Quick question about training the same model at different resolutions

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
martinf
Posts: 27
Joined: Thu Sep 29, 2022 7:58 pm
Been thanked: 3 times

Quick question about training the same model at different resolutions

Post by martinf »

If I successfully train a model at, say 128p, will I get the same results at 256p or 512p given everything else remains the same (facesets, model setting - except for resolution)? I know the time difference will be geometric, but will the results essentially be the same only resolved to more detail? This assumes that my faceset data is high resolution.

User avatar
torzdf
Posts: 2649
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 128 times
Been thanked: 622 times

Re: Quick question about training the same model at different resolutions

Post by torzdf »

Short answer is 'no'. You would not get the same results even if running the same model again with the same data at the same resolution. The random element of ML ensures that results are non-deterministic.

Increasing spatial dimensions also means the model will behave somewhat differently.

My word is final

Locked