config.json

{
  "architectures": [
    "Nemotron3DiarizationAudioModel"
  ],
  "attention_dropout": 0.0,
  "dropout": 0.0,
  "dtype": "float32",
  "hidden_act": "gelu",
  "hidden_size": 32,
  "initializer_range": 0.02,
  "intermediate_size": 32,
  "max_position_embeddings": 5000,
  "model_type": "nemotron3_diarization_audio",
  "num_attention_heads": 2,
  "num_hidden_layers": 2,
  "num_key_value_heads": 2,
  "num_mel_bins": 8,
  "partial_rotary_factor": 1.0,
  "rope_parameters": {
    "partial_rotary_factor": 1.0,
    "rope_theta": 10000.0,
    "rope_type": "default"
  },
  "subsampling_factor": 8,
  "transformers_version": "5.19.0.dev0"
}