# machine-learning
- t
Toasty
05/28/2021, 6:03 PMthe Synthesis model on #841437191073955920 does not do that - p
Pikachu ✓
05/28/2021, 6:03 PMMaybe a bit more training - p
Pikachu ✓
05/28/2021, 6:03 PMHow long did you train it for - p
Pikachu ✓
05/28/2021, 6:03 PMand how many samples did you use? - t
Toasty
05/28/2021, 6:04 PM127 epochs - t
Toasty
05/28/2021, 6:05 PM120 audio files, totally 17 minutes of audio. - t
Toasty
05/28/2021, 6:05 PMI was going to do more, but I had to cut short for finals week. - p
Pikachu ✓
05/28/2021, 6:05 PMTry to train more, maybe till like 200 epochs - t
Toasty
05/28/2021, 6:05 PM200?! - p
Pikachu ✓
05/28/2021, 6:05 PMYeah - p
Pikachu ✓
05/28/2021, 6:06 PM200 epochs - p
Pikachu ✓
05/28/2021, 6:06 PMnot samples - t
Toasty
05/28/2021, 6:06 PMWould that overfit the model? - p
Pikachu ✓
05/28/2021, 6:06 PMI did it on 60 samples, it didn't overfit - p
Pikachu ✓
05/28/2021, 6:06 PMbut try to keep an eye on it and test it every 5-10 minutes - p
Pikachu ✓
05/28/2021, 6:06 PMalso, when testing, redownload the model every time you test - t
Toasty
05/28/2021, 6:07 PMHow? Training it uses all of my GPU - p
Pikachu ✓
05/28/2021, 6:07 PMDo you mean on colab or your own computer? - t
Toasty
05/28/2021, 6:07 PMColab. - t
Toasty
05/28/2021, 6:07 PMWhen I train, 15 GB of the GPU is used. - p
Pikachu ✓
05/28/2021, 6:08 PMThat shouldn't be an issue tho - h
HoodyP
05/28/2021, 6:08 PMOh yeah, I also have this problem, Google Colab doesn’t let me use another notebook when one is used - p
Pikachu ✓
05/28/2021, 6:08 PMI think - h
HoodyP
05/28/2021, 6:08 PMBc of the GPU - h
HoodyP
05/28/2021, 6:08 PMIdk - p
Pikachu ✓
05/28/2021, 6:08 PMyou can only run 1 gpu instance - t
Toasty
05/28/2021, 6:08 PMI don't understand - t
Toasty
05/28/2021, 6:08 PMPi told me the magic number was 100 epochs - t
Toasty
05/28/2021, 6:09 PMbut now 200 epochs is the magic number? - t
Toasty
05/28/2021, 6:09 PMWhats next? 300?