Skip to content

Latest commit

 

History

History
65 lines (45 loc) · 1.95 KB

build_from_source.md

File metadata and controls

65 lines (45 loc) · 1.95 KB

Build BladeDISC from Source

This document introduces how to build BladeDISC from source, to make the installation and configuration easier, we provide a Dockerfile that contains required software to build and test.

Prerequisite

  1. Git for checking out the source code.
  2. Nvidia Docker to launch Docker container on Nvidia GPU host.

Checkout the Source

git clone git@github.com:alibaba/BladeDISC.git
cd BladeDISC && git submodule update --init --recursive

Building BladeDISC for TensorFlow Users

  • step1: launch a development Docker container that runs a development Docker image.

    docker run --gpus all --rm -it -v $PWD:/disc bladedisc/bladedisc:latest-devel-cuda11.0 bash

    please goto this website to find more images with various CUDA versions.

  • step2: build and test tensorflow_bladedisc with an all-in-on bash script.

    bash ./scripts/ci/build_and_test.sh

    the above command generates a wheel Python package on the path: ./build, please free feel to install it with the pip installation toolkit.

Building BladeDISC for PyTorch Users

  • step1: launch a development Docker container that runs a development Docker image.

    docker run --gpus all --rm -it -v $PWD:/disc bladedisc/bladedisc:latest-devel-cuda11.0 bash

  • step2: build and test pytorch_blade with an all-in-one script:

    cd pytorch_blade && bash ./ci_build/build_pytorch_blade.sh

  • step3: build the pytorch_blade Python wheel package.

    python setup.py bdist_wheel

    the above command generates a wheel Python package on the path: pytorch_blade/dist/, please install it with the pip installation toolkit.