FunAudioLLM/SenseVoiceSmall

1encoder: SenseVoiceEncoderSmall2encoder_conf:3    output_size: 5124    attention_heads: 45    linear_units: 20486    num_blocks: 507    tp_blocks: 208    dropout_rate: 0.19    positional_dropout_rate: 0.110    attention_dropout_rate: 0.111    input_layer: pe12    pos_enc_class: SinusoidalPositionEncoder13    normalize_before: true14    kernel_size: 1115    sanm_shfit: 016    selfattention_layer_type: sanm17 18 19model: SenseVoiceSmall20model_conf:21    length_normalized_loss: true22    sos: 123    eos: 224    ignore_id: -125 26tokenizer: SentencepiecesTokenizer27tokenizer_conf:28  bpemodel: null29  unk_symbol: <unk>30  split_with_space: true31 32frontend: WavFrontend33frontend_conf:34    fs: 1600035    window: hamming36    n_mels: 8037    frame_length: 2538    frame_shift: 1039    lfr_m: 740    lfr_n: 641    cmvn_file: null42 43 44dataset: SenseVoiceCTCDataset45dataset_conf:46  index_ds: IndexDSJsonl47  batch_sampler: EspnetStyleBatchSampler48  data_split_num: 3249  batch_type: token50  batch_size: 1400051  max_token_length: 200052  min_token_length: 6053  max_source_length: 200054  min_source_length: 6055  max_target_length: 20056  min_target_length: 057  shuffle: true58  num_workers: 459  sos: ${model_conf.sos}60  eos: ${model_conf.eos}61  IndexDSJsonl: IndexDSJsonl62  retry: 2063 64train_conf:65  accum_grad: 166  grad_clip: 567  max_epoch: 2068  keep_nbest_models: 1069  avg_nbest_model: 1070  log_interval: 10071  resume: true72  validate_interval: 1000073  save_checkpoint_interval: 1000074 75optim: adamw76optim_conf:77  lr: 0.0000278scheduler: warmuplr79scheduler_conf:80  warmup_steps: 2500081 82specaug: SpecAugLFR83specaug_conf:84    apply_time_warp: false85    time_warp_window: 586    time_warp_mode: bicubic87    apply_freq_mask: true88    freq_mask_width_range:89    - 090    - 3091    lfr_rate: 692    num_freq_mask: 193    apply_time_mask: true94    time_mask_width_range:95    - 096    - 1297    num_time_mask: 198