{
  "architectures": [
    "AndreaForCausalLM"
  ],
  "model_type": "andrea",
  "auto_map": {
    "AutoConfig": "configuration_andrea.AndreaConfig",
    "AutoModelForCausalLM": "modeling_andrea.AndreaForCausalLM"
  },
  "vocab_size": 24833,
  "hidden_size": 768,
  "intermediate_size": 3072,
  "num_hidden_layers": 15,
  "num_attention_heads": 12,
  "max_position_embeddings": 1024,
  "hidden_act": "relu",
  "position_embedding_type": "learned_absolute",
  "normalization": "rmsnorm",
  "rms_norm_eps": 1e-05,
  "norm_has_gain": false,
  "norm_after_embedding": true,
  "final_norm": true,
  "qk_norm": true,
  "sliding_window": null,
  "attention_bias": false,
  "mlp_bias": false,
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "bos_token_id": 24832,
  "eos_token_id": 24832,
  "tokenizer_class": "AndreaHarrisTokenizer",
  "andrea": {
    "model_name": "ANDREA-120M-ft",
    "tokenizer": "harris-successor-variety",
    "harris_segments": 24576,
    "trained_steps": 852106,
    "num_params": 145098240,
    "source_checkpoint": "step_852106.bin"
  }
}