Update README.md

Author: IDKiro

README.md

@@ -10,12 +10,12 @@
 
 </div>
 
-> **Abstract:** 
-Recent advancements in diffusion models have positioned them at the forefront of image generation. Despite their superior performance, diffusion models are not without drawbacks; they are characterized by complex architectures and substantial computational demands, resulting in significant latency due to their iterative sampling process. To mitigate these limitations, we introduce a dual approach involving model miniaturization and a reduction in sampling steps, aimed at significantly decreasing model latency. Our methodology leverages knowledge distillation to streamline the U-Net and image decoder architectures, and introduces an innovative one-step DM training technique that utilizes feature matching and score distillation. We present two models, SDXS-512 and SDXS-1024, achieving inference speeds of approximately <b>100 FPS</b> (30x faster than SD v1.5) and <b>30 FPS</b> (60x faster than SDXL) on a single GPU, respectively. Moreover, our training approach offers promising applications in image-conditioned control, facilitating efficient image-to-image translation.
+
+Assuming the image generation time is limited to <b>1 second</b>, then SDXL can only use 16 NFEs to produce a slightly blurry image, while SDXS-1024 can generate 30 clear images. 
 
 ![](images/intro.png)
 
-Assuming the image generation time is limited to <b>1 second</b>, then SDXL can only use 16 NFEs to produce a slightly blurry image, while SDXS-1024 can generate 30 clear images. Besides, our proposed method can also train ControlNet.
+Besides, our proposed method can also train ControlNet.
 
 <p align="left" >
 <img src="images\sketch.gif" width="800" />
@@ -49,6 +49,9 @@ python demo_sketch.py
 
 ## Method
 
+> **Abstract:** 
+Recent advancements in diffusion models have positioned them at the forefront of image generation. Despite their superior performance, diffusion models are not without drawbacks; they are characterized by complex architectures and substantial computational demands, resulting in significant latency due to their iterative sampling process. To mitigate these limitations, we introduce a dual approach involving model miniaturization and a reduction in sampling steps, aimed at significantly decreasing model latency. Our methodology leverages knowledge distillation to streamline the U-Net and image decoder architectures, and introduces an innovative one-step DM training technique that utilizes feature matching and score distillation. We present two models, SDXS-512 and SDXS-1024, achieving inference speeds of approximately <b>100 FPS</b> (30x faster than SD v1.5) and <b>30 FPS</b> (60x faster than SDXL) on a single GPU, respectively. Moreover, our training approach offers promising applications in image-conditioned control, facilitating efficient image-to-image translation.
+
 ### Model Acceleration
 
 We train an extremely light-weight image decoder to mimic the original VAE decoder’s output through a combination of output distillation loss and GAN loss. We also leverage the block removal distillation strategy to efficiently transfer the knowledge from the original U-Net to a more compact version.