Cuda not available after isntalling the environment

Question

Cuda not available after isntalling the environment

CristopherUbotica opened this issue 10 months ago · 4 comments

CristopherUbotica commented 10 months ago

Hello,

I installed the environment from the .yaml file following the procedure described in the readme, and in my case cuda was not available. Is it problem of the environment? Or am I missing something?

Thank you in advance for your time and consideration.

Kind regards,

Cristopher

Answer 1 · 2024-03-22T11:55:47.000Z

Hi Cris,
could you detail the OS, and the steps you have done so we can replicate your bug?

Answer 2 · 2024-03-22T12:45:25.000Z

Hi Roberto,

I am using Ubuntu 22.04. I run the the 'pyraws_install.sh' shell script to set up the environment. After that, I activated the environment and tried torch.cuda.is_available() on Python.

Can the problem be that the cuda version that I have installed in my OS is not compatible with the environment version? Or should the environment be enough to run cuda in pytorch? In that case, the problem wouldn't be part of the repo.

Thank you in advance!

Answer 3 · 2024-03-23T00:36:45.000Z

Seems to me an error related to your cuda version. Can you try to install NVIDIA drivers compatible with CUDA 10.2 or 11.3.
Another easy solution is use the dockerized application.

Answer 4 · 2024-03-23T13:39:16.000Z

@CristopherUbotica, I uploaded a docker-compose, a devcontainer.json, and updated the installation routines. The project has been tested up to cudatoolkit 11.3 and pytorch 1.12.0.

Let me know if that was useful.