【WSL2笔记4】 PyTorch深度学习框架的多环境CUDA安装配置

1、 Conda多环境下的不同CUDA版本实现

1.1 实现思路

1.1.1 根环境(base)安装最新独立版本CUDA ToolKit工具包

NVIDIA CUDA ToolKit
https://developer.nvidia.com/cuda-downloads

安装与显卡驱动匹配的最新版本CUDA ToolKit

详见【WSL2笔记2】 搭建深度学习开发环境
https://blog.csdn.net/fuweipeng2012/article/details/129980680

1.1.2 在不同的Python虚拟子环境，根据需求，安装相应版本的CUDA和cuDDN

conda install cudatoolkit=版本号

Conda 安装的CUDA ToolKit 包含了CUDA和cuDDN
Conda虚拟环境调用的是子环境下的CUDA ToolKit的CUDA/cuDDN,不是调用根环境版本的CUDA工具包
虚拟子环境CUDA ToolKit版本一定要低于本机根环境的独立CUDA版本

2、 Python 3.10 + PyTorch 2.0.0 + CUDA 11.8

2.1 创建Python310环境

创建
conda create -n py310 python=3.10
激活
conda activate py310

(base) xf@VP01:~/ai/env$ conda create -n py310 python=3.10
Collecting package metadata (current_repodata.json): done
Solving environment: done

## Package Plan ##

  environment location: /home/xf/anaconda3/envs/py310

  added / updated specs:
    - python=3.10


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    certifi-2022.12.7          |  py310h06a4308_0         150 KB  defaults
    pip-23.0.1                 |  py310h06a4308_0         2.6 MB  defaults
    python-3.10.10             |       h7a1cb2a_2        26.9 MB  defaults
    setuptools-65.6.3          |  py310h06a4308_0         1.2 MB  defaults
    tzdata-2023c               |       h04d1e81_0         116 KB  defaults
    wheel-0.38.4               |  py310h06a4308_0          64 KB  defaults
    ------------------------------------------------------------
                                           Total:        31.0 MB

The following NEW packages will be INSTALLED:

  _libgcc_mutex      anaconda/pkgs/main/linux-64::_libgcc_mutex-0.1-main
  _openmp_mutex      anaconda/pkgs/main/linux-64::_openmp_mutex-5.1-1_gnu
  bzip2              anaconda/pkgs/main/linux-64::bzip2-1.0.8-h7b6447c_0
  ca-certificates    anaconda/pkgs/main/linux-64::ca-certificates-2023.01.10-h06a4308_0
  certifi            anaconda/pkgs/main/linux-64::certifi-2022.12.7-py310h06a4308_0
  ld_impl_linux-64   anaconda/pkgs/main/linux-64::ld_impl_linux-64-2.38-h1181459_1
  libffi             anaconda/pkgs/main/linux-64::libffi-3.4.2-h6a678d5_6
  libgcc-ng          anaconda/pkgs/main/linux-64::libgcc-ng-11.2.0-h1234567_1
  libgomp            anaconda/pkgs/main/linux-64::libgomp-11.2.0-h1234567_1
  libstdcxx-ng       anaconda/pkgs/main/linux-64::libstdcxx-ng-11.2.0-h1234567_1
  libuuid            anaconda/pkgs/main/linux-64::libuuid-1.41.5-h5eee18b_0
  ncurses            anaconda/pkgs/main/linux-64::ncurses-6.4-h6a678d5_0
  openssl            anaconda/pkgs/main/linux-64::openssl-1.1.1t-h7f8727e_0
  pip                anaconda/pkgs/main/linux-64::pip-23.0.1-py310h06a4308_0
  python             anaconda/pkgs/main/linux-64::python-3.10.10-h7a1cb2a_2
  readline           anaconda/pkgs/main/linux-64::readline-8.2-h5eee18b_0
  setuptools         anaconda/pkgs/main/linux-64::setuptools-65.6.3-py310h06a4308_0
  sqlite             anaconda/pkgs/main/linux-64::sqlite-3.41.1-h5eee18b_0
  tk                 anaconda/pkgs/main/linux-64::tk-8.6.12-h1ccaba5_0
  tzdata             anaconda/pkgs/main/noarch::tzdata-2023c-h04d1e81_0
  wheel              anaconda/pkgs/main/linux-64::wheel-0.38.4-py310h06a4308_0
  xz                 anaconda/pkgs/main/linux-64::xz-5.2.10-h5eee18b_1
  zlib               anaconda/pkgs/main/linux-64::zlib-1.2.13-h5eee18b_0


