fp32/fp16 | - | - | +| 正常模型 | CPU | 1/6 | - | fp32 | 支持 | +| 量化模型 | GPU | 1/6 | int8 | - | - | +| 量化模型 | CPU | 1/6 | - | int8 | 支持 | + +## 2. 测试流程 +运行环境配置请参考[文档](./install.md)的内容配置TIPC的运行环境。 + +### 2.1 功能测试 +先运行`prepare.sh`准备数据和模型，然后运行`test_inference_cpp.sh`进行测试，最终在```test_tipc/output```目录下生成`cpp_infer_*.log`后缀的日志文件。 + +```shell +bash test_tipc/prepare.sh test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt cpp_infer + +# 用法1: +bash test_tipc/test_inference_cpp.sh test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt +# 用法2: 指定GPU卡预测，第三个传入参数为GPU卡号 +bash test_tipc/test_inference_cpp.sh test_tipc/config/ResNet/ResNet50_vd_linux_gpu_normal_normal_infer_cpp_linux_gpu_cpu.txt 1 +``` + +运行预测指令后，在`test_tipc/output`文件夹下自动会保存运行日志，包括以下文件： + +```shell +test_tipc/output/ +|- results_cpp.log # 运行指令状态的日志 +|- cls_cpp_infer_cpu_usemkldnn_False_threads_1_precision_fp32_batchsize_1.log # CPU上不开启Mkldnn，线程数设置为1，测试batch_size=1条件下的预测运行日志 +|- cls_cpp_infer_cpu_usemkldnn_False_threads_6_precision_fp32_batchsize_1.log # CPU上不开启Mkldnn，线程数设置为6，测试batch_size=1条件下的预测运行日志 +|- cls_cpp_infer_gpu_usetrt_False_precision_fp32_batchsize_1.log # GPU上不开启TensorRT，测试batch_size=1的fp32精度预测日志 +|- cls_cpp_infer_gpu_usetrt_True_precision_fp16_batchsize_1.log # GPU上开启TensorRT，测试batch_size=1的fp16精度预测日志 +...... +``` +其中results_cpp.log中包含了每条指令的运行状态，如果运行成功会输出： + +``` +Run successfully with command - ./deploy/cpp/build/clas_system -c inference_cls.yaml 2>&1|tee test_tipc/output/cls_cpp_infer_gpu_usetrt_False_precision_fp32_batchsize_1.log +...... +``` +如果运行失败，会输出： +``` +Run failed with command - ./deploy/cpp/build/clas_system -c inference_cls.yaml 2>&1|tee test_tipc/output/cls_cpp_infer_gpu_usetrt_False_precision_fp32_batchsize_1.log +...... +``` +可以很方便的根据results_cpp.log中的内容判定哪一个指令运行错误。 + + +### 2.2 精度测试 + +使用compare_results.py脚本比较模型预测的结果是否符合预期，主要步骤包括： +- 提取日志中的预测坐标； +- 从本地文件中提取保存好的坐标结果； +- 比较上述两个结果是否符合精度预期，误差大于设置阈值时会报错。 + +#### 使用方式 +运行命令： +```shell +python3.7 test_tipc/compare_results.py --gt_file=./test_tipc/results/cls_cpp_*.txt --log_file=./test_tipc/output/cls_cpp_*.log --atol=1e-3 --rtol=1e-3 +``` + +参数介绍： +- gt_file： 指向事先保存好的预测结果路径，支持*.txt 结尾，会自动索引*.txt格式的文件，文件默认保存在test_tipc/result/ 文件夹下 +- log_file: 指向运行test_tipc/test_inference_cpp.sh 脚本的infer模式保存的预测日志，预测日志中打印的有预测结果，比如：文本框，预测文本，类别等等，同样支持cpp_infer_*.log格式传入 +- atol: 设置的绝对误差 +- rtol: 设置的相对误差 + +#### 运行结果 + +正常运行效果如下图： + + +出现不一致结果时的运行输出： + + + +## 3. 更多教程 + +本文档为功能测试用，更详细的c++预测使用教程请参考：[服务器端C++预测](../../docs/zh_CN/inference_deployment/) diff --git a/exprements/PaddleClas/test_tipc/docs/test_train_inference_python.md b/exprements/PaddleClas/test_tipc/docs/test_train_inference_python.md new file mode 100644 index 00000000..ef25e145 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/docs/test_train_inference_python.md @@ -0,0 +1,154 @@ +# Linux端基础训练预测功能测试 + +Linux端基础训练预测功能测试的主程序为`test_train_inference_python.sh`，可以测试基于Python的模型训练、评估、推理等基本功能，包括裁剪、量化、蒸馏。 + +- Mac端基础训练预测功能测试参考[链接](./mac_test_train_inference_python.md) +- Windows端基础训练预测功能测试参考[链接](./win_test_train_inference_python.md) + +## 1. 测试结论汇总 + +- 训练相关： + +| 算法名称 | 模型名称 | 单机单卡 | 单机多卡 | 多机多卡 | 模型压缩（单机多卡） | +| :---- | :---- | :---- | :---- | :---- | :---- | +| MobileNetV3 | MobileNetV3_large_x1_0 | 正常训练
混合精度 | 正常训练
混合精度 | 正常训练
混合精度 | 正常训练：FPGM裁剪、PACT量化
离线量化（无需训练） | +| ResNet50 | ResNet50_vd | 正常训练
混合精度 | 正常训练
混合精度 | 正常训练
混合精度 | 正常训练：FPGM裁剪、PACT量化
离线量化（无需训练） | + + +- 预测相关：基于训练是否使用量化，可以将训练产出的模型可以分为`正常模型`和`量化模型`，这两类模型对应的预测功能汇总如下， + +| 模型类型 |device | batchsize | tensorrt | mkldnn | cpu多线程 | +| ---- | ---- | ---- | :----: | :----: | :----: | +| 正常模型 | GPU | 1/6 | fp32/fp16 | - | - | +| 正常模型 | CPU | 1/6 | - | fp32/fp16 | 支持 | +| 量化模型 | GPU | 1/6 | int8 | - | - | +| 量化模型 | CPU | 1/6 | - | int8 | 支持 | + + +## 2. 测试流程 + +运行环境配置请参考[文档](./install.md)的内容配置TIPC的运行环境。 + +### 2.1 安装依赖 +- 安装PaddlePaddle >= 2.2 +- 安装PaddleClas依赖 + ``` + pip3 install -r ../requirements.txt + ``` +- 安装autolog（规范化日志输出工具） + ``` + git clone https://github.com/LDOUBLEV/AutoLog + cd AutoLog + pip3 install -r requirements.txt + python3 setup.py bdist_wheel + pip3 install ./dist/auto_log-1.0.0-py3-none-any.whl + cd ../ + ``` +- 安装PaddleSlim (可选) + ``` + # 如果要测试量化、裁剪等功能，需要安装PaddleSlim + pip3 install paddleslim + ``` + + +### 2.2 功能测试 +先运行`prepare.sh`准备数据和模型，然后运行`test_train_inference_python.sh`进行测试，最终在```test_tipc/output```目录下生成`python_infer_*.log`格式的日志文件。 + +`test_train_inference_python.sh`包含5种运行模式，每种模式的运行数据不同，分别用于测试速度和精度，分别是： + +- 模式1：lite_train_lite_infer，使用少量数据训练，用于快速验证训练到预测的走通流程，不验证精度和速度； +```shell +bash test_tipc/prepare.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'lite_train_lite_infer' +bash test_tipc/test_train_inference_python.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'lite_train_lite_infer' +``` + +- 模式2：lite_train_whole_infer，使用少量数据训练，一定量数据预测，用于验证训练后的模型执行预测，预测速度是否合理； +```shell +bash test_tipc/prepare.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'lite_train_whole_infer' +bash test_tipc/test_train_inference_python.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'lite_train_whole_infer' +``` + +- 模式3：whole_infer，不训练，全量数据预测，走通开源模型评估、动转静，检查inference model预测时间和精度; +```shell +bash test_tipc/prepare.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'whole_infer' +# 用法1: +bash test_tipc/test_train_inference_python.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'whole_infer' +# 用法2: 指定GPU卡预测，第三个传入参数为GPU卡号 +bash test_tipc/test_train_inference_python.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'whole_infer' '1' +``` + +- 模式4：whole_train_whole_infer，CE： 全量数据训练，全量数据预测，验证模型训练精度，预测精度，预测速度； +```shell +bash test_tipc/prepare.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'whole_train_whole_infer' +bash test_tipc/test_train_inference_python.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'whole_train_whole_infer' +``` + +- 模式5: klquant_whole_infer，测试离线量化； + +```shell +bash test_tipc/prepare.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'klquant_whole_infer' +bash test_tipc/test_train_inference_python.sh ./test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt 'klquant_whole_infer' +``` + + + +运行相应指令后，在`test_tipc/output`文件夹下自动会保存运行日志。如'lite_train_lite_infer'模式下，会运行训练+inference的链条，因此，在`test_tipc/output`文件夹有以下文件： + +``` +test_tipc/output/ +|- results_python.log # 运行指令状态的日志 +|- norm_train_gpus_0_autocast_null/ # GPU 0号卡上正常训练的训练日志和模型保存文件夹 +|- pact_train_gpus_0_autocast_null/ # GPU 0号卡上量化训练的训练日志和模型保存文件夹 +...... +|- python_infer_cpu_usemkldnn_True_threads_1_batchsize_1.log # CPU上开启Mkldnn线程数设置为1，测试batch_size=1条件下的预测运行日志 +|- python_infer_gpu_usetrt_True_precision_fp16_batchsize_1.log # GPU上开启TensorRT，测试batch_size=1的半精度预测日志 +...... +``` + +其中`results_python.log`中包含了每条指令的运行状态，如果运行成功会输出： +``` +Run successfully with command - python3.7 tools/train.py -c tests/configs/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained Global.use_gpu=True Global.save_model_dir=./tests/output/norm_train_gpus_0_autocast_null Global.epoch_num=1 Train.loader.batch_size_per_card=2 ! +Run successfully with command - python3.7 tools/export_model.py -c tests/configs/det_mv3_db.yml -o Global.pretrained_model=./tests/output/norm_train_gpus_0_autocast_null/latest Global.save_inference_dir=./tests/output/norm_train_gpus_0_autocast_null! +...... +``` +如果运行失败，会输出： +``` +Run failed with command - python3.7 tools/train.py -c tests/configs/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained Global.use_gpu=True Global.save_model_dir=./tests/output/norm_train_gpus_0_autocast_null Global.epoch_num=1 Train.loader.batch_size_per_card=2 ! +Run failed with command - python3.7 tools/export_model.py -c tests/configs/det_mv3_db.yml -o Global.pretrained_model=./tests/output/norm_train_gpus_0_autocast_null/latest Global.save_inference_dir=./tests/output/norm_train_gpus_0_autocast_null! +...... +``` +可以很方便的根据`results_python.log`中的内容判定哪一个指令运行错误。 + + +### 2.3 精度测试 + +使用compare_results.py脚本比较模型预测的结果是否符合预期，主要步骤包括： +- 提取日志中的预测坐标； +- 从本地文件中提取保存好的坐标结果； +- 比较上述两个结果是否符合精度预期，误差大于设置阈值时会报错。 + +#### 使用方式 +运行命令： +```shell +python3.7 test_tipc/compare_results.py --gt_file=./test_tipc/results/python_*.txt --log_file=./test_tipc/output/python_*.log --atol=1e-3 --rtol=1e-3 +``` + +参数介绍： +- gt_file： 指向事先保存好的预测结果路径，支持*.txt 结尾，会自动索引*.txt格式的文件，文件默认保存在test_tipc/result/ 文件夹下 +- log_file: 指向运行test_tipc/test_train_inference_python.sh 脚本的infer模式保存的预测日志，预测日志中打印的有预测结果，比如：文本框，预测文本，类别等等，同样支持python_infer_*.log格式传入 +- atol: 设置的绝对误差 +- rtol: 设置的相对误差 + +#### 运行结果 + +正常运行效果如下图： + + +出现不一致结果时的运行输出： + + + +## 3. 更多教程 +本文档为功能测试用，更丰富的训练预测使用教程请参考： +[模型训练](../../docs/zh_CN/models_training) +[基于Python预测引擎推理](../../docs/zh_CN/inference_deployment/python_deploy.md) diff --git a/exprements/PaddleClas/test_tipc/generate_cpp_yaml.py b/exprements/PaddleClas/test_tipc/generate_cpp_yaml.py new file mode 100644 index 00000000..2e541de3 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/generate_cpp_yaml.py @@ -0,0 +1,80 @@ +import os +import yaml +import argparse + + +def str2bool(v): + if v.lower() == 'true': + return True + else: + return False + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument('--type', required=True, choices=["cls", "shitu"]) + parser.add_argument('--batch_size', type=int, default=1) + parser.add_argument('--mkldnn', type=str2bool, default=True) + parser.add_argument('--gpu', type=str2bool, default=False) + parser.add_argument('--cpu_thread', type=int, default=1) + parser.add_argument('--tensorrt', type=str2bool, default=False) + parser.add_argument('--precision', type=str, choices=["fp32", "fp16"]) + parser.add_argument('--benchmark', type=str2bool, default=True) + parser.add_argument('--gpu_id', type=int, default=0) + parser.add_argument( + '--cls_yaml_path', + type=str, + default="deploy/configs/inference_cls.yaml") + parser.add_argument( + '--shitu_yaml_path', + type=str, + default="deploy/configs/inference_drink.yaml") + parser.add_argument('--data_dir', type=str, required=True) + parser.add_argument('--save_path', type=str, default='./') + parser.add_argument('--cls_model_dir', type=str) + parser.add_argument('--det_model_dir', type=str) + args = parser.parse_args() + return args + + +def main(): + args = parse_args() + if args.type == "cls": + save_path = os.path.join(args.save_path, + os.path.basename(args.cls_yaml_path)) + fd = open(args.cls_yaml_path) + else: + save_path = os.path.join(args.save_path, + os.path.basename(args.shitu_yaml_path)) + fd = open(args.shitu_yaml_path) + config = yaml.load(fd, yaml.FullLoader) + fd.close() + + config["Global"]["batch_size"] = args.batch_size + config["Global"]["use_gpu"] = args.gpu + config["Global"]["enable_mkldnn"] = args.mkldnn + config["Global"]["benchmark"] = args.benchmark + config["Global"]["use_tensorrt"] = args.tensorrt + config["Global"]["use_fp16"] = True if args.precision == "fp16" else False + config["Global"]["gpu_id"] = args.gpu_id + if args.type == "cls": + config["Global"]["infer_imgs"] = args.data_dir + assert args.cls_model_dir + config["Global"]["inference_model_dir"] = args.cls_model_dir + else: + config["Global"]["infer_imgs"] = os.path.join(args.data_dir, + "test_images") + config["IndexProcess"]["index_dir"] = os.path.join(args.data_dir, + "index") + assert args.cls_model_dir + assert args.det_model_dir + config["Global"]["det_inference_model_dir"] = args.det_model_dir + config["Global"]["rec_inference_model_dir"] = args.cls_model_dir + + with open(save_path, 'w') as fd: + yaml.dump(config, fd) + print("Generate new yaml done") + + +if __name__ == "__main__": + main() diff --git a/exprements/PaddleClas/test_tipc/prepare.sh b/exprements/PaddleClas/test_tipc/prepare.sh new file mode 100644 index 00000000..acb2eb05 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/prepare.sh @@ -0,0 +1,184 @@ +#!/bin/bash +FILENAME=$1 + +# MODE be one of ['lite_train_lite_infer' 'lite_train_whole_infer' 'whole_train_whole_infer', +# 'whole_infer', 'klquant_whole_infer', +# 'cpp_infer', 'serving_infer', 'lite_infer'] + +MODE=$2 + +dataline=$(cat ${FILENAME}) +# parser params +IFS=$'\n' +lines=(${dataline}) + +function func_parser_key(){ + strs=$1 + IFS=":" + array=(${strs}) + tmp=${array[0]} + echo ${tmp} +} + +function func_parser_value(){ + strs=$1 + IFS=":" + array=(${strs}) + if [ ${#array[*]} = 2 ]; then + echo ${array[1]} + else + IFS="|" + tmp="${array[1]}:${array[2]}" + echo ${tmp} + fi +} + +function func_get_url_file_name(){ + strs=$1 + IFS="/" + array=(${strs}) + tmp=${array[${#array[@]}-1]} + echo ${tmp} +} + +model_name=$(func_parser_value "${lines[1]}") + +if [ ${MODE} = "cpp_infer" ];then + if [[ $FILENAME == *infer_cpp_linux_gpu_cpu.txt ]];then + cpp_type=$(func_parser_value "${lines[2]}") + cls_inference_model_dir=$(func_parser_value "${lines[3]}") + det_inference_model_dir=$(func_parser_value "${lines[4]}") + cls_inference_url=$(func_parser_value "${lines[5]}") + det_inference_url=$(func_parser_value "${lines[6]}") + + if [[ $cpp_type == "cls" ]];then + eval "wget -nc $cls_inference_url" + tar xf "${model_name}_inference.tar" + eval "mv inference $cls_inference_model_dir" + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_infer.tar + tar xf whole_chain_infer.tar + ln -s whole_chain_infer ILSVRC2012 + cd .. + elif [[ $cpp_type == "shitu" ]];then + eval "wget -nc $cls_inference_url" + tar_name=$(func_get_url_file_name "$cls_inference_url") + model_dir=${tar_name%.