still flame
We're in need of assistance to refine SDXL models for custom product images, so that we can generate product Ads. We tried some lora training but it does not give us good results. So we need someone who has experience in sdxl fine-tuning and controlnet.
Is this something you'd be interested in?
serene groveBOT
Hi @still flame. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
I'm an AI developed by Invoke AI and I'm here to provide information and guidance based on the knowledge sources provided to me. I can certainly provide some guidance on fine-tuning Stable Diffusion (SDXL) models and using ControlNet based on the information I have.
From the knowledge sources, it's clear that Invoke AI supports the use of diffusion format ControlNets. For example, a canny ControlNet can be used and downloaded through the model manager with the repo id: diffusers/controlnet-canny-sdxl-1.0 source (https://github.com/invoke-ai/InvokeAI/issues/4743).
If you're having trouble with specific ControlNet models, it's worth noting that not all models are correctly formatted or intended for general use. For instance, some models uploaded by users might be intended for use in specific extensions and not as a general Diffusers repo source (https://github.com/invoke-ai/InvokeAI/issues/4487).
In terms of fine-tuning SDXL models, it seems there have been issues reported with certain LoRAs (Learning Rate Annealers) during the conversion process source (https://github.com/invoke-ai/InvokeAI/issues/5725). It might be worth exploring different LoRAs or seeking further assistance on this matter.