Merge pull request #679 from PaddlePaddle/chunk_fix

fix chunk default config; tarball ckpt prfix dir;
pull/682/head
Hui Zhang 3 years ago committed by GitHub
commit f8a464c707
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

@ -9,6 +9,16 @@
| conformer | conf/conformer.yaml | spec_aug + shift | test | ctc_prefix_beam_search | - | 0.062196 | | conformer | conf/conformer.yaml | spec_aug + shift | test | ctc_prefix_beam_search | - | 0.062196 |
| conformer | conf/conformer.yaml | spec_aug + shift | test | attention_rescoring | - | 0.054694 | | conformer | conf/conformer.yaml | spec_aug + shift | test | attention_rescoring | - | 0.054694 |
## Chunk Conformer
| Model | Config | Augmentation| Test set | Decode method | Chunk | Loss | WER |
| --- | --- | --- | --- | --- | --- | --- | --- |
| conformer | conf/chunk_conformer.yaml | spec_aug + shift | test | attention | 16 | - | 0.061939 |
| conformer | conf/chunk_conformer.yaml | spec_aug + shift | test | ctc_greedy_search | 16 | - | 0.070806 |
| conformer | conf/chunk_conformer.yaml | spec_aug + shift | test | ctc_prefix_beam_search | 16 | - | 0.070739 |
| conformer | conf/chunk_conformer.yaml | spec_aug + shift | test | attention_rescoring | 16 | - | 0.059400 |
## Transformer ## Transformer
| Model | Config | Augmentation| Test set | Decode method | Loss | WER | | Model | Config | Augmentation| Test set | Decode method | Loss | WER |

@ -3,17 +3,20 @@ data:
train_manifest: data/manifest.train train_manifest: data/manifest.train
dev_manifest: data/manifest.dev dev_manifest: data/manifest.dev
test_manifest: data/manifest.test test_manifest: data/manifest.test
vocab_filepath: data/vocab.txt
unit_type: 'char'
spm_model_prefix: ''
augmentation_config: conf/augmentation.json
batch_size: 32
min_input_len: 0.5 min_input_len: 0.5
max_input_len: 20.0 # second max_input_len: 20.0 # second
min_output_len: 0.0 min_output_len: 0.0
max_output_len: 400.0 max_output_len: 400.0
min_output_input_ratio: 0.05 min_output_input_ratio: 0.05
max_output_input_ratio: 10.0 max_output_input_ratio: 10.0
collator:
vocab_filepath: data/vocab.txt
unit_type: 'char'
spm_model_prefix: ''
augmentation_config: conf/augmentation.json
batch_size: 32
raw_wav: True # use raw_wav or kaldi feature raw_wav: True # use raw_wav or kaldi feature
specgram_type: fbank #linear, mfcc, fbank specgram_type: fbank #linear, mfcc, fbank
feat_dim: 80 feat_dim: 80
@ -30,7 +33,7 @@ data:
keep_transcription_text: False keep_transcription_text: False
sortagrad: True sortagrad: True
shuffle_method: batch_shuffle shuffle_method: batch_shuffle
num_workers: 0 num_workers: 2
# network architecture # network architecture
@ -78,7 +81,7 @@ model:
training: training:
n_epoch: 180 n_epoch: 240
accum_grad: 4 accum_grad: 4
global_grad_clip: 5.0 global_grad_clip: 5.0
optim: adam optim: adam

@ -18,7 +18,13 @@ function clean() {
} }
trap clean EXIT trap clean EXIT
cp ${ckpt_prefix}.* ${output} # ckpt_prfix dir
if [ -d ${ckpt_prefix} ];then
cp -r ${ckpt_prefix} ${output}
fi
# ckpt_prfix.{json,...}
cp ${ckpt_prefix}.* ${output}
# model config, mean std, vocab
cp ${model_config} ${mean_std} ${vocab} ${output} cp ${model_config} ${mean_std} ${vocab} ${output}
tar zcvf release.tar.gz ${output} tar zcvf release.tar.gz ${output}

Loading…
Cancel
Save