From cc36fbcad002f54ecb102d2e1e2335483f56366e Mon Sep 17 00:00:00 2001 From: sssyousen <98675408+sssyousen@users.noreply.github.com> Date: Mon, 1 Aug 2022 11:13:16 +0800 Subject: [PATCH] fixed bugs (#49) * Delete egs_new.py * Delete lr_scheduler_new.py * Delete trainer_new.py * Update trainer_online.py * Update runEcapaXvector_online.py * Update plda_base.py * Update scoreSets.sh * Update extract_embeddings_new.py * Update README.md --- README.md | 2 +- pytorch/pipeline/onestep/extract_embeddings_new.py | 2 +- scoreSets.sh | 4 ++-- 3 files changed, 4 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index e9031c1..ce2d8f5 100644 --- a/README.md +++ b/README.md @@ -100,7 +100,7 @@ Of course, this data pipeline could be also followed to know the basic principle
### Update Pipeline -- **20220707** +- **[20220707](https://mp.weixin.qq.com/s/L1TJdZdUyE1OruqcNOJy8Q)** + Online Datasets is implemented (Including online feature extracting, online VAD, online augmentation and online x-vector extracting) + Supporting mixed precision training. + Runtime module for exporting jit model. diff --git a/pytorch/pipeline/onestep/extract_embeddings_new.py b/pytorch/pipeline/onestep/extract_embeddings_new.py index f6899c6..4284d5b 100644 --- a/pytorch/pipeline/onestep/extract_embeddings_new.py +++ b/pytorch/pipeline/onestep/extract_embeddings_new.py @@ -19,7 +19,7 @@ import libs.support.kaldi_io as kaldi_io import libs.support.kaldi_common as kaldi_common -from libs.egs.egs_new import WavEgsXvector +from libs.egs.egs_online import WavEgsXvector torchaudio_backend = utils.get_torchaudio_backend() torchaudio.set_audio_backend(torchaudio_backend) # Parse diff --git a/scoreSets.sh b/scoreSets.sh index 4d55d0f..96f8398 100755 --- a/scoreSets.sh +++ b/scoreSets.sh @@ -188,8 +188,8 @@ for set in $(echo $allsets | sed 's/ /\n/g' | sed '/^$/d' | sort -u);do errorNum=0 logNum=0 [ -d $vectordir/$set/log ] && logNum=$(find $vectordir/$set/log/ -name "extract.*.log" | wc -l) - [[ "$logNum" -gt 0 ]] && errorNum=$(grep ERROR $vectordir/$set/log/*.log | wc -l) - [[ "$errorNum" -gt 0 ]] && echo "There are some ERRORS in $vectordir/$set/log/*.log and it means you lose many vectors which is so bad thing and I suggest you to extract vectors of this dataset again." && exit 1 + [[ "$logNum" -gt 0 ]] && errorNum=$(grep ERROR $vectordir/$set/log/extract.*.log | wc -l) + [[ "$errorNum" -gt 0 ]] && echo "There are some ERRORS in $vectordir/$set/log/extract.*.log and it means you lose many vectors which is so bad thing and I suggest you to extract vectors of this dataset again." && exit 1 echo -e "name $set\ndata data/$prefix/$set\ndir $vectordir/$set\ninput $vectortype.scp" > $vectordir/$set/config done