diff --git a/Journey/Day04/Day04.md b/Journey/Day04/Day04.md
index c3a1286..becee9c 100644
--- a/Journey/Day04/Day04.md
+++ b/Journey/Day04/Day04.md
@@ -222,7 +222,7 @@ litgpt pretrain --config Experiments/configs/microstories.yaml
 
 我的模型其实已经训练了一段时间，show一下训练过程中的图表：
 
-![wandb](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/image.png)
+![pretrain_wandb](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/pretrain_wandb.png)
 
 ## 小结
 1. 详细介绍了`litgpt`的预训练模型配置文件。
diff --git a/Journey/Day13/Day13.md b/Journey/Day13/Day13.md
index 2cb8b99..7d3a37b 100644
--- a/Journey/Day13/Day13.md
+++ b/Journey/Day13/Day13.md
@@ -107,7 +107,7 @@ if __name__ == "__main__":
 
 训练过程中的指标波动比较大，简单贴个图：
 
-![](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/image.png)
+![](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/dpo_train.png)
 
 `DPO`的原理细节这里就不展开了，大家可以自行参考论文和开源的代码实现。
 
diff --git a/README.md b/README.md
index dfcf9bd..05e261e 100644
--- a/README.md
+++ b/README.md
@@ -79,7 +79,7 @@ Memory: 9347MiB / 64195MiB
 ```
 ### 预训练
 详细参数请参考 [pretrain.yaml](./Experiments/configs/microstories/pretrain.yaml)。
-![](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/image.png)
+![pretrain_wandb](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/pretrain_wandb.png)
 
 ### 指令微调
 详细参数请参考 [sft.yaml](./Experiments/configs/microstories/sft.yaml)。
@@ -87,7 +87,7 @@ Memory: 9347MiB / 64195MiB
 
 ### DPO
 训练脚本参考[dpo_train.py](./Journey/Day13/dpo_train.py)
-![](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/image.png)
+![](https://erxuanyi-1257355350.cos.ap-beijing.myqcloud.com/dpo_train.png)
 
 
 ## 目录结构