Bagheera PRO

it would be nice to mark one as default somehow! currently the first one is.

and small image models! i am working on a small ensemble of experts.. :]

well, the initial point was about making early victory claims and it looks like the model merge by you is ironically scoring higher than the trained checkpoint. do you have a recipe for model training that can reproduce your results?

i never really go there and i don't think it's reflective of the research community here on Hugging Face

as in "this is already the most thoughtful model" feels a bit like a total dismissal of everything the others have put into THEIR models, right? i've trained foundational model from gaussian weights and your words felt like they were intended to make me feel small. i think what you've done is impressive, but it's still an SDXL finetune.

i wasn't trying to bemoan or insult your efforts 🙏🏽 but in my humble opinion when you make these great claims people will itch to correct it. if you approach it with a cautious humbleness though, people generally will want to uplift and support.

after just 717 samples can we make conclusions like it took 4th place?