update README

This commit is contained in:
wangshuai6
2025-04-09 11:41:22 +08:00
parent 9b37dcadde
commit 426803f15b

View File

@@ -2,6 +2,9 @@
## Introduction
We decouple diffusion transformer into encoder-decoder design, and surpresingly that a **more substantial encoder yields performance improvements as model size increases**.
![](./figs/main.png)
* We achieves **1.26 FID** on ImageNet256x256 Benchmark with DDT-XL/2(22en6de).
* We achieves **1.28 FID** on ImageNet512x512 Benchmark with DDT-XL/2(22en6de).
* As a byproduct, our DDT can reuse encoder among adjacent steps to accelerate inference.
## Visualizations
![](./figs/teaser.png)
## Usgae