How to use LoRA correctly?

#4
by max-fofanov - opened

First of all, thank you for your great work. It seems to me, that any lora scale besides 0.1 or less just makes the resulting image worse. Any tips on using it?

I would also very much appreciate it, if you would share the parameters you used for the demo results

First of all, thank you for your great work. It seems to me, that any lora scale besides 0.1 or less just makes the resulting image worse. Any tips on using it?

have you used the code in "README" to test?

First of all, thank you for your great work. It seems to me, that any lora scale besides 0.1 or less just makes the resulting image worse. Any tips on using it?

have you used the code in "README" to test?

Yes, I did and it works good enough, however I see no use of LoRA there

hi, the demo in "README" has used LORA. https://github.com/tencent-ailab/IP-Adapter/blob/main/ip_adapter/ip_adapter_faceid.py#L71-L80

I will update the code soon, and you can use a normal ip-adapter + normal LORA, just as https://github.com/cubiq/ComfyUI_IPAdapter_plus/issues/145

Once again, thank you for your work!

max-fofanov changed discussion status to closed

Sign up or log in to comment