dx8152

/

Flux2-Klein-9B-Enhanced-Details

like 107

i think it should inherit the models license (flux non commercial).

Oh, I just picked randomly, I don't know which one is suitable.

i always pick unknown, because that drop down list is a bit of a pain and i can never find the correct one (which should be flux non commercial, unless its based on Klein4b or Flux1 Schnell).

i always pick unknown, because that drop down list is a bit of a pain and i can never find the correct one (which should be flux non commercial, unless its based on Klein4b or Flux1 Schnell).

That's a bit troublesome, and I don't understand it, but ignoring it should be fine.

i always pick unknown, because that drop down list is a bit of a pain and i can never find the correct one (which should be flux non commercial, unless its based on Klein4b or Flux1 Schnell).

That's a bit troublesome, and I don't understand it, but ignoring it should be fine.

hi, why in workflow.json use klein 9b why 16steps instead 4 steps for 9b klein (not base), im newbie

doesn't really make sense. Some say 6 or so can improve results but 16 will have no advances and likely makes things much worse.

But since you can just slot in base, i assume that was a leftover from when this was selected. and not changed when the workflow was exported. Not an expert but long enough at it to know that a 4step distilles model doesn't benefit from overdoing it.
IF you use the base model but with a lora, you can play around with lower lora weight and more steps.

doesn't really make sense. Some say 6 or so can improve results but 16 will have no advances and likely makes things much worse.

But since you can just slot in base, i assume that was a leftover from when this was selected. and not changed when the workflow was exported. Not an expert but long enough at it to know that a 4step distilles model doesn't benefit from overdoing it.
IF you use the base model but with a lora, you can play around with lower lora weight and more steps.

I forgot to edit it.