K

Koala Lightning 1b

Developed by etri-vilab
KOALA-Lightning-1B is a knowledge distillation model based on SDXL-Lightning. It achieves efficient text-to-image generation by compressing the U-Net structure, with a parameter scale of 1.16B.
Downloads 390
Release Time : 5/29/2024

Model Overview

An efficient text-to-image generation model that compresses the SDXL architecture through self-attention knowledge distillation technology, significantly improving the inference speed while maintaining the generation quality.

Model Features

Efficient U-Net architecture
Compared with SDXL, the U-Net volume is reduced by 54%, and the number of parameters is only 1.16B.
Self-attention knowledge distillation
Maintain the image generation quality by focusing on distilling self-attention features.
Low video memory requirement
Can generate 1024px images on a consumer-grade GPU with 8GB video memory (such as 3060Ti).

Model Capabilities

Text-to-image generation
High-quality image synthesis
Fast inference

Use Cases

Creative design
Fast concept map generation
Provide designers with the ability to generate rapid prototypes.
Generate a 1024x1024 image in 0.66 seconds (NVIDIA 4090)
Resource-constrained environment
Deployment on consumer-grade GPUs
Achieve high-definition image generation on devices with limited video memory.
Can run on a 3060Ti with 8GB video memory
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase