General discovery. Theories, procedures, practices etc. However not model specific.
- CFG-STEP scan: cfg_step.md
- Prior scan: prior.md
- Any theory to compare a model? similarity_cfg1.md
- Prompt dystopia: prompt.md
- My thought of Aesthetic Gradients, which is... interpolation with a confusing scoring: ag.md
- My procedure for making arts: my_procedure.md
- Materials on aesthetic analysis
and philosophy stuffs (Human value / aesthics / ambitious AI etc.): aesthetic.md - Extended use cases on AI art drawings: controlnet.md
- "euler" is actually "momentum sampler" / "algorithm 2" : k_euler.md
- Findings on "Hi-Res" fix: hires_fix.md
- Findings on "Dynamic CFG / CFG rescale": dynamic_cfg.md
- Findings on "Aspect Ratio Bucketing" (moved from CH98): arb.md
- Some record about "LyCORIS": lycoris.md
- Findings on "FreeU": freeu.md
- Findings on CFG along with "PAG" and "CHG": cfg_pag_chg.md
- Findings on precision "NF4": nf4.md
- Quick Survey on the vpred patch (which killed WD1.5 and most SD2.1 models): vpred.md
This session is focus on model merging, which an extended method on model ensembling.
- Any theory about merging model? merge.md
- Maybe any theory about AutoMBW? autombw.md
- Findings on "Git Re-Basin" in SD: rebasin.md
- Findings on "TIES merge" in SD: ties.md
- Findings on "Fisher Merge" in SD: fisher.md
- Findings on "DARE Merge" in SD: dare.md
- Findings on "Model Stock" in SD: modelstock.md
- Findings on "DELLA Merge" in SD: dare.md