*} + eval "tar xf ${tar_name}" + eval "mv ${model_dir} ${cls_inference_model_dir}" + + eval "wget -nc $det_inference_url" + tar_name=$(func_get_url_file_name "$det_inference_url") + model_dir=${tar_name%.*} + eval "tar xf ${tar_name}" + eval "mv ${model_dir} ${det_inference_model_dir}" + cd dataset + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/data/drink_dataset_v1.0.tar + tar -xf drink_dataset_v1.0.tar + else + echo "Wrong cpp type in config file in line 3. only support cls, shitu" + fi + exit 0 + else + echo "use wrong config file" + exit 1 + fi +fi + +model_name=$(func_parser_value "${lines[1]}") +model_url_value=$(func_parser_value "${lines[35]}") +model_url_key=$(func_parser_key "${lines[35]}") + +if [[ $FILENAME == *GeneralRecognition* ]];then + cd dataset + rm -rf Aliproduct + rm -rf train_reg_all_data.txt + rm -rf demo_train + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/tipc_shitu_demo_data.tar + tar -xf tipc_shitu_demo_data.tar + ln -s tipc_shitu_demo_data Aliproduct + ln -s tipc_shitu_demo_data/demo_train.txt train_reg_all_data.txt + ln -s tipc_shitu_demo_data/demo_train demo_train + cd tipc_shitu_demo_data + ln -s demo_test.txt val_list.txt + cd ../../ + eval "wget -nc $model_url_value" + mv general_PPLCNet_x2_5_pretrained_v1.0.pdparams GeneralRecognition_PPLCNet_x2_5_pretrained.pdparams + exit 0 +fi + +if [[ $FILENAME == *use_dali* ]];then + python_name=$(func_parser_value "${lines[2]}") + ${python_name} -m pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/nightly --upgrade nvidia-dali-nightly-cuda102 +fi + +if [ ${MODE} = "lite_train_lite_infer" ] || [ ${MODE} = "lite_train_whole_infer" ];then + # pretrain lite train data + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_little_train.tar + tar xf whole_chain_little_train.tar + ln -s whole_chain_little_train ILSVRC2012 + cd ILSVRC2012 + mv train.txt train_list.txt + mv val.txt val_list.txt + cp -r train/* val/ + cd ../../ +elif [ ${MODE} = "whole_infer" ] || [ ${MODE} = "klquant_whole_infer" ];then + # download data + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_infer.tar + tar xf whole_chain_infer.tar + ln -s whole_chain_infer ILSVRC2012 + cd ILSVRC2012 + mv val.txt val_list.txt + ln -s val_list.txt train_list.txt + cd ../../ + # download inference or pretrained model + eval "wget -nc $model_url_value" + if [[ $model_url_key == *inference* ]]; then + rm -rf inference + tar xf "${model_name}_inference.tar" + fi + if [[ $model_name == "SwinTransformer_large_patch4_window7_224" || $model_name == "SwinTransformer_large_patch4_window12_384" ]];then + cmd="mv ${model_name}_22kto1k_pretrained.pdparams ${model_name}_pretrained.pdparams" + eval $cmd + fi + +elif [ ${MODE} = "whole_train_whole_infer" ];then + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_CIFAR100.tar + tar xf whole_chain_CIFAR100.tar + ln -s whole_chain_CIFAR100 ILSVRC2012 + cd ILSVRC2012 + mv train.txt train_list.txt + mv test.txt val_list.txt + cd ../../ +fi + +if [ ${MODE} = "serving_infer" ];then + # prepare serving env + python_name=$(func_parser_value "${lines[2]}") + ${python_name} -m pip install install paddle-serving-server-gpu==0.6.1.post101 + ${python_name} -m pip install paddle_serving_client==0.6.1 + ${python_name} -m pip install paddle-serving-app==0.6.1 + unset http_proxy + unset https_proxy + cd ./deploy/paddleserving + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet50_vd_infer.tar && tar xf ResNet50_vd_infer.tar +fi + +if [ ${MODE} = "paddle2onnx_infer" ];then + # prepare paddle2onnx env + python_name=$(func_parser_value "${lines[2]}") + ${python_name} -m pip install install paddle2onnx + ${python_name} -m pip install onnxruntime + + # wget model + cd deploy && mkdir models && cd models + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet50_vd_infer.tar && tar xf ResNet50_vd_infer.tar + cd ../../ +fi diff --git a/exprements/PaddleClas/test_tipc/prepare_lite_cpp.sh b/exprements/PaddleClas/test_tipc/prepare_lite_cpp.sh new file mode 100644 index 00000000..b129322d --- /dev/null +++ b/exprements/PaddleClas/test_tipc/prepare_lite_cpp.sh @@ -0,0 +1,93 @@ +#!/bin/bash +source ./test_tipc/common_func.sh +FILENAME=$1 +dataline=$(cat ${FILENAME}) +# parser params +IFS=$'\n' +lines=(${dataline}) +IFS=$'\n' + +inference_cmd=$(func_parser_value "${lines[1]}") +DEVICE=$(func_parser_value "${lines[2]}") +det_lite_model_list=$(func_parser_value "${lines[3]}") +rec_lite_model_list=$(func_parser_value "${lines[4]}") +cls_lite_model_list=$(func_parser_value "${lines[5]}") + +if [[ $inference_cmd =~ "det" ]];then + lite_model_list=${det_lite_model_list} +elif [[ $inference_cmd =~ "rec" ]];then + lite_model_list=(${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) +elif [[ $inference_cmd =~ "system" ]];then + lite_model_list=(${det_lite_model_list[*]} ${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) +else + echo "inference_cmd is wrong, please check." + exit 1 +fi + +if [ ${DEVICE} = "ARM_CPU" ];then + valid_targets="arm" + paddlelite_url="https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10-rc/inference_lite_lib.android.armv8.gcc.c++_shared.with_extra.with_cv.tar.gz" + end_index="66" +elif [ ${DEVICE} = "ARM_GPU_OPENCL" ];then + valid_targets="opencl" + paddlelite_url="https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10-rc/inference_lite_lib.armv8.clang.with_exception.with_extra.with_cv.opencl.tar.gz" + end_index="71" +else + echo "DEVICE only suport ARM_CPU, ARM_GPU_OPENCL." + exit 2 +fi + +# prepare lite .nb model +pip install paddlelite==2.10-rc +current_dir=${PWD} +IFS="|" +model_path=./inference_models + +for model in ${lite_model_list[*]}; do + if [[ $model =~ "PP-OCRv2" ]];then + inference_model_url=https://paddleocr.bj.bcebos.com/PP-OCRv2/chinese/${model}.tar + elif [[ $model =~ "v2.0" ]];then + inference_model_url=https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/${model}.tar + else + echo "Model is wrong, please check." + exit 3 + fi + inference_model=${inference_model_url##*/} + wget -nc -P ${model_path} ${inference_model_url} + cd ${model_path} && tar -xf ${inference_model} && cd ../ + model_dir=${model_path}/${inference_model%.*} + model_file=${model_dir}/inference.pdmodel + param_file=${model_dir}/inference.pdiparams + paddle_lite_opt --model_dir=${model_dir} --model_file=${model_file} --param_file=${param_file} --valid_targets=${valid_targets} --optimize_out=${model_dir}_opt +done + +# prepare test data +data_url=https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015_lite.tar +model_path=./inference_models +inference_model=${inference_model_url##*/} +data_file=${data_url##*/} +wget -nc -P ./inference_models ${inference_model_url} +wget -nc -P ./test_data ${data_url} +cd ./inference_models && tar -xf ${inference_model} && cd ../ +cd ./test_data && tar -xf ${data_file} && rm ${data_file} && cd ../ + +# prepare lite env +paddlelite_zipfile=$(echo $paddlelite_url | awk -F "/" '{print $NF}') +paddlelite_file=${paddlelite_zipfile:0:${end_index}} +wget ${paddlelite_url} && tar -xf ${paddlelite_zipfile} +mkdir -p ${paddlelite_file}/demo/cxx/ocr/test_lite +cp -r ${model_path}/*_opt.nb test_data ${paddlelite_file}/demo/cxx/ocr/test_lite +cp ppocr/utils/ppocr_keys_v1.txt deploy/lite/config.txt ${paddlelite_file}/demo/cxx/ocr/test_lite +cp -r ./deploy/lite/* ${paddlelite_file}/demo/cxx/ocr/ +cp ${paddlelite_file}/cxx/lib/libpaddle_light_api_shared.so ${paddlelite_file}/demo/cxx/ocr/test_lite +cp ${FILENAME} test_tipc/test_lite_arm_cpp.sh test_tipc/common_func.sh ${paddlelite_file}/demo/cxx/ocr/test_lite +cd ${paddlelite_file}/demo/cxx/ocr/ +git clone https://github.com/cuicheng01/AutoLog.git + +# make +make -j +sleep 1 +make -j +cp ocr_db_crnn test_lite && cp test_lite/libpaddle_light_api_shared.so test_lite/libc++_shared.so +tar -cf test_lite.tar ./test_lite && cp test_lite.tar ${current_dir} && cd ${current_dir} +rm -rf ${paddlelite_file}* && rm -rf ${model_path} diff --git a/exprements/PaddleClas/test_tipc/test_inference_cpp.sh b/exprements/PaddleClas/test_tipc/test_inference_cpp.sh new file mode 100644 index 00000000..129e4395 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/test_inference_cpp.sh @@ -0,0 +1,303 @@ +#!/bin/bash +source test_tipc/common_func.sh + +FILENAME=$1 +GPUID=$2 +if [[ ! $GPUID ]];then + GPUID=0 +fi +dataline=$(awk 'NR==1, NR==16{print}' $FILENAME) + +# parser params +IFS=$'\n' +lines=(${dataline}) + +# parser cpp inference model +model_name=$(func_parser_value "${lines[1]}") +cpp_infer_type=$(func_parser_value "${lines[2]}") +cpp_infer_model_dir=$(func_parser_value "${lines[3]}") +cpp_det_infer_model_dir=$(func_parser_value "${lines[4]}") +cpp_infer_is_quant=$(func_parser_value "${lines[7]}") +# parser cpp inference +inference_cmd=$(func_parser_value "${lines[8]}") +cpp_use_gpu_list=$(func_parser_value "${lines[9]}") +cpp_use_mkldnn_list=$(func_parser_value "${lines[10]}") +cpp_cpu_threads_list=$(func_parser_value "${lines[11]}") +cpp_batch_size_list=$(func_parser_value "${lines[12]}") +cpp_use_trt_list=$(func_parser_value "${lines[13]}") +cpp_precision_list=$(func_parser_value "${lines[14]}") +cpp_image_dir_value=$(func_parser_value "${lines[15]}") +cpp_benchmark_value=$(func_parser_value "${lines[16]}") +generate_yaml_cmd=$(func_parser_value "${lines[17]}") +transform_index_cmd=$(func_parser_value "${lines[18]}") + +LOG_PATH="./test_tipc/output" +mkdir -p ${LOG_PATH} +status_log="${LOG_PATH}/results_cpp.log" +# generate_yaml_cmd="python3 test_tipc/generate_cpp_yaml.py" + +function func_shitu_cpp_inference(){ + IFS='|' + _script=$1 + _model_dir=$2 + _log_path=$3 + _img_dir=$4 + _flag_quant=$5 + # inference + + for use_gpu in ${cpp_use_gpu_list[*]}; do + if [ ${use_gpu} = "False" ] || [ ${use_gpu} = "cpu" ]; then + for use_mkldnn in ${cpp_use_mkldnn_list[*]}; do + if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then + continue + fi + for threads in ${cpp_cpu_threads_list[*]}; do + for batch_size in ${cpp_batch_size_list[*]}; do + precision="fp32" + if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then + precison="int8" + fi + _save_log_path="${_log_path}/shitu_cpp_infer_cpu_usemkldnn_${use_mkldnn}_threads_${threads}_precision_${precision}_batchsize_${batch_size}.log" + + command="${generate_yaml_cmd} --type shitu --batch_size ${batch_size} --mkldnn ${use_mkldnn} --gpu ${use_gpu} --cpu_thread ${threads} --tensorrt False --precision ${precision} --data_dir ${_img_dir} --benchmark True --cls_model_dir ${cpp_infer_model_dir} --det_model_dir ${cpp_det_infer_model_dir} --gpu_id ${GPUID}" + eval $command + eval $transform_index_cmd + command="${_script} 2>&1|tee ${_save_log_path}" + eval $command + last_status=${PIPESTATUS[0]} + status_check $last_status "${command}" "${status_log}" + done + done + done + elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then + for use_trt in ${cpp_use_trt_list[*]}; do + for precision in ${cpp_precision_list[*]}; do + if [[ ${_flag_quant} = "False" ]] && [[ ${precision} =~ "int8" ]]; then + continue + fi + if [[ ${precision} =~ "fp16" || ${precision} =~ "int8" ]] && [ ${use_trt} = "False" ]; then + continue + fi + if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [ ${_flag_quant} = "True" ]; then + continue + fi + for batch_size in ${cpp_batch_size_list[*]}; do + _save_log_path="${_log_path}/shitu_cpp_infer_gpu_usetrt_${use_trt}_precision_${precision}_batchsize_${batch_size}.log" + command="${generate_yaml_cmd} --type shitu --batch_size ${batch_size} --mkldnn False --gpu ${use_gpu} --cpu_thread 1 --tensorrt ${use_trt} --precision ${precision} --data_dir ${_img_dir} --benchmark True --cls_model_dir ${cpp_infer_model_dir} --det_model_dir ${cpp_det_infer_model_dir} --gpu_id ${GPUID}" + eval $command + eval $transform_index_cmd + command="${_script} 2>&1|tee ${_save_log_path}" + eval $command + last_status=${PIPESTATUS[0]} + status_check $last_status "${_script}" "${status_log}" + done + done + done + else + echo "Does not support hardware