GPU Troubleshooting¶
Common GPU Issues¶
1. No GPUs Available¶
Symptom: Job stuck in PENDING state
Check:
sinfo --partition=<valid_gpu_partition>
nvidia-smi # on a GPU node
Solutions:
Wait for GPUs to become available
Try a different GPU partition
Request during off-peak hours
2. CUDA Out of Memory¶
Symptom: RuntimeError: CUDA out of memory
Solutions:
Reduce batch size
Clear cache
Spread work across multiple GPUs
3. GPU Not Visible¶
Symptom: torch.cuda.is_available() returns False
Check:
nvidia-smi
python -c "import torch; print(torch.cuda.device_count())"
Solutions:
Verify
--gres=gpu:1in job scriptCheck you’re on a GPU node
Reload CUDA module
4. Wrong GPU Type¶
Symptom: Job runs on wrong GPU or fails
Solution: Specify GPU type explicitly:
#SBATCH --gres=gpu:a100:1
#SBATCH --partition=gpu-a100
5. MIG Configuration Issues¶
Symptom: MIG jobs fail or show wrong memory
Check:
nvidia-smi -q | grep -A 5 "MIG"
Solution: Ensure MIG request matches available configurations