Logic/DDT
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs: update README.md

Browse Source 
surpresingly -> surprisingly
This commit is contained in:
Ikko Eltociear Ashimine
2025-04-15 17:27:35 +09:00
committed by GitHub
parent 45d5e8db5b
commit 39a5a47961
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
README.md
View File

@@ -8,7 +8,7 @@
[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/ddt-decoupled-diffusion-transformer-1/image-generation-on-imagenet-256x256)](https://paperswithcode.com/sota/image-generation-on-imagenet-256x256?p=ddt-decoupled-diffusion-transformer-1)
## Introduction
We decouple diffusion transformer into encoder-decoder design, and surpresingly that a **more substantial encoder yields performance improvements as model size increases**.
We decouple diffusion transformer into encoder-decoder design, and surprisingly 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).
@@ -122,4 +122,4 @@ python main.py fit -c configs/repa_improved_ddt_xlen22de6_256.yaml
```
## Acknowledgement
The code is mainly built upon [FlowDCN](https://github.com/MCG-NJU/FlowDCN), we also borrow ideas from the [REPA](https://github.com/sihyun-yu/REPA), [MAR](https://github.com/LTH14/mar) and [SiT](https://github.com/willisma/SiT).
The code is mainly built upon [FlowDCN](https://github.com/MCG-NJU/FlowDCN), we also borrow ideas from the [REPA](https://github.com/sihyun-yu/REPA), [MAR](https://github.com/LTH14/mar) and [SiT](https://github.com/willisma/SiT).