Best Training settings batch size steps etc

[exxis]

Is there somewhere a board discussion where people talk about the "best" training settings, or at least what worked for them.
How many steps are optimal which batch sizes. For example. I've around 40 minutes of audio as sample which I will beforehand run through ozone to get the dataset, but then it's the questing what setting to use for a dataset like that to get a perfect/good outcome.