encoder_config.json

{
  "chunk_samples": 714240,
  "feature_extractors": {
    "muencoder_config": {
      "conformer_config": {
        "activation_dropout": 0.1,
        "attention_dropout": 0.1,
        "conformer_conv_dropout": 0.1,
        "conv_depthwise_kernel_size": 31,
        "hidden_act": "swish",
        "hidden_dropout": 0.1,
        "hidden_size": 1024,
        "intermediate_size": 4096,
        "layer_norm_eps": 1e-05,
        "layerdrop": 0.0,
        "max_source_positions": 5000,
        "num_attention_heads": 16,
        "num_conv_pos_embedding_groups": 16,
        "num_conv_pos_embeddings": 128,
        "num_hidden_layers": 12,
        "position_embeddings_type": "rotary",
        "rotary_embedding_base": 10000
      },
      "conv_dim": 512,
      "hop_length": 240,
      "mean": 6.768444971712967,
      "n_fft": 2048,
      "n_mels": 128,
      "sample_rate": 24000,
      "std": 18.417922652295623
    },
    "wavlm_config": {
      "activation_dropout": 0.0,
      "add_adapter": false,
      "attention_dropout": 0.1,
      "conv_bias": false,
      "conv_dim": [
        512,
        512,
        512,
        512,
        512,
        512,
        512
      ],
      "conv_kernel": [
        10,
        3,
        3,
        3,
        3,
        2,
        2
      ],
      "conv_stride": [
        5,
        2,
        2,
        2,
        2,
        2,
        2
      ],
      "do_stable_layer_norm": false,
      "feat_extract_activation": "gelu",
      "feat_extract_norm": "group",
      "feat_proj_dropout": 0.1,
      "hidden_act": "gelu",
      "hidden_dropout": 0.1,
      "hidden_size": 768,
      "intermediate_size": 3072,
      "layer_norm_eps": 1e-05,
      "layerdrop": 0.05,
      "max_bucket_distance": 800,
      "num_attention_heads": 12,
      "num_buckets": 320,
      "num_conv_pos_embedding_groups": 16,
      "num_conv_pos_embeddings": 128,
      "num_hidden_layers": 12
    },
    "whisper_config": {
      "activation_dropout": 0.0,
      "activation_function": "gelu",
      "attention_dropout": 0.0,
      "d_model": 768,
      "dropout": 0.0,
      "encoder_attention_heads": 12,
      "encoder_ffn_dim": 3072,
      "encoder_layerdrop": 0.0,
      "encoder_layers": 12,
      "max_source_positions": 1500,
      "num_mel_bins": 80,
      "scale_embedding": false
    }
  },
  "query": {
    "dim": 512,
    "head_dim": 128,
    "interval": 2,
    "num_layers": 6
  },
  "version": "vqv12"
}