# machine-learning
- p
pi
05/13/2021, 11:52 PMwhat dataset? - u
user
05/13/2021, 11:53 PMwdym? - p
pi
05/13/2021, 11:53 PMread the faq - s
SidPlays_144p
05/13/2021, 11:54 PMif you don't know what a dataset is, then no wonder it doesn't sound like it doesn't sound like the person you want it to - s
SidPlays_144p
05/13/2021, 11:54 PMno offense - u
user
05/13/2021, 11:54 PMi know what it is but wdym by that - p
pi
05/13/2021, 11:54 PMI want to listen to your dataset - p
pi
05/13/2021, 11:55 PMI helped a lot of people and just asking directly for the dataset saves a lot of time - u
user
05/13/2021, 11:55 PMwait hold on is there a minimum amount of files needed for the training to work - p
pi
05/13/2021, 11:55 PMREAD THE FAQ - u
user
05/13/2021, 11:57 PMthe least it says on the FAQ is more than the amount in the tutorial video - p
pi
05/13/2021, 11:58 PMI wrote that and I stand by it: https://pieksend.github.io/posts/2021-05-09-tacotron-faq/ - t
Toasty
05/14/2021, 12:29 AMMight retrain Hammond now that I should focus on epochs and not hours of training - t
Toasty
05/14/2021, 12:46 AMJESUS CHRIST - t
Toasty
05/14/2021, 12:46 AMIt's gonna take 9 minutes to train one epoch - p
pi
05/14/2021, 12:47 AMthat's very long - p
pi
05/14/2021, 12:47 AMwhy? - t
Toasty
05/14/2021, 12:47 AMWell, that was the default batch number on the uberduck training notebook - t
Toasty
05/14/2021, 12:47 AM18 - p
pi
05/14/2021, 12:47 AMhuge dataset and only a t4? - t
Toasty
05/14/2021, 12:47 AMHuge Dataset and a P100 - t
Toasty
05/14/2021, 12:47 AM(1,506 files, 2 hours) - p
pi
05/14/2021, 12:47 AMoh ok - p
pi
05/14/2021, 12:48 AMi'd say that's mid size - p
pi
05/14/2021, 12:48 AMhuge is double digit hours - t
Toasty
05/14/2021, 12:48 AMAh. - p
pi
05/14/2021, 12:48 AMno, I mean dataset size - t
Toasty
05/14/2021, 12:48 AMI get it - t
Toasty
05/14/2021, 12:49 AMSo, if the numbers are consistent, it's gonna take 15 hours to train Hammond to 100 epochs 😆 - p
pi
05/14/2021, 12:49 AMwhy does that take so long tough