license: openrail++
tags:
- text-to-image
- stable-diffusion
library_name: diffusers
SDXL-Lightning
SDXL-Lightning is a lightning-fast text-to-image generation model. It can generate high-quality 1024px images in a few steps. For more information, please refer to our paper: SDXL-Lightning: Progressive Adversarial Diffusion Distillation. The models are released for research purposes only.
Our models are distilled from stabilityai/stable-diffusion-xl-base-1.0. This repository contains checkpoints for 1-step, 2-step, 4-step, and 8-step distilled models. The generation quality of our 2-step, 4-step, and 8-step model is amazing. Our 1-step model is more experimental.
We provide both full UNet and LoRA checkpoints. The full UNet models have the best quality while the LoRA models can be applied to other base models.
Diffusers Usage
Please always use the correct checkpoint for the corresponding inference steps.
2-Step, 4-Step, 8-Step UNet
import torch
from diffusers import StableDiffusionXLPipeline, EulerDiscreteScheduler
from huggingface_hub import hf_hub_download
base = "stabilityai/stable-diffusion-xl-base-1.0"
repo = "ByteDance/SDXL-Lightning"
ckpt = "sdxl_lightning_4step_unet.pth" # Use the correct ckpt for your step setting!
# Load model.
pipe = StableDiffusionXLPipeline.from_pretrained(base, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.unet.load_state_dict(torch.load(hf_hub_download(repo, ckpt), map_location="cuda"))
# Ensure sampler uses "trailing" timesteps.
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
# Ensure using the same inference steps as the loaded model and CFG set to 0.
pipe("A girl smiling", num_inference_steps=4, guidance_scale=0).images[0].save("output.png")
2-Step, 4-Step, 8-Step LoRA
import torch
from diffusers import StableDiffusionXLPipeline, EulerDiscreteScheduler
from huggingface_hub import hf_hub_download
base = "stabilityai/stable-diffusion-xl-base-1.0"
repo = "ByteDance/SDXL-Lightning"
ckpt = "sdxl_lightning_4step_lora.pth" # Use the correct ckpt for your step setting!
# Load model.
pipe = StableDiffusionXLPipeline.from_pretrained(base, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.load_lora_weights(hf_hub_download(repo, ckpt))
pipe.fuse_lora()
# Ensure sampler uses "trailing" timesteps.
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
# Ensure using the same inference steps as the loaded model and CFG set to 0.
pipe("A girl smiling", num_inference_steps=4, guidance_scale=0).images[0].save("output.png")
1-Step UNet
The 1-step model uses "sample" prediction instead of "epsilon" prediction! The scheduler needs to be configured correctly.
import torch
from diffusers import StableDiffusionXLPipeline, EulerDiscreteScheduler
from huggingface_hub import hf_hub_download
base = "stabilityai/stable-diffusion-xl-base-1.0"
repo = "ByteDance/SDXL-Lightning"
ckpt = "sdxl_lightning_1step_unet_x0.pth" # Use the correct ckpt for your step setting!
# Load model.
pipe = StableDiffusionXLPipeline.from_pretrained(base, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.unet.load_state_dict(torch.load(hf_hub_download(repo, ckpt), map_location="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=1, guidance_scale=0).images[0].save("output.png")
ComfyUI Usage
Please always use the correct checkpoint for the corresponding inference steps. Please use Euler sampler with sgm_uniform scheduler.
2-Step, 4-Step, 8-Step UNet
- Download the UNet checkpoint to
/ComfyUI/models/unet. - Download our ComfyUI UNet workflow.
2-Step, 4-Step, 8-Step LoRA
- Download the LoRA checkpoint to
/ComfyUI/models/loras - Download our ComfyUI LoRA workflow.
1-Step UNet
ComfyUI does not support changing model formulation to x0-prediction, so it is not usable in ComfyUI yet. Hopefully ComfyUI gets updated soon.