zuiwanting/oneflow

OneFlow is a performance-centered and open-source deep learning framework.

OneFlow

OneFlow is a performance-centered and open-source deep learning framework.

Latest News

• Version 0.7.0 is out!
  • Introducing global tensor
  • Semi-auto parallelization has landed
  • Full changelog

Publication

• OneFlow: Redesign the Distributed Deep Learning Framework from Scratch

• Bibtex Citation

  @misc{yuan2021oneflow,
        title={OneFlow: Redesign the Distributed Deep Learning Framework from Scratch},
        author={Jinhui Yuan and Xinqi Li and Cheng Cheng and Juncheng Liu and Ran Guo and Shenghang Cai and Chi Yao and Fei Yang and Xiaodong Yi and Chuan Wu and Haoran Zhang and Jie Zhao},
        year={2021},
        eprint={2110.15032},
        archivePrefix={arXiv},
        primaryClass={cs.DC}
  }
  

Install OneFlow

System Requirements

• Python 3.6, 3.7, 3.8, 3.9, 3.10

• (Highly recommended) Upgrade pip

  python3 -m pip install --upgrade pip #--user
  

• CUDA Toolkit Linux x86_64 Driver

  • CUDA runtime is statically linked into OneFlow. OneFlow will work on a minimum supported driver, and any driver beyond. For more information, please refer to CUDA compatibility documentation.

  • Please upgrade your Nvidia driver to version 440.33 or above and install OneFlow for CUDA 10.2 if possible.

Install with Pip Package

• To install latest stable release of OneFlow with CUDA support:

  python3 -m pip install -f https://release.oneflow.info oneflow==0.7.0+cu102

• To install nightly release of OneFlow with CUDA support:

  python3 -m pip install --pre oneflow -f https://staging.oneflow.info/branch/master/cu102

• To install other available builds for different variants:

  • Stable

    python3 -m pip install --find-links https://release.oneflow.info oneflow==0.7.0+[PLATFORM]

  • Nightly

    python3 -m pip install --pre oneflow -f https://staging.oneflow.info/branch/master/[PLATFORM]
    

  • All available [PLATFORM]:

    Platform	CUDA Driver Version	Supported GPUs
    cu112	>= 450.80.02	GTX 10xx, RTX 20xx, A100, RTX 30xx
    cu102	>= 440.33	GTX 10xx, RTX 20xx
    cpu	N/A	N/A

• If you are in China, you could run this to have pip download packages from domestic mirror of pypi:

  python3 -m pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
  

  For more information on this, please refer to pypi 镜像使用帮助

Use docker image

docker pull oneflowinc/oneflow:nightly-cuda10.2
docker pull oneflowinc/oneflow:nightly-cuda11.2

Build from Source

Clone Source Code

• Option 1: Clone source code from GitHub

  git clone https://github.com/Oneflow-Inc/oneflow --depth=1

• Option 2: Download from Aliyun

  If you are in China, please download OneFlow source code from: https://oneflow-public.oss-cn-beijing.aliyuncs.com/oneflow-src.zip

  curl https://oneflow-public.oss-cn-beijing.aliyuncs.com/oneflow-src.zip -o oneflow-src.zip
  unzip oneflow-src.zip

Build OneFlow

• Option 1: Build with Conda (recommended)

  Please refer to this repo

• Option 2: Build in docker container (recommended)

  • Pull the docker image:

    docker pull oneflowinc/manylinux2014_x86_64_cuda11.2

  • Follow the instructions in the bare metal build guide below.

• Option 3: Build on bare metal

  • Install dependencies (not required if you are using docker):

    • on Ubuntu 20.04, run:

      sudo apt install -y libopenblas-dev nasm g++ gcc python3-pip cmake autoconf libtool
      

    • on macOS, run:

      brew install nasm
      

  • In the root directory of OneFlow source code, run:

    mkdir build
    cd build
    

  • Config the project, inside build directory:

    • If you are in China

      run this to config for CUDA:

      cmake .. -C ../cmake/caches/cn/cuda.cmake
      

      run this to config for CPU-only:

      cmake .. -C ../cmake/caches/cn/cpu.cmake
      

    • If you are not in China

      run this to config for CUDA:

      cmake .. -C ../cmake/caches/international/cuda.cmake
      

      run this to config for CPU-only:

      cmake .. -C ../cmake/caches/international/cpu.cmake
      

  • Build the project, inside build directory, run:

    make -j$(nproc)
    

  • Add oneflow to your PYTHONPATH, inside build directory, run:

    source source.sh
    

    Please note that this change is not permanent.

  • Simple validation

    python3 -m oneflow --doctor
    

Troubleshooting

Please refer to troubleshooting for common issues you might encounter when compiling and running OneFlow.

Advanced features

XRT

• You can check this doc to obtain more details about how to use XLA and TensorRT with OneFlow.

Getting Started

• Please refer to QUICKSTART
• 中文版请参见 快速上手

Documentation

• API Reference
• Usage & Design Docs
• System Design

Model Zoo and Benchmark

• Libai(Toolbox for Parallel Training Large-Scale Transformer Models)
  • BERT-large
  • GPT
  • T5
  • VisionTransformer
  • SwinTransformer
• FlowVision(Toolbox for Computer Vision Datasets, SOTA Models and Utils)
• OneFlow-Models(Examples of How to Implement Models in Various Fields with OneFlow)
  • ResNet-50
  • Wide&Deep
• OneFlow-Benchmark(Outdated)

Communication

• GitHub issues: any install, bug, feature issues.

• www.oneflow.org: brand related information.

• 中文

  • QQ 群: 331883
  • 微信号（加好友入交流群）: OneFlowXZS
  • 知乎

• International

  • Discord
  • Twitter
  • LinkedIn
  • Medium

The Team

OneFlow was originally developed by OneFlow Inc and Zhejiang Lab.

License

Apache License 2.0