S

Svdq Int4 Flux.1 Depth Dev

Developed by mit-han-lab
INT4 quantized version of FLUX.1-Depth-dev, capable of generating images from text descriptions while adhering to the structure of the input image. Compared to the original BF16 model, this version saves approximately 4x memory and improves runtime speed by 2-3x.
Downloads 9,085
Release Time : 2/4/2025

Model Overview

This model is an INT4 quantized version based on FLUX.1-Depth-dev, primarily used for image generation tasks. It can generate images from text descriptions while preserving the structure of the input image.

Model Features

Efficient Quantization
Utilizes SVDQuant method for INT4 quantization, significantly reducing memory usage and improving runtime speed.
Structure Preservation
Generates images from text descriptions while preserving the structure of the input image.
High-Performance Inference
Optimized via the Nunchaku engine to reduce data movement overhead and enhance inference efficiency.

Model Capabilities

Text-to-Image Generation
Depth-to-Image Conversion
Image Structure Preservation
Efficient Quantized Inference

Use Cases

Creative Design
Concept Art Generation
Generates concept art from text descriptions while preserving the structure of the input image.
Produces high-quality concept art with rich details and accurate structure.
Image Editing
Image Style Transfer
Transforms input images into different styles while maintaining the original structure.
Style-transferred images retain the original structure with diverse and natural styles.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase