30系显卡笔记本使用pytorch(win10)

win10下30系笔记本显卡使用pytorch

1. 介绍背景:新换的笔记本3070laptop显卡(技嘉15G 2021)

cmd中输入“nvidia-smi”

30系显卡笔记本使用pytorch(win10)_第1张图片

2. 安装Anaconda3-2019.10-Windows-x86_64。官网下载慢的很,这里放一下百度网盘。

链接:https://pan.baidu.com/s/1Za4FVY6QhP1gKtz8GwE8wA 
提取码:kmax 

过程参考大佬Bubbliiiing的博客:神经网络学习小记录48——windows下的torch=1.2.0环境配置

https://blog.csdn.net/weixin_44791964/article/details/106037141?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522161241711716780269853447%2522%252C%2522scm%2522%253A%252220140713.130102334.pc%255Fblog.%2522%257D&request_id=161241711716780269853447&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~blog~first_rank_v2~rank_v29-2-106037141.pc_v2_rank_blog_default&utm_term=%E7%8E%AF%E5%A2%83&spm=1018.2226.3001.4450

3. 安装cudatoolkit和对应的cudnn。官网下载慢的很,这里放一下百度网盘。

链接:https://pan.baidu.com/s/1Za4FVY6QhP1gKtz8GwE8wA 
提取码:kmax 

过程参考大佬Bubbliiiing的博客:神经网络学习小记录48——windows下的torch=1.2.0环境配置

https://blog.csdn.net/weixin_44791964/article/details/106037141?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522161241711716780269853447%2522%252C%2522scm%2522%253A%252220140713.130102334.pc%255Fblog.%2522%257D&request_id=161241711716780269853447&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~blog~first_rank_v2~rank_v29-2-106037141.pc_v2_rank_blog_default&utm_term=%E7%8E%AF%E5%A2%83&spm=1018.2226.3001.4450

4. 安装适配cuda11.0对应的pytorch(兼容cuda11.1的驱动)

30系显卡笔记本使用pytorch(win10)_第2张图片

链接:https://pan.baidu.com/s/1Za4FVY6QhP1gKtz8GwE8wA 
提取码:kmax 

过程参考大佬Bubbliiiing的博客:神经网络学习小记录48——windows下的torch=1.2.0环境配置

https://blog.csdn.net/weixin_44791964/article/details/106037141?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522161241711716780269853447%2522%252C%2522scm%2522%253A%252220140713.130102334.pc%255Fblog.%2522%257D&request_id=161241711716780269853447&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~blog~first_rank_v2~rank_v29-2-106037141.pc_v2_rank_blog_default&utm_term=%E7%8E%AF%E5%A2%83&spm=1018.2226.3001.4450

注:安装pytorch时共有三个部分,其中“torch1.7.1”有1.9G比较大,百度网盘里有,无需再去官网龟速下载。

经测试,兼容pytorch=1.2版本下的YOLOv3预测代码,且能发挥30系显卡实力。

总结:

cuda_11.1.0_456.43_win10.exe + cudnn-11.1-windows-x64-v8.0.4.30 + torch-1.7.1+cu110-cp37-cp37m-win_amd64.whl + torchvision-0.8.2+cu110-cp37-cp37m-win_amd64.whl

你可能感兴趣的:(神经网络,深度学习,cuda,anaconda,pytorch)