SDXL Flash i collaboratio with Project Fluetly

Itroducig the ew fast model SDXL Flash, we leared that all fast XL models work fast, but the quality decreases, ad we also made a fast model, but it is ot as fast as LCM, Turbo, Lightig ad Hyper, but the quality is higher. Below you will see the study with steps ad cfg.

Steps ad CFG (Guidace)

Optimal settigs

• Steps: 6-9
• CFG Scale: 2.5-3.5
• Sampler: DPM++ SDE

Diffusers usage

pip istall torch diffusers

import torch
from diffusers import StableDiffusioXLPipelie, DPMSolverSiglestepScheduler

# Load model.
pipe = StableDiffusioXLPipelie.from_pretraied("sd-commuity/sdxl-flash", torch_dtype=torch.float16, variat="fp16").to("cuda")

# Esure sampler uses "trailig" timesteps.
pipe.scheduler = DPMSolverSiglestepScheduler.from_cofig(pipe.scheduler.cofig, timestep_spacig="trailig")

# Image geeratio.
pipe("a happy dog, suy day, realism", um_iferece_steps=7, guidace_scale=3).images[0].save("output.pg")

SDXL Flash in collaboration with Project Fluently Introducing the new fast model SDXL Flash, we lea