Create quantized models from Hugging Face repos

Calculate VRAM for GGUF models using GPU layers and context