submit.sh

#!/bin/bash
#SBATCH --partition=h100
#SBATCH --gres=gpu:1
#SBATCH --cpus-per-task=24
#SBATCH --time=24:00:00
#SBATCH --job-name=no_TSDL_old_mixtures
#SBATCH --output=no_TSDL_old_mixtures.%j.out

echo "Hello from $(hostname)!"
source "$(conda info --base)/etc/profile.d/conda.sh"
conda activate semhear_emma2

echo "Job running on node: $(hostname)"
echo "SLURM_JOB_ID=$SLURM_JOB_ID"

srun --gres=gpu:1 bash -lc '
  source "$(conda info --base)/etc/profile.d/conda.sh"
  conda activate semhear_emma2

  which python
  python -V

  echo CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES
  python - <<PY
import numpy as np
import torch
print("numpy", np.__version__)
print("torch", torch.__version__)
print("cuda avail", torch.cuda.is_available())
print("device count", torch.cuda.device_count())
PY

  python -m src.training.train experiments/no_TSDL_old_mixtures --use_cuda
' || echo "Command failed, keeping job alive for debugging..."