Proceed ([y]/n)? y


Downloading and Extracting Packages

Preparing transaction: done
Verifying transaction: done
Executing transaction: done
#
# To activate this environment, use
#
#     $ conda activate py310
#
# To deactivate an active environment, use
#
#     $ conda deactivate

(base) xf@VP01:~/ai/env$ conda activate py310
(py310) xf@VP01:~/ai/env$

2.2 安装PyTorch200 GPU版本

2.2.1 官网链接

PyTorch最新版
https://pytorch.org/get-started/locally/

Lunix系统PyTorch CUDA 对Python版本的要求 >= Python 3.7

Installing on Linux
Python

Python 3.7 or greater is generally installed by default on any of our supported Linux distributions, which meets our recommendation.

Win系统PyTorch CUDA 对Python版本的要求 >= Python 3.7 <= Python 3.9

Installing on Windows
Python

Currently, PyTorch on Windows only supports Python 3.7-3.9; Python 2.x is not supported.

2.2.2 Conda安装方法

conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

package                      |      build
-----------------------------|-----------------
> pytorch-2.0.0              |      py3.10_cuda11.8_cudnn8.7.0_0        1.41 GB  pytorch
...
> torchaudio-2.0.0           |      py310_cu118                         7.5 MB  pytorch
...
> torchvision-0.15.0         |      py310_cu118                         8.0 MB  pytorch

在安装包列表中了解即将安装的是否CUDA版本, 第2列版本build中带_cu118，说明是CUDA加速版本

(py310) xf@VP01:~/ai/env$ conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: done

## Package Plan ##

  environment location: /home/xf/anaconda3/envs/py310

  added / updated specs:
    - pytorch
    - pytorch-cuda=11.8
    - torchaudio
    - torchvision


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    blas-1.0                   |              mkl           6 KB  defaults
    brotlipy-0.7.0             |py310h7f8727e_1002         712 KB  defaults
    cffi-1.15.1                |  py310h5eee18b_3         243 KB  defaults
    cryptography-39.0.1        |  py310h9ce1e76_0         1.4 MB  defaults
    filelock-3.9.0             |  py310h06a4308_0          18 KB  defaults
    flit-core-3.8.0            |  py310h06a4308_0          84 KB  defaults
    gmpy2-2.1.2                |  py310heeb90bb_0         517 KB  defaults
    idna-3.4                   |  py310h06a4308_0          97 KB  defaults
    jinja2-3.1.2               |  py310h06a4308_0         215 KB  defaults
    markupsafe-2.1.1           |  py310h7f8727e_0          34 KB  defaults
    mkl-service-2.4.0          |  py310h7f8727e_0         177 KB  defaults
    mkl_fft-1.3.1              |  py310hd6ae3a3_0         567 KB  defaults
    mkl_random-1.2.2           |  py310h00e6091_0        1009 KB  defaults
    mpmath-1.2.1               |  py310h06a4308_0         769 KB  defaults
    networkx-2.8.4             |  py310h06a4308_1         2.7 MB  defaults
    numpy-1.23.5               |  py310hd5efca6_0          10 KB  defaults
    numpy-base-1.23.5          |  py310h8e6c178_0         6.7 MB  defaults
    pillow-9.4.0               |  py310h6a678d5_0         730 KB  defaults
    pyopenssl-23.0.0           |  py310h06a4308_0          97 KB  defaults
    pysocks-1.7.1              |  py310h06a4308_0          28 KB  defaults
    pytorch-2.0.0              |py3.10_cuda11.8_cudnn8.7.0_0        1.41 GB  pytorch
    requests-2.28.1            |  py310h06a4308_1         100 KB  defaults
    sympy-1.11.1               |  py310h06a4308_0        11.8 MB  defaults
    torchaudio-2.0.0           |      py310_cu118         7.5 MB  pytorch
    torchtriton-2.0.0          |            py310        62.6 MB  pytorch
    torchvision-0.15.0         |      py310_cu118         8.0 MB  pytorch
    typing_extensions-4.4.0    |  py310h06a4308_0          46 KB  defaults
    urllib3-1.26.15            |  py310h06a4308_0         199 KB  defaults
    ------------------------------------------------------------