text_encoder/hf_quant_config.json

{
    "producer": {
        "name": "modelopt",
        "version": "0.45.0rc1.dev28+gc88b62bee.d20260625"
    },
    "quantization": {
        "quant_algo": "FP8_PER_CHANNEL_PER_TOKEN",
        "kv_cache_quant_algo": null,
        "exclude_modules": [
            "model.language_model.embed_tokens",
            "model.language_model.layers.17.mlp.down_proj",
            "model.language_model.layers.18.mlp.down_proj",
            "model.language_model.layers.19.mlp.down_proj",
            "model.language_model.layers.21.mlp.down_proj",
            "model.language_model.layers.22.mlp.down_proj",
            "model.language_model.layers.23.mlp.down_proj",
            "model.language_model.layers.24.mlp.down_proj",
            "model.language_model.layers.24.self_attn.o_proj",
            "model.language_model.layers.25.mlp.down_proj",
            "model.language_model.layers.25.self_attn.o_proj",
            "model.language_model.layers.26.mlp.down_proj",
            "model.language_model.layers.27.mlp.down_proj",
            "model.language_model.layers.28.mlp.down_proj",
            "model.language_model.layers.29.mlp.down_proj",
            "model.language_model.layers.30.mlp.down_proj",
            "model.language_model.layers.30.self_attn.o_proj",
            "model.language_model.layers.31.self_attn.o_proj",
            "model.language_model.layers.34.mlp.down_proj",
            "model.language_model.layers.35.self_attn.o_proj",
            "model.language_model.layers.36.mlp.down_proj",
            "model.language_model.layers.37.self_attn.o_proj",
            "model.language_model.layers.38.self_attn.o_proj",
            "model.language_model.layers.40.self_attn.o_proj",
            "model.language_model.layers.42.self_attn.o_proj",
            "model.language_model.layers.43.self_attn.o_proj",
            "model.language_model.layers.44.self_attn.o_proj",
            "model.language_model.layers.45.self_attn.o_proj",
            "model.language_model.layers.46.mlp.down_proj",
            "model.language_model.layers.46.self_attn.o_proj",
            "model.language_model.layers.47.self_attn.o_proj",
            "model.language_model.layers.48.mlp.down_proj",
            "model.language_model.layers.48.self_attn.o_proj",
            "model.language_model.layers.49.mlp.down_proj",
            "model.language_model.layers.49.self_attn.o_proj"
        ]
    }
}