conda 安装pytorch with cuda 失败问题@pytorch历史版本安装问题
文章目录
-
- conda 安装pytorch with cuda 失败问题
-
- 使用pip安装
- 使用conda安装pytorch with cuda
-
- 正确的安装组合@适用于安装最新版
- 检查cuda可用性
- 安装耗时
- condarc配置文件示例
- 清华源
- 阿里源
- conda的相关使用参考
- FAQ
-
- 安装完cuda依然无法调用GPU:错误的版本搭配
- 历史版本的安装
-
- 通道问题@Channel
- COMMANDS FOR VERSIONS >= 1.0.0
-
- v1.13.1
-
- Conda
-
- OSX
- Linux and Windows
- Wheel
-
- OSX
- Linux and Windows
conda 安装pytorch with cuda 失败问题
-
激活环境(本例假设环境为
pytorch_ser)PS D:\repos\PythonLearn> conda activate pytorch_ser -
尝试直接运行pytorch官网给出的conda安装命令,发现解析操作迟迟无法结束
-
Solving environment: failed with initial frozen solve. Retrying with flexible solve. Collecting package metadata (repodata.json): done .... Solving environment: .... -
原因可能是:
- 我将默认的源换成清华源,而清华源的镜像没有能够满足要安装的配套组件
- 网络环境问题,更换网络重试
- 服务器问题,更改时段再试
-
使用pip安装
-
(d:\condaPythonEnvs\pytorch_ser) PS D:\repos\blogs> pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu117 Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu117 Requirement already satisfied: torch in d:\condapythonenvs\pytorch_ser\lib\site-packages (1.13.1) Requirement already satisfied: torchvision in d:\condapythonenvs\pytorch_ser\lib\site-packages (0.14.1) Requirement already satisfied: torchaudio in d:\condapythonenvs\pytorch_ser\lib\site-packages (0.13.1) Requirement already satisfied: typing_extensions in d:\condapythonenvs\pytorch_ser\lib\site-packages (from torch) (4.4.0) Requirement already satisfied: numpy in d:\condapythonenvs\pytorch_ser\lib\site-packages (from torchvision) (1.23.5) Requirement already satisfied: requests in d:\condapythonenvs\pytorch_ser\lib\site-packages (from torchvision) (2.28.1) Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in d:\condapythonenvs\pytorch_ser\lib\site-packages (from torchvision) (9.3.0) Requirement already satisfied: certifi>=2017.4.17 in d:\condapythonenvs\pytorch_ser\lib\site-packages (from requests->torchvision) (2022.12.7) Requirement already satisfied: charset-normalizer<3,>=2 in d:\condapythonenvs\pytorch_ser\lib\site-packages (from requests->torchvision) (2.0.4) Requirement already satisfied: idna<4,>=2.5 in d:\condapythonenvs\pytorch_ser\lib\site-packages (from requests->torchvision) (3.4) Requirement already satisfied: urllib3<1.27,>=1.21.1 in d:\condapythonenvs\pytorch_ser\lib\site-packages (from requests->torchvision) (1.26.13) -
从上面的输出上看,pip似乎无法完成cuda组件的安装
使用conda安装pytorch with cuda
正确的安装组合@适用于安装最新版
-
如果之前安装过cpu only 版本的pytorch,导致pytorch基础组件和cuda pytorch 组件不能够配合工作
-
所以再在一个新的环境中重新安装cuda版pytorch
conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia
-
(d:\condaPythonEnvs\pytorch_ser) PS C:\Users\cxxu\Desktop> conda activate py310 (py310) PS C:\Users\cxxu\Desktop> conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -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: C:\Users\cxxu\miniconda3\envs\py310 added / updated specs: - pytorch - pytorch-cuda=11.7 - torchaudio - torchvision The following packages will be downloaded: package | build ---------------------------|----------------- pytorch-1.13.1 |py3.10_cuda11.7_cudnn8_0 1.10 GB pytorch pytorch-mutex-1.0 | cuda 3 KB pytorch torchaudio-0.13.1 | py310_cu117 4.7 MB pytorch torchvision-0.14.1 | py310_cu117 7.5 MB pytorch ------------------------------------------------------------ Total: 1.11 GB The following NEW packages will be INSTALLED: brotlipy anaconda/pkgs/main/win-64::brotlipy-0.7.0-py310h2bbff1b_1002 cffi anaconda/pkgs/main/win-64::cffi-1.15.1-py310h2bbff1b_3 charset-normalizer anaconda/pkgs/main/noarch::charset-normalizer-2.0.4-pyhd3eb1b0_0 cryptography anaconda/pkgs/main/win-64::cryptography-38.0.1-py310h21b164f_0 cuda nvidia/win-64::cuda-11.7.1-0 cuda-cccl nvidia/win-64::cuda-cccl-11.7.91-0 