cuda与pytorch版本不匹配的问题及解决


问题

cuda与pytorch版本不匹配的问题：

1
2
3
4
5
6
### 输入torch.ones((2,3)).cuda()时，报错。
RuntimeError: The NVIDIA driver on your system is too old (found version 10010).

### 运行代码时，报错。
RuntimeError: CUDA error: CUDA driver version is insufficient for CUDA runtime version
Exception raised from getDevice at /pytorch/c10/cuda/impl/CUDAGuardImpl.h:37 (most recent call first):

解决

  1. 查看torch版本
1
2
3
import torch
print(torch.__version__)  # '1.7.0'
print(torch.version.cuda)  # 10.2
  1. 查看cuda版本
1
2
3
4
5
6
7
8
$ nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Fri_Feb__8_19:08:17_PST_2019
Cuda compilation tools, release 10.1, V10.1.105

$ cat /usr/local/cuda/version.txt
CUDA Version 10.1.105
  1. 卸载pytorch
1
2
pip uninstall pytorch
pip uninstall pytorchvision

  1. 安装新版本torch1.7+cu10.1
1
pip install torch==1.7.0+cu101 torchvision==0.8.1+cu101 torchaudio===0.7.0 -f https://download.pytorch.org/whl/torch_stable.html
  1. 检测
1
2
3
import torch
print(torch.__version__)  # '1.7.0'
print(torch.version.cuda)  # 10.1

完成。