pixelddt-t2i update
This commit is contained in:
@@ -13,6 +13,9 @@ We decouple diffusion transformer into encoder-decoder design, and surprisingly
|
|||||||
* We achieves **1.26 FID** on ImageNet256x256 Benchmark with DDT-XL/2(22en6de).
|
* 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).
|
* 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.
|
* As a byproduct, our DDT can reuse encoder among adjacent steps to accelerate inference.
|
||||||
|
## Update 5/6/2025
|
||||||
|
* PixelDDT-XXL/16-R1024-T2I achieves **66.7** without prompt rewriting and **71.2** with prompt rewriting on GenEval benchmark
|
||||||
|
* Pixel Space Text-to-image models(PixelDDT-XXL/16) will be released soon.
|
||||||
## Visualizations
|
## Visualizations
|
||||||

|

|
||||||
## Checkpoints
|
## Checkpoints
|
||||||
|
|||||||
Reference in New Issue
Block a user