I named my simple algorithm that generate new merging recipe near current and let user choose which is better to "learn" the best weight merging ratios with a exaggerated name "RMHF - Reinforcement Merging on Human Feedback".<a target="_blank" rel="ugc" href="https://github.com/TkskKurumi/DiffusersFastAPI/blob/main/rmhf_v2.py">https://github.com/TkskKurumi/DiffusersFastAPI/blob/main/rmhf_v2.py</a>

RMHF

recipe:<a target="_blank" rel="ugc" href="https://huggingface.co/TkskKurumi/RMHF/blob/main/RMHF-V6.2/rmhf.log">https://huggingface.co/TkskKurumi/RMHF/blob/main/RMHF-V6.2/rmhf.log</a>

Files with SHA256: f03e45e6329d76d26812a4f18f0145cdfa334d0071f8c74041311ab045b4522b