Yep!

This once in a while happens every 10 secs actually

But maybe it goes slower with more wav files?

Yeah i would've said "every epoch" but i wasn't sure if you would get that

That is correct

What's an epoch?

Amount of times a model goes through the entire dataset

So if I don't stop it, it's going to continue till 69 thousand?

It's the trainer going through the entire dataset once So it goes through iterations (combing through the dataset, larger or smaller based on the batch size) and then once it gets to enough iterations to go through the entire dataset, that is called an epoch

Yeah. If you think a specifc epoch sounds better than others you can stop it early, otherwise you risk overtraining or undertrainung it

Does this number (69,420) represent something? Is it like the limit/the best this model could do? Or is it more of a setting that doesn't mean much?

It's a setting that doesn't mean much

In fact, it's a joke about 69 (sex position) and 420 (weed)

Oh lmao

I thought it was purely a coincidence that it turned out to be this number for me XD

It's just an arbitrary large number You're always expected to stop early

I mean, unless you have a decades worth of audio that needs that many epochs ¯\_(ツ)_/¯

I don't see something called Synth notebook in #841437191073955920. Do you mean this one?

https://colab.research.google.com/drive/1wXWuhnw2pdfFy1L-pUzHfopW10W2GiJS this is the one you clicked, righ?

Yes

That's the right one

Can I test without stopping the epoch-ing process first?

Yes

Are the pt files in /tacotron on my Drive the models that have been made so far a.k.a. the files I'm looking for?

Every 20 epochs it saves a model, so you can test it then

A saved epoch wile show up in the folder path

Hm, I can only find those three pt files on my Drive

No epoch files

Pt files are what you're looking for

Okay