feature: template

Author: cnstark

README_CN.md

@@ -54,11 +54,29 @@ pip install -r requirements.txt
 
 ## 开始使用
 
-### 初始化git仓库
+### 新建项目文件夹并初始化git仓库
+
+```shell
+mkdir my_deeplearning_project
+cd my_deeplearning_project
+git init
+```
 
 ### 添加easytorch子仓库
 
-### 
+```shell
+git submodule add https://github.com/cnstark/easytorch.git
+git add .
+git commit -m "init by easytorch"
+```
+
+### 复制EasyTorch模板至工作目录
+
+```shell
+cp easytorch/examples/template/* .
+```
+
+*接下来就可以使用EasyTorch构建你的深度学习项目。*
 
 ## README 徽章
 

examples/template/configs/__init__.py

@@ -0,0 +1,39 @@
+import os
+from easydict import EasyDict
+
+from runners.runner_template import RunnerTemplate
+
+CFG = EasyDict()
+
+CFG.DESC = 'config template'
+CFG.RUNNER = RunnerTemplate
+CFG.GPU_NUM = 1
+
+CFG.MODEL = EasyDict()
+CFG.MODEL.NAME = 'model_template'
+
+CFG.TRAIN = EasyDict()
+
+CFG.TRAIN.NUM_EPOCHS = 100
+CFG.TRAIN.CKPT_SAVE_DIR = os.path.join(
+    'checkpoints',
+    '_'.join([CFG.MODEL.NAME, str(CFG.TRAIN.NUM_EPOCHS)])
+)
+CFG.TRAIN.CKPT_SAVE_STRATEGY = None
+
+CFG.TRAIN.OPTIM = EasyDict()
+CFG.TRAIN.OPTIM.TYPE = 'SGD'
+CFG.TRAIN.OPTIM.PARAM = {
+    'lr': 0.002,
+    'momentum': 0.1,
+}
+
+CFG.TRAIN.DATA = EasyDict()
+CFG.TRAIN.DATA.BATCH_SIZE = 4
+CFG.TRAIN.DATA.SHUFFLE = True
+
+CFG.VAL = EasyDict()
+
+CFG.VAL.INTERVAL = 1
+
+CFG.VAL.DATA = EasyDict()

examples/template/configs/config_template.py

@@ -0,0 +1,7 @@
+from .model_template import ModelTemplate
+
+
+MODEL_DICT = {
+    'model_template': ModelTemplate
+    # other models...
+}

examples/template/models/__init__.py

@@ -0,0 +1,13 @@
+from torch import nn
+
+
+class ModelTemplate(nn.Module):
+    def __init__(self):
+        super().__init__()
+        self.op = nn.Sequential(
+            nn.Conv2d(1, 20, 5),
+            nn.ReLU(inplace=True)
+        )
+
+    def forward(self, x):
+        return self.op(x)

examples/template/models/model_template.py

@@ -0,0 +1,62 @@
+import torch
+from torch import nn
+from torch.utils.data import Dataset
+
+from easytorch import Runner
+
+from ..models import MODEL_DICT
+
+
+class RunnerTemplate(Runner):
+    def __init__(self, cfg: dict):
+        super().__init__(cfg)
+
+    def init_training(self, cfg: dict):
+        super().init_training(cfg)
+
+        # init loss
+        # e.g.
+        # self.loss = nn.MSELoss()
+        # self.loss = self.to_running_device(self.loss)
+
+        # register meters by calling:
+        # self.register_epoch_meter('train_loss', 'train', '{:.2f}')
+
+    def init_validation(self, cfg: dict):
+        super().init_validation(cfg)
+
+        # self.register_epoch_meter('val_acc', 'val', '{:.2f}%')
+
+    @staticmethod
+    def define_model(cfg: dict) -> nn.Module:
+        return MODEL_DICT[cfg['MODEL']['NAME']](**cfg['MODEL'].get('PARAM', {}))
+
+    @staticmethod
+    def build_train_dataset(cfg: dict) -> Dataset:
+        # return your train Dataset
+        pass
+
+    @staticmethod
+    def build_val_dataset(cfg: dict):
+        # return your val Dataset
+        pass
+
+    def train_iters(self, epoch: int, iter_index: int, data: torch.Tensor or tuple) -> torch.Tensor:
+        # forward and compute loss
+        # update meters if necessary
+        # return loss (will be auto backward and update params) or don't return any thing
+
+        # e.g.
+        # _input, _target = data
+        # _input = self.to_running_device(_input)
+        # _target = self.to_running_device(_target)
+        #
+        # output = self.model(_input)
+        # loss = self.loss(output, _target)
+        # self.update_epoch_meter('train_loss', loss.item())
+        # return loss
+        pass
+
+    def val_iters(self, iter_index: int, data: torch.Tensor or tuple):
+        # forward and update meters
+        pass

examples/template/runners/__init__.py

@@ -0,0 +1,17 @@
+from argparse import ArgumentParser
+
+from easytorch.easytorch import launch_training
+
+
+def parse_args():
+    parser = ArgumentParser(description='Welcome to EasyTorch!')
+    parser.add_argument('-c', '--cfg', help='training config', required=True)
+    parser.add_argument('--gpus', help='visible gpus', type=str)
+    parser.add_argument('--tf32', help='enable tf32 on Ampere device', action='store_true')
+    return parser.parse_args()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+
+    launch_training(args.cfg, args.gpus, args.tf32)