{
    "model_name": "meta-llama--Llama-3.2-3B-Instruct",
    "model_type": "llm-llama",
    "vm_cfg": null,
    "mm_cfg": null,
    "lm_cfg": {
        "model_type": "llama",
        "data_type": "bfloat16",
        "arch": "llama",
        "token_cfg": {
            "vocab_size": 128256
        },
        "rope_cfg": {
            "rope_theta": 500000.0,
            "rope_local_base_freq": 500000.0,
            "rope_dimension_count": 128,
            "sliding_rope_dimension_count": null,
            "rope_scaling": {
                "factor": 32.0,
                "low_freq_factor": 1.0,
                "high_freq_factor": 4.0,
                "original_max_position_embeddings": 8192,
                "long_factor": null,
                "short_factor": null,
                "rope_type": "llama3",
                "mrope_section": null,
                "mrope_interleaved": false
            }
        },
        "attn_cfg": {
            "num_attention_heads": 24,
            "num_key_value_heads": 8,
            "head_dim": 128,
            "sliding_head_dim": null,
            "swa_enable": false,
            "sliding_window": 0,
            "attention_bias": false,
            "attention_dropout": 0.0,
            "query_pre_attn_scalar": 0
        },
        "mlp_cfg": {
            "intermediate_size": 8192,
            "act": "silu",
            "num_layers": 3,
            "mlp_bias": false
        },
        "hidden_size": 3072,
        "num_hidden_layers": 28,
        "max_position_embeddings": 2048,
        "rms_norm_eps": 1e-05,
        "rms_norm_unit_offset": false,
        "layer_types": [
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention",
            "full_attention"
        ],
        "hidden_size_per_layer_input": 0,
        "num_kv_shared_layers": 0,
        "use_double_wide_mlp": false,
        "attn_logit_softcapping": null,
        "final_logit_softcapping": null,
        "lm_head_num_splits": 2,
        "lm_head_split_dim": 64128,
        "draft_vocab_size": 0,
        "conv_L_cache": 3,
        "conv_bias": false,
        "lora_cfg": null,
        "speculative_decoding_cfg": null
    },
    "pipeline_cfg": {
        "system_prompt": null,
        "chat_template": null,
        "max_num_tokens": 2048,
        "input_token_group_size": 320,
        "input_token_group_offsets": [
            0,
            320,
            640,
            704,
            960,
            1280,
            1600,
            1920
        ],
        "future_token_mask_size": 128,
        "return_logits": true,
        "enable_filter_sharing": false,
        "quantize_embeddings": false,
        "quantize_kv_cache": false,
        "split_mlp": true
    },
    "language_model_name": "meta-llama--Llama-3.2-3B-Instruct_language"
}