File size: 1,564 Bytes
c7014a1
353f31c
 
 
 
 
 
 
 
 
 
c7014a1
353f31c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1ed3c1c
353f31c
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
---
license: openrail++
base_model: stabilityai/stable-diffusion-xl-refiner-1.0
language:
  - en
tags:
  - stable-diffusion
  - stable-diffusion-xl
  - onnxruntime
  - onnx
  - text-to-image
---


# Stable Diffusion XL 1.0 for ONNX Runtime CUDA provider

## Introduction

This repository hosts the optimized versions of **Stable Diffusion XL Refiner 1.0** to accelerate inference with ONNX Runtime CUDA execution provider.

The models are generated by [Olive](https://github.com/microsoft/Olive/tree/main/examples/stable_diffusion) with command like the following:
```
python stable_diffusion_xl.py --provider cuda --optimize --use_fp16_fixed_vae --model_id stabilityai/stable-diffusion-xl-refiner-1.0
```

## Model Description

- **Developed by:** Stability AI
- **Model type:** Diffusion-based text-to-image generative model
- **License:** [CreativeML Open RAIL++-M License](https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/blob/main/LICENSE.md)
- **Model Description:** This is a conversion of the [SDXL refiner 1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0) models for [ONNX Runtime](https://github.com/microsoft/onnxruntime) inference with CUDA execution provider.

The VAE decoder is converted from [sdxl-vae-fp16-fix](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix). There are slight discrepancies between its output and that of the original VAE, but the decoded images should be [close enough for most purposes](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/discussions/7#64c5c0f8e2e5c94bd04eaa80).