tsurumeso / vocal-remover

Vocal Remover using Deep Neural Networks
MIT License
1.47k stars 215 forks source link

ZeroDivisionError: division by zero #86

Open ThatGuythatlikesRubyLanguage opened 3 years ago

ThatGuythatlikesRubyLanguage commented 3 years ago

My goal was to extract all instruments from this video: https://www.youtube.com/watch?v=lsQAoDsH0Ws I gathered similiar sounding audio from freesound.org mixed it into a 10 minute music (i designed it just for the ai its not supposed to be normal music) to train my model on. So not general model just for specific instruments. I got this error: Traceback (most recent call last): File "train.py", line 224, in main() File "train.py", line 200, in main batchsize=args.batchsize) File "train.py", line 48, in train_inner_epoch return sum_loss / len(X) ZeroDivisionError: division by zero Even if the ai wouldnt train the best on such a small dataset, is there a fix?

tsurumeso commented 3 years ago

The size of the dataset should be at least 5 pairs when using the default --val_rate.