cuda-command-line~ nvidia/win-64::cuda-command-line-tools-11.7.1-0 cuda-compiler nvidia/win-64::cuda-compiler-11.7.1-0 .... cuda-tools nvidia/win-64::cuda-tools-11.7.1-0 cuda-visual-tools nvidia/win-64::cuda-visual-tools-11.7.1-0 flit-core anaconda/pkgs/main/noarch::flit-core-3.6.0-pyhd3eb1b0_0 freetype anaconda/pkgs/main/win-64::freetype-2.12.1-ha860e81_0 idna anaconda/pkgs/main/win-64::idna-3.4-py310haa95532_0 jpeg anaconda/pkgs/main/win-64::jpeg-9e-h2bbff1b_0 lerc anaconda/pkgs/main/win-64::lerc-3.0-hd77b12b_0 .... pytorch-mutex pytorch/noarch::pytorch-mutex-1.0-cuda requests anaconda/pkgs/main/win-64::requests-2.28.1-py310haa95532_0 torchaudio pytorch/win-64::torchaudio-0.13.1-py310_cu117 torchvision pytorch/win-64::torchvision-0.14.1-py310_cu117 typing_extensions anaconda/pkgs/main/win-64::typing_extensions-4.4.0-py310haa95532_0 urllib3 anaconda/pkgs/main/win-64::urllib3-1.26.13-py310haa95532_0 win_inet_pton anaconda/pkgs/main/win-64::win_inet_pton-1.1.0-py310haa95532_0 zstd anaconda/pkgs/main/win-64::zstd-1.5.2-h19a0ad4_0 Proceed ([y]/n)? y Downloading and Extracting Packages torchaudio-0.13.1 | 4.7 MB | ############################################################################ | 100% pytorch-mutex-1.0 | 3 KB | ############################################################################ | 100% pytorch-1.13.1 | 1.10 GB | ###########################################################################9 | 100% torchvision-0.14.1 | 7.5 MB | ############################################################################ | 100% GB | ######################################################## Preparing transaction: done Verifying transaction: done Executing transaction: done (py310) PS C:\Users\cxxu\Desktop>
检查cuda可用性
-
import torch as torch import torch as th print(th.__version__) print(th.version.cuda) print(th.cuda.is_available()) -
(py310) PS D:\repos\CCSER> python Python 3.10.8 | packaged by conda-forge | (main, Nov 24 2022, 14:07:00) [MSC v.1916 64 bit (AMD64)] on win32 Type "help", "copyright", "credits" or "license" for more information. >>> import torch as torch >>> import torch as th >>> print(th.__version__) 1.13.1 >>> print(th.version.cuda) 11.7 >>> print(th.cuda.is_available()) True
安装耗时
-
安装的源用的清华源,宽带500M,在几分钟内(5分钟)可以完成安装
-
nvidia驱动版本和cuda驱动版本(CUDA Version: 12.0 )
-
cuda驱动版本要高于cuda运行时版本
-
如果驱动版本过旧,到nvidia官方下载更新
-
官方驱动 | NVIDIA
-
PS C:\Users\cxxu\Desktop> nvidia-smi.exe Sun Jan 8 17:15:39 2023 +-----------------------------------------------------------------------------+ | NVIDIA-SMI 527.56 Driver Version: 527.56 CUDA Version: 12.0 | |-------------------------------+----------------------+----------------------+ | GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |===============================+======================+======================| | 0 NVIDIA GeForce ... WDDM | 00000000:02:00.0 Off | N/A | | N/A 45C P0 N/A / N/A | 0MiB / 2048MiB | 0% Default | | | | N/A | +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=============================================================================| | No running processes found | +-----------------------------------------------------------------------------+- 玩具显卡,但是不影响过程演示
-
-
condarc配置文件示例
-
Using the .condarc conda configuration file — conda 23.3.0.post2+8419c02f5 documentation
- creating and editing
-
You can find information about your
.condarcfile by typingconda infoin your terminal or Anaconda Prompt.-
This will give you information about your
.condarcfile, including where it is located. -
