Fine-Tuning of Llama-2 7B Chat for Python Code Generation: Using QLoRA, SFTTrainer, and Gradient Checkpointing on the Alpaca-14k Dataset