config.json

{
  "architectures": [
    "ParakeetCTC"
  ],
  "blank_token_id": 0,
  "bos_token_id": 1,
  "ctc_loss_reduction": "mean",
  "ctc_zero_infinity": true,
  "eos_token_id": 2,
  "fastconformer_config": {
    "activation_dropout": 0.1,
    "activation_function": "silu",
    "architectures": [
      "FastConformerModel"
    ],
    "attention_bias": false,
    "attention_dropout": 0.1,
    "conv_kernel_size": 9,
    "d_model": 1024,
    "dropout": 0.1,
    "dropout_emb": 0.0,
    "encoder_attention_heads": 8,
    "encoder_ffn_dim": 4096,
    "encoder_layerdrop": 0.1,
    "encoder_layers": 42,
    "head_dim": 128,
    "hidden_act": "silu",
    "hidden_dropout_prob": 0.1,
    "hidden_size": 1024,
    "initializer_range": 0.02,
    "intermediate_size": 4096,
    "max_position_embeddings": 2048,
    "mlp_bias": false,
    "model_type": "fastconformer",
    "nemo_decoder_type": "nemo.collections.asr.modules.ConvASRDecoder",
    "nemo_encoder_type": "nemo.collections.asr.modules.ConformerEncoder",
    "nemo_model_type": "parakeet_ctc",
    "num_attention_heads": 8,
    "num_hidden_layers": 42,
    "num_key_value_heads": 8,
    "num_mel_bins": 80,
    "pretraining_tp": 1,
    "rms_norm_eps": 1e-06,
    "rope_scaling": null,
    "rope_theta": 10000.0,
    "subsampling_conv_channels": 256,
    "subsampling_factor": 8,
    "use_bias": true,
    "use_cache": true,
    "vocab_size": 1024,
    "xscaling": true
  },
  "model_type": "parakeet_ctc",
  "pad_token_id": 0,
  "torch_dtype": "float32",
  "transformers_version": "4.53.0.dev0",
  "vocab_size": 1025
}