cross_attention_kwargs ['scale'] are not expected by XFormersAttnProcessor and will be ignored.

[wb315]

Describe the bug

when i want to inference the trainde model, i use the "scale" params in cross_attention_kwargs, there has the info as "cross_attention_kwargs ['scale'] are not expected by XFormersAttnProcessor and will be ignored."。 I what to konw the "scale" is effective ? Looking forward to your reply！

Reproduction

model_path = "sd-pokemon-model-lora-sdxl" 
pipe.load_lora_weights(model_path)
prompt = "Charmander from Pokémon is playing in the swimming pool."
image = pipe(prompt, num_inference_steps=30, guidance_scale=7.5, cross_attention_kwargs={"scale":0.5}).images[0]

Logs

System Info

diffusers==0.27.0.dev0 xformers==0.0.24+cu118 torch==2.2.0+cu118 torchvision==0.17.0+cu118

Who can help?

No response

[sayakpaul]

It's not a bug. peft takes care of scale already.

[YuzhiChen001]

Describe the bug

when i want to inference the trainde model, i use the "scale" params in cross_attention_kwargs, there has the info as "cross_attention_kwargs ['scale'] are not expected by XFormersAttnProcessor and will be ignored."。 I what to konw the "scale" is effective ? Looking forward to your reply！

Reproduction

model_path = "sd-pokemon-model-lora-sdxl" 
pipe.load_lora_weights(model_path)
prompt = "Charmander from Pokémon is playing in the swimming pool."
image = pipe(prompt, num_inference_steps=30, guidance_scale=7.5, cross_attention_kwargs={"scale":0.5}).images[0]

Logs

System Info

diffusers==0.27.0.dev0 xformers==0.0.24+cu118 torch==2.2.0+cu118 torchvision==0.17.0+cu118

Who can help?

No response

Have you solved this problem?