anaconda命令行下检查:
|
1
2
3
4
5
6
7
8
9
10
11
12
13
|
(base) PS C:\Users\chenxuqi> conda deactivatePS C:\Users\chenxuqi> conda activate ssd(ssd) PS C:\Users\chenxuqi> pythonPython 3.6.12 |Anaconda, Inc.| (default, Sep 9 2020, 00:29:25) [MSC v.1916 64 bit (AMD64)] on win32Type "help", "copyright", "credits" or "license" for more information.>>> import torch>>> print(torch.__version__)0.4.0>>> print(torch.cuda.is_available())True>>>>>>>>> |
cmd命令行下检查cuda安装:
|
1
2
3
4
5
6
7
8
9
10
11
|
Microsoft Windows [版本 10.0.18363.1139](c) 2019 Microsoft Corporation。保留所有权利。C:\Users\chenxuqi>nvcc --versionnvcc: NVIDIA (R) Cuda compiler driverCopyright (c) 2005-2017 NVIDIA CorporationBuilt on Fri_Sep__1_21:08:32_Central_Daylight_Time_2017Cuda compilation tools, release 9.0, V9.0.176C:\Users\chenxuqi>C:\Users\chenxuqi> |
执行使用GPU的代码:
|
1
2
3
4
5
6
7
8
9
10
11
12
|
import timeimport torch ##################################################for i in range(1,10): start = time.time() a = torch.FloatTensor(i*100,1000,1000) a = a.cuda() #a = a a = torch.matmul(a,a) end = time.time() - start print(end) |
执行结果:
注意，这里显存太小，溢出了...但是安装是成功的...
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
|
Windows PowerShell尝试新的跨平台 PowerShell https://aka.ms/pscore6PS C:\Users\chenxuqi\Desktop\新建文件夹> & 'D:\Anaconda3\envs\ssd\python.exe' 'c:\Users\chenxuqi\.vscode\extensions\ms-python.python-2020.10.332292344\pythonFiles\lib\python\debugpy\launcher' '50571' '--' 'c:\Users\chenxuqi\Desktop\新建文件夹\testGPU.py'3.62603592872619630.63051700592041020.90559673309326171.31999874114990231.59791398048400882.0483360290527344THCudaCheck FAIL file=c:\programdata\miniconda3\conda-bld\pytorch_1524549877902\work\aten\src\thc\generic/THCStorage.cu line=58 error=2 : out of memoryTraceback (most recent call last): File "c:\Users\chenxuqi\Desktop\新建文件夹\testGPU.py", line 10, in <module> a = torch.matmul(a,a)RuntimeError: cuda runtime error (2) : out of memory at c:\programdata\miniconda3\conda-bld\pytorch_1524549877902\work\aten\src\thc\generic/THCStorage.cu:58PS C:\Users\chenxuqi\Desktop\新建文件夹> conda activate ssdPS C:\Users\chenxuqi\Desktop\新建文件夹> |
补充：pytorch离线安装，验证gpu版安装成功
使用conda 命令在线安装pytorch会下载中断，添加pip清华大学镜像源https://pypi.tuna.tsinghua.edu.cn/simple/中下好torch,安装包，然后使用pip命令
|
1
|
pip install "下在的安装包的路径" |
例如：
|
1
|
pip install "C:\Users\28614\Desktop\pytorch-nightly-cpu-1.0.0.dev20181222-py3.7_cpu_0.tar.bz2" |
安装gpu版，需要安装cuda,和cudnn。
验证gpu版是否安装成功
|
1
2
|
import torchprint(torch.cuda.is_available()) |
返回True，则安装成功
以上为个人经验，希望能给大家一个参考，也希望大家多多支持服务器之家。如有错误或未考虑完全的地方，望不吝赐教。
原文链接：https://blog.csdn.net/m0_46653437/article/details/109575389
