使用手动的方式安装Torch+GPU

安装过程记录

(pytorch) C:\Users\wangguowei>cd Desktop

(pytorch) C:\Users\wangguowei\Desktop>pip install "torch-1.7.0+cu110-cp37-cp37m-win_amd64.whl"
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Processing c:\users\wangguowei\desktop\torch-1.7.0+cu110-cp37-cp37m-win_amd64.whl
Requirement already satisfied: numpy in d:\wangguowei\anaconda3\envs\pytorch\lib\site-packages (from torch==1.7.0+cu110) (1.19.5)
Requirement already satisfied: dataclasses in d:\wangguowei\anaconda3\envs\pytorch\lib\site-packages (from torch==1.7.0+cu110) (0.6)
Requirement already satisfied: typing-extensions in d:\wangguowei\anaconda3\envs\pytorch\lib\site-packages (from torch==1.7.0+cu110) (3.7.4.3)
Requirement already satisfied: future in d:\wangguowei\anaconda3\envs\pytorch\lib\site-packages (from torch==1.7.0+cu110) (0.18.3)
Installing collected packages: torch
Successfully installed torch-1.7.0+cu110

(pytorch) C:\Users\wangguowei\Desktop>a

通过conda 或者pip安装都会产生意想不到的问题

• 我们关键使用的东西是torch
• 不是torchvision
• 也不是torchtext

所以一定要根据灵活一点，让自己的环境实时处于可用状态