Llama 30b model using lora tuning for NeurIPS Challenge