# machine-learning
- w
WeegeeFan1
11/10/2022, 5:34 PMYeah I don't know why I said that - w
WeegeeFan1
11/10/2022, 5:34 PMI'll keep note of what I have and what I need to either find or create - w
WeegeeFan1
11/10/2022, 5:34 PMI though it was just more data=better results - h
hecko
11/10/2022, 5:34 PMno need, just pitch everything up a semitone and then down a semitone - w
WeegeeFan1
11/10/2022, 5:34 PMBut being that what I've added doesn't really do much - c
Couch
11/10/2022, 5:34 PMdataset augmentation (is legal) they do it with pictures - w
WeegeeFan1
11/10/2022, 5:34 PMOh I never thought of that - w
WeegeeFan1
11/10/2022, 5:35 PMXD - h
hecko
11/10/2022, 5:35 PMwell this is more data, so what if it's not natural in the biz we call it data augmentation yeah - w
WeegeeFan1
11/10/2022, 5:35 PMI'm very slowly learning how to train locally - w
WeegeeFan1
11/10/2022, 5:35 PMUsing jupyter and such - h
hecko
11/10/2022, 5:35 PMooh - h
hecko
11/10/2022, 5:35 PMwhat gpu do you have - w
WeegeeFan1
11/10/2022, 5:35 PMSo I won't have to worry about the stupid colab stuff - w
WeegeeFan1
11/10/2022, 5:35 PMRTX 2080 - h
hecko
11/10/2022, 5:35 PMnice - w
WeegeeFan1
11/10/2022, 5:35 PMEven if it's too slow I'll keep it as a long-term training setup - w
WeegeeFan1
11/10/2022, 5:35 PMIs it going to be slow? - h
hecko
11/10/2022, 5:37 PM2080 should be decent - w
WeegeeFan1
11/10/2022, 5:37 PMWhat do you have? - w
WeegeeFan1
11/10/2022, 5:37 PMAnd have you trained locally? - h
hecko
11/10/2022, 5:37 PM1050 - h
hecko
11/10/2022, 5:37 PMtook me a hundred hours to train a model with xvatrainer - w
WeegeeFan1
11/10/2022, 5:37 PM1050 of what - w
WeegeeFan1
11/10/2022, 5:37 PMI don't think I've ever used xvatrainer for anything exept transcriptions - h
hecko
11/10/2022, 5:38 PMgtx - w
WeegeeFan1
11/10/2022, 5:38 PMAh - c
Couch
11/10/2022, 5:38 PMit took me about 11 hours to train a model on legacy on a gtx 1650 - c
Couch
11/10/2022, 5:38 PMback when it was The Popular Choice - w
WeegeeFan1
11/10/2022, 5:38 PMWell I'