Repository navigation
Support for LoRA / adapters in AceStepPipeline #14060
Description
Hello,
First of all, thank you for adding the ACE-Step 1.5 pipeline. It's working nicely!
I noticed that LoRA support doesn't seem to be implemented yet in AceStepPipeline.
When trying to load a LoRA I get this error:
'AceStepPipeline' object has no attribute 'set_adapters'
(using the standard Diffusers method: load_lora_weights + set_adapters)
Additionally, this was also reported in the Amuse project (which uses Diffusers under the hood):
saddam213/AmuseAI#54
I'm using the XL SFT and Turbo variants on an AMD RX 7900 XTX, both in Amuse and directly with Diffusers.
Would it be possible to add LoRA support for this pipeline in the future? It would be very useful for style transfer and personalization.
Happy to provide more details, logs, or help test if needed.
Thank you!
Activity
riyaarah commentedon Jun 24, 2026
I'd like to work on adding LoRA support for AceStepPipeline. Planning to follow the WanLoraLoaderMixin pattern ,transformer only LoRA, no text encoder. Will add AceStepLoraLoaderMixin to lora_pipeline.py and wire it into the pipeline class.
Nib-git commentedon Jun 24, 2026
Thank you! Looking forward to it.
Happy to test on AMD RX 7900 XTX whenever you have something ready.
chenyangzhu1 commentedon Jul 15, 2026
#14193
I opened the PR to support this.
@Nib-git
Nib-git commentedon Jul 15, 2026
Thank you for working on this!
I'm following the progress and happy to test on AMD RX 7900 XTX (Windows) whenever there's something ready.
Nib-git commentedon Jul 19, 2026
@chenyangzhu1 @sayakpaul Thanks a ton.