config.json
962 B · 38 lines · json Raw
1 {
2 "architectures": [
3 "ParakeetForCTC"
4 ],
5 "ctc_loss_reduction": "mean",
6 "ctc_zero_infinity": true,
7 "dtype": "bfloat16",
8 "encoder_config": {
9 "activation_dropout": 0.1,
10 "attention_bias": true,
11 "attention_dropout": 0.1,
12 "conv_kernel_size": 9,
13 "dropout": 0.1,
14 "dropout_positions": 0.0,
15 "hidden_act": "silu",
16 "hidden_size": 1024,
17 "initializer_range": 0.02,
18 "intermediate_size": 4096,
19 "layerdrop": 0.1,
20 "max_position_embeddings": 5000,
21 "model_type": "parakeet_encoder",
22 "num_attention_heads": 8,
23 "num_hidden_layers": 42,
24 "num_key_value_heads": 8,
25 "num_mel_bins": 80,
26 "scale_input": true,
27 "subsampling_conv_channels": 256,
28 "subsampling_conv_kernel_size": 3,
29 "subsampling_conv_stride": 2,
30 "subsampling_factor": 8
31 },
32 "initializer_range": 0.02,
33 "model_type": "parakeet_ctc",
34 "pad_token_id": 1024,
35 "transformers_version": "4.57.0.dev0",
36 "vocab_size": 1025
37 }
38