{
  "model_type": "ana-lm",
  "architectures": [
    "ANALMForCausalLM"
  ],
  "auto_map": {
    "AutoConfig": "model.ANALMConfig",
    "AutoModelForCausalLM": "model.ANALMForCausalLM"
  },
  "vocab_size": 32000,
  "layers": 12,
  "num_hidden_layers": 12,
  "d": 448,
  "hidden_size": 448,
  "h": 8,
  "num_attention_heads": 8,
  "max_l": 512,
  "max_position_embeddings": 512,
  "key_mask": false,
  "use_gates": false,
  "gate_init": 1.0,
  "gate_channels": false,
  "ffn_mult": 2.0,
  "attn_impl": "sdpa",
  "qk_norm": true,
  "attn_softcap": null,
  "z_loss_coef": 0.0,
  "diff_attn": "none",
  "architecture": "full",
  "use_output_scaling": true,
  "bos_token_id": 1,
  "eos_token_id": 2,
  "pad_token_id": 2,
  "torch_dtype": "float16",
  "decode_defaults": {
    "max_new_tokens": 64,
    "temperature": 0.65,
    "top_k": 24,
    "top_p": 0.9,
    "repetition_penalty": 1.1,
    "frequency_penalty": 0.03,
    "presence_penalty": 0.0,
    "no_repeat_ngram": 3,
    "history_scope": "generated",
    "history_window": 96,
    "ban_special_tokens": true,
    "min_new_before_eos": 16,
    "stop_eos": true,
    "strategy": "sample",
    "beam_size": 4,
    "beam_top_k": 8,
    "beam_score_alpha": 1.0
  }
}