hf_quant_config.json

{
  "config_groups": {
    "group_0": {
      "input_activations": {
        "dynamic": false,
        "num_bits": 4,
        "type": "float",
        "group_size": 16
      },
      "weights": {
        "dynamic": false,
        "num_bits": 4,
        "type": "float",
        "group_size": 16
      },
      "targets": [
        "Linear"
      ]
    }
  },
  "ignore": [
    "audio_encoder*",
    "speech_embeddings*",
    "visual*",
    "lm_head",
    "model.embed_tokens",
    "model.norm",
    "model.rotary_emb*",
    "model.swa_rotary_emb*",
    "model.mtp*",
    "model.layers.0*",
    "model.layers.1.self_attn*",
    "model.layers.1.mlp.gate*",
    "model.layers.1.mlp.shared_expert*",
    "model.layers.1.mlp.shared_expert_gate*",
    "model.layers.2.self_attn*",
    "model.layers.2.mlp.gate*",
    "model.layers.2.mlp.shared_expert*",
    "model.layers.2.mlp.shared_expert_gate*",
    "model.layers.3.self_attn*",
    "model.layers.3.mlp.gate*",
    "model.layers.3.mlp.shared_expert*",
    "model.layers.3.mlp.shared_expert_gate*",
    "model.layers.4.self_attn*",
    "model.layers.4.mlp.gate*",
    "model.layers.4.mlp.shared_expert*",
    "model.layers.4.mlp.shared_expert_gate*",
    "model.layers.5.self_attn*",
    "model.layers.5.mlp.gate*",
    "model.layers.5.mlp.shared_expert*",
    "model.layers.5.mlp.shared_expert_gate*",
    "model.layers.6.self_attn*",
    "model.layers.6.mlp.gate*",
    "model.layers.6.mlp.shared_expert*",
    "model.layers.6.mlp.shared_expert_gate*",
    "model.layers.7.self_attn*",
    "model.layers.7.mlp.gate*",
    "model.layers.7.mlp.shared_expert*",
    "model.layers.7.mlp.shared_expert_gate*",
    "model.layers.8.self_attn*",
    "model.layers.8.mlp.gate*",
    "model.layers.8.mlp.shared_expert*",
    "model.layers.8.mlp.shared_expert_gate*",
    "model.layers.9.self_attn*",
    "model.layers.9.mlp.gate*",
    "model.layers.9.mlp.shared_expert*",
    "model.layers.9.mlp.shared_expert_gate*",
    "model.layers.10.self_attn*",
    "model.layers.10.mlp.gate*",
    "model.layers.10.mlp.shared_expert*",
    "model.layers.10.mlp.shared_expert_gate*",
    "model.layers.11.self_attn*",
    "model.layers.11.mlp.gate*",
    "model.layers.11.mlp.shared_expert*",
    "model.layers.11.mlp.shared_expert_gate*",
    "model.layers.12.self_attn*",
    "model.layers.12.mlp.gate*",
    "model.layers.12.mlp.shared_expert*",
    "model.layers.12.mlp.shared_expert_gate*",
    "model.layers.13.self_attn*",
    "model.layers.13.mlp.gate*",
    "model.layers.13.mlp.shared_expert*",
    "model.layers.13.mlp.shared_expert_gate*",
    "model.layers.14.self_attn*",
    "model.layers.14.mlp.gate*",
    "model.layers.14.mlp.shared_expert*",
    "model.layers.14.mlp.shared_expert_gate*",
    "model.layers.15.self_attn*",
    "model.layers.15.mlp.gate*",
    "model.layers.15.mlp.shared_expert*",
    "model.layers.15.mlp.shared_expert_gate*",
    "model.layers.16.self_attn*",
    "model.layers.16.mlp.gate*",
    "model.layers.16.mlp.shared_expert*",
    "model.layers.16.mlp.shared_expert_gate*",
    "model.layers.17.self_attn*",
    "model.layers.17.mlp.gate*",
    "model.layers.17.mlp.shared_expert*",
    "model.layers.17.mlp.shared_expert_gate*",
    "model.layers.18.self_attn*",
    "model.layers.18.mlp.gate*",
    "model.layers.18.mlp.shared_expert*",
    "model.layers.18.mlp.shared_expert_gate*",
    "model.layers.19.self_attn*",
    "model.layers.19.mlp.gate*",
    "model.layers.19.mlp.shared_expert*",
    "model.layers.19.mlp.shared_expert_gate*",
    "model.layers.20.self_attn*",
    "model.layers.20.mlp.gate*",
    "model.layers.20.mlp.shared_expert*",
    "model.layers.20.mlp.shared_expert_gate*",
    "model.layers.21.self_attn*",
    "model.layers.21.mlp.gate*",
    "model.layers.21.mlp.shared_expert*",
    "model.layers.21.mlp.shared_expert_gate*",
    "model.layers.22.self_attn*",
    "model.layers.22.mlp.gate*",
    "model.layers.22.mlp.shared_expert*",
    "model.layers.22.mlp.shared_expert_gate*",
    "model.layers.23.self_attn*",
