{
  "model_type": "custom",
  "architectures": ["IndicF5Hinglish"],
  "model_name": "IndicF5-Hinglish",
  "version": "1.0",
  "base_model": "ai4bharat/IndicF5",
  
  "model_config": {
    "dim": 1024,
    "depth": 22,
    "heads": 16,
    "ff_mult": 2,
    "text_dim": 512,
    "conv_layers": 4,
    "text_num_embeds": 2546,
    "mel_dim": 100
  },
  
  "mel_spec_kwargs": {
    "n_fft": 1024,
    "hop_length": 256,
    "win_length": 1024,
    "n_mel_channels": 100,
    "target_sample_rate": 24000,
    "mel_spec_type": "vocos"
  },
  
  "training": {
    "dataset": "OpenSLR-104",
    "language": "Hindi-English (Hinglish) code-switched",
    "step": 32990
  },
  
  "card": {
    "language": ["hi", "en"],
    "license": "apache-2.0",
    "task_species": "text-to-speech"
  }
}