Update README.md

README.md

@@ -19,12 +19,13 @@ base model: google/ddpm-celebahq-256
 
 fine-tuning dataset: vintageface
 
-> Although loss is not very stable, it seems that the training results show that it can add bright colors to the excellent portrait performance of the base model without changing the hue of the character
-> It's AMAZING.
+> The loss is not very stable, and the output is not very satisfactory.
+> But since it's a study case, you can't ask for too much.
 
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/2ilkZjEo0SW-JdoJIK3kR.png)
 
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/1Aj7dR12dbBBosSuz_pY3.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/QF-nZ18Q251cAiA4Ksi9O.png)
+
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/gkqTtkcIwConyJJaQU4zi.png)
 
 ### Usage
 
@@ -46,12 +47,12 @@ image
 
 微调数据集：Norod78/Vintage-Faces-FFHQAligned
 
-> 虽然 loss 已经不是很稳定，但是看起来训练结果表明它能够在底模优秀的人像表现中，加入鲜明的色彩而不使人物的色相发生变化。
-> 太神奇了。
+> loss 不是很稳定，输出结果也不太令人满意。
+> 不过既然是学习案例，那也不能要求太多。
 
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/2ilkZjEo0SW-JdoJIK3kR.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/QF-nZ18Q251cAiA4Ksi9O.png)
 
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/1Aj7dR12dbBBosSuz_pY3.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/6462500d2538819c729dc355/gkqTtkcIwConyJJaQU4zi.png)
 
 ### 模型使用