{
  "architectures": [
    "NemotronDenseAudexForConditionalGeneration"
  ],
  "audio_config": {
    "activation_function": "gelu",
    "d_model": 1280,
    "encoder_attention_heads": 20,
    "encoder_ffn_dim": 5120,
    "encoder_layers": 32,
    "max_source_positions": 1500,
    "model_type": "qwen2_audio_encoder",
    "num_mel_bins": 128,
    "scale_embedding": false
  },
  "audio_encoder_hidden_size": 1280,
  "audio_model_type": "NV-Whisper",
  "audio_preprocessor_path": "audio_preprocessor",
  "audio_projector_activation": "relu2",
  "audio_projector_intermediate_size": 4096,
  "audio_projector_norm_eps": 1e-05,
  "auto_map": {
    "AutoConfig": "configuration_nemotron_h_audio.NemotronDenseAudexConfig",
    "AutoModel": "modeling_nemotron_h_audio.NemotronDenseAudexForConditionalGeneration",
    "AutoModelForCausalLM": "modeling_nemotron_h_audio.NemotronDenseAudexForConditionalGeneration"
  },
  "base_model_tp_plan": null,
  "dtype": "float32",
  "eos_token_id": 11,
  "head_dim": 128,
  "hidden_act": "relu2",
  "hidden_size": 2048,
  "intermediate_size": 9216,
  "max_position_embeddings": 131072,
  "model_type": "nemotron_dense_audex",
  "norm_eps": 1e-05,
  "num_attention_heads": 16,
  "num_hidden_layers": 28,
  "num_key_value_heads": 8,
  "rope_parameters": {
    "partial_rotary_factor": 1.0,
    "rope_theta": 100000000,
    "rope_type": "default"
  },
  "sound_clip_duration": 30.0,
  "sound_embedding_size": 750,
  "sound_end_token": "<so_end>",
  "sound_end_token_id": 31,
  "sound_model_type": "hf:///nv-whisper",
  "sound_start_token": "<so_start>",
  "sound_start_token_id": 30,
  "sound_target_rate": 16000,
  "sound_token": "<so_embedding>",
  "sound_token_id": 29,
  "tie_word_embeddings": false,
  "tp_plan": null,
  "transformers_version": "5.4.0",
  "vocab_size": 205312
}