Record the torch encountered by win10 cuda. is_ False problem in available()

This article computer configuration
Win10+Anaconda+CUDA10.1+cudnn7.6.5

I used the following command on the official website

conda install pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=10.1

Loading PyTorch Encountered after running ：

AssertionError: Torch not compiled with CUDA enabled

I found some solutions on the Internet ：
1. Upgrade the graphics card driver ;
2. l / drop torch edition

But it didn't work out , Finally, I use conda list After the discovery

pytorch                   1.10.0              py3.7_cpu_0

This shows that what I installed is cpu edition
Combined with the posts of other bloggers , I guess it occurred to me because conda Automatic installation is cpu edition , But I didn't find it in the mirror website cudnn7.6 Version correspondence PyTorch, I'm lazy again , Not reinstalled cudnn, So in PyTorch Continue to read the installation instructions , I found that in addition to conda command , You can also use pip Mode of installation , So I just use the command to install

pip install torch==1.7.1+cu101 torchvision==0.8.2+cu101 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html

Specific connections ：PyTorch Official website

Then test

>>> import torch
>>>> print(torch.cuda.is_available())
True

Successful installation gpu Version of PyTorch！

I hope it's never too late to learn , Now learn from Mu Shen PyTorch, I hope the scientific research can go smoothly ！