{
  "_local_rope_theta_note": "local_rope_theta is DEPRECATED/unused: acoustic decoder now uses a learned slot-position embedding, not RoPE.",
  "_reserved_tokens_note": "ids 13..42 are reserved/preserve slots (random-init), inserted right after emotion_4 (12); unk and phoneme vocab shifted +30.",
  "architectures": [
    "VieNeuV3TurboForTTS"
  ],
  "attention_dropout": 0.0,
  "audio_pad_token_id": 1024,
  "audio_ref_slot_token_id": 7,
  "audio_sample_rate": 48000,
  "audio_tokenizer_pretrained_name_or_path": "OpenMOSS-Team/MOSS-Audio-Tokenizer-Nano",
  "audio_vocab_size": 1024,
  "bos_token_id": 1,
  "dtype": "float32",
  "emotion_0_token_id": 8,
  "emotion_1_token_id": 9,
  "emotion_2_token_id": 10,
  "emotion_3_token_id": 11,
  "emotion_4_token_id": 12,
  "eos_token_id": 2,
  "head_dim": 64,
  "hidden_size": 768,
  "intermediate_size": 3072,
  "local_intermediate_size": 2048,
  "local_num_attention_heads": 8,
  "local_num_hidden_layers": 2,
  "local_rope_theta": 10000.0,
  "max_position_embeddings": 1024,
  "model_type": "vieneu_v3_turbo",
  "n_vq": 16,
  "num_attention_heads": 12,
  "num_hidden_layers": 12,
  "num_key_value_heads": 4,
  "num_reserved_tokens": 30,
  "pad_token_id": 0,
  "reserved_token_start": 13,
  "rms_norm_eps": 1e-06,
  "rope_theta": 10000.0,
  "speech_generation_end_token_id": 6,
  "speech_generation_start_token_id": 5,
  "text_prompt_end_token_id": 4,
  "text_prompt_start_token_id": 3,
  "text_vocab_size": 419,
  "tie_word_embeddings": false,
  "transformers_version": "4.56.0",
  "unk_token_id": 43
}