安装CUDA Toolkit

异常信息

在执行pip install flash_attn，安装一个推理加速库的时候，遇到如下异常：

Looking in indexes: https://mirrors.aliyun.com/pypi/simple/
Collecting flash_attn
  Downloading https://mirrors.aliyun.com/pypi/packages/72/94/06f618bb338ec7203b48ac542e73087362b7750f9c568b13d213a3f181bb/flash_attn-2.5.8.tar.gz (2.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.5/2.5 MB 1.6 MB/s eta 0:00:00
  Preparing metadata (setup.py) ... error
  error: subprocess-exited-with-error
  
  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [20 lines of output]
      fatal: not a git repository (or any of the parent directories): .git
      /tmp/pip-install-fg7pt8f4/flash-attn_1e4c76d3ba9f4a5d968930613e3c4bd7/setup.py:78: UserWarning: flash_attn was requested, but nvcc was not found.  Are you sure your environment has nvcc available?  If you're installing within a container from https://hub.docker.com/r/pytorch/pytorch, only images whose names contain 'devel' will provide nvcc.
        warnings.warn(
      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "/tmp/pip-install-fg7pt8f4/flash-attn_1e4c76d3ba9f4a5d968930613e3c4bd7/setup.py", line 134, in <module>
          CUDAExtension(
        File "/usr/local/program/miniconda3/envs/llama3/lib/python3.10/site-packages/torch/utils/cpp_extension.py", line 1077, in CUDAExtension
          library_dirs += library_paths(cuda=True)
        File "/usr/local/program/miniconda3/envs/llama3/lib/python3.10/site-packages/torch/utils/cpp_extension.py", line 1204, in library_paths
          if (not os.path.exists(_join_cuda_home(lib_dir)) and
        File "/usr/local/program/miniconda3/envs/llama3/lib/python3.10/site-packages/torch/utils/cpp_extension.py", line 2419, in _join_cuda_home
          raise OSError('CUDA_HOME environment variable is not set. '
      OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root.
      
      
      torch.__version__  = 2.3.0+cu121
      
      
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed

× Encountered error while generating package metadata.
╰─> See above for output.

note: This is an issue with the package mentioned above, not pip.
hint: See above for details.

分析

首先操作系统已经安装了驱动，并且驱动自带CUDA，可通过nvidia-smi命令查看

注意：

当时看到这里是有疑惑的，GPU显卡上已经有了CUDA，为何还提示需要CUDA？

原因如下：

首先CUDA有两个主要的API，runtime API和driver API。显然GPU显卡中的CUDA对应driver API，那么此时出现这个异常提示需要CUDA信息，很显然这个CUDA需要的就是runtime API，因此为了支持runtime API，就需要额外再安装CUDA Toolkit

解决异常：

CUDA Toolkit的安装路径通常在usr/local/路径下，经检查发现该路径下确实不存在CUDA Toolkit的安装目录

既然没有安装CUDA Toolkit，那么直接安装CUDA Toolkit来尝试解决这个问题。

下载CUDA

CUDA Toolkit是CUDA的工具包，安装CUDA其实就是安装CUDA Toolkit。

访问https://developer.nvidia.com/cuda-toolkit-archive，选择需要的CUDA版本

为了兼容性，执行nvidia-smi命令，查看GPU的驱动与CUDA版本

由于GPU自身CUDA版本是12.2，因此这里选择下载CUDA Toolkit 12.2。

这里选择：Linux系统、x86_64架构、Ubuntu系统、系统版本22.04、runfile(local)安装方式

同时页面下方也给出了安装说明

执行安装

wget https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda_12.2.0_535.54.03_linux.run

sudo sh cuda_12.2.0_535.54.03_linux.run

选择Continue后回车

输入accept接受

因为安装了Drive驱动，所以取消安装，默认勾选（x），取消后选择Install进行安装。

出现如下日志，表示安装成功

===========
= Summary =
===========

Driver:   Not Selected
Toolkit:  Installed in /usr/local/cuda-12.2/

Please make sure that
 -   PATH includes /usr/local/cuda-12.2/bin
 -   LD_LIBRARY_PATH includes /usr/local/cuda-12.2/lib64, or, add /usr/local/cuda-12.2/lib64 to /etc/ld.so.conf and run ldconfig as root

To uninstall the CUDA Toolkit, run cuda-uninstaller in /usr/local/cuda-12.2/bin
***WARNING: Incomplete installation! This installation did not install the CUDA Driver. A driver of version at least 535.00 is required for CUDA 12.2 functionality to work.
To install the driver using this installer, run the following command, replacing <CudaInstaller> with the name of this run file:
    sudo <CudaInstaller>.run --silent --driver

Logfile is /var/log/cuda-installer.log

配置环境变量

编辑vim ~/.bashrc文件,配置环境变量，参考官方文档： Environment Setup

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64
export PATH=$PATH:/usr/local/cuda/bin
export CUDA_HOME=$CUDA_HOME:/usr/local/cuda

验证

执行nvcc -V命令，查看cuda是否安装成功

CUDA NVCC就是CUDA的编译器，可以从CUDA Toolkit的/bin目录中获取，类似于gcc就是c语言的编译器

root@master:~# nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Tue_Jun_13_19:16:58_PDT_2023
Cuda compilation tools, release 12.2, V12.2.91
Build cuda_12.2.r12.2/compiler.32965470_0