DoRA is a better way to represent weight diffs. I've trained an example DoRA and would like to know what the support is across platforms. A1111 only supports on the latest dev branch (Apr 5), for instance. This is SDXL, metadata is wrong and I can't change it lol.
I'd like people to test on comfyui, diffusers, hosted services, etc. The goal of this is to see (and maybe push) the current level of DoRA support.
Publish successful images to here https://civitai.com/models/384623/sad-to-happy-dora-slidersntcaixyz?modelVersionId=429247 . Each image will get some buzz. Up to 2 images per person.
Publish unsuccessful attempts as comments. Up to 2000 buzz per platform.
I will be spending at least 10k buzz total on this. Technical details at this amazing blog post https://sebastianraschka.com/blog/2024/lora-dora.html . Here is my implementation https://github.com/ntc-ai/sliders-conceptmod/blob/main/conceptmod/textsliders/dora.py . If you find a bug in the implementation you get buzz too.
Test DoRA on various platforms
10,000 BUZZ
Awarded
7
2
3
4
Started: Apr 5, 2024
Bounty Type | Image Creation |
Base Model | SD 1.5 |
Started | Apr 5, 2024 |
Deadline | Apr 15, 2024 |
