Skip to content

Latest commit

 

History

History
384 lines (300 loc) · 19.7 KB

README.md

File metadata and controls

384 lines (300 loc) · 19.7 KB

TFG-PointCloud Custom Ops

This repository is based on the TensorFlow-custom-op repository, for a detailed guide about how to add an op we refer to this template.

Setting up the environment

The c++-toolchains are dependent on the latest tensorflow-custom-op docker container, to set it up run (tested with tensorflow/tensorflow:2.2.0-custom-op-gpu-ubuntu16)

  docker pull tensorflow/tensorflow:custom-op-gpu-ubuntu16
  sudo docker run --gpus all --privileged -it -v ${PWD}:/working_dir -w /working_dir tensorflow/tensorflow:custom-op-gpu-ubuntu16
  root@docker: ./configure.sh

and answer the first question with yes y, and specify the TensorFlow version you want to build for. (tested with TF version 2.1, 2.2, 2.3)

Building the PIP-Package

To compile the pip package run

  bazel build build_pip_pkg
  bazel-bin/build_pip_pkg artifacts

The package .whl is located in artifacts/, by default it should be a python3 package. (tested with python3.6)

To install the package via pip run

  pip3 install artifacts/*.whl

To test out the package run

  cd ..
  python3 -c 'import tfg_custom_ops'

Additional Information

You may use this software under the Apache 2.0 License.