本人水平有限,如有问题请多指正
笔记本系统:Windows 10 64位
显卡:NVIDIA GeForce MX150
显卡驱动程序版本:512.78
显卡驱动程序支持CUDA版本:11.6.134及以下
安装CUDA:11.6.0
安装cuDNN:8.7.0
一、安装CUDA
1、确定CUDA版本
查看本机驱动程序版本。打开“NVIDIA 控制面板”,点击“帮助”,“系统信息”。
驱动程序版本“512.78”
![](https://aitechtogether.com/wp-content/uploads/2023/04/4cebd812-9f9d-4a1d-9fe5-cf8b34ba7063.webp)
1.1、显卡驱动支持的CUDA版本
安装CUDA工具包,对显卡驱动版本有最低要求
查看显卡驱动版本最低要求:本机驱动程序版本512.78,可安装CUDA12.0以下版本
https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html#cuda-major-component-versions__table-cuda-toolkit-driver-versions
Table 2. CUDA Toolkit and Minimum Required Driver Version for CUDA Minor Version Compatibility
![](https://aitechtogether.com/wp-content/uploads/2023/04/66e0f9bf-6be0-4b99-9cc8-43dc901c586e.webp)
1.2、确定CUDA可安装最高版本
NVCUDA64.DLL后面参数显示,支持11.6.134及以下版本CUDA
结合上面支持12.0以下版本CUDA,这里选择CUDA 11.6.0版本
![](https://aitechtogether.com/wp-content/uploads/2023/04/c935b8bb-b544-41ab-b324-cb09929b1825.webp)
2、下载CUDA 11.6.0
cuda 下载网址:https://developer.nvidia.com/cuda-toolkit-archive
![](https://aitechtogether.com/wp-content/uploads/2023/04/f8c622c1-b47c-452c-9fad-1023ae533409.webp)
![](https://aitechtogether.com/wp-content/uploads/2023/04/6ac93ddc-f801-4d00-a620-ed3c7d496d11.webp)
3、安装
error: Could not create file”D:\Program Files\CUDA\GFExperience\chrome_elf.d”.拒绝访问。
![](https://aitechtogether.com/wp-content/uploads/2023/04/a28005cf-d006-494c-ad93-76b6d1d862bc.webp)
解法办法:退出360,重新安装
![](https://aitechtogether.com/wp-content/uploads/2023/04/a2efb5b7-194d-4d98-88f1-60e65026667e.webp)
error:安装程序无法继续
![](https://aitechtogether.com/wp-content/uploads/2023/04/de497a39-0aeb-47b7-b3d5-d0d726bf16d3.webp)
解决方法:重启电脑,退出360,重新安装
选择“自定义安装选项”,将“Visual Studio Integration”前面的框勾选掉,否则会报错
![](https://aitechtogether.com/wp-content/uploads/2023/04/9ec35aa7-1d9c-46d9-8085-b309bccd2ca0.webp)
这里修改了安装路径,可以不修改
![](https://aitechtogether.com/wp-content/uploads/2023/04/965a2aaa-48cb-4833-bbc6-bc46dbe260ac.webp)
正在安装,等一会
![](https://aitechtogether.com/wp-content/uploads/2023/04/a3a7bf50-0747-4797-bf54-4f51fdb38351.webp)
正在安装,等一会
![](https://aitechtogether.com/wp-content/uploads/2023/04/f79325b6-8dfe-4007-8f48-a1990a02c68b.webp)
点击”下一步”
![](https://aitechtogether.com/wp-content/uploads/2023/04/1cd42f1c-3964-47e3-a98f-51314998eb4d.webp)
点击”关闭”
![](https://aitechtogether.com/wp-content/uploads/2023/04/abc3c08a-09ca-4132-b3c2-fd1c009f1438.webp)
4、检验。打开命令行,输入”nvcc -V”,显示下面类似界面就安装好了
![](https://aitechtogether.com/wp-content/uploads/2023/04/9571d02c-568d-4f38-9cf8-b255575f48bd.webp)
二、安装cuDNN
1、下载
注册账号,输入邮箱,验证一下邮箱, 再填几个信息就行了。注册完成会跳转到下载页面
本机安装了CUDA11,对应的cuDNN版本是v8.7.0
