Open iuliaturc opened 5 months ago
When setting refiner=base_image_refiner, the refine_steps argument promises to control how many denoising steps will be performed with the refiner:
refiner=base_image_refiner
refine_steps
However, in practice, this is only true if you additional set prompt_strength=1.0. The reason is that:
prompt_strength=1.0
cog-sdxl
num_inference_steps=refiner_steps
diffusers
init_step
num_inference_steps
strength
This results in surprising behavior, where you set refiner_steps=N only to see that it ran for fewer steps. I think you could either:
refiner_steps=N
strength=1
When setting
refiner=base_image_refiner
, therefine_steps
argument promises to control how many denoising steps will be performed with the refiner:However, in practice, this is only true if you additional set
prompt_strength=1.0
. The reason is that:cog-sdxl
setsnum_inference_steps=refiner_steps
here, anddiffusers
computes theinit_step
for refinement based both onnum_inference_steps
andstrength
here.This results in surprising behavior, where you set
refiner_steps=N
only to see that it ran for fewer steps. I think you could either:strength=1
here