| accum_grad: 4 | |
| cmvn: global_cmvn | |
| cmvn_conf: | |
| cmvn_file: data/train_libri/global_cmvn | |
| is_json_cmvn: true | |
| ctc: ctc | |
| ctc_conf: | |
| ctc_blank_id: 0 | |
| dataset: asr | |
| dataset_conf: | |
| batch_conf: | |
| batch_size: 8 | |
| batch_type: dynamic | |
| max_frames_in_batch: 120000 | |
| pad_feat: true | |
| fbank_conf: | |
| dither: 1.0 | |
| frame_length: 25 | |
| frame_shift: 10 | |
| num_mel_bins: 80 | |
| filter_conf: | |
| max_length: 40960 | |
| min_length: 0 | |
| token_max_length: 400 | |
| token_min_length: 1 | |
| resample_conf: | |
| resample_rate: 16000 | |
| shuffle: true | |
| shuffle_conf: | |
| shuffle_size: 1000 | |
| sort: false | |
| sort_conf: | |
| sort_size: 2000 | |
| spec_aug: true | |
| spec_aug_conf: | |
| max_f: 10 | |
| max_t: 50 | |
| num_f_mask: 2 | |
| num_t_mask: 2 | |
| spec_sub: false | |
| spec_sub_conf: | |
| max_t: 30 | |
| num_t_sub: 3 | |
| speed_perturb: true | |
| decoder: bitransformer | |
| decoder_conf: | |
| attention_heads: 4 | |
| dropout_rate: 0.1 | |
| linear_units: 2048 | |
| num_blocks: 3 | |
| positional_dropout_rate: 0.1 | |
| r_num_blocks: 3 | |
| self_attention_dropout_rate: 0.1 | |
| src_attention_dropout_rate: 0.1 | |
| dtype: fp16 | |
| encoder: chunkformer | |
| encoder_conf: | |
| activation_type: swish | |
| attention_dropout_rate: 0.1 | |
| attention_heads: 4 | |
| cnn_module_kernel: 15 | |
| cnn_module_norm: layer_norm | |
| dropout_rate: 0.1 | |
| dynamic_chunk_sizes: | |
| - -1 | |
| - -1 | |
| - 64 | |
| - 128 | |
| - 256 | |
| dynamic_conv: true | |
| dynamic_left_context_sizes: | |
| - 64 | |
| - 128 | |
| - 256 | |
| dynamic_right_context_sizes: | |
| - 64 | |
| - 128 | |
| - 256 | |
| input_layer: dw_striding | |
| linear_units: 2048 | |
| normalize_before: true | |
| num_blocks: 12 | |
| output_size: 256 | |
| pos_enc_layer_type: chunk_rel_pos | |
| positional_dropout_rate: 0.1 | |
| selfattention_layer_type: chunk_rel_seflattn | |
| use_cnn_module: true | |
| grad_clip: 5 | |
| input_dim: 80 | |
| log_interval: 100 | |
| max_epoch: 200 | |
| model: asr_model | |
| model_conf: | |
| ctc_weight: 0.3 | |
| length_normalized_loss: false | |
| lsm_weight: 0.1 | |
| reverse_weight: 0.3 | |
| model_dir: exp/v11 | |
| optim: adamw | |
| optim_conf: | |
| lr: 0.001 | |
| output_dim: 4992 | |
| save_states: model_only | |
| scheduler: warmuplr | |
| scheduler_conf: | |
| warmup_steps: 25000 | |
| tokenizer: bpe | |
| tokenizer_conf: | |
| bpe_path: data/lang_char/train_libri_bpe5000.model | |
| is_multilingual: false | |
| non_lang_syms_path: null | |
| num_languages: 1 | |
| special_tokens: | |
| <blank>: 0 | |
| <eos>: 2 | |
| <sos>: 2 | |
| <unk>: 1 | |
| split_with_space: false | |
| symbol_table_path: data/lang_char/train_libri_bpe5000_units.txt | |
| train_engine: torch_ddp | |
| use_amp: true | |
| vocab_size: 4992 | |