docs: proper model card with base-model link + inference notes

fix: coerce quantization_config.bits to int + add bits_per_weight

Upload folder using huggingface_hub

initial commit