llm Fine-tuning Zephyr 7B GPTQ with 4-Bit Quantization for Custom Data and Inference Harnessing Zephyr's Breeze: DPO Training on Mistral-7B-GPTQ for Language Model Alignment