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
Now that Kohya SS is supporting ClipL and T5xxl for Lora training, having clip input/output for Flux Lora loaders have become a thing again. Are there plans to update this node or a recommende workaround? I've tried just loading the clip through a seperate Lora loader but I'm not sure if each block has also it's own Lora weight.
The text was updated successfully, but these errors were encountered:
Your workaround sounds fine, the transformer blocks aren't tied to the text encoder blocks, the text encoder does have it's own set of blocks/layers though. I don't know if there would be any point to try adjusting those.
Now that Kohya SS is supporting ClipL and T5xxl for Lora training, having clip input/output for Flux Lora loaders have become a thing again. Are there plans to update this node or a recommende workaround? I've tried just loading the clip through a seperate Lora loader but I'm not sure if each block has also it's own Lora weight.
The text was updated successfully, but these errors were encountered: