This was recently re-trained with the exact same settings with some experimental extra lines added to my ai-toolkit config. You can read more here.
A Flux Dev version of my Schnell LoRA.