other than CPU and GPU Currently!" + fi + done +} + +function func_cls_cpp_inference(){ + IFS='|' + _script=$1 + _model_dir=$2 + _log_path=$3 + _img_dir=$4 + _flag_quant=$5 + # inference + + for use_gpu in ${cpp_use_gpu_list[*]}; do + if [ ${use_gpu} = "False" ] || [ ${use_gpu} = "cpu" ]; then + for use_mkldnn in ${cpp_use_mkldnn_list[*]}; do + if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then + continue + fi + for threads in ${cpp_cpu_threads_list[*]}; do + for batch_size in ${cpp_batch_size_list[*]}; do + precision="fp32" + if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then + precison="int8" + fi + _save_log_path="${_log_path}/cls_cpp_infer_cpu_usemkldnn_${use_mkldnn}_threads_${threads}_precision_${precision}_batchsize_${batch_size}.log" + + command="${generate_yaml_cmd} --type cls --batch_size ${batch_size} --mkldnn ${use_mkldnn} --gpu ${use_gpu} --cpu_thread ${threads} --tensorrt False --precision ${precision} --data_dir ${_img_dir} --benchmark True --cls_model_dir ${cpp_infer_model_dir} --gpu_id ${GPUID}" + eval $command + command1="${_script} 2>&1|tee ${_save_log_path}" + eval ${command1} + last_status=${PIPESTATUS[0]} + status_check $last_status "${command1}" "${status_log}" + done + done + done + elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then + for use_trt in ${cpp_use_trt_list[*]}; do + for precision in ${cpp_precision_list[*]}; do + if [[ ${_flag_quant} = "False" ]] && [[ ${precision} =~ "int8" ]]; then + continue + fi + if [[ ${precision} =~ "fp16" || ${precision} =~ "int8" ]] && [ ${use_trt} = "False" ]; then + continue + fi + if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [ ${_flag_quant} = "True" ]; then + continue + fi + for batch_size in ${cpp_batch_size_list[*]}; do + _save_log_path="${_log_path}/cls_cpp_infer_gpu_usetrt_${use_trt}_precision_${precision}_batchsize_${batch_size}.log" + command="${generate_yaml_cmd} --type cls --batch_size ${batch_size} --mkldnn False --gpu ${use_gpu} --cpu_thread 1 --tensorrt ${use_trt} --precision ${precision} --data_dir ${_img_dir} --benchmark True --cls_model_dir ${cpp_infer_model_dir} --gpu_id ${GPUID}" + eval $command + command="${_script} 2>&1|tee ${_save_log_path}" + eval $command + last_status=${PIPESTATUS[0]} + status_check $last_status "${command}" "${status_log}" + done + done + done + else + echo "Does not support hardware other than CPU and GPU Currently!" + fi + done +} + + +if [[ $cpp_infer_type == "cls" ]]; then + cd deploy/cpp +elif [[ $cpp_infer_type == "shitu" ]]; then + cd deploy/cpp_shitu +else + echo "Only support cls and shitu" + exit 0 +fi + +if [[ $cpp_infer_type == "shitu" ]]; then + echo "################### update cmake ###################" + wget -nc https://github.com/Kitware/CMake/releases/download/v3.22.0/cmake-3.22.0.tar.gz + tar xf cmake-3.22.0.tar.gz + cd ./cmake-3.22.0 + export root_path=$PWD + export install_path=${root_path}/cmake + eval "./bootstrap --prefix=${install_path}" + make -j + make install + export PATH=${install_path}/bin:$PATH + cd .. + echo "################### update cmake done ###################" + + echo "################### build faiss ###################" + apt-get install -y libopenblas-dev + git clone https://github.com/facebookresearch/faiss.git + cd faiss + export faiss_install_path=$PWD/faiss_install + eval "cmake -B build . -DFAISS_ENABLE_PYTHON=OFF -DCMAKE_INSTALL_PREFIX=${faiss_install_path}" + make -C build -j faiss + make -C build install + cd .. +fi + +if [ -d "opencv-3.4.7/opencv3/" ] && [ $(md5sum opencv-3.4.7.tar.gz | awk -F ' ' '{print $1}') = "faa2b5950f8bee3f03118e600c74746a" ];then + echo "################### build opencv skipped ###################" +else + echo "################### build opencv ###################" + rm -rf opencv-3.4.7.tar.gz opencv-3.4.7/ + wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/opencv-3.4.7.tar.gz + tar -xf opencv-3.4.7.tar.gz + + cd opencv-3.4.7/ + install_path=$(pwd)/opencv3 + + rm -rf build + mkdir build + cd build + + cmake .. \ + -DCMAKE_INSTALL_PREFIX=${install_path} \ + -DCMAKE_BUILD_TYPE=Release \ + -DBUILD_SHARED_LIBS=OFF \ + -DWITH_IPP=OFF \ + -DBUILD_IPP_IW=OFF \ + -DWITH_LAPACK=OFF \ + -DWITH_EIGEN=OFF \ + -DCMAKE_INSTALL_LIBDIR=lib64 \ + -DWITH_ZLIB=ON \ + -DBUILD_ZLIB=ON \ + -DWITH_JPEG=ON \ + -DBUILD_JPEG=ON \ + -DWITH_PNG=ON \ + -DBUILD_PNG=ON \ + -DWITH_TIFF=ON \ + -DBUILD_TIFF=ON + + make -j + make install + cd ../../ + echo "################### build opencv finished ###################" +fi + +echo "################### build PaddleClas demo ####################" +OPENCV_DIR=$(pwd)/opencv-3.4.7/opencv3/ +# LIB_DIR=/work/project/project/test/paddle_inference/ +LIB_DIR=$(pwd)/Paddle/build/paddle_inference_install_dir/ +CUDA_LIB_DIR=$(dirname `find /usr -name libcudart.so`) +CUDNN_LIB_DIR=$(dirname `find /usr -name libcudnn.so`) + +BUILD_DIR=build +rm -rf ${BUILD_DIR} +mkdir ${BUILD_DIR} +cd ${BUILD_DIR} +if [[ $cpp_infer_type == cls ]]; then + cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DWITH_GPU=ON \ + -DWITH_STATIC_LIB=OFF \ + -DWITH_TENSORRT=OFF \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + -DTENSORRT_DIR=${TENSORRT_DIR} +else + cmake ..\ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DWITH_GPU=ON \ + -DWITH_STATIC_LIB=OFF \ + -DWITH_TENSORRT=OFF \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + -DTENSORRT_DIR=${TENSORRT_DIR} \ + -DFAISS_DIR=${faiss_install_path} \ + -DFAISS_WITH_MKL=OFF +fi +make -j +cd ../../../ +# cd ../../ +echo "################### build PaddleClas demo finished ###################" + + +# set cuda device +# GPUID=$2 +# if [ ${#GPUID} -le 0 ];then +# env="export CUDA_VISIBLE_DEVICES=0" +# else +# env="export CUDA_VISIBLE_DEVICES=${GPUID}" +# fi +# set CUDA_VISIBLE_DEVICES +# eval $env + + +echo "################### run test ###################" +export Count=0 +IFS="|" +infer_quant_flag=(${cpp_infer_is_quant}) +for infer_model in ${cpp_infer_model_dir[*]}; do + #run inference + is_quant=${infer_quant_flag[Count]} + if [[ $cpp_infer_type == "cls" ]]; then + func_cls_cpp_inference "${inference_cmd}" "${infer_model}" "${LOG_PATH}" "${cpp_image_dir_value}" ${is_quant} + else + func_shitu_cpp_inference "${inference_cmd}" "${infer_model}" "${LOG_PATH}" "${cpp_image_dir_value}" ${is_quant} + fi + Count=$(($Count + 1)) +done diff --git a/exprements/PaddleClas/test_tipc/test_inference_jeston.sh b/exprements/PaddleClas/test_tipc/test_inference_jeston.sh new file mode 100644 index 00000000..2fd76e1e --- /dev/null +++ b/exprements/PaddleClas/test_tipc/test_inference_jeston.sh @@ -0,0 +1,87 @@ +#!/bin/bash +source test_tipc/common_func.sh +source test_tipc/test_train_inference_python.sh + +FILENAME=$1 +# MODE be one of ['whole_infer'] +MODE=$2 + +dataline=$(awk 'NR==1, NR==17{print}' $FILENAME) + +# parser params +IFS=$'\n' +lines=(${dataline}) + +model_name=$(func_parser_value "${lines[1]}") +python=$(func_parser_value "${lines[2]}") + +infer_model_dir_list=$(func_parser_value "${lines[3]}") +infer_export_list=$(func_parser_value "${lines[4]}") +infer_is_quant=$(func_parser_value "${lines[5]}") +# parser inference +inference_py=$(func_parser_value "${lines[6]}") +use_gpu_key=$(func_parser_key "${lines[7]}") +use_gpu_list=$(func_parser_value "${lines[7]}") +use_mkldnn_key=$(func_parser_key "${lines[8]}") +use_mkldnn_list=$(func_parser_value "${lines[8]}") +cpu_threads_key=$(func_parser_key "${lines[9]}") +cpu_threads_list=$(func_parser_value "${lines[9]}") +batch_size_key=$(func_parser_key "${lines[10]}") +batch_size_list=$(func_parser_value "${lines[10]}") +use_trt_key=$(func_parser_key "${lines[11]}") +use_trt_list=$(func_parser_value "${lines[11]}") +precision_key=$(func_parser_key "${lines[12]}") +precision_list=$(func_parser_value "${lines[12]}") +infer_model_key=$(func_parser_key "${lines[13]}") +image_dir_key=$(func_parser_key "${lines[14]}") +infer_img_dir=$(func_parser_value "${lines[14]}") +save_log_key=$(func_parser_key "${lines[15]}") +benchmark_key=$(func_parser_key "${lines[16]}") +benchmark_value=$(func_parser_value "${lines[16]}") +infer_key1=$(func_parser_key "${lines[17]}") +infer_value1=$(func_parser_value "${lines[17]}") + + +LOG_PATH="./test_tipc/output" +mkdir -p ${LOG_PATH} +status_log="${LOG_PATH}/results_python.log" + + +if [ ${MODE} = "whole_infer" ]; then + GPUID=$3 + if [ ${#GPUID} -le 0 ];then + env=" " + else + env="export CUDA_VISIBLE_DEVICES=${GPUID}" + fi + # set CUDA_VISIBLE_DEVICES + eval $env + export Count=0 + IFS="|" + infer_run_exports=(${infer_export_list}) + infer_quant_flag=(${infer_is_quant}) + for infer_model in ${infer_model_dir_list[*]}; do + # run export + if [ ${infer_run_exports[Count]} != "null" ];then + save_infer_dir=$(dirname $infer_model) + set_export_weight=$(func_set_params "${export_weight}" "${infer_model}") + set_save_infer_key=$(func_set_params "${save_infer_key}" "${save_infer_dir}") + export_cmd="${python} ${infer_run_exports[Count]} ${set_export_weight} ${set_save_infer_key}" + echo ${infer_run_exports[Count]} + echo $export_cmd + eval $export_cmd + status_export=$? + status_check $status_export "${export_cmd}" "${status_log}" + else + save_infer_dir=${infer_model} + fi + #run inference + is_quant=${infer_quant_flag[Count]} + if [ ${MODE} = "klquant_infer" ]; then + is_quant="True" + fi + func_inference "${python}" "${inference_py}" "${save_infer_dir}" "${LOG_PATH}" "${infer_img_dir}" ${is_quant} + Count=$(($Count + 1)) + done +fi + diff --git a/exprements/PaddleClas/test_tipc/test_lite_arm_cpp.sh b/exprements/PaddleClas/test_tipc/test_lite_arm_cpp.sh new file mode 100644 index 00000000..c071a236 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/test_lite_arm_cpp.sh @@ -0,0 +1,159 @@ +#!/bin/bash +source ./common_func.sh +export LD_LIBRARY_PATH=${PWD}:$LD_LIBRARY_PATH + +FILENAME=$1 +dataline=$(cat $FILENAME) +# parser params +IFS=$'\n' +lines=(${dataline}) + +# parser lite inference +inference_cmd=$(func_parser_value "${lines[1]}") +runtime_device=$(func_parser_value "${lines[2]}") +det_model_list=$(func_parser_value "${lines[3]}") +rec_model_list=$(func_parser_value "${lines[4]}") +cls_model_list=$(func_parser_value "${lines[5]}") +cpu_threads_list=$(func_parser_value "${lines[6]}") +det_batch_size_list=$(func_parser_value "${lines[7]}") +rec_batch_size_list=$(func_parser_value "${lines[8]}") +infer_img_dir_list=$(func_parser_value "${lines[9]}") +config_dir=$(func_parser_value "${lines[10]}") +rec_dict_dir=$(func_parser_value "${lines[11]}") +benchmark_value=$(func_parser_value "${lines[12]}") + +if [[ $inference_cmd =~ "det" ]]; then + lite_model_list=${det_lite_model_list} +elif [[ $inference_cmd =~ "rec" ]]; then + lite_model_list=(${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) +elif [[ $inference_cmd =~ "system" ]]; then + lite_model_list=(${det_lite_model_list[*]} ${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) +else + echo "inference_cmd is wrong, please check." + exit 1 +fi + +LOG_PATH="./output" +mkdir -p ${LOG_PATH} +status_log="${LOG_PATH}/results.log" + + +function func_test_det(){ + IFS='|' + _script=$1 + _det_model=$2 + _log_path=$3 + _img_dir=$4 + _config=$5 + if [[ $_det_model =~ "slim" ]]; then + precision="INT8" + else + precision="FP32" + fi + + # lite inference + for num_threads in ${cpu_threads_list[*]}; do + for det_batchsize in ${det_batch_size_list[*]}; do + _save_log_path="${_log_path}/lite_${_det_model}_runtime_device_${runtime_device}_precision_${precision}_det_batchsize_${det_batchsize}_threads_${num_threads}.log" + command="${_script} ${_det_model} ${runtime_device} ${precision} ${num_threads} ${det_batchsize} ${_img_dir} ${_config} ${benchmark_value} > ${_save_log_path} 2>&1" + eval ${command} + status_check $? "${command}" "${status_log}" + done + done +} + +function func_test_rec(){ + IFS='|' + _script=$1 + _rec_model=$2 + _cls_model=$3 + _log_path=$4 + _img_dir=$5 + _config=$6 + _rec_dict_dir=$7 + + if [[ $_det_model =~ "slim" ]]; then + _precision="INT8" + else + _precision="FP32" + fi + + # lite inference + for num_threads in ${cpu_threads_list[*]}; do + for rec_batchsize in ${rec_batch_size_list[*]}; do + _save_log_path="${_log_path}/lite_${_rec_model}_${cls_model}_runtime_device_${runtime_device}_precision_${_precision}_rec_batchsize_${rec_batchsize}_threads_${num_threads}.log" + command="${_script} ${_rec_model} ${_cls_model} ${runtime_device} ${_precision} ${num_threads} ${rec_batchsize} ${_img_dir} ${_config} ${_rec_dict_dir} ${benchmark_value} > ${_save_log_path} 2>&1" + eval ${command} + status_check $? "${command}" "${status_log}" + done + done +} + +function func_test_system(){ + IFS='|' + _script=$1 + _det_model=$2 + _rec_model=$3 + _cls_model=$4 + _log_path=$5 + _img_dir=$6 + _config=$7 + _rec_dict_dir=$8 + if [[ $_det_model =~ "slim" ]]; then + _precision="INT8" + else + _precision="FP32" + fi + + # lite inference + for num_threads in ${cpu_threads_list[*]}; do + for det_batchsize in ${det_batch_size_list[*]}; do + for rec_batchsize in ${rec_batch_size_list[*]}; do + _save_log_path="${_log_path}/lite_${_det_model}_${_rec_model}_${_cls_model}_runtime_device_${runtime_device}_precision_${_precision}_det_batchsize_${det_batchsize}_rec_batchsize_${rec_batchsize}_threads_${num_threads}.log" + command="${_script} ${_det_model} ${_rec_model} ${_cls_model} ${runtime_device} ${_precision} ${num_threads} ${det_batchsize} ${_img_dir} ${_config} ${_rec_dict_dir} ${benchmark_value} > ${_save_log_path} 2>&1" + eval ${command} + status_check $? "${command}" "${status_log}" + done + done + done +} + + +echo "################### run test ###################" + +if [[ $inference_cmd =~ "det" ]]; then + IFS="|" + det_model_list=(${det_model_list[*]}) + + for i in {0..1}; do + #run lite inference + for img_dir in ${infer_img_dir_list[*]}; do + func_test_det "${inference_cmd}" "${det_model_list[i]}_opt.nb" "${LOG_PATH}" "${img_dir}" "${config_dir}" + done + done + +elif [[ $inference_cmd =~ "rec" ]]; then + IFS="|" + rec_model_list=(${rec_model_list[*]}) + cls_model_list=(${cls_model_list[*]}) + + for i in {0..1}; do + #run lite inference + for img_dir in ${infer_img_dir_list[*]}; do + func_test_rec "${inference_cmd}" "${rec_model}_opt.nb" "${cls_model_list[i]}_opt.nb" "${LOG_PATH}" "${img_dir}" "${rec_dict_dir}" "${config_dir}" + done + done + +elif [[ $inference_cmd =~ "system" ]]; then + IFS="|" + det_model_list=(${det_model_list[*]}) + rec_model_list=(${rec_model_list[*]}) + cls_model_list=(${cls_model_list[*]}) + + for i in {0..1}; do + #run lite inference + for img_dir in ${infer_img_dir_list[*]}; do + func_test_system "${inference_cmd}" "${det_model_list[i]}_opt.nb" "${rec_model_list[i]}_opt.nb" "${cls_model_list[i]}_opt.nb" "${LOG_PATH}" "${img_dir}" "${config_dir}" "${rec_dict_dir}" + done + done +fi diff --git a/exprements/PaddleClas/test_tipc/test_paddle2onnx.sh b/exprements/PaddleClas/test_tipc/test_paddle2onnx.sh new file mode 100644 index 00000000..850fc904 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/test_paddle2onnx.sh @@ -0,0 +1,78 @@ +#!/bin/bash +source test_tipc/common_func.sh + +FILENAME=$1 + +dataline=$(cat ${FILENAME}) +lines=(${dataline}) +# common params +model_name=$(func_parser_value "${lines[1]}") +python=$(func_parser_value "${lines[2]}") + + +# parser params +dataline=$(awk 'NR==1, NR==14{print}' $FILENAME) +IFS=$'\n' +lines=(${dataline}) + +# parser paddle2onnx +model_name=$(func_parser_value "${lines[1]}") +python=$(func_parser_value "${lines[2]}") +padlle2onnx_cmd=$(func_parser_value "${lines[3]}") +infer_model_dir_key=$(func_parser_key "${lines[4]}") +infer_model_dir_value=$(func_parser_value "${lines[4]}") +model_filename_key=$(func_parser_key "${lines[5]}") +model_filename_value=$(func_parser_value "${lines[5]}") +params_filename_key=$(func_parser_key "${lines[6]}") +params_filename_value=$(func_parser_value "${lines[6]}") +save_file_key=$(func_parser_key "${lines[7]}") +save_file_value=$(func_parser_value "${lines[7]}") +opset_version_key=$(func_parser_key "${lines[8]}") +opset_version_value=$(func_parser_value "${lines[8]}") +enable_onnx_checker_key=$(func_parser_key "${lines[9]}") +enable_onnx_checker_value=$(func_parser_value "${lines[9]}") +# parser onnx inference +inference_py=$(func_parser_value "${lines[10]}") +use_onnx_key=$(func_parser_key "${lines[11]}") +use_onnx_value=$(func_parser_value "${lines[11]}") +inference_model_dir_key=$(func_parser_key "${lines[12]}") +inference_model_dir_value=$(func_parser_value "${lines[12]}") +inference_hardware_key=$(func_parser_key "${lines[13]}") +inference_hardware_value=$(func_parser_value "${lines[13]}") + +LOG_PATH="./test_tipc/output" +mkdir -p ./test_tipc/output +status_log="${LOG_PATH}/results_paddle2onnx.log" + + +function func_paddle2onnx(){ + IFS='|' + _script=$1 + + # paddle2onnx + _save_log_path=".${LOG_PATH}/paddle2onnx_infer_cpu.log" + set_dirname=$(func_set_params "${infer_model_dir_key}" "${infer_model_dir_value}") + set_model_filename=$(func_set_params "${model_filename_key}" "${model_filename_value}") + set_params_filename=$(func_set_params "${params_filename_key}" "${params_filename_value}") + set_save_model=$(func_set_params "${save_file_key}" "${save_file_value}") + set_opset_version=$(func_set_params "${opset_version_key}" "${opset_version_value}") + set_enable_onnx_checker=$(func_set_params "${enable_onnx_checker_key}" "${enable_onnx_checker_value}") + trans_model_cmd="${padlle2onnx_cmd} ${set_dirname} ${set_model_filename} ${set_params_filename} ${set_save_model} ${set_opset_version} ${set_enable_onnx_checker}" + eval $trans_model_cmd + last_status=${PIPESTATUS[0]} + status_check $last_status "${trans_model_cmd}" "${status_log}" + # python inference + set_model_dir=$(func_set_params "${inference_model_dir_key}" "${inference_model_dir_value}") + set_use_onnx=$(func_set_params "${use_onnx_key}" "${use_onnx_value}") + set_hardware=$(func_set_params "${inference_hardware_key}" "${inference_hardware_value}") + infer_model_cmd="cd deploy && ${python} ${inference_py} -o ${set_model_dir} -o ${set_use_onnx} -o ${set_hardware} >${_save_log_path} 2>&1 && cd ../" + eval $infer_model_cmd + status_check $last_status "${infer_model_cmd}" "${status_log}" +} + + +echo "################### run test ###################" + +export Count=0 +IFS="|" +func_paddle2onnx \ No newline at end of file diff --git a/exprements/PaddleClas/test_tipc/test_serving.sh b/exprements/PaddleClas/test_tipc/test_serving.sh new file mode 100644 index 00000000..c36935a6 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/test_serving.sh @@ -0,0 +1,168 @@ +#!/bin/bash +source test_tipc/common_func.sh + +FILENAME=$1 +dataline=$(awk 'NR==1, NR==18{print}' $FILENAME) + +# parser params +IFS=$'\n' +lines=(${dataline}) + +# parser serving +model_name=$(func_parser_value "${lines[1]}") +python=$(func_parser_value "${lines[2]}") +trans_model_py=$(func_parser_value "${lines[3]}") +infer_model_dir_key=$(func_parser_key "${lines[4]}") +infer_model_dir_value=$(func_parser_value "${lines[4]}") +model_filename_key=$(func_parser_key "${lines[5]}") +model_filename_value=$(func_parser_value "${lines[5]}") +params_filename_key=$(func_parser_key "${lines[6]}") +params_filename_value=$(func_parser_value "${lines[6]}") +serving_server_key=$(func_parser_key "${lines[7]}") +serving_server_value=$(func_parser_value "${lines[7]}") +serving_client_key=$(func_parser_key "${lines[8]}") +serving_client_value=$(func_parser_value "${lines[8]}") +serving_dir_value=$(func_parser_value "${lines[9]}") +web_service_py=$(func_parser_value "${lines[10]}") +web_use_gpu_key=$(func_parser_key "${lines[11]}") +web_use_gpu_list=$(func_parser_value "${lines[11]}") +web_use_mkldnn_key=$(func_parser_key "${lines[12]}") +web_use_mkldnn_list=$(func_parser_value "${lines[12]}") +web_cpu_threads_key=$(func_parser_key "${lines[13]}") +web_cpu_threads_list=$(func_parser_value "${lines[13]}") +web_use_trt_key=$(func_parser_key "${lines[14]}") +web_use_trt_list=$(func_parser_value "${lines[14]}") +web_precision_key=$(func_parser_key "${lines[15]}") +web_precision_list=$(func_parser_value "${lines[15]}") +pipeline_py=$(func_parser_value "${lines[16]}") +image_dir_key=$(func_parser_key "${lines[17]}") +image_dir_value=$(func_parser_value "${lines[17]}") + +LOG_PATH="../../test_tipc/output" +mkdir -p ./test_tipc/output +status_log="${LOG_PATH}/results_serving.log" + +function func_serving(){ + IFS='|' + _python=$1 + _script=$2 + _model_dir=$3 + # pdserving + set_dirname=$(func_set_params "${infer_model_dir_key}" "${infer_model_dir_value}") + set_model_filename=$(func_set_params "${model_filename_key}" "${model_filename_value}") + set_params_filename=$(func_set_params "${params_filename_key}" "${params_filename_value}") + set_serving_server=$(func_set_params "${serving_server_key}" "${serving_server_value}") + set_serving_client=$(func_set_params "${serving_client_key}" "${serving_client_value}") + set_image_dir=$(func_set_params "${image_dir_key}" "${image_dir_value}") + trans_model_cmd="${python} ${trans_model_py} ${set_dirname} ${set_model_filename} ${set_params_filename} ${set_serving_server} ${set_serving_client}" + eval $trans_model_cmd + cd ${serving_dir_value} + echo $PWD + unset https_proxy + unset http_proxy + for python in ${python[*]}; do + if [ ${python} = "cpp"]; then + for use_gpu in ${web_use_gpu_list[*]}; do + if [ ${use_gpu} = "null" ]; then + web_service_cpp_cmd="${python} -m paddle_serving_server.serve --model ppocr_det_mobile_2.0_serving/ ppocr_rec_mobile_2.0_serving/ --port 9293" + eval $web_service_cmd + sleep 2s + _save_log_path="${LOG_PATH}/server_infer_cpp_cpu_pipeline_usemkldnn_False_threads_4_batchsize_1.log" + pipeline_cmd="${python} ocr_cpp_client.py ppocr_det_mobile_2.0_client/ ppocr_rec_mobile_2.0_client/" + eval $pipeline_cmd + status_check $last_status "${pipeline_cmd}" "${status_log}" + sleep 2s + ps ux | grep -E 'web_service|pipeline' | awk '{print $2}' | xargs kill -s 9 + else + web_service_cpp_cmd="${python} -m paddle_serving_server.serve --model ppocr_det_mobile_2.0_serving/ ppocr_rec_mobile_2.0_serving/ --port 9293 --gpu_id=0" + eval $web_service_cmd + sleep 2s + _save_log_path="${LOG_PATH}/server_infer_cpp_cpu_pipeline_usemkldnn_False_threads_4_batchsize_1.log" + pipeline_cmd="${python} ocr_cpp_client.py ppocr_det_mobile_2.0_client/ ppocr_rec_mobile_2.0_client/" + eval $pipeline_cmd + status_check $last_status "${pipeline_cmd}" "${status_log}" + sleep 2s + ps ux | grep -E 'web_service|pipeline' | awk '{print $2}' | xargs kill -s 9 + fi + done + else + # python serving + for use_gpu in ${web_use_gpu_list[*]}; do + echo ${ues_gpu} + if [ ${use_gpu} = "null" ]; then + for use_mkldnn in ${web_use_mkldnn_list[*]}; do + if [ ${use_mkldnn} = "False" ]; then + continue + fi + for threads in ${web_cpu_threads_list[*]}; do + set_cpu_threads=$(func_set_params "${web_cpu_threads_key}" "${threads}") + web_service_cmd="${python} ${web_service_py} ${web_use_gpu_key}=${use_gpu} ${web_use_mkldnn_key}=${use_mkldnn} ${set_cpu_threads} &" + eval $web_service_cmd + sleep 2s + for pipeline in ${pipeline_py[*]}; do + _save_log_path="${LOG_PATH}/server_infer_cpu_${pipeline%_client*}_usemkldnn_${use_mkldnn}_threads_${threads}_batchsize_1.log" + pipeline_cmd="${python} ${pipeline} ${set_image_dir} > ${_save_log_path} 2>&1 " + eval $pipeline_cmd + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${pipeline_cmd}" "${status_log}" + sleep 2s + done + ps ux | grep -E 'web_service|pipeline' | awk '{print $2}' | xargs kill -s 9 + done + done + elif [ ${use_gpu} = "0" ]; then + for use_trt in ${web_use_trt_list[*]}; do + for precision in ${web_precision_list[*]}; do + if [[ ${_flag_quant} = "False" ]] && [[ ${precision} =~ "int8" ]]; then + continue + fi + if [[ ${precision} =~ "fp16" || ${precision} =~ "int8" ]] && [ ${use_trt} = "False" ]; then + continue + fi + if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [[ ${_flag_quant} = "True" ]]; then + continue + fi + set_tensorrt=$(func_set_params "${web_use_trt_key}" "${use_trt}") + set_precision=$(func_set_params "${web_precision_key}" "${precision}") + web_service_cmd="${python} ${web_service_py} ${web_use_gpu_key}=${use_gpu} ${set_tensorrt} ${set_precision} & " + eval $web_service_cmd + + sleep 2s + for pipeline in ${pipeline_py[*]}; do + _save_log_path="${LOG_PATH}/server_infer_gpu_${pipeline%_client*}_usetrt_${use_trt}_precision_${precision}_batchsize_1.log" + pipeline_cmd="${python} ${pipeline} ${set_image_dir}> ${_save_log_path} 2>&1" + eval $pipeline_cmd + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${pipeline_cmd}" "${status_log}" + sleep 2s + done + ps ux | grep -E 'web_service|pipeline' | awk '{print $2}' | xargs kill -s 9 + done + done + else + echo "Does not support hardware other than CPU and GPU Currently!" + fi + done + fi + done +} + + +# set cuda device +GPUID=$2 +if [ ${#GPUID} -le 0 ];then + env=" " +else + env="export CUDA_VISIBLE_DEVICES=${GPUID}" +fi +set CUDA_VISIBLE_DEVICES +eval $env + + +echo "################### run test ###################" + +export Count=0 +IFS="|" +func_serving "${web_service_cmd}" diff --git a/exprements/PaddleClas/test_tipc/test_train_inference_python.sh b/exprements/PaddleClas/test_tipc/test_train_inference_python.sh new file mode 100644 index 00000000..39503dd0 --- /dev/null +++ b/exprements/PaddleClas/test_tipc/test_train_inference_python.sh @@ -0,0 +1,340 @@ +#!