| #SBATCH --job-name=qwen-lora-finetune # Descriptive job name | |
| #SBATCH --partition=h200_normal_q # or a100_normal_q depending on GPU availability | |
| #SBATCH --account=ece_6514 # Your ARC account | |
| #SBATCH --output=logs/qwen_lora_%j.out | |
| #SBATCH --error=logs/qwen_lora_%j.err | |
| #SBATCH --time=1:00:00 # 24-hour wall time | |
| #SBATCH --gres=gpu:1 # Request 1 GPU | |
| #SBATCH --mem=16G | |
| #SBATCH --ntasks-per-node=2 # Two CPU tasks for dataloader | |
| # Load necessary modules and activate conda environment | |
| module purge | |
| # Load modules | |
| module --ignore_cache load Miniconda3/24.7.1-0 | |
| module --ignore_cache load PyTorch/2.1.2-foss-2023a | |
| # Source conda for script | |
| source /cm/local/apps/Miniconda3/24.7.1-0/etc/profile.d/conda.sh | |
| conda activate cvenv | |
| # Safety: unset PYTHONPATH to avoid HPC NumPy conflicts | |
| unset PYTHONPATH | |
| # Start the fine-tuning process | |
| echo "🚀 Starting fine-tuning job..." | |
| python sft_lora_finetuner.py | |
| echo "✅ Fine-tuning completed." |