config.json

{
  "architectures": [
    "BarkModel"
  ],
  "coarse_acoustics_config": {
    "block_size": 256,
    "bos_token_id": 32,
    "eos_token_id": 32,
    "hidden_size": 16,
    "input_vocab_size": 300,
    "model_type": "coarse_acoustics",
    "num_heads": 2,
    "num_layers": 2,
    "output_vocab_size": 300,
    "pad_token_id": 32
  },
  "codec_config": {
    "audio_channels": 2,
    "chunk_in_sec": null,
    "codebook_dim": 32,
    "codebook_size": 64,
    "hidden_size": 32,
    "model_type": "encodec",
    "num_filters": 8,
    "num_lstm_layers": 1,
    "upsampling_ratios": [
      8,
      4
    ]
  },
  "fine_acoustics_config": {
    "block_size": 256,
    "bos_token_id": 32,
    "eos_token_id": 32,
    "hidden_size": 16,
    "input_vocab_size": 300,
    "model_type": "fine_acoustics",
    "num_heads": 2,
    "num_layers": 2,
    "output_vocab_size": 300,
    "pad_token_id": 32
  },
  "initializer_range": 0.02,
  "model_type": "bark",
  "semantic_config": {
    "block_size": 256,
    "bos_token_id": 32,
    "eos_token_id": 32,
    "hidden_size": 16,
    "input_vocab_size": 300,
    "model_type": "semantic",
    "num_heads": 2,
    "num_layers": 2,
    "output_vocab_size": 300,
    "pad_token_id": 32
  },
  "torch_dtype": "float32",
  "transformers_version": "4.32.0.dev0"
}