README.md

metadata

license: openrail++
language:
  - en
tags:
  - stable-diffusion
  - tensorrt
  - text-to-image

Stable Diffusion XL 1.0 TensorRT

Introduction

This repository contains Stable Diffusion XL 1.0 ONNX models compatible with TensorRT.

Source models:

• SDXL base 1.0
• SDXL refiner 1.0

Benchmark

Timings at 1024x1024

	CLIP	UNet 40 steps	VAE decode	Pipline	Throughput
A10	8.98 ms	12576.53 ms	0.00 ms	12588.26 ms	0.08 image/s
A100	5.99 ms	3358.87 ms	0.00 ms	3367.04 ms	0.30 image/s
H100	4.70 ms	1772.29 ms	0.00 ms	1779.01 ms	0.56 image/s

Model Description

• Developed by: Stability AI
• Model type: Diffusion-based text-to-image generative model
• License: CreativeML Open RAIL++-M License
• Model Description: This is a conversion of the Stable Diffusion XL 1.0 model for NVIDIA TensorRT optimized inference