PS D:\repos\blogs\python> conda info active environment : None user config file : C:\Users\cxxu\.condarc populated config files : C:\Users\cxxu\.condarc conda version : 23.1.0 conda-build version : not installed python version : 3.9.5.final.0 virtual packages : __archspec=1=x86_64 __cuda=12.0=0 __win=0=0 base environment : C:\Users\cxxu\miniconda3 (writable) conda av data dir : C:\Users\cxxu\miniconda3\etc\conda conda av metadata url : None channel URLs : https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/win-64 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/noarch https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/r/win-64 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/r/noarch https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/msys2/win-64 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/msys2/noarch package cache : C:\Users\cxxu\miniconda3\pkgs C:\Users\cxxu\.conda\pkgs C:\Users\cxxu\AppData\Local\conda\conda\pkgs envs directories : d:\condaPythonEnvs C:\Users\cxxu\miniconda3\envs C:\Users\cxxu\.conda\envs C:\Users\cxxu\AppData\Local\conda\conda\envs platform : win-64 user-agent : conda/23.1.0 requests/2.28.1 CPython/3.9.5 Windows/10 Windows/10.0.22621 administrator : False netrc file : None offline mode : False
-
-
本人的配置文件样例如下:
-
channels: - defaults show_channel_urls: true default_channels: - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/r - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/msys2 custom_channels: conda-forge: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud msys2: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud bioconda: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud menpo: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud pytorch: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud pytorch-lts: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud simpleitk: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud auto_activate_base: false
清华源
- anaconda | 镜像站使用帮助 | 清华大学开源软件镜像站 | Tsinghua Open Source Mirror
阿里源
-
anaconda镜像_anaconda下载地址_anaconda安装教程-阿里巴巴开源镜像站 (aliyun.com)
-
channels: - defaults show_channel_urls: true default_channels: - http://mirrors.aliyun.com/anaconda/pkgs/main - http://mirrors.aliyun.com/anaconda/pkgs/r - http://mirrors.aliyun.com/anaconda/pkgs/msys2 custom_channels: conda-forge: http://mirrors.aliyun.com/anaconda/cloud msys2: http://mirrors.aliyun.com/anaconda/cloud bioconda: http://mirrors.aliyun.com/anaconda/cloud menpo: http://mirrors.aliyun.com/anaconda/cloud pytorch: http://mirrors.aliyun.com/anaconda/cloud simpleitk: http://mirrors.aliyun.com/anaconda/cloud
conda的相关使用参考
- conda发行版比较@python环境管理@conda命令的基本操作@配置conda_xuchaoxin1375的博客-CSDN博客
FAQ
安装完cuda依然无法调用GPU:错误的版本搭配
-
最初本人尝试安装
pytorch with cuda,发现无法安装(具体表现为:不停的解析,而无法顺利结束) -
于是我尝试安装一遍
pytorch cpu only,发现可以顺利安装 -
过了若干天,想体验GPU加速,重试,发现可以安装
pytorch with cuda(此期间没有修改condarc配置文件) -
安装过程
-
(d:\condaPythonEnvs\pytorch_ser) PS D:\repos\blogs> conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -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: d:\condaPythonEnvs\pytorch_ser added / updated specs: - pytorch - pytorch-cuda=11.7 - torchaudio - torchvision The following packages will be downloaded: package | build ---------------------------|----------------- cuda-11.7.1 | 0 1 KB nvidia cuda-cccl-11.7.91 | 0 1.2 MB nvidia cuda-command-line-tools-11.7.1| 0 1 KB nvidia cuda-compiler-11.7.1 | 0 1 KB nvidia cuda-cudart-11.7.99 | 0 1.4 MB nvidia cuda-cudart-dev-11.7.99 | 0 711 KB nvidia cuda-cuobjdump-11.7.91 | 0 2.5 MB nvidia cuda-cupti-11.7.101 | 0 10.2 MB nvidia cuda-cuxxfilt-11.7.91 | 0 165 KB nvidia .... cuda-toolkit-11.7.1 | 0 1 KB nvidia cuda-tools-11.7.1 | 0 1 KB nvidia cuda-visual-tools-11.7.1 | 0 1 KB nvidia libcublas-11.10.3.66 | 0 24 KB nvidia libcublas-dev-11.10.3.66 | 0 282.4 MB nvidia