are you testing the lora on the same
are you testing the lora on the same base model?
4 Replies
Yes I used the same checkpoint, no errors during generation. I trained on juggernaut XL 8, getting this message in cmd in onetrainer but I am getting that with SD XL BASE too.
Some weights of the model checkpoint were not used when initializing CLIPTextModel:
['text_model.embeddings.position_ids']
Some weights of the model checkpoint were not used when initializing CLIPTextModelWithProjection:
['text_model.embeddings.position_ids']
could be auto1111 error
Sadly it's not working with ComfyUI or Fooocus either.
Now I found my mistake, I did not realise Onetrainer was training the lora into a base model, not into a lora, that's fine tuning. So the results was always a Checkpoint. That is the reason it was not triggering, because it is not just the lora.
Does your settings work too, when I would use "lora"-mode instead of "fine tuning"-mode?
i never tried lora there
my config is checkpoint