下载官网
查看 win 10 的cuda版本参考此百度经验
我们需要在此网站安装这个版本的pytorch-1.9.1-py3.8_cuda11.1_cudnn8_0.tar.bz2
或者参考此篇直接安装
1. CPU版本
先下载一下pytorch-1.9.1-py3.9_cpu_0.tar.bz2
试一下,共189MB,将安装包移动到...\Anaconda3\pkgs
中,解压一下,
下面专门新建(克隆)一个深度学习的虚拟环境:
conda create -n deep_learning --clone env1
conda activate deep_learning
cd
到...\Anaconda3\pkgs
中
conda install --use-local pytorch-1.9.1-py3.8_cpu_0.tar.bz2
Downloading and Extracting Packages
############################################################################################################### | 100%
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
在pycharm中切换到相应的环境,输入:
import torch
print(torch.__version__)
print(torch.cuda.is_available())
print(torch.cuda.get_device_name(0))
报错:
OSError: [WinError 126] 找不到指定的模块。 Error loading "C:\Users\...\Anaconda3\envs\deep_learning\lib\site-packages\torch\lib\torch_python.dll" or one of its dependencies.
Microsoft Visual C++ 2015-2019 Redistributable (x64) - 14.29.30135,重启一下也不行,重新打开官网,输入:
conda install pytorch torchvision torchaudio cpuonly -c pytorch
1.9.1
False
Traceback (most recent call last):
File "C:/Users/HCSI/PycharmProjects/pythonProject1/main.py", line 5, in <module>
print(torch.cuda.get_device_name(0))
File "C:\Users\HCSI\Anaconda3\envs\deep_learning\lib\site-packages\torch\cuda\__init__.py", line 279, in get_device_name
return get_device_properties(device).name
File "C:\Users\HCSI\Anaconda3\envs\deep_learning\lib\site-packages\torch\cuda\__init__.py", line 309, in get_device_properties
_lazy_init() # will define _get_device_properties
File "C:\Users\HCSI\Anaconda3\envs\deep_learning\lib\site-packages\torch\cuda\__init__.py", line 166, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled
安装成功,这是一个CPU的版本的pytorch
2. GPU版本
换一台有GPU的电脑,下载pytorch-1.9.1-py3.8_cuda11.1_cudnn8_0.tar.bz2
,共1.6GB
查看所有环境:
conda info -e
# conda environments:
#
base * E:\Python\Anaconda3
env1 E:\Python\Anaconda3\envs\env1
和上面一样,克隆环境并激活:
(base) C:\Users\SiCheng Yang>conda activate deep_learning
(deep_learning) C:\Users\SiCheng Yang>
将下载好的包放到...\Anaconda3\pkgs
中,注意要在urls
文件中添加下载路径:https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/win-64/pytorch-1.9.1-py3.8_cuda11.1_cudnn8_0.tar.bz2
同样的:
conda install --use-local pytorch-1.9.1-py3.8_cuda11.1_cudnn8_0.tar.bz2
解压过程大概10分钟左右:
(deep_learning) E:\Python\Anaconda3\pkgs>conda install --use-local pytorch-1.9.1-py3.8_cuda11.1_cudnn8_0.tar.bz2
Downloading and Extracting Packages
############################################################################################################### | 100%
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
测试一下:
>>> import torch
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "E:\Python\Anaconda3\envs\deep_learning\lib\site-packages\torch\__init__.py", line 135, in <module>
raise err
OSError: [WinError 126] 找不到指定的模块。 Error loading "E:\Python\Anaconda3\envs\deep_learning\lib\site-packages\torch\lib\caffe2_detectron_ops_gpu.dll" or one of its dependencies.
再进官网,输入:
conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch -c conda-forge
原来需要重新下载一个1.2GB的cudatoolkit=11.1.1,那还不如直接在官网进行下载和安装呢…
测试结果如下:
>>> import torch
>>> print(torch.__version__)
1.9.1
>>> print(torch.cuda.is_available())
True
>>> print(torch.cuda.get_device_name(0))
NVIDIA GeForce GTX 1050 Ti
总结
能在win10电脑上顺利安装torch并运行