fix: define model_version so load_tokenizer resolves it

fix: restore generation_config fields lost on quantized save

Upload folder using huggingface_hub

initial commit