forked from espnet/espnet
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge branch 'espnet:master' into master
- Loading branch information
Showing
9 changed files
with
321 additions
and
14 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,30 +1,28 @@ | ||
| ## End to End Speech Recognition | ||
| ## End to End Speech Recognition with How2-2000h | ||
|
|
||
| This recipe can be used to build E2E Speech Summarization models using restricted self-attention on the HowTo corpus of instructional videos. | ||
|
|
||
| HowTo 2000h fbank-pitch features have been released to enable reproduction of this recipe. | ||
|
|
||
| #Results on ASR | ||
|
|
||
| # Results on ASR | ||
|
|
||
|
|
||
| ## asr_base_conformer_lf_mix | ||
|
|
||
| ### WER | ||
|
|
||
| |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| | ||
| |---|---|---|---|---|---|---|---|---| | ||
| |decode_asr_model_valid.acc.best/dev5_test|3016|55215|93.1|4.8|2.1|1.9|8.8|56.7| | ||
| |decode_asr_model_valid.acc.best/held_out_test|2761|47348|92.7|5.0|2.3|2.2|9.5|54.6| | ||
|
|
||
| ### CER | ||
|
|
||
| |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| | ||
| |---|---|---|---|---|---|---|---|---| | ||
| |decode_asr_model_valid.acc.best/dev5_test|3016|276377|97.1|1.1|1.9|1.9|4.8|56.7| | ||
| |decode_asr_model_valid.acc.best/held_out_test|2761|236575|96.8|1.2|2.0|2.1|5.4|54.6| | ||
|
|
||
| ### TER | ||
|
|
||
| |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| | ||
| |---|---|---|---|---|---|---|---|---| | ||
| |decode_asr_model_valid.acc.best/dev5_test|3016|82484|94.1|3.5|2.4|2.2|8.0|56.7| | ||
| |decode_asr_model_valid.acc.best/held_out_test|2761|70264|93.9|3.7|2.4|2.7|8.9|54.6| |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,3 +1,3 @@ | ||
| lm_weight: 0.6 | ||
| ctc_weight: 0.4 | ||
| beam_size: 60 | ||
| ctc_weight: 0.3 | ||
| lm_weight: 0.6 |
76 changes: 76 additions & 0 deletions
76
egs2/librispeech/asr1/conf/tuning/train_asr_conformer8.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,76 @@ | ||
| # Trained with Tesla V100 (32GB) x 3 GPUs. It takes about 3 days. | ||
| 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 | ||
|
|
||
| frontend_conf: | ||
| n_fft: 512 | ||
| hop_length: 256 | ||
|
|
||
| use_amp: true | ||
| num_workers: 4 | ||
| batch_type: numel | ||
| batch_bins: 35000000 | ||
| accum_grad: 4 | ||
| 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 |
90 changes: 90 additions & 0 deletions
90
egs2/librispeech/asr1/conf/tuning/train_asr_conformer9_layerdrop0.1_last6.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,90 @@ | ||
| # Trained with Tesla V100 (32GB) x 3 GPUs. It takes about 3 days. | ||
| 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 | ||
| stochastic_depth_rate: | ||
| - 0.0 | ||
| - 0.0 | ||
| - 0.0 | ||
| - 0.0 | ||
| - 0.0 | ||
| - 0.0 | ||
| - 0.1 | ||
| - 0.1 | ||
| - 0.1 | ||
| - 0.1 | ||
| - 0.1 | ||
| - 0.1 | ||
|
|
||
| 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 | ||
|
|
||
| frontend_conf: | ||
| n_fft: 512 | ||
| hop_length: 256 | ||
|
|
||
| unused_parameters: true # due to layer dropout, some layers are not used | ||
| use_amp: true | ||
| num_workers: 4 | ||
| batch_type: numel | ||
| batch_bins: 35000000 | ||
| accum_grad: 4 | ||
| 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 |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.