libcufft-10.7.2.124 | 0 6 KB nvidia libcufft-dev-10.7.2.124 | 0 250.1 MB nvidia libcurand-10.3.1.50 | 0 3 KB nvidia libcurand-dev-10.3.1.50 | 0 50.0 MB nvidia libcusolver-11.4.0.1 | 0 29 KB nvidia libcusolver-dev-11.4.0.1 | 0 76.5 MB nvidia libcusparse-11.7.4.91 | 0 13 KB nvidia libcusparse-dev-11.7.4.91 | 0 149.6 MB nvidia libnpp-11.7.4.75 | 0 294 KB nvidia libnpp-dev-11.7.4.75 | 0 125.6 MB nvidia libnvjpeg-11.8.0.2 | 0 4 KB nvidia libnvjpeg-dev-11.8.0.2 | 0 1.7 MB nvidia nsight-compute-2022.4.0.15 | 0 598.6 MB nvidia pytorch-cuda-11.7 | h67b0de4_1 3 KB pytorch ------------------------------------------------------------ Total: 1.82 GB The following NEW packages will be INSTALLED: cuda nvidia/win-64::cuda-11.7.1-0 cuda-cccl nvidia/win-64::cuda-cccl-11.7.91-0 cuda-command-line~ nvidia/win-64::cuda-command-line-tools-11.7.1-0 cuda-compiler nvidia/win-64::cuda-compiler-11.7.1-0 cuda-cudart nvidia/win-64::cuda-cudart-11.7.99-0 cuda-cudart-dev nvidia/win-64::cuda-cudart-dev-11.7.99-0 cuda-cuobjdump nvidia/win-64::cuda-cuobjdump-11.7.91-0 cuda-cupti nvidia/win-64::cuda-cupti-11.7.101-0 ... cuda-tools nvidia/win-64::cuda-tools-11.7.1-0 cuda-visual-tools nvidia/win-64::cuda-visual-tools-11.7.1-0 libcublas nvidia/win-64::libcublas-11.10.3.66-0 libcublas-dev nvidia/win-64::libcublas-dev-11.10.3.66-0 libcufft nvidia/win-64::libcufft-10.7.2.124-0 libcufft-dev nvidia/win-64::libcufft-dev-10.7.2.124-0 libcurand nvidia/win-64::libcurand-10.3.1.50-0 libcurand-dev nvidia/win-64::libcurand-dev-10.3.1.50-0 libcusolver nvidia/win-64::libcusolver-11.4.0.1-0 libcusolver-dev nvidia/win-64::libcusolver-dev-11.4.0.1-0 libcusparse nvidia/win-64::libcusparse-11.7.4.91-0 libcusparse-dev nvidia/win-64::libcusparse-dev-11.7.4.91-0 libnpp nvidia/win-64::libnpp-11.7.4.75-0 libnpp-dev nvidia/win-64::libnpp-dev-11.7.4.75-0 libnvjpeg nvidia/win-64::libnvjpeg-11.8.0.2-0 libnvjpeg-dev nvidia/win-64::libnvjpeg-dev-11.8.0.2-0 nsight-compute nvidia/win-64::nsight-compute-2022.4.0.15-0 pytorch-cuda pytorch/noarch::pytorch-cuda-11.7-h67b0de4_1 Proceed ([y]/n)? y Downloading and Extracting Packages cuda-cudart-dev-11.7 | 711 KB | ############################################################################################################################################### | 100% cuda-memcheck-11.8.8 | 183 KB | ############################################################################################################################################### | 100% cuda-cudart-11.7.99 | 1.4 MB | ############################################################################################################################################### | 100% libnvjpeg-11.8.0.2 | 4 KB | ############################################################################################################################################### | 100% pytorch-cuda-11.7 | 3 KB | ############################################################################################################################################### | 100% ........ ####################################################################################################################5 | 81% cuda-cupti-11.7.101 | 10.2 MB | ############################################################################################################################################### | 100% cuda-demo-suite-12.0 | 4.7 MB | ############################################################################################################################################### | 100%
-
历史版本的安装
- Previous PyTorch Versions | PyTorch
通道问题@Channel
- 对于
conda install命令而言,-c参数指定的Channel对于安装操作是至关重要的 - 特别是对于复杂或大型的框架的安装,更加容易因为指定的通道不合适而导致安装失败
COMMANDS FOR VERSIONS >= 1.0.0
-
