[doc] add new s0 aishell non-streaming result (#143) (#144)

Co-authored-by: Di Wu <1176705630@qq.com>

Author: robin1001

examples/aishell/s0/README.md

@@ -3,17 +3,17 @@
 ## Conformer Result
 
 * Feature info: using fbank feature, dither=0, cmvn, speed perturb
-* Training info: lr 0.002, batch size 16, 8 gpu, acc_grad 4, 200 epochs, dither 0.0
-* Decoding info: ctc_weight 0.6, average_num 30
-* Git hash: 132954a9ce27d0381ed3879c4f43cc158860167e
-* Model link: http://mobvoi-speech-public.ufile.ucloud.cn/public/wenet/aishell/20210116_conformer_exp.tar.gz
+* Training info: lr 0.002, batch size 18, 4 gpu, acc_grad 4, 240 epochs, dither 0.1
+* Decoding info: ctc_weight 0.5, average_num 20
+* Git hash: 919f07c4887ac500168ba84b39b535fd8e58918a
+* Model link: http://mobvoi-speech-public.ufile.ucloud.cn/public/wenet/aishell/20210204_conformer_exp.tar.gz
 
 | decoding mode          | CER  |
 |------------------------|------|
-| attention decoder      | 5.36 |
-| ctc greedy search      | 5.14 |
-| ctc prefix beam search | 5.14 |
-| attention rescoring    | 4.77 |
+| attention decoder      | 5.18 |
+| ctc greedy search      | 4.94 |
+| ctc prefix beam search | 4.94 |
+| attention rescoring    | 4.61 |
 
 ## Unified Conformer Result
 
@@ -33,16 +33,17 @@
 ## Transformer Result
 
 * Feature info: using fbank feature, dither=0, with cmvn, no speed perturb.
-* Training info: lr 0.002, batch size 16, 8 gpu, acc_grad 1, 120 epochs, dither 0.0
-* Git hash: fb8e0f8c12b5d547fc22e62365e1e114f059c609
-* Model link: http://mobvoi-speech-public.ufile.ucloud.cn/public/wenet/aishell/20210120_transformer_exp.tar.gz
+* Training info: lr 0.002, batch size 26, 4 gpu, acc_grad 4, 240 epochs, dither 0.1
+* Decoding info: ctc_weight 0.5, average_num 20
+* Git hash: 919f07c4887ac500168ba84b39b535fd8e58918a
+* Model link: http://mobvoi-speech-public.ufile.ucloud.cn/public/wenet/aishell/20210204_transformer_exp.tar.gz
 
 | decoding mode          | CER  |
 |------------------------|------|
-| attention decoder      | 5.76 |
-| ctc greedy search      | 6.21 |
-| ctc prefix beam search | 6.21 |
-| attention rescoring    | 5.47 |
+| attention decoder      | 5.69 |
+| ctc greedy search      | 5.92 |
+| ctc prefix beam search | 5.91 |
+| attention rescoring    | 5.30 |
 
 ## Unified Transformer Result
 

examples/aishell/s0/conf/train_conformer.yaml

@@ -41,10 +41,10 @@ raw_wav: true
 collate_conf:
     # waveform level config
     wav_distortion_conf:
-        wav_dither: 0.0
+        wav_dither: 0.1
         wav_distortion_rate: 0.0
         distortion_methods: []
-    speed_perturb: false
+    speed_perturb: true
     feature_extraction_conf:
         feature_type: 'fbank'
         mel_bins: 80
@@ -74,7 +74,7 @@ dataset_conf:
 
 grad_clip: 5
 accum_grad: 4
-max_epoch: 200
+max_epoch: 240
 log_interval: 100
 
 optim: adam

examples/aishell/s0/conf/train_transformer.yaml

@@ -36,18 +36,17 @@ raw_wav: true
 collate_conf:
     # waveform level config
     wav_distortion_conf:
-        wav_dither: 0.0
+        wav_dither: 0.1
         wav_distortion_rate: 0.0
         distortion_methods: []
-    speed_perturb: false
+    speed_perturb: true
     feature_extraction_conf:
         feature_type: 'fbank'
         mel_bins: 80
         frame_shift: 10
         frame_length: 25
         using_pitch: false
     # spec level config
-    # spec_swap: false
     feature_dither: 0.0 # add dither [-feature_dither,feature_dither] on fbank feature
     spec_aug: true
     spec_aug_conf:
@@ -64,12 +63,12 @@ dataset_conf:
     max_length: 40960
     min_length: 0
     batch_type: 'static' # static or dynamic
-    batch_size: 16
+    batch_size: 26
     sort: true
 
 grad_clip: 5
 accum_grad: 1
-max_epoch: 120
+max_epoch: 240
 log_interval: 100
 
 optim: adam