Updated version of Serenity V2.
Pretty much it works the same as before for inference but there are two crucial differences between this and 2.0:
1) The structural issue is gone, it can now be used as a base for training (kohya_ss and others).
2) Some additional pruning was done, not it has the standard size of 2 GBs.
Updated version of Serenity V2.
Pretty much it works the same as before for inference but there are two crucial differences between this and 2.0:
1) The structural issue is gone, it can now be used as a base for training (kohya_ss and others).
2) Some additional pruning was done, not it has the standard size of 2 GBs.