No CUDA runtime is found, using CUDA_HOME='/usr'

[franzhd]

Hello everyone, i have this error: "No CUDA runtime is found, using CUDA_HOME='/usr' "

each time i import stuff from slayer like "import lava.lib.dl.slayer as slayer" . Asking torch.cuda.is_available() in a python3 shell inside the env does tell me that CUDA is available. pytest pass all necessary steps. Anyone has suggestion on how to resolve this problem? thanks for your attention!

Objective of issue:

Lava DL version:

• [ ] 0.3.0 (feature release)
• [ ] 0.2.1 (bug fixes)
• [x] 0.2.0 (current version)
• [ ] 0.1.2

Lava version:

• [ ] 0.4.0 (feature release)
• [ ] 0.3.1 (bug fixes)
• [x] 0.3.0 (current version)
• [ ] 0.2.0
• [ ] 0.1.2

I'm submitting a ...

• [x] bug report
• [ ] feature request
• [ ] documentation request

Current behavior:

Expected behavior:

Steps to reproduce:

Related code:

insert short code snippets here

Other information:

insert the output from lava debug here

[bamsumit]

Can you post the outputs of these commands

nvidia-smi | grep CUDA
nvcc --version
python -c "import torch; print(torch.__version__)"

[franzhd]

there are mismatching version of CUDA, but as i know the version in the nvidia-smi command is only the latest version supported by that version of nvidia drivers, while the version suggested by nvcc is the actual version installed. is it correct?

[bamsumit]

As long as nvcc and torch CUDA version match, it should be okay.

Can you try this:

1. Find cuda installation path: which nvcc
2. Make sure <cuda_path>/home exists (try ls ...)
3. export CUDA_HOME=<cuda_path>/home

and try running the unittests. python -m unittest

[bamsumit]

Hi @franzhd any progress?

[bamsumit]

Closed as STALE