diff --git a/split-tacotron2.sh b/split-tacotron2.sh new file mode 100755 index 0000000..4d07881 --- /dev/null +++ b/split-tacotron2.sh @@ -0,0 +1,13 @@ +#!/bin/bash + +cut --complement -f 2 -d\| metadata.csv > metadata-cut.csv +awk -F '|' -v OFS='|' '$1 {print "mefwavs/wavs/" $1 ".wav" "|" $2}' metadata-cut.csv > ljs_audio_text.txt +shuf metadata-cut.csv > metadata-cut-shuf.csv +awk -F '|' -v OFS='|' '$1 {print "mefwavs/wavs/" $1 ".wav" "|" $2}' metadata-cut-shuf.csv > ljs_audio_text-shuf.txt +sed -n '1,50p' ljs_audio_text-shuf.txt > ljs_audio_text_val_filelist.txt +sed -n '51,200p' ljs_audio_text-shuf.txt > ljs_audio_text_test_filelist.txt +sed -n '201,$ p' ljs_audio_text-shuf.txt > ljs_audio_text_train_filelist.txt +awk -F '|' -v OFS='|' '$1 {print "mefwavs/mels/" $1 ".pt" "|" $2}' metadata-cut-shuf.csv > ljs_mel_text-shuf.txt +sed -n '1,50p' ljs_mel_text-shuf.txt > ljs_mel_text_val_filelist.txt +sed -n '51,200p' ljs_mel_text-shuf.txt > ljs_mel_text_test_filelist.txt +sed -n '201,$ p' ljs_mel_text-shuf.txt > ljs_mel_text_train_filelist.txt