You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Some LAION checkpoints, the large variant for example, use different mean and std values for image normalization. Figure out a way to, preferably, encode what values to use, or introduce another funcion to preprocess with these values.
Yes, another issue is, there seems to be no programatic way to infer what values to use apart from the notes in open-clip's readme. So I'm afraid we need to leave it to the user to specify whether to use custom mean-std values with a flag when converting models. Maybe a better way?
Activity
[-]Support custom means-std normalization[/-][+]Support custom mean-std normalization[/+]Green-Sky commentedon Jul 7, 2023
yay, more hyper parameters
cant wait for gguf
monatis commentedon Jul 8, 2023
Yes, another issue is, there seems to be no programatic way to infer what values to use apart from the notes in open-clip's readme. So I'm afraid we need to leave it to the user to specify whether to use custom mean-std values with a flag when converting models. Maybe a better way?
monatis commentedon Sep 27, 2023
Released in #75