Text-to-Image
Diffusers
stable-diffusion

how to run with UNET 8 step with diffusers? getting noise

#10
by froilo - opened

when i try this

import torch
from diffusers import StableDiffusionXLPipeline, UNet2DConditionModel, EulerDiscreteScheduler
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file

base = "stabilityai/stable-diffusion-xl-base-1.0"
repo = "ByteDance/SDXL-Lightning"
ckpt = "sdxl_lightning_8step_unet.safetensors" # Use the correct ckpt for your step setting!

# Load model.
unet = UNet2DConditionModel.from_config(base, subfolder="unet").to("cuda", torch.float16)
unet.load_state_dict(load_file(hf_hub_download(repo, ckpt), device="cuda"))
pipe = StableDiffusionXLPipeline.from_pretrained(base, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")

# Ensure sampler uses "trailing" timesteps and "sample" prediction type.
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing", prediction_type="sample")

# Ensure using the same inference steps as the loaded model and CFG set to 0.
pipe("A girl smiling", num_inference_steps=8, guidance_scale=0).images[0].save("output.png")

when running lora 8 step its fine.

output.png

ok
solution

...
unet = UNet2DConditionModel.from_config(base, subfolder="unet").to("cuda", torch.float16)
unet.load_state_dict(load_file(hf_hub_download(repo, ckpt), device="cuda"))
pipe = StableDiffusionXLPipeline.from_pretrained(base, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")
froilo changed discussion status to closed
ByteDance org
edited Feb 23

From this line

pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing", prediction_type="sample")

Remove prediction_type="sample". This is only used for 1-step model. You copied the wrong snippet.

Sign up or log in to comment