Finetunning AI in containers on HPC