4. Why is quantization essential for improving the computational efficiency of generative AI models? What are the different types of quantization techniques used in generative AI models? Please calculate the GPU VRAM requirements for a Large Language Model (LLM) with 7 billion parameters using three different data formats: FP32, INT8, and INT4. Provide the specific VRAM values in gigabytes for each format and explain how the memory requirement changes with each data format. (20%)