config.json

{
  "model_type": "text_sync_mimi",
  "architectures": ["TextSyncMimi"],
  "model_version": "v1.6",
  "mimi_model_id": "kyutai/mimi",
  "vocab_size": 128256,
  "alpha": 1.0,
  "cross_attention_layers": 4,
  "causal_attention_layers": 4,
  "bce_threshold": 0.1,
  "end_token_threshold": 0.5,
  "max_z_tokens": 50,
  "hidden_size": 512,
  "text_embedding_dim": 4096,
  "sample_rate": 24000,
  "frame_rate": 12.5,
  "initializer_range": 0.02,
  "transformers_version": "4.47.0",
  "torch_dtype": "float32",
  "auto_map": {
    "AutoConfig": "configuration_text_sync_mimi.TextSyncMimiConfig",
    "AutoModel": "modeling_text_sync_mimi.TextSyncMimi"
  }
}