hf_quant_config.json

{
    "producer": {
        "name": "modelopt",
        "version": "0.47.0.dev45+g9b8caf623"
    },
    "quantization": {
        "quant_algo": "MXFP8",
        "kv_cache_quant_algo": "FP8",
        "group_size": 32,
        "exclude_modules": [
            "lm_head",
            "model.embed_tokens"
        ]
    }
}