Are Iterations the same?

Want to understand the training process better? Got tips for which model to use and when? This is the place for you


Forum rules

Read the FAQs and search the forum before posting a new topic.

This forum is for discussing tips and understanding the process involved with Training a Faceswap model.

If you have found a bug are having issues with the Training process not working, then you should post in the Training Support forum.

Please mark any answers that fixed your problems so others can find the solutions.

Locked
User avatar
MaxHunter
Posts: 193
Joined: Thu May 26, 2022 6:02 am
Has thanked: 177 times
Been thanked: 13 times

Are Iterations the same?

Post by MaxHunter »

This is probably a basic question, but are all Iterations the same?

What I mean by this is I noticed my iterations slow down the higher my batch size is. For example, a batch size of 1 may produce 60k ITs over 6 hours, however a batch size of 4 of the same model will produce 25k ITs over the same amount of time. At first I thought there was a problem with having multiple gpus, but then I thought the BS 1 is seeing 120k faces (2 faces per iteration), whereas BS 4 is actually seeing 200k faces (8 faces per iteration) over the same amount of time. So judging by ITs can be a bit misleading. Is that correct? Wouldn't having a face count rather than IT count be more standardized? (Sorry for such a basic question.)

Last edited by MaxHunter on Sat Jan 28, 2023 6:04 pm, edited 2 times in total.
User avatar
torzdf
Posts: 2649
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 128 times
Been thanked: 623 times

Re: Are Iterations the same?

Post by torzdf »

Yes. It's why we use EG/sec. Iteration is entirely dependant on batch size. 1 iteration = 1 batch through the model.

My word is final

Locked