config.json

{
  "architectures": [
    "OmniVoice"
  ],
  "audio_codebook_weights": [
    8,
    8,
    6,
    6,
    4,
    4,
    2,
    2
  ],
  "audio_mask_id": 1024,
  "audio_vocab_size": 1025,
  "bos_token_id": null,
  "dtype": "float32",
  "eos_token_id": 151645,
  "llm_config": {
    "_name_or_path": "",
    "architectures": [
      "Qwen3ForCausalLM"
    ],
    "attention_bias": false,
    "attention_dropout": 0.0,
    "bos_token_id": 151643,
    "chunk_size_feed_forward": 0,
    "dtype": "float32",
    "eos_token_id": 151645,
    "head_dim": 32,
    "hidden_act": "silu",
    "hidden_size": 8,
    "id2label": {
      "0": "LABEL_0",
      "1": "LABEL_1"
    },
    "initializer_range": 0.02,
    "intermediate_size": 32,
    "is_encoder_decoder": false,
    "label2id": {
      "LABEL_0": 0,
      "LABEL_1": 1
    },
    "layer_types": [
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention"
    ],
    "max_position_embeddings": 40960,
    "max_window_layers": 2,
    "model_type": "qwen3",
    "num_attention_heads": 8,
    "num_hidden_layers": 4,
    "num_key_value_heads": 4,
    "output_attentions": false,
    "output_hidden_states": false,
    "pad_token_id": null,
    "problem_type": null,
    "return_dict": true,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "rope_theta": 1000000,
      "rope_type": "default"
    },
    "sliding_window": null,
    "tie_word_embeddings": true,
    "use_cache": true,
    "use_sliding_window": false,
    "vocab_size": 151676
  },
  "model_type": "omnivoice",
  "num_audio_codebook": 8,
  "pad_token_id": 151643,
  "transformers_version": "5.5.0"
}