| accum_grad: 4 | |
| cmvn: null | |
| cmvn_conf: | |
| cmvn_file: null | |
| is_json_cmvn: null | |
| ctc: ctc | |
| ctc_conf: | |
| ctc_blank_id: 50363 | |
| dataset: asr | |
| dataset_conf: | |
| batch_conf: | |
| batch_type: dynamic | |
| max_frames_in_batch: 8000 | |
| cycle: 6 | |
| feats_type: log_mel_spectrogram | |
| filter_conf: | |
| max_length: 2700 | |
| min_length: 0 | |
| token_max_length: 448 | |
| token_min_length: 1 | |
| log_mel_spectrogram_conf: | |
| hop_length: 160 | |
| n_fft: 400 | |
| num_mel_bins: 80 | |
| padding: 0 | |
| resample_conf: | |
| resample_rate: 16000 | |
| shuffle: true | |
| shuffle_conf: | |
| shuffle_size: 1500 | |
| sort: true | |
| sort_conf: | |
| sort_size: 500 | |
| spec_aug: true | |
| spec_aug_conf: | |
| max_f: 10 | |
| max_t: 50 | |
| num_f_mask: 2 | |
| num_t_mask: 2 | |
| spec_sub: true | |
| spec_sub_conf: | |
| max_t: 30 | |
| num_t_sub: 3 | |
| spec_trim: false | |
| speed_perturb: true | |
| decoder: transformer | |
| decoder_conf: | |
| activation_type: gelu | |
| attention_heads: 16 | |
| dropout_rate: 0.1 | |
| gradient_checkpointing: true | |
| input_layer: embed_learnable_pe | |
| key_bias: false | |
| linear_units: 4096 | |
| normalize_before: true | |
| num_blocks: 24 | |
| positional_dropout_rate: 0.1 | |
| self_attention_dropout_rate: 0.0 | |
| src_attention: true | |
| src_attention_dropout_rate: 0.0 | |
| src_key_bias: false | |
| tie_word_embedding: true | |
| use_output_layer: true | |
| dtype: fp32 | |
| encoder: transformer | |
| encoder_conf: | |
| activation_type: gelu | |
| attention_dropout_rate: 0.0 | |
| attention_heads: 16 | |
| dropout_rate: 0.0 | |
| gradient_checkpointing: true | |
| input_layer: conv1d2 | |
| key_bias: false | |
| linear_units: 4096 | |
| normalize_before: true | |
| num_blocks: 24 | |
| output_size: 1024 | |
| pos_enc_layer_type: abs_pos_whisper | |
| positional_dropout_rate: 0.1 | |
| static_chunk_size: -1 | |
| use_dynamic_chunk: false | |
| use_dynamic_left_chunk: false | |
| grad_clip: 5 | |
| input_dim: 80 | |
| log_interval: 100 | |
| max_epoch: 6 | |
| model: whisper | |
| model_conf: | |
| ctc_weight: 0.0 | |
| length_normalized_loss: false | |
| lsm_weight: 0.1 | |
| model_dir: exp/whisper_medium_fintune_wenetspeech_yue_opensource | |
| optim: adam | |
| optim_conf: | |
| lr: 1.0e-05 | |
| output_dim: 51865 | |
| save_interval: 2000 | |
| save_states: model_only | |
| scheduler: warmuplr | |
| scheduler_conf: | |
| warmup_steps: 12000 | |
| tokenizer: whisper | |
| tokenizer_conf: | |
| bpe_path: null | |
| is_multilingual: true | |
| non_lang_syms_path: null | |
| num_languages: 99 | |
| special_tokens: | |
| eot: 50257 | |
| no_speech: 50362 | |
| no_timestamps: 50363 | |
| sot: 50258 | |
| sot_prev: 50361 | |
| timestamp_begin: 50364 | |
| transcribe: 50359 | |
| translate: 50358 | |
| split_with_space: false | |
| symbol_table_path: null | |
| train_engine: torch_ddp | |
| use_amp: false | |
| vocab_size: 51865 | |