wav2vec-u Common Voice Swedish - vad

Original here

%cd /tmp

/tmp

!git clone https://github.com/pytorch/fairseq/

%cd fairseq/examples/wav2vec/unsupervised

/tmp/fairseq/examples/wav2vec/unsupervised

!git clone https://github.com/zhenghuatan/rVADfast

!cat scripts/vads.py|sed -e 's!/path/to/rVADfast_py_2.0!/tmp/fairseq/examples/wav2vec/unsupervised/rVADfast!' > tmp
!mv tmp scripts/vads.py

!for i in train valid test;do cat /kaggle/input/fork-of-wav2vec-u-cv-swedish-tsv/$i.tsv|python scripts/vads.py > /kaggle/working/$i.vads;done

100%|█████████████████████████████████████| 2331/2331 [1:01:46<00:00,  1.59s/it]
100%|███████████████████████████████████████| 2019/2019 [53:39<00:00,  1.59s/it]
100%|███████████████████████████████████████| 2027/2027 [57:26<00:00,  1.70s/it]

!mkdir /kaggle/working/wav
!mkdir /kaggle/working/wav/train
!mkdir /kaggle/working/wav/test
!mkdir /kaggle/working/wav/valid

!for i in train test valid; do python scripts/remove_silence.py --tsv /kaggle/input/fork-of-wav2vec-u-cv-swedish-tsv/$i.tsv --vads /kaggle/working/$i.vads --out /kaggle/working/wav/$i;done