A model trained on 3,000+ PonyXL images - these were tagged manually with descriptions and danbooru tags.I've only uploaded fp8 and nf4 because those are the only models I can make on my measly 4080S &lt;3This was created by merging <a target="_blank" rel="ugc" href="https://civarchive.com/models/891281/ponyish-flux-sfwnsfw">https://civarchive.com/models/891281/ponyish-flux-sfwnsfw</a> into Flux Dev fp8. The intention in future releases to is to merge additional LoRAs to avoid "retraining" on large datasets. Currently it takes me 5 days to train the equivalent LoRA, so it's time to make the jump to a checkpoint.Recommended Settings: CFG: 2.5 Steps: 20-30

Files with SHA256: dc7f16a1655614c745c65f0e657b3cbd643317d4c1ba3e97bedd5855c678f1ef