Add full MLX mixed-4/8-bit quantization (5 shards)
Upload README.md with huggingface_hub
initial commit