Create quantized models from Hugging Face

Calculate VRAM for GGUF models using GPU layers and context