Hi, guys,
I heard from the Internet that, the eps of 1e-8 in torch.optim.AdamW
be rounded to zero when open torch.amp
, is this true?
Your answer and guide will be appreciated!
Hi, guys,
I heard from the Internet that, the eps of 1e-8 in torch.optim.AdamW
be rounded to zero when open torch.amp
, is this true?
Your answer and guide will be appreciated!
Could you post a reference you saw as I would expect to see some discussion about it there, too?