![](https://aitechtogether.com/wp-content/uploads/2023/04/f2f61eb7-6489-4ba8-8172-ff90c87cdf69.webp)
2、解压
![](https://aitechtogether.com/wp-content/uploads/2023/04/1b57d085-02e4-401a-8b9e-f2a5abb7cdf1.webp)
3、安装。复制到CUDA安装目录下
![](https://aitechtogether.com/wp-content/uploads/2023/04/76fb782c-8f5a-4674-8a3d-01f17945bdd0.webp)
4、检验。
打开命令行,进入“\extras\demo_suite文件夹”,
输入”.\bandwidthTest.exe”运行bandwidthTest.exe,
显示“Result = PASS”,说明cuDNN安装好了
![](https://aitechtogether.com/wp-content/uploads/2023/04/4c25b1d8-a972-46ec-9f26-8e25912a0c7c.webp)
三、安装Pytorch
1、设置虚拟环境pytorch
在Anaconda下设置虚拟环境,命名为pytorch,使用“activate pytorch”进入环境,“exit()”退出。
使用conda创建pytorch环境,python为3.7版本
conda create -n pytorch python=3.7
查看当前环境
conda info --envs
![](https://aitechtogether.com/wp-content/uploads/2023/04/779b5a28-e49f-43c1-84a8-9231a11b4ae8.webp)
2、确定安装torch和torchvison版本
见链接:https://github.com/pytorch/vision#installation
本机的pytorch环境下,python为3.7.16,对应的torch版本是1.11或1.12,对应的torchvision版本是0.12.0或0.13.0。参考https://blog.csdn.net/love_respect/article/details/124681233。
![windows10操作系统 显卡MX150 安装CUDA+cuDNN+pytorch](https://img-blog.csdnimg.cn/img_convert/a21729b01f3e537b171c7a02fc6be851.png)
3、下载。
下载到本地,从左到右,torch-1.12.0意思torch是1.12版本,cu116说明cuda是11.6版本,cp37意思是python3.7,win_amd64意思windows操作系统64位
![](https://aitechtogether.com/wp-content/uploads/2023/04/02620077-e08d-4fa9-8371-69b669e82124.webp)
4、安装。进入存放上面文件的文件夹,pip install
![](https://aitechtogether.com/wp-content/uploads/2023/04/08946f90-c538-42a0-b13f-a7ce58799b95.webp)
5、检验。
打开Anaconda,进入pytorch环境,输入“pip list”,显示下图说明安装好了。
![](https://aitechtogether.com/wp-content/uploads/2023/04/8a7d5ccc-a64b-40be-832b-73036f82db5d.webp)
四、验证。
打开Anaconda,进入pytorch环境,打开python,输入如下代码,
显示“True”,以上步骤全部成功。
import torch
torch.cuda.is_available()
![](https://aitechtogether.com/wp-content/uploads/2023/04/66de8056-1638-4004-bc85-fd00002e2f6e.webp)
感谢:https://blog.csdn.net/qq_43651945/article/details/123183926?spm=1001.2101.3001.6650.4&utm_medium=distribute.pc_relevant.none-task-blog-2~default~CTRLIST~Rate-4-123183926-blog-84452560.pc_relevant_3mothn_strategy_and_data_recovery&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2~default~CTRLIST~Rate-4-123183926-blog-84452560.pc_relevant_3mothn_strategy_and_data_recovery&utm_relevant_index=7
https://blog.csdn.net/marvel1014/article/details/84452560
https://blog.csdn.net/nini321123/article/details/105286380/
https://www.cnblogs.com/linglanhuakai/p/16334008.html
https://blog.csdn.net/Elon15/article/details/125678410
https://www.jianshu.com/p/50bb9698effd
https://blog.csdn.net/love_respect/article/details/124681233
https://blog.csdn.net/qq_45956730/article/details/126600028
文章出处登录后可见!