/bin/bash +FILENAME=$1 +source test_tipc/common_func.sh + +# MODE be one of ['lite_train_lite_infer' 'lite_train_whole_infer' 'whole_train_whole_infer', 'whole_infer', 'klquant_whole_infer'] +MODE=$2 + +dataline=$(cat ${FILENAME}) + +# parser params +IFS=$'\n' +lines=(${dataline}) + +# The training params +model_name=$(func_parser_value "${lines[1]}") +python=$(func_parser_value "${lines[2]}") +gpu_list=$(func_parser_value "${lines[3]}") +train_use_gpu_key=$(func_parser_key "${lines[4]}") +train_use_gpu_value=$(func_parser_value "${lines[4]}") +autocast_list=$(func_parser_value "${lines[5]}") +autocast_key=$(func_parser_key "${lines[5]}") +epoch_key=$(func_parser_key "${lines[6]}") +epoch_num=$(func_parser_params "${lines[6]}") +save_model_key=$(func_parser_key "${lines[7]}") +train_batch_key=$(func_parser_key "${lines[8]}") +train_batch_value=$(func_parser_value "${lines[8]}") +pretrain_model_key=$(func_parser_key "${lines[9]}") +pretrain_model_value=$(func_parser_value "${lines[9]}") +train_model_name=$(func_parser_value "${lines[10]}") +train_infer_img_dir=$(func_parser_value "${lines[11]}") +train_param_key1=$(func_parser_key "${lines[12]}") +train_param_value1=$(func_parser_value "${lines[12]}") + +trainer_list=$(func_parser_value "${lines[14]}") +trainer_norm=$(func_parser_key "${lines[15]}") +norm_trainer=$(func_parser_value "${lines[15]}") +pact_key=$(func_parser_key "${lines[16]}") +pact_trainer=$(func_parser_value "${lines[16]}") +fpgm_key=$(func_parser_key "${lines[17]}") +fpgm_trainer=$(func_parser_value "${lines[17]}") +distill_key=$(func_parser_key "${lines[18]}") +distill_trainer=$(func_parser_value "${lines[18]}") +trainer_key1=$(func_parser_key "${lines[19]}") +trainer_value1=$(func_parser_value "${lines[19]}") +trainer_key2=$(func_parser_key "${lines[20]}") +trainer_value2=$(func_parser_value "${lines[20]}") + +eval_py=$(func_parser_value "${lines[23]}") +eval_key1=$(func_parser_key "${lines[24]}") +eval_value1=$(func_parser_value "${lines[24]}") + +save_infer_key=$(func_parser_key "${lines[27]}") +export_weight=$(func_parser_key "${lines[28]}") +norm_export=$(func_parser_value "${lines[29]}") +pact_export=$(func_parser_value "${lines[30]}") +fpgm_export=$(func_parser_value "${lines[31]}") +distill_export=$(func_parser_value "${lines[32]}") +kl_quant_cmd_key=$(func_parser_key "${lines[33]}") +kl_quant_cmd_value=$(func_parser_value "${lines[33]}") +export_key2=$(func_parser_key "${lines[34]}") +export_value2=$(func_parser_value "${lines[34]}") + +# parser inference model +infer_model_dir_list=$(func_parser_value "${lines[36]}") +infer_export_flag=$(func_parser_value "${lines[37]}") +infer_is_quant=$(func_parser_value "${lines[38]}") + +# parser inference +inference_py=$(func_parser_value "${lines[39]}") +use_gpu_key=$(func_parser_key "${lines[40]}") +use_gpu_list=$(func_parser_value "${lines[40]}") +use_mkldnn_key=$(func_parser_key "${lines[41]}") +use_mkldnn_list=$(func_parser_value "${lines[41]}") +cpu_threads_key=$(func_parser_key "${lines[42]}") +cpu_threads_list=$(func_parser_value "${lines[42]}") +batch_size_key=$(func_parser_key "${lines[43]}") +batch_size_list=$(func_parser_value "${lines[43]}") +use_trt_key=$(func_parser_key "${lines[44]}") +use_trt_list=$(func_parser_value "${lines[44]}") +precision_key=$(func_parser_key "${lines[45]}") +precision_list=$(func_parser_value "${lines[45]}") +infer_model_key=$(func_parser_key "${lines[46]}") +image_dir_key=$(func_parser_key "${lines[47]}") +infer_img_dir=$(func_parser_value "${lines[47]}") +save_log_key=$(func_parser_key "${lines[48]}") +benchmark_key=$(func_parser_key "${lines[49]}") +benchmark_value=$(func_parser_value "${lines[49]}") +infer_key1=$(func_parser_key "${lines[50]}") +infer_value1=$(func_parser_value "${lines[50]}") +if [ ! $epoch_num ]; then + epoch_num=2 +fi +LOG_PATH="./test_tipc/output" +mkdir -p ${LOG_PATH} +status_log="${LOG_PATH}/results_python.log" + +function func_inference(){ + IFS='|' + _python=$1 + _script=$2 + _model_dir=$3 + _log_path=$4 + _img_dir=$5 + _flag_quant=$6 + # inference + for use_gpu in ${use_gpu_list[*]}; do + if [ ${use_gpu} = "False" ] || [ ${use_gpu} = "cpu" ]; then + for use_mkldnn in ${use_mkldnn_list[*]}; do + if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then + continue + fi + for threads in ${cpu_threads_list[*]}; do + for batch_size in ${batch_size_list[*]}; do + _save_log_path="${_log_path}/infer_cpu_usemkldnn_${use_mkldnn}_threads_${threads}_batchsize_${batch_size}.log" + set_infer_data=$(func_set_params "${image_dir_key}" "${_img_dir}") + set_benchmark=$(func_set_params "${benchmark_key}" "${benchmark_value}") + set_batchsize=$(func_set_params "${batch_size_key}" "${batch_size}") + set_cpu_threads=$(func_set_params "${cpu_threads_key}" "${threads}") + set_model_dir=$(func_set_params "${infer_model_key}" "${_model_dir}") + set_infer_params1=$(func_set_params "${infer_key1}" "${infer_value1}") + command="${_python} ${_script} ${use_gpu_key}=${use_gpu} ${use_mkldnn_key}=${use_mkldnn} ${set_cpu_threads} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} ${set_infer_params1} > ${_save_log_path} 2>&1 " + eval $command + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${command}" "../${status_log}" + done + done + done + elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then + for use_trt in ${use_trt_list[*]}; do + for precision in ${precision_list[*]}; do + if [ ${precision} = "True" ] && [ ${use_trt} = "False" ]; then + continue + fi + if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [ ${_flag_quant} = "True" ]; then + continue + fi + for batch_size in ${batch_size_list[*]}; do + _save_log_path="${_log_path}/infer_gpu_usetrt_${use_trt}_precision_${precision}_batchsize_${batch_size}.log" + set_infer_data=$(func_set_params "${image_dir_key}" "${_img_dir}") + set_benchmark=$(func_set_params "${benchmark_key}" "${benchmark_value}") + set_batchsize=$(func_set_params "${batch_size_key}" "${batch_size}") + set_tensorrt=$(func_set_params "${use_trt_key}" "${use_trt}") + set_precision=$(func_set_params "${precision_key}" "${precision}") + set_model_dir=$(func_set_params "${infer_model_key}" "${_model_dir}") + command="${_python} ${_script} ${use_gpu_key}=${use_gpu} ${set_tensorrt} ${set_precision} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} > ${_save_log_path} 2>&1 " + eval $command + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${command}" "../${status_log}" + done + done + done + else + echo "Does not support hardware other than CPU and GPU Currently!" + fi + done +} + +if [ ${MODE} = "whole_infer" ] || [ ${MODE} = "klquant_whole_infer" ]; then + IFS="|" + infer_export_flag=(${infer_export_flag}) + if [ ${infer_export_flag} != "null" ] && [ ${infer_export_flag} != "False" ]; then + rm -rf ${infer_model_dir_list/..\//} + export_cmd="${python} ${norm_export} -o Global.pretrained_model=${model_name}_pretrained -o Global.save_inference_dir=${infer_model_dir_list/..\//}" + eval $export_cmd + fi +fi + +if [ ${MODE} = "whole_infer" ]; then + GPUID=$3 + if [ ${#GPUID} -le 0 ];then + env=" " + else + env="export CUDA_VISIBLE_DEVICES=${GPUID}" + fi + # set CUDA_VISIBLE_DEVICES + eval $env + export Count=0 + cd deploy + for infer_model in ${infer_model_dir_list[*]}; do + #run inference + is_quant=${infer_quant_flag[Count]} + echo "is_quant: ${is_quant}" + func_inference "${python}" "${inference_py}" "${infer_model}" "../${LOG_PATH}" "${infer_img_dir}" ${is_quant} + Count=$(($Count + 1)) + done + cd .. + +elif [ ${MODE} = "klquant_whole_infer" ]; then + # for kl_quant + if [ ${kl_quant_cmd_value} != "null" ] && [ ${kl_quant_cmd_value} != "False" ]; then + echo "kl_quant" + command="${python} ${kl_quant_cmd_value}" + eval $command + last_status=${PIPESTATUS[0]} + status_check $last_status "${command}" "${status_log}" + cd inference/quant_post_static_model + ln -s __model__ inference.pdmodel + ln -s __params__ inference.pdiparams + cd ../../deploy + is_quant=True + func_inference "${python}" "${inference_py}" "${infer_model_dir_list}/quant_post_static_model" "../${LOG_PATH}" "${infer_img_dir}" ${is_quant} + cd .. + fi +else + IFS="|" + export Count=0 + USE_GPU_KEY=(${train_use_gpu_value}) + for gpu in ${gpu_list[*]}; do + train_use_gpu=${USE_GPU_KEY[Count]} + Count=$(($Count + 1)) + ips="" + if [ ${gpu} = "-1" ];then + env="" + elif [ ${#gpu} -le 1 ];then + env="export CUDA_VISIBLE_DEVICES=${gpu}" + eval ${env} + elif [ ${#gpu} -le 15 ];then + IFS="," + array=(${gpu}) + env="export CUDA_VISIBLE_DEVICES=${array[0]}" + IFS="|" + else + IFS=";" + array=(${gpu}) + ips=${array[0]} + gpu=${array[1]} + IFS="|" + env=" " + fi + for autocast in ${autocast_list[*]}; do + for trainer in ${trainer_list[*]}; do + flag_quant=False + if [ ${trainer} = ${pact_key} ]; then + run_train=${pact_trainer} + run_export=${pact_export} + flag_quant=True + elif [ ${trainer} = "${fpgm_key}" ]; then + run_train=${fpgm_trainer} + run_export=${fpgm_export} + elif [ ${trainer} = "${distill_key}" ]; then + run_train=${distill_trainer} + run_export=${distill_export} + elif [ ${trainer} = ${trainer_key1} ]; then + run_train=${trainer_value1} + run_export=${export_value1} + elif [[ ${trainer} = ${trainer_key2} ]]; then + run_train=${trainer_value2} + run_export=${export_value2} + else + run_train=${norm_trainer} + run_export=${norm_export} + fi + + if [ ${run_train} = "null" ]; then + continue + fi + + set_autocast=$(func_set_params "${autocast_key}" "${autocast}") + set_epoch=$(func_set_params "${epoch_key}" "${epoch_num}") + set_pretrain=$(func_set_params "${pretrain_model_key}" "${pretrain_model_value}") + set_batchsize=$(func_set_params "${train_batch_key}" "${train_batch_value}") + set_train_params1=$(func_set_params "${train_param_key1}" "${train_param_value1}") + set_use_gpu=$(func_set_params "${train_use_gpu_key}" "${train_use_gpu_value}") + if [ ${#ips} -le 26 ];then + save_log="${LOG_PATH}/${trainer}_gpus_${gpu}_autocast_${autocast}" + nodes=1 + else + IFS="," + ips_array=(${ips}) + IFS="|" + nodes=${#ips_array[@]} + save_log="${LOG_PATH}/${trainer}_gpus_${gpu}_autocast_${autocast}_nodes_${nodes}" + fi + + # load pretrain from norm training if current trainer is pact or fpgm trainer + # if [ ${trainer} = ${pact_key} ] || [ ${trainer} = ${fpgm_key} ]; then + # set_pretrain="${load_norm_train_model}" + # fi + + set_save_model=$(func_set_params "${save_model_key}" "${save_log}") + if [ ${#gpu} -le 2 ];then # train with cpu or single gpu + cmd="${python} ${run_train} ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1} " + elif [ ${#ips} -le 26 ];then # train with multi-gpu + cmd="${python} -m paddle.distributed.launch --gpus=${gpu} ${run_train} ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1}" + else # train with multi-machine + cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} ${set_use_gpu} ${set_save_model} ${set_pretrain} ${set_epoch} ${set_autocast} ${set_batchsize} ${set_train_params1}" + fi + # run train + eval "unset CUDA_VISIBLE_DEVICES" + export FLAGS_cudnn_deterministic=True + sleep 5 + eval $cmd + status_check $? "${cmd}" "${status_log}" + sleep 5 + + if [[ $FILENAME == *GeneralRecognition* ]]; then + set_eval_pretrain=$(func_set_params "${pretrain_model_key}" "${save_log}/RecModel/${train_model_name}") + else + set_eval_pretrain=$(func_set_params "${pretrain_model_key}" "${save_log}/${model_name}/${train_model_name}") + fi + # save norm trained models to set pretrain for pact training and fpgm training + if [ ${trainer} = ${trainer_norm} ]; then + load_norm_train_model=${set_eval_pretrain} + fi + # run eval + if [ ${eval_py} != "null" ]; then + set_eval_params1=$(func_set_params "${eval_key1}" "${eval_value1}") + eval_cmd="${python} ${eval_py} ${set_eval_pretrain} ${set_use_gpu} ${set_eval_params1}" + eval $eval_cmd + status_check $? "${eval_cmd}" "${status_log}" + sleep 5 + fi + # run export model + if [ ${run_export} != "null" ]; then + # run export model + save_infer_path="${save_log}" + if [[ $FILENAME == *GeneralRecognition* ]]; then + set_eval_pretrain=$(func_set_params "${pretrain_model_key}" "${save_log}/RecModel/${train_model_name}") + else + set_export_weight=$(func_set_params "${export_weight}" "${save_log}/${model_name}/${train_model_name}") + fi + set_save_infer_key=$(func_set_params "${save_infer_key}" "${save_infer_path}") + export_cmd="${python} ${run_export} ${set_export_weight} ${set_save_infer_key}" + eval $export_cmd + status_check $? "${export_cmd}" "${status_log}" + + #run inference + eval $env + save_infer_path="${save_log}" + cd deploy + func_inference "${python}" "${inference_py}" "../${save_infer_path}" "../${LOG_PATH}" "${infer_img_dir}" "${flag_quant}" + cd .. + fi + eval "unset CUDA_VISIBLE_DEVICES" + done # done with: for trainer in ${trainer_list[*]}; do + done # done with: for autocast in ${autocast_list[*]}; do + done # done with: for gpu in ${gpu_list[*]}; do +fi # end if [ ${MODE} = "infer" ]; then diff --git a/exprements/PaddleClas/tests/README.md b/exprements/PaddleClas/tests/README.md new file mode 100644 index 00000000..55b3690b --- /dev/null +++ b/exprements/PaddleClas/tests/README.md @@ -0,0 +1,70 @@ + +# 从训练到推理部署工具链测试方法介绍 + +test.sh和config文件夹下的txt文件配合使用，完成Clas模型从训练到预测的流程测试。 + +# 安装依赖 +- 安装PaddlePaddle >= 2.0 +- 安装PaddleClass依赖 + ``` + pip3 install -r ../requirements.txt + ``` +- 安装autolog + ``` + git clone https://github.com/LDOUBLEV/AutoLog + cd AutoLog + pip3 install -r requirements.txt + python3 setup.py bdist_wheel + pip3 install ./dist/auto_log-1.0.0-py3-none-any.whl + cd ../ + ``` + +# 目录介绍 + +```bash +tests/ +├── config # 测试模型的参数配置文件 +| |--- *.txt +└── prepare.sh # 完成test.sh运行所需要的数据和模型下载 +└── test.sh # 测试主程序 +``` + +# 使用方法 + +test.sh包四种运行模式，每种模式的运行数据不同，分别用于测试速度和精度，分别是： + +- 模式1：lite_train_infer，使用少量数据训练，用于快速验证训练到预测的走通流程，不验证精度和速度； +```shell +bash tests/prepare.sh ./tests/config/ResNet50_vd.txt 'lite_train_infer' +bash tests/test.sh ./tests/config/ResNet50_vd.txt 'lite_train_infer' +``` + +- 模式2：whole_infer，使用少量数据训练，一定量数据预测，用于验证训练后的模型执行预测，预测速度是否合理； +```shell +bash tests/prepare.sh ./tests/config/ResNet50_vd.txt 'whole_infer' +bash tests/test.sh ./tests/config/ResNet50_vd.txt 'whole_infer' +``` + +- 模式3：infer 不训练，全量数据预测，走通开源模型评估、动转静，检查inference model预测时间和精度; +```shell +bash tests/prepare.sh ./tests/config/ResNet50_vd.txt 'infer' +# 用法1: +bash tests/test.sh ./tests/config/ResNet50_vd.txt 'infer' +``` + +需注意的是，模型的离线量化需使用`infer`模式进行测试 + +- 模式4：whole_train_infer , CE： 全量数据训练，全量数据预测，验证模型训练精度，预测精度，预测速度； +```shell +bash tests/prepare.sh ./tests/config/ResNet50_vd.txt 'whole_train_infer' +bash tests/test.sh ./tests/config/ResNet50_vd.txt 'whole_train_infer' +``` + +- 模式5：cpp_infer , CE： 验证inference model的c++预测是否走通； +```shell +bash tests/prepare.sh ./tests/config/ResNet50_vd.txt 'cpp_infer' +bash tests/test.sh ./tests/config/ResNet50_vd.txt 'cpp_infer' +``` + +# 日志输出 +最终在```tests/output```目录下生成.log后缀的日志文件 diff --git a/exprements/PaddleClas/tests/config/DarkNet53.txt b/exprements/PaddleClas/tests/config/DarkNet53.txt new file mode 100644 index 00000000..9b9b95c8 --- /dev/null +++ b/exprements/PaddleClas/tests/config/DarkNet53.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:DarkNet53 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +infer_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/DarkNet53_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/GhostNet_x0_5.txt b/exprements/PaddleClas/tests/config/GhostNet_x0_5.txt new file mode 100644 index 00000000..b752da46 --- /dev/null +++ b/exprements/PaddleClas/tests/config/GhostNet_x0_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:GhostNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/GhostNet_x1_0.txt b/exprements/PaddleClas/tests/config/GhostNet_x1_0.txt new file mode 100644 index 00000000..c927b2bf --- /dev/null +++ b/exprements/PaddleClas/tests/config/GhostNet_x1_0.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:GhostNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/GhostNet_x1_3.txt b/exprements/PaddleClas/tests/config/GhostNet_x1_3.txt new file mode 100644 index 00000000..5766332f --- /dev/null +++ b/exprements/PaddleClas/tests/config/GhostNet_x1_3.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:GhostNet_x1_3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/HRNet_W18_C.txt b/exprements/PaddleClas/tests/config/HRNet_W18_C.txt new file mode 100644 index 00000000..c5334e38 --- /dev/null +++ b/exprements/PaddleClas/tests/config/HRNet_W18_C.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:HRNet_W18_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/HRNet_W18_C_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/LeViT_128S.txt b/exprements/PaddleClas/tests/config/LeViT_128S.txt new file mode 100644 index 00000000..6ddc0cd9 --- /dev/null +++ b/exprements/PaddleClas/tests/config/LeViT_128S.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:LeViT_128S +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +infer_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/LeViT_128S_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|Fasle +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV1.txt b/exprements/PaddleClas/tests/config/MobileNetV1.txt new file mode 100644 index 00000000..319f852d --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV1.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/MobileNetV1_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV1_x0_25.txt b/exprements/PaddleClas/tests/config/MobileNetV1_x0_25.txt new file mode 100644 index 00000000..a46a8e30 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV1_x0_25.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV1_x0_5.txt b/exprements/PaddleClas/tests/config/MobileNetV1_x0_5.txt new file mode 100644 index 00000000..bca197b7 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV1_x0_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV1_x0_75.txt b/exprements/PaddleClas/tests/config/MobileNetV1_x0_75.txt new file mode 100644 index 00000000..ba81e35c --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV1_x0_75.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV2.txt b/exprements/PaddleClas/tests/config/MobileNetV2.txt new file mode 100644 index 00000000..d4cf3fc5 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV2.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +infer_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/MobileNetV2_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV2_x0_25.txt b/exprements/PaddleClas/tests/config/MobileNetV2_x0_25.txt new file mode 100644 index 00000000..5368aef9 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV2_x0_25.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV2_x0_5.txt b/exprements/PaddleClas/tests/config/MobileNetV2_x0_5.txt new file mode 100644 index 00000000..ba35c3af --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV2_x0_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV2_x0_75.txt b/exprements/PaddleClas/tests/config/MobileNetV2_x0_75.txt new file mode 100644 index 00000000..0f5a8ab8 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV2_x0_75.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV2_x1_5.txt b/exprements/PaddleClas/tests/config/MobileNetV2_x1_5.txt new file mode 100644 index 00000000..55da65b1 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV2_x1_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV2_x1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV2_x2_0.txt b/exprements/PaddleClas/tests/config/MobileNetV2_x2_0.txt new file mode 100644 index 00000000..cd32244f --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV2_x2_0.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV2_x2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_35.txt b/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_35.txt new file mode 100644 index 00000000..bc6e77c0 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_35.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_5.txt b/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_5.txt new file mode 100644 index 00000000..693423da --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_75.txt b/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_75.txt new file mode 100644 index 00000000..24b35de5 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_large_x0_75.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_large_x1_0.txt b/exprements/PaddleClas/tests/config/MobileNetV3_large_x1_0.txt new file mode 100644 index 00000000..873072f2 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_large_x1_0.txt @@ -0,0 +1,58 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.save_inference_dir=./inference +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/MobileNetV3_large_x1_0_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================cpp_infer_params=========================== +use_gpu:0|1 +cpu_threads:1|6 +use_mkldnn:0|1 +use_tensorrt:0|1 +use_fp16:0|1 diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_large_x1_25.txt b/exprements/PaddleClas/tests/config/MobileNetV3_large_x1_25.txt new file mode 100644 index 00000000..34057be5 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_large_x1_25.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_35.txt b/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_35.txt new file mode 100644 index 00000000..0f8b75e9 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_35.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_5.txt b/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_5.txt new file mode 100644 index 00000000..0693c086 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_75.txt b/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_75.txt new file mode 100644 index 00000000..ce424273 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_small_x0_75.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_small_x1_0.txt b/exprements/PaddleClas/tests/config/MobileNetV3_small_x1_0.txt new file mode 100644 index 00000000..a1f2b8b7 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_small_x1_0.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/MobileNetV3_small_x1_25.txt b/exprements/PaddleClas/tests/config/MobileNetV3_small_x1_25.txt new file mode 100644 index 00000000..49831b66 --- /dev/null +++ b/exprements/PaddleClas/tests/config/MobileNetV3_small_x1_25.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x1_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/PPLCNet_x0_25.txt b/exprements/PaddleClas/tests/config/PPLCNet_x0_25.txt new file mode 100644 index 00000000..0777cfa4 --- /dev/null +++ b/exprements/PaddleClas/tests/config/PPLCNet_x0_25.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/PPLCNet_x0_35.txt b/exprements/PaddleClas/tests/config/PPLCNet_x0_35.txt new file mode 100644 index 00000000..0ce679e4 --- /dev/null +++ b/exprements/PaddleClas/tests/config/PPLCNet_x0_35.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/PPLCNet_x0_5.txt b/exprements/PaddleClas/tests/config/PPLCNet_x0_5.txt new file mode 100644 index 00000000..fbee882a --- /dev/null +++ b/exprements/PaddleClas/tests/config/PPLCNet_x0_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/PPLCNet_x1_0.txt b/exprements/PaddleClas/tests/config/PPLCNet_x1_0.txt new file mode 100644 index 00000000..1e991164 --- /dev/null +++ b/exprements/PaddleClas/tests/config/PPLCNet_x1_0.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/ResNeXt101_vd_64x4d.txt b/exprements/PaddleClas/tests/config/ResNeXt101_vd_64x4d.txt new file mode 100644 index 00000000..aab18d56 --- /dev/null +++ b/exprements/PaddleClas/tests/config/ResNeXt101_vd_64x4d.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:ResNeXt101_vd_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/ResNeXt101_64x4d_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/ResNet50_vd.txt b/exprements/PaddleClas/tests/config/ResNet50_vd.txt new file mode 100644 index 00000000..ed706085 --- /dev/null +++ b/exprements/PaddleClas/tests/config/ResNet50_vd.txt @@ -0,0 +1,58 @@ +===========================train_params=========================== +model_name:ResNet50_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train|pact_train|fpgm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +fpgm_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./inference +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/ResNet50_vd_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================cpp_infer_params=========================== +use_gpu:0|1 +cpu_threads:1|6 +use_mkldnn:0|1 +use_tensorrt:0|1 +use_fp16:0|1 diff --git a/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_25.txt b/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_25.txt new file mode 100644 index 00000000..1c80e4f4 --- /dev/null +++ b/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_25.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_33.txt b/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_33.txt new file mode 100644 index 00000000..34e813fb --- /dev/null +++ b/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_33.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_33 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_5.txt b/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_5.txt new file mode 100644 index 00000000..b918a780 --- /dev/null +++ b/exprements/PaddleClas/tests/config/ShuffleNetV2_x0_5.