{
  "architectures": [
    "LlamaForCausalLMEagle3"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": 1,
  "draft_vocab_size": 32000,
  "dtype": "bfloat16",
  "eagle_config": {
    "eagle_aux_hidden_state_layer_ids": [
      1,
      21,
      40
    ],
    "use_aux_hidden_state": true,
    "use_input_layernorm_in_first_layer": true,
    "use_last_layernorm": true,
    "use_mtp_layernorm": false
  },
  "eos_token_id": 2,
  "fc_norm": true,
  "head_dim": 128,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "initializer_range": 0.02,
  "intermediate_size": 12288,
  "max_position_embeddings": 16384,
  "mlp_bias": false,
  "model_type": "llama",
  "norm_output": true,
  "num_attention_heads": 32,
  "num_aux_hidden_states": 3,
  "num_hidden_layers": 1,
  "num_key_value_heads": 8,
  "pretraining_tp": 1,
  "rms_norm_eps": 1e-06,
  "rope_scaling": null,
  "rope_theta": 1000000,
  "target_hidden_size": 4096,
  "target_model_type": "deepseek_v4",
  "tie_word_embeddings": false,
  "transformers_version": "4.57.6",
  "use_cache": true,
  "vocab_size": 129280
}