Retraining 20 LoRAs every new model drop: lessons from the SDXL era
Inevitable_Pen9043 · reddit · 2026-11-20
A image-model finetuner describes the pain of retraining 20 LoRAs each time a new model drops, tuning learning rates, steps, dataset sizes, and choosing between LoRA/LoKR/DoRA plus samplers and schedulers.
Key lessons: the LoRA itself is often not the problem—the base model matters. SDXL LoRAs trained on the base model never converged well, but a year later ones trained on fine-tunes performed far better. Some initially bad models turned out great with time; promising ones like Flux 1 never took off. Some models excel at people but not art styles, others suit editing LoRAs.
More from Multimodal
- Google ships 8 updates in 2 weeks: Gemini 4 Neon/Hydrogen/Carbon, Genie 4, Robotics 2.5 — GoogleAI · 2026-11-21
- CapCut plugin connects ChatGPT to auto-cut gameplay highlights, then human picks the payoff — iamfakhrealam · 2026-11-21
- Posterior Matching Distillation preserves teacher diversity in few-step generation — ArashVahdat · 2026-11-21
- Qwen3 TTS goes live on Together AI: 9 voices, plain-language control, 10 languages — togethercompute · 2026-11-21
- 3DGS vs Point Cloud comparison looks great; surface reconstruction for sim collisions next — jonstephens85 · 2026-11-21
- AI video showing 25+ celebrities 'before and after' plastic surgery goes viral — evisoft · 2026-11-21