{
    "producer": {
        "name": "modelopt",
        "version": "0.25.0"
    },
    "architecture": "LlamaForCausalLM",
    "dtype": "float16",
    "logits_dtype": "float16",
    "num_hidden_layers": 28,
    "num_attention_heads": 24,
    "num_key_value_heads": 8,
    "hidden_size": 3072,
    "norm_epsilon": 1e-05,
    "vocab_size": 128256,
    "max_position_embeddings": 131072,
    "hidden_act": "silu",
    "use_parallel_embedding": true,
    "embedding_sharding_dim": 0,
    "head_size": 128,
    "intermediate_size": 8192,
    "position_embedding_type": "rope_gpt_neox",
    "share_embedding_table": false,
    "residual_mlp": false,
    "bias": false,
    "rotary_pct": 1.0,
    "rank": 0,
    "decoder": "llama",
    "rmsnorm": true,
    "lm_head_bias": false,
    "mlp_bias": false,
    "attn_bias": false,
    "rotary_base": 500000.0,
    "rotary_scaling": {
        "factor": 32.0,
        "high_freq_factor": 4.0,
        "low_freq_factor": 1.0,
        "original_max_position_embeddings": 8192,
        "rope_type": "llama3"
    },
    "disable_weight_only_quant_plugin": false,
    "moe": {
        "num_experts": 0,
        "shared_expert_intermediate_size": 0,
        "top_k": 0,
        "normalization_mode": null,
        "sparse_mixer_epsilon": 0.01,
        "tp_mode": 0,
        "device_limited_n_group": 0,
        "device_limited_topk_group": 0,
        "device_limited_routed_scaling_factor": 1.0
    },
    "remove_duplicated_kv_heads": false,
    "fc_after_embed": false,
    "use_input_layernorm_in_first_layer": true,
    "use_last_layernorm": true,
    "layer_idx_offset": 0,
    "embedding_multiplier": 1.0,
    "attention_multiplier": 1.0,
    "residual_multiplier": 1.0,
    "output_multiplier_scale": 1.0,
    "runtime_defaults": null,
    "mapping": {
        "world_size": 1,
        "gpus_per_node": 8,
        "cp_size": 1,
        "tp_size": 1,
        "pp_size": 1,
        "moe_tp_size": 1,
        "moe_ep_size": 1,
        "auto_parallel": false
    },
    "quantization": {
        "quant_algo": "W4A16_AWQ",
        "kv_cache_quant_algo": "INT8",
        "group_size": 128,
        "smoothquant_val": 0.5,
        "clamp_val": null,
        "use_meta_recipe": false,
        "has_zero_point": false,
        "pre_quant_scale": true,
        "exclude_modules": [
            "transformer.layers.17.input_layernorm",
            "transformer.layers.1.input_layernorm",
            "transformer.layers.24.input_layernorm",
            "transformer.layers.1.post_layernorm",
            "transformer.layers.6.input_layernorm",
            "transformer.layers.14.post_layernorm",
            "lm_head",
            "transformer.layers.19.post_layernorm",
            "transformer.layers.22.input_layernorm",
            "transformer.layers.13.input_layernorm",
            "transformer.layers.4.post_layernorm",
            "transformer.layers.16.post_layernorm",
            "transformer.layers.8.input_layernorm",
            "transformer.layers.13.post_layernorm",
            "transformer.layers.20.post_layernorm",
            "transformer.layers.5.input_layernorm",
            "transformer.layers.12.post_layernorm",
            "transformer.layers.27.input_layernorm",
            "transformer.layers.14.input_layernorm",
            "transformer.layers.23.input_layernorm",
            "transformer.layers.10.input_layernorm",
            "transformer.layers.4.input_layernorm",
            "transformer.layers.7.post_layernorm",
            "transformer.layers.18.post_layernorm",
            "transformer.layers.15.post_layernorm",
            "transformer.layers.25.input_layernorm",
            "transformer.layers.21.input_layernorm",
            "transformer.layers.9.post_layernorm",
            "transformer.layers.11.post_layernorm",
            "transformer.layers.24.post_layernorm",
            "transformer.layers.25.post_layernorm",
            "transformer.layers.27.post_layernorm",
            "transformer.layers.26.post_layernorm",
            "transformer.layers.0.post_layernorm",
            "transformer.layers.9.input_layernorm",
            "transformer.layers.10.post_layernorm",
            "transformer.layers.20.input_layernorm",
            "transformer.layers.15.input_layernorm",
            "transformer.layers.7.input_layernorm",
            "transformer.layers.2.input_layernorm",
            "transformer.layers.12.input_layernorm",
            "transformer.layers.5.post_layernorm",
            "transformer.layers.3.post_layernorm",
            "transformer.layers.6.post_layernorm",
            "transformer.layers.17.post_layernorm",
            "transformer.layers.8.post_layernorm",
            "transformer.layers.18.input_layernorm",
            "transformer.layers.2.post_layernorm",
            "transformer.layers.11.input_layernorm",
            "transformer.layers.22.post_layernorm",
            "transformer.ln_f",
            "transformer.layers.21.post_layernorm",
            "transformer.layers.26.input_layernorm",
            "transformer.layers.19.input_layernorm",
            "transformer.layers.23.post_layernorm",
            "transformer.layers.3.input_layernorm",
            "transformer.layers.16.input_layernorm",
            "transformer.layers.0.input_layernorm",
            "transformer.vocab_embedding"
        ]
    },
    "qk_layernorm": false,
    "rotary_embedding_dim": 128,
    "tie_word_embeddings": true,
    "model_type": "llama"
}