latent-consistency / lcm-ssd-1b

Latent Consistency Model (LCM): SSD-1B

Latent Consistency Model (LCM) was proposed in Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference by Simian Luo, Yiqin Tan et al. and Simian Luo , Suraj Patil , and Daniel Gu succesfully applied the same approach to create LCM for SDXL.

This checkpoint is a LCM distilled version of segmind/SSD-1B that allows to reduce the number of inference steps to only between 2 - 8 steps .

Usage

LCM SDXL is supported in 🤗 Hugging Face Diffusers library from version v0.23.0 onwards. To run the model, first install the latest version of the Diffusers library as well as peft , accelerate and transformers . audio dataset from the Hugging Face Hub:

pip install --upgrade pip
pip install --upgrade diffusers transformers accelerate peft

Text-to-Image

The model can be loaded with it's base pipeline segmind/SSD-1B . Next, the scheduler needs to be changed to LCMScheduler and we can reduce the number of inference steps to just 2 to 8 steps.

from diffusers import UNet2DConditionModel, DiffusionPipeline, LCMScheduler
import torch

unet = UNet2DConditionModel.from_pretrained("latent-consistency/lcm-ssd-1b", torch_dtype=torch.float16, variant="fp16")
pipe = DiffusionPipeline.from_pretrained("segmind/SSD-1B", unet=unet, torch_dtype=torch.float16, variant="fp16")

pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
pipe.to("cuda")

prompt = "a close-up picture of an old man standing in the rain"

image = pipe(prompt, num_inference_steps=4, guidance_scale=1.0).images[0]

Image-to-Image

Works as well! TODO docs

Inpainting

Works as well! TODO docs

ControlNet

Works as well! TODO docs

T2I Adapter

Works as well! TODO docs

Speed Benchmark

TODO

Training

TODO