search

yasamin-med / MEDDAP

Official implementation of "MEDDAP: Medical Dataset Enhancement via Diversified Augmentation Pipeline"
4 stars 0 forks source link

ValueError: The deprecation tuple ('LoRAAttnProcessor', '0.30.0', 'Using LoRAAttnProcessor is deprecated. Please use the PEFT backend for all things LoRA. You can install PEFT by running `pip install peft`.') should be removed since diffusers' version 0.30.0.dev0 is >= 0.30.0 #3

Open qingpowuwu opened 2 weeks ago

qingpowuwu commented 2 weeks ago

Hi, Thank you for your work. I am trying to run train_text_to_image_lora.py on my own dataset, but I encountered the following error:

ValueError: The deprecation tuple ('LoRAAttnProcessor', '0.30.0', 'Using LoRAAttnProcessor is deprecated. Please use the PEFT backend for all things LoRA. You can install PEFT by runningpip install peft.') should be removed since diffusers' version 0.30.0.dev0 is >= 0.30.0

I suspect this might be related to the version of the diffusers library, my diffuser version is 0.30.0. Could you please let me know which version of diffusers you are using? Any assistance in resolving this issue would be greatly appreciated.

yasamin-med commented 2 weeks ago

Hi, I will update my code based on the new update of hugging face since some of the function are not supported by the new version anymore. I suggest you to use their new code which is located in https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora.py. I run their code and I think they had problem with unscaling gradients when using FP16 but it must be solved now but if you encounter this error use float32 instead of that.