progress/github/SpecForge/configs/deepseek-v3-671b-eagle3.json

{
  "architectures": [
    "LlamaForCausalLMEagle3"
  ],
  "eagle_config": {
    "eagle_aux_hidden_state_layer_ids": [
      1,
      29,
      57
    ],
    "use_aux_hidden_state": true
  },
  "bos_token_id": 151643,
  "eos_token_id": 151645,
  "hidden_act": "silu",
  "hidden_size": 7168,
  "initializer_range": 0.02,
  "intermediate_size": 40960,
  "max_position_embeddings": 163840,
  "model_type": "llama",
  "num_attention_heads": 56,
  "num_key_value_heads": 8,
  "num_hidden_layers": 1,
  "pad_token_id": 0,
  "rms_norm_eps": 1e-05,
  "tie_word_embeddings": false,
  "torch_dtype": "float16",
  "transformers_version": "4.51.0",
  "use_cache": true,
  "vocab_size": 129280,
  "draft_vocab_size": 32000
}