FluxBooru is a series of models that are trained on a collection of SFW booru images, aesthetic photos, and artistic uh.. anatomy. It doesn't know the kind of positioning information that 🐴 does, if you get my drift. This model was not trained on tags. The tags were used as ground truth for captioning via multiple VLMs.What FluxBooru tries to offer is a better base model for finetuning and inferencing illustrative artworks. This has been a problem area for tuning the base Dev and Schnell models, which resist illustrative training like it's got a counter-intelligence operation going on inside it.These models received serious amount of compute. 32x H100s were used with 16 of them dedicated to multi-node training, and two other nodes split between various training jobs.Approx. 2500 H100 GPU hours have been burned on these weights.<ul><li>FluxBooru v0.1 is the first release (Dev-based)<ul><li>Requires classifier-free guidance (CFG) at inference time<ul><li>This slows down (ComfyUI impl) or slows it down less but requires more VRAM (batched CFG in Diffusers)</li><li>Use a value around 5.0 but lower values look less cooked, w/ anatomy issues</li><li>The Flux Guidance value can't be set to 1.0 but still has some kind of impact.</li><li>Has obvious issues and mostly serves as an artifact for continued training in the same direction, since it is less biased than v0.2 and was trained in full.</li></ul></li><li>Requires ~20-25 steps</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/terminusresearch/flux-booru-CFG3.5">https://huggingface.co/terminusresearch/flux-booru-CFG3.5</a></li></ul></li><li>An attempt was made at de-distilling Schnell<ul><li>but even it seems, 32x H100 isn't enough for that job, and it was abandoned.</li><li>Not worth testing, doesn't really function at all.</li><li>Offered incase anybody looking to do the same would like somewhat of a head start.</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/terminusresearch/flux-schnellbooru/tree/bfcf16e7a0d28b0ad394ffa4ab5f62f75eef8ab5/transformer">https://huggingface.co/terminusresearch/flux-schnellbooru/tree/bfcf16e7a0d28b0ad394ffa4ab5f62f75eef8ab5/transformer</a></li></ul></li><li>FluxBooru v0.2 is a continued training run of v0.1<ul><li>Requires classifier-free guidance (CFG) at inference time<ul><li>Same caveats about speed as with v0.1</li><li>CFG remains restricted to ~5.0-6.0 with slight improvements in acceptable range</li><li>Much better results while still not much character name knowledge learnt in such a relatively short training session (1.6 million seen samples, not even 1 full epoch)</li></ul></li><li>Requires 20-25 steps</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/terminusresearch/flux-booru-v0.2/tree/main">https://huggingface.co/terminusresearch/flux-booru-v0.2/tree/main</a></li></ul></li><li>FluxBooru v0.2-LoKr 1.6B is a LyCORIS LoKr adapter trained to target the full FluxBooru-v0.2 model<ul><li>Requires classifier-free guidance (CFG) at inference time<ul><li>But can be merged into Flux.1 Dev, and then use without CFG</li></ul></li><li>Works in ComfyUI, Diffusers</li><li>Can be merged into the base FluxBooru-v0.2 model</li><li>Has much improved reliability over the base v0.2 model as the LoKr adapter trained much faster and more robustly than the full finetuning had</li><li>Works with an expanded CFG range</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/terminusresearch/FluxBooru-v0.2-LoKr-1.6B">https://huggingface.co/terminusresearch/FluxBooru-v0.2-LoKr-1.6B</a><ul><li>The validation samples here are with too-low of a CFG value and are not representative of the final model's quality</li></ul></li></ul></li><li>FluxBooru v0.1-LoKr-1.3B is a LyCORIS LoKr adapter trained against the Flux.1 Dev base model<ul><li>Does not require classifier-free guidance at inference time<ul><li>The fastest option out of all of the above.</li></ul></li><li>Works in ComfyUI, Diffusers</li><li>Can be merged into the base Flux.1-Dev or Schnell models</li><li>Weaker impact and not as fun to use. Might have had an issue during training.</li></ul></li></ul>

FluxBooru

Merged v0.2 finetuned LoKr into v0.2.

fluxbooru_v03CFGFullcheckpoint.safetensors

Files with SHA256: 353c7c2430acc0e74c639ff7380b303c8a63327e28473f06df137a616fd0e341