    "model.layers.23.mlp.gate*",
    "model.layers.23.mlp.shared_expert*",
    "model.layers.23.mlp.shared_expert_gate*",
    "model.layers.24.self_attn*",
    "model.layers.24.mlp.gate*",
    "model.layers.24.mlp.shared_expert*",
    "model.layers.24.mlp.shared_expert_gate*",
    "model.layers.25.self_attn*",
    "model.layers.25.mlp.gate*",
    "model.layers.25.mlp.shared_expert*",
    "model.layers.25.mlp.shared_expert_gate*",
    "model.layers.26.self_attn*",
    "model.layers.26.mlp.gate*",
    "model.layers.26.mlp.shared_expert*",
    "model.layers.26.mlp.shared_expert_gate*",
    "model.layers.27.self_attn*",
    "model.layers.27.mlp.gate*",
    "model.layers.27.mlp.shared_expert*",
    "model.layers.27.mlp.shared_expert_gate*",
    "model.layers.28.self_attn*",
    "model.layers.28.mlp.gate*",
    "model.layers.28.mlp.shared_expert*",
    "model.layers.28.mlp.shared_expert_gate*",
    "model.layers.29.self_attn*",
    "model.layers.29.mlp.gate*",
    "model.layers.29.mlp.shared_expert*",
    "model.layers.29.mlp.shared_expert_gate*",
    "model.layers.30.self_attn*",
    "model.layers.30.mlp.gate*",
    "model.layers.30.mlp.shared_expert*",
    "model.layers.30.mlp.shared_expert_gate*",
    "model.layers.31.self_attn*",
    "model.layers.31.mlp.gate*",
    "model.layers.31.mlp.shared_expert*",
    "model.layers.31.mlp.shared_expert_gate*",
    "model.layers.32.self_attn*",
    "model.layers.32.mlp.gate*",
    "model.layers.32.mlp.shared_expert*",
    "model.layers.32.mlp.shared_expert_gate*",
    "model.layers.33.self_attn*",
    "model.layers.33.mlp.gate*",
    "model.layers.33.mlp.shared_expert*",
    "model.layers.33.mlp.shared_expert_gate*",
    "model.layers.34.self_attn*",
    "model.layers.34.mlp.gate*",
    "model.layers.34.mlp.shared_expert*",
    "model.layers.34.mlp.shared_expert_gate*",
    "model.layers.35.self_attn*",
    "model.layers.35.mlp.gate*",
    "model.layers.35.mlp.shared_expert*",
    "model.layers.35.mlp.shared_expert_gate*",
    "model.layers.36.self_attn*",
    "model.layers.36.mlp.gate*",
    "model.layers.36.mlp.shared_expert*",
    "model.layers.36.mlp.shared_expert_gate*",
    "model.layers.37.self_attn*",
    "model.layers.37.mlp.gate*",
    "model.layers.37.mlp.shared_expert*",
    "model.layers.37.mlp.shared_expert_gate*",
    "model.layers.38.self_attn*",
    "model.layers.38.mlp.gate*",
    "model.layers.38.mlp.shared_expert*",
    "model.layers.38.mlp.shared_expert_gate*",
    "model.layers.39.self_attn*",
    "model.layers.39.mlp.gate*",
    "model.layers.39.mlp.shared_expert*",
    "model.layers.39.mlp.shared_expert_gate*",
    "model.layers.40.self_attn*",
    "model.layers.40.mlp.gate*",
    "model.layers.40.mlp.shared_expert*",
    "model.layers.40.mlp.shared_expert_gate*",
    "model.layers.41.self_attn*",
    "model.layers.41.mlp.gate*",
    "model.layers.41.mlp.shared_expert*",
    "model.layers.41.mlp.shared_expert_gate*",
    "model.layers.42.self_attn*",
    "model.layers.42.mlp.gate*",
    "model.layers.42.mlp.shared_expert*",
    "model.layers.42.mlp.shared_expert_gate*",
    "model.layers.43.self_attn*",
    "model.layers.43.mlp.gate*",
    "model.layers.43.mlp.shared_expert*",
    "model.layers.43.mlp.shared_expert_gate*",
    "model.layers.44.self_attn*",
    "model.layers.44.mlp.gate*",
    "model.layers.44.mlp.shared_expert*",
    "model.layers.44.mlp.shared_expert_gate*",
    "model.layers.45.self_attn*",
    "model.layers.45.mlp.gate*",
    "model.layers.45.mlp.shared_expert*",
    "model.layers.45.mlp.shared_expert_gate*",
    "model.layers.46.self_attn*",
    "model.layers.46.mlp.gate*",
    "model.layers.46.mlp.shared_expert*",
    "model.layers.46.mlp.shared_expert_gate*",
    "model.layers.47.self_attn*",
    "model.layers.47.mlp.gate*",
    "model.layers.47.mlp.shared_expert*",
    "model.layers.47.mlp.shared_expert_gate*"
  ],
  "quant_algo": "NVFP4",
  "producer": {
    "name": "modelopt",
    "version": "0.39.0.dev290+gf9d9a71de.d20260407"
  },
  "quant_method": "modelopt"
}