config.json

{
  "architectures": [
    "UltravoxModel"
  ],
  "audio_config": {
    "_name_or_path": "facebook/wav2vec2-base-960h",
    "architectures": [
      "Wav2Vec2ForCTC"
    ],
    "feat_extract_dropout": 0.0,
    "feat_proj_dropout": 0.1,
    "gradient_checkpointing": false,
    "hidden_dropout_prob": 0.1,
    "model_type": "wav2vec2"
  },
  "audio_model_id": "facebook/wav2vec2-base-960h",
  "audio_model_lora_config": {
    "lora_alpha": 8,
    "r": 0,
    "target_modules": [
      "k_proj",
      "q_proj",
      "linear_k",
      "linear_q"
    ]
  },
  "audio_token_index": 32000,
  "hidden_size": 4096,
  "ignore_index": -100,
  "initializer_range": 0.02,
  "model_type": "ultravox",
  "norm_init": 0.4,
  "projector_act": "swiglu",
  "stack_factor": 8,
  "text_config": {
    "_name_or_path": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
    "architectures": [
      "LlamaForCausalLM"
    ],
    "hidden_size": 2048,
    "intermediate_size": 5632,
    "model_type": "llama",
    "num_hidden_layers": 22,
    "num_key_value_heads": 4,
    "rms_norm_eps": 1e-05,
    "torch_dtype": "bfloat16"
  },
  "text_model_id": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
  "text_model_lora_config": {
    "lora_alpha": 8,
    "r": 0,
    "target_modules": [
      "k_proj",
      "q_proj",
      "linear_k",
      "linear_q"
    ]
  },
  "torch_dtype": "float32",
  "transformers_version": "4.41.2",
  "vocab_size": 32000
}