submit best ckpt

.gitattributes

@@ -42,3 +42,12 @@ validation_images/.ipynb_checkpoints/500-checkpoint.png filter=lfs diff=lfs merg
 validation_images/.ipynb_checkpoints/200-checkpoint.png filter=lfs diff=lfs merge=lfs -text
 validation_images/.ipynb_checkpoints/300-checkpoint.png filter=lfs diff=lfs merge=lfs -text
 validation_images/.ipynb_checkpoints/100-checkpoint.png filter=lfs diff=lfs merge=lfs -text
+validation_images/1000.png filter=lfs diff=lfs merge=lfs -text
+validation_images/900.png filter=lfs diff=lfs merge=lfs -text
+validation_images/800.png filter=lfs diff=lfs merge=lfs -text
+validation_images/700.png filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/700-checkpoint.png filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/800-checkpoint.png filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/600-checkpoint.png filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/1000-checkpoint.png filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/900-checkpoint.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -14,8 +14,8 @@ inference: false
     
 # LoRA DreamBooth - zhl1bug/sword_1bug
 本仓库的 LoRA 权重是基于 stabilityai/stable-diffusion-2-1 训练而来的，我们采用[DreamBooth](https://dreambooth.github.io/)的技术并使用 sword_1bug 文本进行了训练。 下面是在训练过程中生成的一些图片。
- ![img_0](validation_images/500.png)
-![img_0](validation_images/400.png)
-![img_0](validation_images/300.png)
-![img_0](validation_images/200.png)
+ ![img_0](validation_images/1000.png)
+![img_0](validation_images/900.png)
+![img_0](validation_images/800.png)
+![img_0](validation_images/700.png)
 

paddle_lora_weights.pdparams

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fb359f16050bf11bba0118dfbdfcec2ae519c46f9d37af481682f33b116efc31
+oid sha256:b4a57c1b1d0c57808f3095929750a596abc95bb39c01b2ad0bdf4e04a4cc155b
 size 3357150