Chris Malone

> There's a guide [here](https://github.com/bghira/SimpleTuner/blob/main/documentation/quickstart/FLUX.md) by the SimpleTuner devs for creating FLUX Loras. Edit: They recommend 512px training over 1024px > It would be great if could train with 12GB...

> > I think that onetrainer might finally need to implement multigpu for full flux support, because anything over rank 16 lora will probably be untrainable with the 24gb that...

The same model works fine in the latest Automatic1111 with Euler, but I get the coloured blobs if I use DPM++ SDE. In comfyui, no samplers work.

> > I get the coloured blobs if I use DPM++ SDE> > > because DPM++ SDE does not work with SD3 Yeah I know, but ComfyUI blobs resemble the...

No keys ignored that I could see. Tried both Automatic and Automatic(fp16). For characters and objects, the unet should be the only thing that makes a real difference, but for...

I've had incredible early success with this model in combination with ademix8bit. In 326 steps it achieved the bulk of what it took 40,000 to get to on distilled adamw8bit....

I changed line 152 in library/flux_train_utils: `scale = prompt_dict.get("scale", 1.0)` but the images are still completely distorted. I'm guessing the problem is the model itself. Maybe you'd need to switch...

I was excited for Framepack, but it turned out to be a lot of hype tbh. It's promising for the future once they train a WAN version, but since it's...

Not that I've heard. Development seems to have come to a complete halt with Flux. It's a shame because comfyui is this insanely tiring noodle soup you have to navigate...

Looks like I'm having the same issue #1464 Training with 326 images + 2669 reg images. My loss rate has been around 0.435 for all 5 epochs. Are you using...