Why 'torch.cuda.is_available()' still returned 'False' even I installed CUDA 9.0.176.....Help

1

Hi,

I just got a very strange problem:

I used the commands: “print(torch. version )
print(torch.version.cuda)
print(torch.bac:sob:kends.cudnn.version())”, the output were:

0.4.1.post2 (PyTorch version)
9.0.176 (CUDA version)
7.1.2 (cuDNN version)

I also used the “nvcc --version”, the results were:
Capture4

It was very strange, it seems that the CUDA version installed in Linux system is 9.0.176, and the CUDA that the PyTorch needed is also 9.0.176, but, the “cuda.is_available()” still returns " False ".

Could anyone help me with this?:sob: Many thanks!

Environment

  • PyTorch Version (e.g., 1.0): 0.4.1.post2
  • OS (e.g., Linux): Linux
  • Python version: 3.6.5 : : Anaconda, Inc.
  • CUDA/cuDNN version: 9.0.176 (CUDA version); 7.1.2 (cuDNN version)
1 Like
2

Can you try PyTorch 1.0.0, i.e. the latest version of PyTorch?

Also, is the nvidia driver the latest version on your machine?

3

Hi Smth,

Many thansk for your help. As I am using the University’s HPC machine-I don’t have permission to install any software unless report this issue to staff and they may update it after Christmas holiday.:sob:

In addition, I also use "nvidia-smi’ command to check the NVIDIA driver version, however, it shown: command not found… I am not sure the version of NVIDIA driver.

May I ask if this issue mainly caused by the incompatible versions?

Thanks.

4

Hi Smth,

Thanks for your reply. May I ask if there is any other possible solution for this?

Thanks.

5

Hi Smth,

Many thansk for your help. As I am using the University’s HPC machine-I don’t have permission to install any software unless report this issue to staff and they may update it after Christmas holiday.:sob:

In addition, I also use "nvidia-smi’ command to check the NVIDIA driver version, however, it shown: command not found… I am not sure the version of NVIDIA driver.

May I ask if this issue mainly caused by the incompatible versions?

Thanks.

7

One solution is to try the CUDA8 version of PyTorch that you can install by selecting it from https://pytorch.org

torch.cuda.is_available() can return false if you dont have a new enough NVIDIA driver that your CUDA version needs.

1 Like
8

Hi Smth,

Many thanks for your help. I will try it.:slightly_smiling_face:

9

Hello,
I am facing the same problem and tried all the options discussed here, But did not work. Can somebody please help me to solve this problem.

OS: linux
pytorchVesion: 1.0.1
Python version : 3.7.3
Cuda : 9.0.176
Driver vesion: 396.24

Thank you

10

@TMY
Answered here.

1 Like
11

really helped a lot! after upgrading to the latest NVIDIA driver, torch.cuda.is_available() returns True!