InternLM/doc/install.md

1.7 KiB
Raw Blame History

InternLM项目的依赖安装

环境准备

首先,需要安装的依赖包及对应版本列表如下:

  • Python == 3.10
  • GCC == 10.2.0
  • MPFR == 4.1.0
  • CUDA == 11.7
  • Pytorch == 1.13.1+cu117
  • Transformers >= 4.25.1
  • Flash-Attention == 23.05
  • Ampere或者Hopper架构的GPU (例如H100, A100)
  • Linux OS

以上依赖包安装完成后,需要更新配置系统环境变量:

export CUDA_PATH={path_of_cuda_11.7}
export GCC_HOME={path_of_gcc_10.2.0}
export MPFR_HOME={path_of_mpfr_4.1.0}
export LD_LIBRARY_PATH=${GCC_HOME}/lib64:${MPFR_HOME}/lib:${CUDA_PATH}/lib64:$LD_LIBRARY_PATH
export PATH=${GCC_HOME}/bin:${CUDA_PATH}/bin:$PATH
export CC=${GCC_HOME}/bin/gcc
export CXX=${GCC_HOME}/bin/c++

环境安装

将项目internlm及其依赖子模块,从 github 仓库中 clone 下来,命令如下:

git clone git@github.com:InternLM/InternLM.git --recurse-submodules

推荐使用 conda 构建一个 Python-3.10 的虚拟环境, 并基于requirements/文件安装项目所需的依赖包:

conda create --name internlm-env python=3.10 -y
conda activate internlm-env
cd internlm
pip install -r requirements/torch.txt 
pip install -r requirements/runtime.txt 

安装 flash-attention (version v1.0.5)

cd ./third_party/flash-attention
python setup.py install
cd ./csrc
cd fused_dense_lib && pip install -v .
cd ../xentropy && pip install -v .
cd ../rotary && pip install -v .
cd ../layer_norm && pip install -v .
cd ../../../../

安装 Apex (version 23.05)

cd ./third_party/apex
pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
cd ../../