若使用conda安裝，可將CUDA一併下載。
安裝前要先確認安裝版本，python目前版本與PyTorch、CUDA的版本相容、自身電腦可安裝最高版本的CUDA，都要先確認，否則會無法安裝。另外，不建議安裝最高版本，務必安裝已釋出一段時間且穩定的版本。

---

## 手動下載安裝檔(建議)
到[Link](https://download.pytorch.org/whl/torch_stable.html)下載需要的版本

我使用的版本為
`PyTorch==1.12.0`、`torchvision==0.13.0`、`cuda==11.3`(cu113)、`python==3.10`(cp310)

或`PyTorch==1.11.0`、`torchvision==0.12.0`、`cuda==11.3`(cu113)、`python==3.9`(cp39)
cu113/torch-1.11.0%2Bcu113-cp39-cp39-win_amd64.whl
cu113/torchvision-0.12.0%2Bcu113-cp39-cp39-win_amd64.whl

### 用pip安裝

pip install "torch-1.12.0+cu113-cp310-cp310-win_amd64.whl"
pip install "torchvision-0.13.0+cu113-cp310-cp310-win_amd64.whl"


---

## 測試

python
import torch
torch.cuda.is_available()
torch.cuda.get_device_name(0)


---

## Issue

---

## Reference
https://pytorch.org/get-started/previous-versions/
https://download.pytorch.org/whl/torch_stable.html
https://blog.csdn.net/qq_45956730/article/details/126600028
https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html#install-windows
https://medium.com/ching-i/win10-%E5%AE%89%E8%A3%9D-cuda-cudnn-%E6%95%99%E5%AD%B8-c617b3b76deb

https://blog.csdn.net/qimo601/article/details/125146818

todo: cuda toolkit
https://developer.nvidia.com/cuda-toolkit-archive

https://developer.nvidia.com/rdp/cudnn-archive
https://stackoverflow.max-everyday.com/2022/08/anaconda-cuda-cudnn-win11/

安裝
https://www.it145.com/9/191725.html

---
curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-11-3_11.3.0-1_amd64.deb

curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-libraries-11-3_11.3.0-1_amd64.deb
curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-demo-suite-11-3_11.3.58-1_amd64.deb
curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-toolkit-11-3_11.3.0-1_amd64.deb
curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-runtime-11-3_11.3.0-1_amd64.deb

sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/7fa2af80.pub

sudo dpkg -i ./cuda-libraries-11-3_11.3.0-1_amd64.deb
sudo dpkg -i ./cuda-runtime-11-3_11.3.0-1_amd64.deb
sudo dpkg -i ./cuda-toolkit-11-3_11.3.0-1_amd64.deb
sudo dpkg -i ./cuda-demo-suite-11-3_11.3.58-1_amd64.deb

sudo dpkg -i ./cuda-11-3_11.3.0-1_amd64.deb

sudo apt-get update
sudo apt-get install ??

curl -O https://download.pytorch.org/whl/cu113/torch-1.12.1%2Bcu113-cp310-cp310-linux_x86_64.whl
curl -O https://download.pytorch.org/whl/cu113/torchvision-0.13.0%2Bcu113-cp310-cp310-linux_x86_64.whl

pip install "./torch-1.12.1%2Bcu113-cp310-cp310-linux_x86_64.whl"
pip install "./torchvision-0.13.0%2Bcu113-cp310-cp310-linux_x86_64.whl"