torch.cuda.is_available() == False | Runpod | Page 1

low blaze Mar 26, 2026, 12:28 AM

#

I created a brand new pod and proceeded to have this interaction:

root@86a83fe38a36:/# python
Python 3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
/usr/local/lib/python3.12/dist-packages/torch/cuda/__init__.py:182: UserWarning: CUDA initialization: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero. (Triggered internally at /pytorch/c10/cuda/CUDAFunctions.cpp:109.)
  return torch._C._cuda_getDeviceCount() > 0
False

It seems that every new pod I make exhibits this behavior

📎 message.txt

#

keen coral Mar 26, 2026, 4:20 AM

#

can you run this and paste here the output?

echo "CUDA_VISIBLE_DEVICES='${CUDA_VISIBLE_DEVICES:-}'" echo "NVIDIA_VISIBLE_DEVICES='${NVIDIA_VISIBLE_DEVICES:-}'" echo "NVIDIA_DRIVER_CAPABILITIES='${NVIDIA_DRIVER_CAPABILITIES:-}'" ls -la /dev/nvidia* 2>/dev/null || true

low blaze Mar 26, 2026, 4:24 PM

#

I'm gonna be so honest, I don't want to keep starting pods until I know that this just will work. If Runpod wants to give me the credits to investigate this sure, but otherwise this should just work on a brand new pod and I don't want to waste the money

quiet nacelle Mar 26, 2026, 5:15 PM

#

@low blaze I just credited you for the pod - you may want to try starting a new one to work around it, we'll look into this machine

#torch.cuda.is_available() == False