Update README.md

README.md

@@ -9,10 +9,6 @@ datasets:
 This is the example model of [Distill SDXL](https://github.com/okotaku/diffengine/tree/main/configs/distill_sd).
 The training is based on [DiffEngine](https://github.com/okotaku/diffengine), the open-source toolbox for training state-of-the-art Diffusion Models with diffusers and mmengine.
 
-Paper: [On Architectural Compression of Text-to-Image Diffusion Models](https://arxiv.org/abs/2305.15798)
-
-Unofficial implementation: https://github.com/segmind/distill-sd
-
 # Training
 
 ```
@@ -34,7 +30,7 @@ import torch
 from diffusers import DiffusionPipeline, UNet2DConditionModel, AutoencoderKL
 
 checkpoint = 'takuoko/tiny_sd_xl_pokemon_blip'
-prompt = 'a picture of a pink and yellow pokemon with a sword'
+prompt = 'a very cute looking pokemon with a hat on its head'
 
 unet = UNet2DConditionModel.from_pretrained(
     checkpoint, torch_dtype=torch.bfloat16
@@ -57,6 +53,12 @@ image.save('demo.png')
 
 # Example result
 
-prompt = 'a picture of a pink and yellow pokemon with a sword'
+prompt = 'a very cute looking pokemon with a hat on its head'
+
+![image](demo.png)
 
-![image](demo.png)
+# Reference
+
+Paper: [On Architectural Compression of Text-to-Image Diffusion Models](https://arxiv.org/abs/2305.15798)
+
+Unofficial implementation: https://github.com/segmind/distill-sd