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/ShuffleNetV2_x1_0.txt b/exprements/PaddleClas/tests/config/ShuffleNetV2_x1_0.txt new file mode 100644 index 00000000..1055910a --- /dev/null +++ b/exprements/PaddleClas/tests/config/ShuffleNetV2_x1_0.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/ShuffleNetV2_x1_0_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/SwinTransformer_tiny_patch4_window7_224.txt b/exprements/PaddleClas/tests/config/SwinTransformer_tiny_patch4_window7_224.txt new file mode 100644 index 00000000..94acb150 --- /dev/null +++ b/exprements/PaddleClas/tests/config/SwinTransformer_tiny_patch4_window7_224.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:SwinTransformer_tiny_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_infer=2|whole_train_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +export1:null +export2:null +inference_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/whole_chain/SwinTransformer_tiny_patch4_window7_224_inference.tar +infer_model:../inference/ +infer_export:null +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/exprements/PaddleClas/tests/config/cpp_config.txt b/exprements/PaddleClas/tests/config/cpp_config.txt new file mode 100644 index 00000000..2f128644 --- /dev/null +++ b/exprements/PaddleClas/tests/config/cpp_config.txt @@ -0,0 +1,19 @@ +# model load config +gpu_id 0 +gpu_mem 2000 + +# whole chain test will add following config +# use_gpu 0 +# cpu_threads 10 +# use_mkldnn 1 +# use_tensorrt 0 +# use_fp16 0 + +# cls config +cls_model_path ../inference/inference.pdmodel +cls_params_path ../inference/inference.pdiparams +resize_short_size 256 +crop_size 224 + +# for log env info +benchmark 1 diff --git a/exprements/PaddleClas/tests/prepare.sh b/exprements/PaddleClas/tests/prepare.sh new file mode 100644 index 00000000..57e9a949 --- /dev/null +++ b/exprements/PaddleClas/tests/prepare.sh @@ -0,0 +1,137 @@ +#!/bin/bash +FILENAME=$1 +# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer'] +MODE=$2 + +dataline=$(cat ${FILENAME}) +# parser params +IFS=$'\n' +lines=(${dataline}) + +function func_parser_key(){ + strs=$1 + IFS=":" + array=(${strs}) + tmp=${array[0]} + echo ${tmp} +} + +function func_parser_value(){ + strs=$1 + IFS=":" + array=(${strs}) + if [ ${#array[*]} = 2 ]; then + echo ${array[1]} + else + IFS="|" + tmp="${array[1]}:${array[2]}" + echo ${tmp} + fi +} +model_name=$(func_parser_value "${lines[1]}") +model_url_value=$(func_parser_value "${lines[35]}") +model_url_key=$(func_parser_key "${lines[35]}") + +if [ ${MODE} = "lite_train_infer" ] || [ ${MODE} = "whole_infer" ];then + # pretrain lite train data + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_little_train.tar + tar xf whole_chain_little_train.tar + ln -s whole_chain_little_train ILSVRC2012 + cd ILSVRC2012 + mv train.txt train_list.txt + mv val.txt val_list.txt + if [ ${MODE} = "lite_train_infer" ];then + cp -r train/* val/ + fi + cd ../../ +elif [ ${MODE} = "infer" ] || [ ${MODE} = "cpp_infer" ];then + # download data + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_infer.tar + tar xf whole_chain_infer.tar + ln -s whole_chain_infer ILSVRC2012 + cd ILSVRC2012 + mv val.txt val_list.txt + ln -s val_list.txt train_list.txt + cd ../../ + # download inference or pretrained model + eval "wget -nc $model_url_value" + if [[ $model_url_key == *inference* ]]; then + rm -rf inference + tar xf "${model_name}_inference.tar" + fi + +elif [ ${MODE} = "whole_train_infer" ];then + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_CIFAR100.tar + tar xf whole_chain_CIFAR100.tar + ln -s whole_chain_CIFAR100 ILSVRC2012 + cd ILSVRC2012 + mv train.txt train_list.txt + mv val.txt val_list.txt + cd ../../ +fi + +if [ ${MODE} = "cpp_infer" ];then + cd deploy/cpp + echo "################### build opencv ###################" + rm -rf 3.4.7.tar.gz opencv-3.4.7/ + wget https://github.com/opencv/opencv/archive/3.4.7.tar.gz + tar -xf 3.4.7.tar.gz + install_path=$(pwd)/opencv-3.4.7/opencv3 + cd opencv-3.4.7/ + + rm -rf build + mkdir build + cd build + cmake .. \ + -DCMAKE_INSTALL_PREFIX=${install_path} \ + -DCMAKE_BUILD_TYPE=Release \ + -DBUILD_SHARED_LIBS=OFF \ + -DWITH_IPP=OFF \ + -DBUILD_IPP_IW=OFF \ + -DWITH_LAPACK=OFF \ + -DWITH_EIGEN=OFF \ + -DCMAKE_INSTALL_LIBDIR=lib64 \ + -DWITH_ZLIB=ON \ + -DBUILD_ZLIB=ON \ + -DWITH_JPEG=ON \ + -DBUILD_JPEG=ON \ + -DWITH_PNG=ON \ + -DBUILD_PNG=ON \ + -DWITH_TIFF=ON \ + -DBUILD_TIFF=ON + make -j + make install + cd ../../ + echo "################### build opencv finished ###################" + + echo "################### build PaddleClas demo ####################" + OPENCV_DIR=$(pwd)/opencv-3.4.7/opencv3/ + LIB_DIR=$(pwd)/Paddle/build/paddle_inference_install_dir/ + CUDA_LIB_DIR=$(dirname `find /usr -name libcudart.so`) + CUDNN_LIB_DIR=$(dirname `find /usr -name libcudnn.so`) + + BUILD_DIR=build + rm -rf ${BUILD_DIR} + mkdir ${BUILD_DIR} + cd ${BUILD_DIR} + cmake .. \ + -DPADDLE_LIB=${LIB_DIR} \ + -DWITH_MKL=ON \ + -DDEMO_NAME=clas_system \ + -DWITH_GPU=OFF \ + -DWITH_STATIC_LIB=OFF \ + -DWITH_TENSORRT=OFF \ + -DTENSORRT_DIR=${TENSORRT_DIR} \ + -DOPENCV_DIR=${OPENCV_DIR} \ + -DCUDNN_LIB=${CUDNN_LIB_DIR} \ + -DCUDA_LIB=${CUDA_LIB_DIR} \ + + make -j + echo "################### build PaddleClas demo finished ###################" +fi diff --git a/exprements/PaddleClas/tests/test.sh b/exprements/PaddleClas/tests/test.sh new file mode 100644 index 00000000..56af96ac --- /dev/null +++ b/exprements/PaddleClas/tests/test.sh @@ -0,0 +1,448 @@ +#!/bin/bash +FILENAME=$1 +# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer', 'cpp_infer'] +MODE=$2 + +dataline=$(cat ${FILENAME}) + +# parser params +IFS=$'\n' +lines=(${dataline}) + +function func_parser_key(){ + strs=$1 + IFS=":" + array=(${strs}) + tmp=${array[0]} + echo ${tmp} +} +function func_parser_value(){ + strs=$1 + IFS=":" + array=(${strs}) + tmp=${array[1]} + echo ${tmp} +} +function func_set_params(){ + key=$1 + value=$2 + if [ ${key} = "null" ];then + echo " " + elif [[ ${value} = "null" ]] || [[ ${value} = " " ]] || [ ${#value} -le 0 ];then + echo " " + else + echo "${key}=${value}" + fi +} +function func_parser_params(){ + strs=$1 + IFS=":" + array=(${strs}) + key=${array[0]} + tmp=${array[1]} + IFS="|" + res="" + for _params in ${tmp[*]}; do + IFS="=" + array=(${_params}) + mode=${array[0]} + value=${array[1]} + if [[ ${mode} = ${MODE} ]]; then + IFS="|" + #echo $(func_set_params "${mode}" "${value}") + echo $value + break + fi + IFS="|" + done + echo ${res} +} +function status_check(){ + last_status=$1 # the exit code + run_command=$2 + run_log=$3 + if [ $last_status -eq 0 ]; then + echo -e "\033[33m Run successfully with command - ${run_command}! \033[0m" | tee -a ${run_log} + else + echo -e "\033[33m Run failed with command - ${run_command}! \033[0m" | tee -a ${run_log} + fi +} + +IFS=$'\n' +# The training params +model_name=$(func_parser_value "${lines[1]}") +python=$(func_parser_value "${lines[2]}") +gpu_list=$(func_parser_value "${lines[3]}") +train_use_gpu_key=$(func_parser_key "${lines[4]}") +train_use_gpu_value=$(func_parser_value "${lines[4]}") +autocast_list=$(func_parser_value "${lines[5]}") +autocast_key=$(func_parser_key "${lines[5]}") +epoch_key=$(func_parser_key "${lines[6]}") +epoch_num=$(func_parser_params "${lines[6]}") +save_model_key=$(func_parser_key "${lines[7]}") +train_batch_key=$(func_parser_key "${lines[8]}") +train_batch_value=$(func_parser_params "${lines[8]}") +pretrain_model_key=$(func_parser_key "${lines[9]}") +pretrain_model_value=$(func_parser_value "${lines[9]}") +train_model_name=$(func_parser_value "${lines[10]}") +train_infer_img_dir=$(func_parser_value "${lines[11]}") +train_param_key1=$(func_parser_key "${lines[12]}") +train_param_value1=$(func_parser_value "${lines[12]}") + +trainer_list=$(func_parser_value "${lines[14]}") +trainer_norm=$(func_parser_key "${lines[15]}") +norm_trainer=$(func_parser_value "${lines[15]}") +pact_key=$(func_parser_key "${lines[16]}") +pact_trainer=$(func_parser_value "${lines[16]}") +fpgm_key=$(func_parser_key "${lines[17]}") +fpgm_trainer=$(func_parser_value "${lines[17]}") +distill_key=$(func_parser_key "${lines[18]}") +distill_trainer=$(func_parser_value "${lines[18]}") +trainer_key1=$(func_parser_key "${lines[19]}") +trainer_value1=$(func_parser_value "${lines[19]}") +trainer_key2=$(func_parser_key "${lines[20]}") +trainer_value2=$(func_parser_value "${lines[20]}") + +eval_py=$(func_parser_value "${lines[23]}") +eval_key1=$(func_parser_key "${lines[24]}") +eval_value1=$(func_parser_value "${lines[24]}") + +save_infer_key=$(func_parser_key "${lines[27]}") +export_weight=$(func_parser_key "${lines[28]}") +norm_export=$(func_parser_value "${lines[29]}") +pact_export=$(func_parser_value "${lines[30]}") +fpgm_export=$(func_parser_value "${lines[31]}") +distill_export=$(func_parser_value "${lines[32]}") +kl_quant_cmd_key=$(func_parser_key "${lines[33]}") +kl_quant_cmd_value=$(func_parser_value "${lines[33]}") +export_key2=$(func_parser_key "${lines[34]}") +export_value2=$(func_parser_value "${lines[34]}") + +# parser inference model +infer_model_dir_list=$(func_parser_value "${lines[36]}") +infer_export_flag=$(func_parser_value "${lines[37]}") +infer_is_quant=$(func_parser_value "${lines[38]}") +# parser inference +inference_py=$(func_parser_value "${lines[39]}") +use_gpu_key=$(func_parser_key "${lines[40]}") +use_gpu_list=$(func_parser_value "${lines[40]}") +use_mkldnn_key=$(func_parser_key "${lines[41]}") +use_mkldnn_list=$(func_parser_value "${lines[41]}") +cpu_threads_key=$(func_parser_key "${lines[42]}") +cpu_threads_list=$(func_parser_value "${lines[42]}") +batch_size_key=$(func_parser_key "${lines[43]}") +batch_size_list=$(func_parser_value "${lines[43]}") +use_trt_key=$(func_parser_key "${lines[44]}") +use_trt_list=$(func_parser_value "${lines[44]}") +precision_key=$(func_parser_key "${lines[45]}") +precision_list=$(func_parser_value "${lines[45]}") +infer_model_key=$(func_parser_key "${lines[46]}") +image_dir_key=$(func_parser_key "${lines[47]}") +infer_img_dir=$(func_parser_value "${lines[47]}") +save_log_key=$(func_parser_key "${lines[48]}") +benchmark_key=$(func_parser_key "${lines[49]}") +benchmark_value=$(func_parser_value "${lines[49]}") +infer_key1=$(func_parser_key "${lines[50]}") +infer_value1=$(func_parser_value "${lines[50]}") + +if [ ${MODE} = "cpp_infer" ]; then + cpp_use_gpu_key=$(func_parser_key "${lines[53]}") + cpp_use_gpu_list=$(func_parser_value "${lines[53]}") + cpp_cpu_threads_key=$(func_parser_key "${lines[54]}") + cpp_cpu_threads_list=$(func_parser_value "${lines[54]}") + cpp_use_mkldnn_key=$(func_parser_key "${lines[55]}") + cpp_use_mkldnn_list=$(func_parser_value "${lines[55]}") + cpp_use_tensorrt_key=$(func_parser_key "${lines[56]}") + cpp_use_tensorrt_list=$(func_parser_value "${lines[56]}") + cpp_use_fp16_key=$(func_parser_key "${lines[57]}") + cpp_use_fp16_list=$(func_parser_value "${lines[57]}") +fi + +LOG_PATH="./tests/output" +mkdir -p ${LOG_PATH} +status_log="${LOG_PATH}/results.log" + +function func_cpp_inference(){ + IFS='|' + _script=$1 + _log_path=$2 + _img_dir=$3 + # inference + for use_gpu in ${cpp_use_gpu_list[*]}; do + if [ ${use_gpu} = "0" ] || [ ${use_gpu} = "cpu" ]; then + for use_mkldnn in ${cpp_use_mkldnn_list[*]}; do + if [ ${use_mkldnn} = "0" ] && [ ${_flag_quant} = "True" ]; then + continue + fi + for threads in ${cpp_cpu_threads_list[*]}; do + _save_log_path="${_log_path}/cpp_infer_cpu_usemkldnn_${use_mkldnn}_threads_${threads}.log" + set_infer_data=$(func_set_params "${cpp_image_dir_key}" "${_img_dir}") + cp ../tests/config/cpp_config.txt cpp_config.txt + echo "${cpp_use_gpu_key} ${use_gpu}" >> cpp_config.txt + echo "${cpp_cpu_threads_key} ${threads}" >> cpp_config.txt + echo "${cpp_use_mkldnn_key} ${use_mkldnn}" >> cpp_config.txt + echo "${cpp_use_tensorrt_key} 0" >> cpp_config.txt + echo "${cpp_use_fp16_key} 0" >> cpp_config.txt + command="${_script} cpp_config.txt ${_img_dir} > ${_save_log_path} 2>&1 " + eval $command + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${command}" "${status_log}" + done + done + elif [ ${use_gpu} = "1" ] || [ ${use_gpu} = "gpu" ]; then + for use_trt in ${cpp_use_tensorrt_list[*]}; do + for precision in ${cpp_use_fp16_list[*]}; do + if [[ ${precision} =~ "fp16" || ${precision} =~ "int8" ]] && [ ${use_trt} = "False" ]; then + continue + fi + if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [ ${_flag_quant} = "True" ]; then + continue + fi + _save_log_path="${_log_path}/cpp_infer_gpu_usetrt_${use_trt}_precision_${precision}_batchsize_${batch_size}.log" + cp ../tests/config/cpp_config.txt cpp_config.txt + echo "${cpp_use_gpu_key} ${use_gpu}" >> cpp_config.txt + echo "${cpp_cpu_threads_key} ${threads}" >> cpp_config.txt + echo "${cpp_use_mkldnn_key} ${use_mkldnn}" >> cpp_config.txt + echo "${cpp_use_tensorrt_key} ${use_trt}" >> cpp_config.txt + echo "${cpp_use_fp16_key} ${precision}" >> cpp_config.txt + command="${_script} cpp_config.txt ${_img_dir} > ${_save_log_path} 2>&1 " + eval $command + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${command}" "${status_log}" + + done + done + else + echo "Does not support hardware other than CPU and GPU Currently!" + fi + done +} + +function func_inference(){ + IFS='|' + _python=$1 + _script=$2 + _model_dir=$3 + _log_path=$4 + _img_dir=$5 + _flag_quant=$6 + # inference + for use_gpu in ${use_gpu_list[*]}; do + if [ ${use_gpu} = "False" ] || [ ${use_gpu} = "cpu" ]; then + for use_mkldnn in ${use_mkldnn_list[*]}; do + if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then + continue + fi + for threads in ${cpu_threads_list[*]}; do + for batch_size in ${batch_size_list[*]}; do + _save_log_path="${_log_path}/infer_cpu_usemkldnn_${use_mkldnn}_threads_${threads}_batchsize_${batch_size}.log" + set_infer_data=$(func_set_params "${image_dir_key}" "${_img_dir}") + set_benchmark=$(func_set_params "${benchmark_key}" "${benchmark_value}") + set_batchsize=$(func_set_params "${batch_size_key}" "${batch_size}") + set_cpu_threads=$(func_set_params "${cpu_threads_key}" "${threads}") + set_model_dir=$(func_set_params "${infer_model_key}" "${_model_dir}") + set_infer_params1=$(func_set_params "${infer_key1}" "${infer_value1}") + command="${_python} ${_script} ${use_gpu_key}=${use_gpu} ${use_mkldnn_key}=${use_mkldnn} ${set_cpu_threads} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} ${set_infer_params1} > ${_save_log_path} 2>&1 " + eval $command + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${command}" "../${status_log}" + done + done + done + elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then + for use_trt in ${use_trt_list[*]}; do + for precision in ${precision_list[*]}; do + if [ ${precision} = "True" ] && [ ${use_trt} = "False" ]; then + continue + fi + if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [ ${_flag_quant} = "True" ]; then + continue + fi + for batch_size in ${batch_size_list[*]}; do + _save_log_path="${_log_path}/infer_gpu_usetrt_${use_trt}_precision_${precision}_batchsize_${batch_size}.log" + set_infer_data=$(func_set_params "${image_dir_key}" "${_img_dir}") + set_benchmark=$(func_set_params "${benchmark_key}" "${benchmark_value}") + set_batchsize=$(func_set_params "${batch_size_key}" "${batch_size}") + set_tensorrt=$(func_set_params "${use_trt_key}" "${use_trt}") + set_precision=$(func_set_params "${precision_key}" "${precision}") + set_model_dir=$(func_set_params "${infer_model_key}" "${_model_dir}") + command="${_python} ${_script} ${use_gpu_key}=${use_gpu} ${set_tensorrt} ${set_precision} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} > ${_save_log_path} 2>&1 " + eval $command + last_status=${PIPESTATUS[0]} + eval "cat ${_save_log_path}" + status_check $last_status "${command}" "../${status_log}" + + done + done + done + else + echo "Does not support hardware other than CPU and GPU Currently!" + fi + done +} + +if [ ${MODE} = "infer" ]; then + GPUID=$3 + if [ ${#GPUID} -le 0 ];then + env=" " + else + env="export CUDA_VISIBLE_DEVICES=${GPUID}" + fi + # set CUDA_VISIBLE_DEVICES + eval $env + export Count=0 + IFS="|" + infer_export_flag=(${infer_export_flag}) + infer_quant_flag=(${infer_is_quant}) + if [ ${infer_export_flag} != "null" ] && [ ${infer_export_flag} != "False" ]; then + rm -rf ${infer_model_dir_list/..\//} + export_cmd="${python} ${norm_export} -o Global.pretrained_model=${model_name}_pretrained -o Global.save_inference_dir=${infer_model_dir_list/..\//}" + eval $export_cmd + fi + cd deploy + for infer_model in ${infer_model_dir_list[*]}; do + #run inference + is_quant=${infer_quant_flag[Count]} + echo "is_quant: ${is_quant}" + func_inference "${python}" "${inference_py}" "${infer_model}" "../${LOG_PATH}" "${infer_img_dir}" ${is_quant} + Count=$(($Count + 1)) + done + cd .. + + # for kl_quant + if [ ${kl_quant_cmd_value} != "null" ] && [ ${kl_quant_cmd_value} != "False" ]; then + echo "kl_quant" + command="${python} ${kl_quant_cmd_value}" + eval $command + last_status=${PIPESTATUS[0]} + status_check $last_status "${command}" "${status_log}" + cd inference/quant_post_static_model + ln -s __model__ inference.pdmodel + ln -s __params__ inference.pdiparams + cd ../../deploy + is_quant=True + func_inference "${python}" "${inference_py}" "${infer_model}/quant_post_static_model" "../${LOG_PATH}" "${infer_img_dir}" ${is_quant} + cd .. + fi +elif [ ${MODE} = "cpp_infer" ]; then + cd deploy + func_cpp_inference "./cpp/build/clas_system" "../${LOG_PATH}" "${infer_img_dir}" + cd .. + +else + IFS="|" + export Count=0 + USE_GPU_KEY=(${train_use_gpu_value}) + for gpu in ${gpu_list[*]}; do + use_gpu=${USE_GPU_KEY[Count]} + Count=$(($Count + 1)) + if [ ${gpu} = "-1" ];then + env="" + elif [ ${#gpu} -le 1 ];then + env="export CUDA_VISIBLE_DEVICES=${gpu}" + eval ${env} + elif [ ${#gpu} -le 15 ];then + IFS="," + array=(${gpu}) + env="export CUDA_VISIBLE_DEVICES=${array[0]}" + IFS="|" + else + IFS=";" + array=(${gpu}) + ips=${array[0]} + gpu=${array[1]} + IFS="|" + env=" " + fi + for autocast in ${autocast_list[*]}; do + for trainer in ${trainer_list[*]}; do + flag_quant=False + if [ ${trainer} = ${pact_key} ]; then + run_train=${pact_trainer} + run_export=${pact_export} + flag_quant=True + elif [ ${trainer} = "${fpgm_key}" ]; then + run_train=${fpgm_trainer} + run_export=${fpgm_export} + elif [ ${trainer} = "${distill_key}" ]; then + run_train=${distill_trainer} + run_export=${distill_export} + elif [ ${trainer} = ${trainer_key1} ]; then + run_train=${trainer_value1} + run_export=${export_value1} + elif [[ ${trainer} = ${trainer_key2} ]]; then + run_train=${trainer_value2} + run_export=${export_value2} + else + run_train=${norm_trainer} + run_export=${norm_export} + fi + + if [ ${run_train} = "null" ]; then + continue + fi + + set_autocast=$(func_set_params "${autocast_key}" "${autocast}") + set_epoch=$(func_set_params "${epoch_key}" "${epoch_num}") + set_pretrain=$(func_set_params "${pretrain_model_key}" "${pretrain_model_value}") + set_batchsize=$(func_set_params "${train_batch_key}" "${train_batch_value}") + set_train_params1=$(func_set_params "${train_param_key1}" "${train_param_value1}") + set_use_gpu=$(func_set_params "${train_use_gpu_key}" "${train_use_gpu_value}") + save_log="${LOG_PATH}/${trainer}_gpus_${gpu}_autocast_${autocast}" + + # load pretrain from norm training if current trainer is pact or fpgm trainer + if [ ${trainer} = ${pact_key} ] || [ ${trainer} = ${fpgm_key} ]; then + set_pretrain="${load_norm_train_model}" + fi + + set_save_model=$(func_set_params "${save_model_key}" "${save_log}") + if [ ${#gpu} -le 2 ];then # train with cpu or single gpu + cmd="${python} ${run_train} ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1} " + elif [ ${#gpu} -le 15 ];then # train with multi-gpu + cmd="${python} -m paddle.distributed.launch --gpus=${gpu} ${run_train} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1}" + else # train with multi-machine + cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} ${set_save_model} ${set_pretrain} ${set_epoch} ${set_autocast} ${set_batchsize} ${set_train_params1}" + fi + # run train + eval "unset CUDA_VISIBLE_DEVICES" + export FLAGS_cudnn_deterministic=True + eval $cmd + status_check $? "${cmd}" "${status_log}" + + set_eval_pretrain=$(func_set_params "${pretrain_model_key}" "${save_log}/${$model_name}/${train_model_name}") + # save norm trained models to set pretrain for pact training and fpgm training + if [ ${trainer} = ${trainer_norm} ]; then + load_norm_train_model=${set_eval_pretrain} + fi + # run eval + if [ ${eval_py} != "null" ]; then + set_eval_params1=$(func_set_params "${eval_key1}" "${eval_value1}") + eval_cmd="${python} ${eval_py} ${set_eval_pretrain} ${set_use_gpu} ${set_eval_params1}" + eval $eval_cmd + status_check $? "${eval_cmd}" "${status_log}" + fi + # run export model + if [ ${run_export} != "null" ]; then + # run export model + save_infer_path="${save_log}" + set_export_weight=$(func_set_params "${export_weight}" "${save_log}/${model_name}/${train_model_name}") + set_save_infer_key=$(func_set_params "${save_infer_key}" "${save_infer_path}") + export_cmd="${python} ${run_export} ${set_export_weight} ${set_save_infer_key}" + eval $export_cmd + status_check $? "${export_cmd}" "${status_log}" + + #run inference + eval $env + save_infer_path="${save_log}" + cd deploy + func_inference "${python}" "${inference_py}" "../${save_infer_path}" "../${LOG_PATH}" "${infer_img_dir}" "${flag_quant}" + cd .. + fi + eval "unset CUDA_VISIBLE_DEVICES" + done # done with: for trainer in ${trainer_list[*]}; do + done # done with: for autocast in ${autocast_list[*]}; do + done # done with: for gpu in ${gpu_list[*]}; do +fi # end if [ ${MODE} = "infer" ]; then diff --git a/exprements/PaddleClas/tools/__init__.py b/exprements/PaddleClas/tools/__init__.py new file mode 100644 index 00000000..cdefb805 --- /dev/null +++ b/exprements/PaddleClas/tools/__init__.py @@ -0,0 +1,15 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from . import infer \ No newline at end of file diff --git a/exprements/PaddleClas/tools/eval.py b/exprements/PaddleClas/tools/eval.py new file mode 100644 index 00000000..e086da1b --- /dev/null +++ b/exprements/PaddleClas/tools/eval.py @@ -0,0 +1,31 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +from ppcls.utils import config +from ppcls.engine.engine import Engine + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config( + args.config, overrides=args.override, show=False) + engine = Engine(config, mode="eval") + engine.eval() diff --git a/exprements/PaddleClas/tools/eval.sh b/exprements/PaddleClas/tools/eval.sh new file mode 100644 index 00000000..c13ea6d0 --- /dev/null +++ b/exprements/PaddleClas/tools/eval.sh @@ -0,0 +1,7 @@ +#!/usr/bin/env bash + +# for single card eval +# python3.7 tools/eval.py -c ./ppcls/configs/ImageNet/ResNet/ResNet50.yaml + +# for multi-cards eval +python3.7 -m paddle.distributed.launch --gpus="0,1,2,3" tools/eval.py -c ./ppcls/configs/ImageNet/ResNet/ResNet50.yaml diff --git a/exprements/PaddleClas/tools/export_model.py b/exprements/PaddleClas/tools/export_model.py new file mode 100644 index 00000000..01aba06c --- /dev/null +++ b/exprements/PaddleClas/tools/export_model.py @@ -0,0 +1,34 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +import paddle +import paddle.nn as nn + +from ppcls.utils import config +from ppcls.engine.engine import Engine + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config( + args.config, overrides=args.override, show=False) + engine = Engine(config, mode="export") + engine.export() diff --git a/exprements/PaddleClas/tools/infer.py b/exprements/PaddleClas/tools/infer.py new file mode 100644 index 00000000..4f6bf927 --- /dev/null +++ b/exprements/PaddleClas/tools/infer.py @@ -0,0 +1,31 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +from ppcls.utils import config +from ppcls.engine.engine import Engine + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config( + args.config, overrides=args.override, show=False) + engine = Engine(config, mode="infer") + engine.infer() diff --git a/exprements/PaddleClas/tools/train.py b/exprements/PaddleClas/tools/train.py new file mode 100644 index 00000000..e7c9d7bc --- /dev/null +++ b/exprements/PaddleClas/tools/train.py @@ -0,0 +1,32 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../'))) + +from ppcls.utils import config +from ppcls.engine.engine import Engine + +if __name__ == "__main__": + args = config.parse_args() + config = config.get_config( + args.config, overrides=args.override, show=False) + config.profiler_options = args.profiler_options + engine = Engine(config, mode="train") + engine.train() diff --git a/exprements/PaddleClas/tools/train.sh b/exprements/PaddleClas/tools/train.sh new file mode 100644 index 00000000..412f1081 --- /dev/null +++ b/exprements/PaddleClas/tools/train.sh @@ -0,0 +1,8 @@ +#!/usr/bin/env bash + +# for single card train +# python3.7 tools/train.py -c ./ppcls/configs/ImageNet/ResNet/ResNet50.yaml + +# for multi-cards train +export CUDA_VISIBLE_DEVICES=0,1,2,3 +python3.7 -m paddle.distributed.launch --gpus="0,1,2,3" tools/train.py -c ./ppcls/configs/ImageNet/ResNet/ResNet50.yaml