在这里不得不吐槽以下pytorch的历史版本页面提供的命令,竟然无法工作
-
后来对比最新版命令才发现,是Previous PyTorch Versions | PyTorch页面将
-c nvidia参数错误的写成-nvida-
导致的一个直接问题是,conda命令是没有
-nvidia这样的参数,而且会别识别为-n vidia,也就是识别为一个名为vidia的环境conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -nvidia(是一个错误的命令)
-
刚开始我不知道这个参数是个Channel的名称,就把它删除掉在运行,发现会报一些莫名奇妙的依赖
-
Package pytorch-cuda conflicts for: torchaudio==0.13.1 -> pytorch-cuda[version='11.6.*|11.7.*'] pytorch-cuda=11.6 torchaudio==0.13.1 -> pytorch==1.13.1 -> pytorch-cuda[version='>=11.6,<11.7|>=11.7,<11.8'] pytorch==1.13.1 -> pytorch-cuda[version='>=11.6,<11.7|>=11.7,<11.8'] Package pytorch conflicts for: pytorch==1.13.1 torchaudio==0.13.1 -> pytorch==1.13.1 -
而我们自己检查发现其实依赖是没有问题的,这些版本也都是官网提供的
-
-
将通道修改正确
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia-
得到正确的反馈
-
(d:\condaPythonEnvs\d2l) PS D:\repos\blogs> conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia Collecting package metadata (current_repodata.json): done Solving environment: failed with initial frozen solve. Retrying with flexible solve. Collecting package metadata (repodata.json): done Solving environment: done ## Package Plan ## environment location: d:\condaPythonEnvs\d2l added / updated specs: - pytorch-cuda=11.7 - pytorch==1.13.1 - torchaudio==0.13.1 - torchvision==0.14.1 The following packages will be downloaded: package | build ---------------------------|----------------- cuda-cccl-12.1.55 | 0 1.2 MB nvidia libcurand-10.3.2.56 | 0 3 KB nvidia libcurand-dev-10.3.2.56 | 0 50.0 MB nvidia pytorch-cuda-11.7 | h16d0643_3 7 KB pytorch ------------------------------------------------------------ Total: 51.2 MB The following NEW packages will be INSTALLED: blas anaconda/pkgs/main/win-64::blas-1.0-mkl brotlipy anaconda/pkgs/main/win-64::brotlipy-0.7.0-py310h2bbff1b_1002 bzip2 anaconda/pkgs/main/win-64::bzip2-1.0.8-he774522_0 ... cuda-cupti nvidia/win-64::cuda-cupti-11.7.101-0 cuda-libraries nvidia/win-64::cuda-libraries-11.7.1-0 cuda-libraries-dev nvidia/win-64::cuda-libraries-dev-11.7.1-0 ... 32_0 vc anaconda/pkgs/main/win-64::vc-14.2-h21ff451_1 vs2015_runtime anaconda/pkgs/main/win-64::vs2015_runtime-14.27.29016-h5e58377_2 wheel anaconda/pkgs/main/win-64::wheel-0.38.4-py310haa95532_0 win_inet_pton anaconda/pkgs/main/win-64::win_inet_pton-1.1.0-py310haa95532_0 wincertstore anaconda/pkgs/main/win-64::wincertstore-0.2-py310haa95532_2 xz anaconda/pkgs/main/win-64::xz-5.2.10-h8cc25b3_1 zlib anaconda/pkgs/main/win-64::zlib-1.2.13-h8cc25b3_0 zstd anaconda/pkgs/main/win-64::zstd-1.5.2-h19a0ad4_0 Proceed ([y]/n)? y Downloading and Extracting Packages Preparing transaction: done Verifying transaction: done Executing transaction: done -
可以看到,这次下载量很小,是因为之前我在其他环境用
conda install安装过一次pytorch==1.13.1及其配套依赖,所以这次需要下载的内容比较少,其他内容可以从本地的conda缓存中读取即可
-
-
v1.13.1
Conda
OSX
# conda
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 -c pytorch
Linux and Windows
# CUDA 11.6
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.6 -c pytorch -nvidia
# CUDA 11.7
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -nvidia
# CPU Only
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 cpuonly -c pytorch
Wheel
OSX
pip install torch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1
Linux and Windows
# ROCM 5.2 (Linux only)
pip3 install torch torchvision torchaudio --extra-index-url
pip install torch==1.13.1+rocm5.2 torchvision==0.14.1+rocm5.2 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/rocm5.2
# CUDA 11.6
pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116
# CUDA 11.7
pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117
# CPU only
pip install torch==1.13.1+cpu torchvision==0.14.1+cpu torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cpu