config.json

{
  "architectures": [
    "ParakeetForCTC"
  ],
  "blank_token_id": 1024,
  "bos_token_id": 1,
  "ctc_loss_reduction": "mean",
  "ctc_zero_infinity": true,
  "encoder_config": {
    "activation_dropout": 0.1,
    "architectures": [
      "ParakeetEncoder"
    ],
    "attention_dropout": 0.1,
    "attention_probs_dropout_prob": 0.1,
    "bos_token_id": 1,
    "conv_kernel_size": 9,
    "dropout": 0.1,
    "dropout_emb": 0.0,
    "dropout_positions": 0.0,
    "eos_token_id": 2,
    "hidden_act": "silu",
    "hidden_dropout_prob": 0.1,
    "hidden_size": 1024,
    "initializer_range": 0.02,
    "intermediate_size": 4096,
    "layerdrop": 0.1,
    "max_position_embeddings": 5000,
    "model_type": "parakeet_encoder",
    "nemo_decoder_type": "nemo.collections.asr.modules.ConvASRDecoder",
    "nemo_encoder_type": "nemo.collections.asr.modules.ConformerEncoder",
    "nemo_model_type": "parakeet",
    "num_attention_heads": 8,
    "num_hidden_layers": 42,
    "num_mel_bins": 80,
    "pad_token_id": 0,
    "scale_input": true,
    "subsampling_conv_channels": 256,
    "subsampling_conv_kernel_size": 3,
    "subsampling_conv_stride": 2,
    "subsampling_factor": 8,
    "use_bias": true,
    "vocab_size": 1024,
    "xscaling": true
  },
  "eos_token_id": 2,
  "initializer_range": 0.02,
  "model_type": "parakeet",
  "pad_token_id": 1024,
  "torch_dtype": "float32",
  "transformers_version": "4.55.0.dev0",
  "use_bias": true,
  "vocab_size": 1025
}