GTX 3090安装python，torch环境的版本

青旅半醒 2022-10-09 12:17 321阅读 0赞

1、python 3.6.8 [https://www.python.org/downloads/windows/][https_www.python.org_downloads_windows]

2、TensorFlow-gpu ；1.14.0

3、torch 1.6.0 版本无所谓，cmd输入 nvidia-smi 查看自己的CUDN的信息![20210624165914832.png][]

比如这里是10.1就去官网下载 torch-1.6.0+**cu101**\+win.....以及对应的torchvision

[https://pytorch.org/get-started/locally/][https_pytorch.org_get-started_locally]官网的命令能不能下载全靠网速

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1h1X3NhX3Nh_size_16_color_FFFFFF_t_70][]

4、如果代码报CUDN不可用，去英伟达光放下载对应的cudn安装就行

[https://developer.nvidia.com/cuda-10.2-download-archive][https_developer.nvidia.com_cuda-10.2-download-archive]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1h1X3NhX3Nh_size_16_color_FFFFFF_t_70 1][]网页显示的download都要下

cudnn下载网址：[https://developer.nvidia.com/cudnn][https_developer.nvidia.com_cudnn]

最后torch.version.cuda和nvcc-V显示的版本一致， torch.cuda.is\_available()就会是true

查看cuda版本：torch.version.cuda

查看cuda是否可用： torch.cuda.is\_available()

可用GPU的数量：torch.cuda.device\_count()

查看当前使用的GPU序号：torch.cuda.current\_device()

查看nvidia信息：nvidia-smi

nvcc -V

[https_www.python.org_downloads_windows]: https://www.python.org/downloads/windows/
[20210624165914832.png]: /images/20221005/ef2ac5725fcc45d7abcd0e701899dd87.png
[https_pytorch.org_get-started_locally]: https://pytorch.org/get-started/locally/
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1h1X3NhX3Nh_size_16_color_FFFFFF_t_70]: /images/20221005/588317274f0e4c50a74b75eff570529b.png
[https_developer.nvidia.com_cuda-10.2-download-archive]: https://developer.nvidia.com/cuda-10.2-download-archive
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1h1X3NhX3Nh_size_16_color_FFFFFF_t_70 1]: /images/20221005/940d7ab3278141a6bb8c0b63b8e00947.png
[https_developer.nvidia.com_cudnn]: https://developer.nvidia.com/cudnn