/Tf-Docker-Build

Build up-to-date tensorflow with docker

Primary LanguageDockerfileMIT LicenseMIT

Tensorflow Docker Building Setup

A helper for setting up up-to-date Tensorflow building environment with Docker. Provided with both CPU and GPU support.

Notification

  1. This repository was tested under:
    • tensorflow version: 1.13.1
    • python version: 3.7.3
    • bazel version: 0.24.1
  2. To get the whole picture of building tensorflow, check the official site.

Pre-build image available

It's better approach to build up-to-date images by yourself, by using the Dockerfile inside this repository, though, there is pre-built image available in the docker hub, which may slightly shorten the build time.

Check here to learn more.

Overall build process

  1. Make folder to store built wheel
  2. Pull pre-built image, or build the docker image locally
  3. Run the docker container
  4. Build tensorflow inside the container
  5. Install tensorflow for the host machine

Detailed build process

Check here for detailed process!

Trouble shooting

Wheel file not found in the out folder

You need to copy file from container to host, use docker cp command:

docker cp <containerId>:/file/path/within/container /host/path/target