From e422c6197b5bcada0429986500d8d5ca4ffcb3e4 Mon Sep 17 00:00:00 2001 From: 嘉渊 <wangjiaming.wjm@alibaba-inc.com> Date: 星期三, 10 五月 2023 19:23:37 +0800 Subject: [PATCH] update repo --- egs/librispeech_100h/conf/train_asr_conformer.yaml | 80 ++++++++++++++++++++++++++ egs/librispeech_100h/conf/train_asr_conformer_uttnorm.yaml | 80 ++++++++++++++++++++++++++ egs/librispeech_100h/conf/decode_asr_transformer.yaml | 6 ++ 3 files changed, 166 insertions(+), 0 deletions(-) diff --git a/egs/librispeech_100h/conf/decode_asr_transformer.yaml b/egs/librispeech_100h/conf/decode_asr_transformer.yaml new file mode 100644 index 0000000..a147fa7 --- /dev/null +++ b/egs/librispeech_100h/conf/decode_asr_transformer.yaml @@ -0,0 +1,6 @@ +beam_size: 10 +penalty: 0.0 +maxlenratio: 0.0 +minlenratio: 0.0 +ctc_weight: 0.5 +lm_weight: 0.7 diff --git a/egs/librispeech_100h/conf/train_asr_conformer.yaml b/egs/librispeech_100h/conf/train_asr_conformer.yaml new file mode 100644 index 0000000..68b127f --- /dev/null +++ b/egs/librispeech_100h/conf/train_asr_conformer.yaml @@ -0,0 +1,80 @@ +encoder: conformer +encoder_conf: + output_size: 512 + attention_heads: 8 + linear_units: 2048 + num_blocks: 12 + dropout_rate: 0.1 + positional_dropout_rate: 0.1 + attention_dropout_rate: 0.1 + input_layer: conv2d + normalize_before: true + macaron_style: true + rel_pos_type: latest + pos_enc_layer_type: rel_pos + selfattention_layer_type: rel_selfattn + activation_type: swish + use_cnn_module: true + cnn_module_kernel: 31 + +decoder: transformer +decoder_conf: + attention_heads: 8 + linear_units: 2048 + num_blocks: 6 + dropout_rate: 0.1 + positional_dropout_rate: 0.1 + self_attention_dropout_rate: 0.1 + src_attention_dropout_rate: 0.1 + +model_conf: + ctc_weight: 0.3 + lsm_weight: 0.1 + length_normalized_loss: false + +accum_grad: 2 +max_epoch: 50 +patience: none +init: none +best_model_criterion: +- - valid + - acc + - max +keep_nbest_models: 10 + +optim: adam +optim_conf: + lr: 0.0025 + weight_decay: 0.000001 +scheduler: warmuplr +scheduler_conf: + warmup_steps: 40000 + +specaug: specaug +specaug_conf: + apply_time_warp: true + time_warp_window: 5 + time_warp_mode: bicubic + apply_freq_mask: true + freq_mask_width_range: + - 0 + - 27 + num_freq_mask: 2 + apply_time_mask: true + time_mask_width_ratio_range: + - 0. + - 0.05 + num_time_mask: 10 + +dataset_conf: + shuffle: True + shuffle_conf: + shuffle_size: 1024 + sort_size: 500 + batch_conf: + batch_type: token + batch_size: 10000 + num_workers: 8 + +log_interval: 50 +normalize: None \ No newline at end of file diff --git a/egs/librispeech_100h/conf/train_asr_conformer_uttnorm.yaml b/egs/librispeech_100h/conf/train_asr_conformer_uttnorm.yaml new file mode 100644 index 0000000..16b7cc0 --- /dev/null +++ b/egs/librispeech_100h/conf/train_asr_conformer_uttnorm.yaml @@ -0,0 +1,80 @@ +encoder: conformer +encoder_conf: + output_size: 512 + attention_heads: 8 + linear_units: 2048 + num_blocks: 12 + dropout_rate: 0.1 + positional_dropout_rate: 0.1 + attention_dropout_rate: 0.1 + input_layer: conv2d + normalize_before: true + macaron_style: true + rel_pos_type: latest + pos_enc_layer_type: rel_pos + selfattention_layer_type: rel_selfattn + activation_type: swish + use_cnn_module: true + cnn_module_kernel: 31 + +decoder: transformer +decoder_conf: + attention_heads: 8 + linear_units: 2048 + num_blocks: 6 + dropout_rate: 0.1 + positional_dropout_rate: 0.1 + self_attention_dropout_rate: 0.1 + src_attention_dropout_rate: 0.1 + +model_conf: + ctc_weight: 0.3 + lsm_weight: 0.1 + length_normalized_loss: false + +accum_grad: 2 +max_epoch: 50 +patience: none +init: none +best_model_criterion: +- - valid + - acc + - max +keep_nbest_models: 10 + +optim: adam +optim_conf: + lr: 0.0025 + weight_decay: 0.000001 +scheduler: warmuplr +scheduler_conf: + warmup_steps: 40000 + +specaug: specaug +specaug_conf: + apply_time_warp: true + time_warp_window: 5 + time_warp_mode: bicubic + apply_freq_mask: true + freq_mask_width_range: + - 0 + - 27 + num_freq_mask: 2 + apply_time_mask: true + time_mask_width_ratio_range: + - 0. + - 0.05 + num_time_mask: 10 + +dataset_conf: + shuffle: True + shuffle_conf: + shuffle_size: 1024 + sort_size: 500 + batch_conf: + batch_type: token + batch_size: 10000 + num_workers: 8 + +log_interval: 50 +normalize: utterance_mvn \ No newline at end of file -- Gitblit v1.9.1