how to change the sampling method when using hugging face diffusion library?
#5
by
chirag11
- opened
how to change the sampling method when using hugging face diffusion library?
Check my example:
from diffusers import DPMSolverMultistepScheduler
from diffusers import StableDiffusionImg2ImgPipeline
device = "cuda"
pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
"runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16, safety_checker=None
)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
pipe.to(device)
Related to this, does anyone know the default settings of runwayml/stable-diffusion-v1-5 or where to find these? For example what is the default sampling method when a model is run using runwayml/stable-diffusion-v1-5 in addition to the steps and so forth? Really appreciate it!