LCM-LoRA-SSD-1B Open Source Model - Accelerate segmind/SSD-1B Inference in Just 2

Lcm Lora Ssd 1b

Developed by latent-consistency

A distilled consistency adapter designed for segmind/SSD-1B, reducing inference steps to just 2-8 steps

Text-to-Image #Fast Inference #Text-to-Image Generation #LoRA Fine-Tuning

Downloads 1,512

Release Time : 11/9/2023

Model Overview

Latent Consistency Model (LCM) LoRA is a universal stable diffusion acceleration module specifically designed for the segmind/SSD-1B model, significantly reducing inference steps for text-to-image generation tasks.

Model Features

Efficient Inference

Reduces inference steps to just 2-8 steps, significantly improving generation speed

Strong Compatibility

Designed for segmind/SSD-1B but can be applied to other stable diffusion models

Lightweight Adaptation

Uses LoRA technology, requiring only a small number of parameters to accelerate the model

Model Capabilities

Text-to-Image Generation

Image-to-Image Generation

Image Inpainting

ControlNet Control

T2I Adapter

Use Cases

Artistic Creation

Rapid Concept Art Generation

Artists can quickly generate multiple concept art proposals

High-quality concept art generated within 4 inference steps

Design Assistance

Product Design Visualization

Designers can quickly transform text descriptions into visual representations

Rapid iteration of design solutions

🚀 Latent Consistency Model (LCM) LoRA: SSD-1B

Latent Consistency Model (LCM) LoRA is a distilled consistency adapter that significantly reduces the inference steps for text - to - image generation, enhancing efficiency.

Latent Consistency Model (LCM) LoRA was proposed in LCM-LoRA: A universal Stable-Diffusion Acceleration Module by Simian Luo, Yiqin Tan, Suraj Patil, Daniel Gu et al.

It is a distilled consistency adapter for segmind/SSD-1B that allows to reduce the number of inference steps to only between 2 - 8 steps.

Model	Params / M
lcm-lora-sdv1-5	67.5
lcm-lora-ssd-1b	105
lcm-lora-sdxl	197M

🚀 Quick Start

LCM-LoRA 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.

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

💻 Usage Examples

Basic Usage

Let's load the base model segmind/SSD-1B first. Next, the scheduler needs to be changed to LCMScheduler and we can reduce the number of inference steps to just 2 to 8 steps. Please make sure to either disable guidance_scale or use values between 1.0 and 2.0.

import torch
from diffusers import LCMScheduler, AutoPipelineForText2Image

model_id = "segmind/SSD-1B"
adapter_id = "latent-consistency/lcm-lora-ssd-1b"

pipe = AutoPipelineForText2Image.from_pretrained(model_id, torch_dtype=torch.float16, variant="fp16")
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
pipe.to("cuda")

# load and fuse lcm lora
pipe.load_lora_weights(adapter_id)
pipe.fuse_lora()


prompt = "Self-portrait oil painting, a beautiful cyborg with golden hair, 8k"

# disable guidance_scale by passing 0
image = pipe(prompt=prompt, num_inference_steps=4, guidance_scale=0).images[0]