在linux上安裝torch時， 如果環境安裝好了CUDA環境， 那么安裝torch時不用刻意指定帶cuda的版本， 最終安裝的也是支持GPU的torch版本。但是仍然有一些小的區別，主要就是支持CUDA版本的不同。

(leo_py37) pinefield@edge-gpu-01:/data/codes/OpenPCDet$ pip3 install torch==1.11
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Collecting torch==1.11Downloading https://pypi.tuna.tsinghua.edu.cn/packages/94/32/96a0955e5d6ed8a837eda5ca095dd2694c4617cfa70ca599660cd5ff7447/torch-1.11.0-cp37-cp37m-manylinux1_x86_64.whl (750.6 MB)━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 750.6/750.6 MB 840.4 kB/s eta 0:00:00
Collecting typing-extensionsUsing cached https://pypi.tuna.tsinghua.edu.cn/packages/ec/6b/63cc3df74987c36fe26157ee12e09e8f9db4de771e0f3404263117e75b95/typing_extensions-4.7.1-py3-none-any.whl (33 kB)
Installing collected packages: typing-extensions, torch
Successfully installed torch-1.11.0 typing-extensions-4.7.1
(leo_py37) pinefield@edge-gpu-01:/data/codes/OpenPCDet$ python3 
Python 3.7.5 (default, Oct 25 2019, 15:51:11) 
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch 
/home/pinefield/anaconda3/envs/leo_py37/lib/python3.7/site-packages/torch/_masked/__init__.py:223: UserWarning: Failed to initialize NumPy: numpy.core.multiarray failed to import (Triggered internally at  ../torch/csrc/utils/tensor_numpy.cpp:68.)example_input = torch.tensor([[-3, -2, -1], [0, 1, 2]])
>>> torch.__version__
'1.11.0+cu102'
>>> 
(leo_py37) pinefield@edge-gpu-01:/data/joyiot/liyong/codes/OpenPCDet$ pip3 list |grep torch 
torch             1.11.0

如上圖， 安裝torch時未指定需要安裝gpu版本， 但是實際安裝的也是支持gpu的版本。只是安裝了支持cuda 10.2的版本。 實際上系統的cuda版本是11.5.

如果明確指定cuda版本， 則安裝的就是對應的版本。

(leo_py37) pinefield@edge-gpu-01:/data/codes/OpenPCDet$ pip3 install  torch==1.11.0+cu115
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
ERROR: Could not find a version that satisfies the requirement torch==1.11.0+cu115 (from versions: 1.0.0, 1.0.1, 1.0.1.post2, 1.1.0, 1.2.0, 1.3.0, 1.3.1, 1.4.0, 1.5.0, 1.5.1, 1.6.0, 1.7.0, 1.7.1, 1.8.0, 1.8.1, 1.9.0, 1.9.1, 1.10.0, 1.10.1, 1.10.2, 1.11.0, 1.12.0, 1.12.1, 1.13.0, 1.13.1)
ERROR: No matching distribution found for torch==1.11.0+cu115
(leo_py37) pinefield@edge-gpu-01:/data/joyiot/liyong/codes/OpenPCDet$ pip3 install  torch==1.11.0+cu115   --extra-index-url https://download.pytorch.org/whl/cu115
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple, https://download.pytorch.org/whl/cu115
Collecting torch==1.11.0+cu115Using cached https://download.pytorch.org/whl/cu115/torch-1.11.0%2Bcu115-cp37-cp37m-linux_x86_64.whl (2138.4 MB)
Requirement already satisfied: typing-extensions in /home/pinefield/anaconda3/envs/leo_py37/lib/python3.7/site-packages (from torch==1.11.0+cu115) (4.7.1)
Installing collected packages: torchAttempting uninstall: torchFound existing installation: torch 1.11.0Uninstalling torch-1.11.0:Successfully uninstalled torch-1.11.0
Successfully installed torch-1.11.0+cu115
(leo_py37) pinefield@edge-gpu-01:/data/codes/OpenPCDet$ python3 
Python 3.7.5 (default, Oct 25 2019, 15:51:11) 
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch 
/home/pinefield/anaconda3/envs/leo_py37/lib/python3.7/site-packages/torch/_masked/__init__.py:223: UserWarning: Failed to initialize NumPy: numpy.core.multiarray failed to import (Triggered internally at  ../torch/csrc/utils/tensor_numpy.cpp:68.)example_input = torch.tensor([[-3, -2, -1], [0, 1, 2]])
>>> torch.__version__
'1.11.0+cu115'
>>> 
(leo_py37) pinefield@edge-gpu-01:/data/codes/OpenPCDet$ pip3 list |grep torch 
torch             1.11.0+cu115

注意， 直接安裝可能會找不到對應版本， 需要通過 --extra-index-url https://download.pytorch.org/whl/cu115指定torch官方的下載地址。