Kromoss roadmap: Next checkpoint, higher resolution training, and future improvements?
Hey! Really enjoying Kromoss so far. I went through the model card and had a few questions about its future development:
Next checkpoint: Is there a rough ETA for when the next checkpoint might be released?
Higher resolution training: I see the current fine tune is trained at 1024px. Are there any plans to introduce higher resolution training in future stages? I typically generate at 1344×1792 (~2.4MP), so I'm curious whether that's something you're considering or if you expect the current training approach to generalize well at those resolutions.
Future improvements: Are there particular areas you're focusing on improving with upcoming checkpoints, such as photorealism, anatomy/hands, prompt adherence, or overall image quality?
No pressure for exact timelines. Just interested in the longer term direction of the project and what we might expect from future updates.
Thanks for all the work you're putting into Kromoss!
OPD training atm
which is basically training lots and lots of dedicated loras to be distill back to the model to fill in the gaps
current throughput on my machine is 40s/lora
and my aim is to train 100K unique concepts, arts, anatomy, adherence on the base model back
sooo it's going to take a while