fix transformer version

cn_clip/training/main.py

@@ -261,8 +261,12 @@ def main():
         try:
             teacher_model = Model.from_pretrained(args.teacher_model_name)
         except Exception as e:
-            print("An error occurred while loading the model:", e)
-            print("Maybe the transformer version is not compatible, recommend to use transformers >= 4.10.0 and <= 4.30.2")
+            error_message = (
+                "An error occurred while loading the model: {}\n"
+                "Maybe the transformer version is not compatible, "
+                "recommend to use transformers >= 4.10.0 and <= 4.30.2".format(e)
+            )
+            raise RuntimeError(error_message)
 
         for k, v in teacher_model.state_dict().items():
             v.requires_grad = False

distillation.md

@@ -11,6 +11,7 @@
 + Pytorch 1.12及以上版本。
 + [requirements.txt](requirements.txt)要求的其他依赖项
 + **ModelScope**：通过执行`pip install modelscope`安装ModelScope。
++ 为了配合**ModelScope**的使用，**transformers**版本最好控制在(4.10.0, 4.30.2)之间。
 
 ## 在Chinese-CLIP中用起来！
 

distillation_En.md

@@ -9,8 +9,9 @@ Here we provide an example of knowledge distillation for Chinese-CLIP fine-tunin
 + Nvidia GPUs **with Turning, Ampere, Ada or Hopper architecture** (such as H100, A100, RTX 3090, T4, and RTX 2080). Please refer to [this document](https://en.wikipedia.org/wiki/CUDA#GPUs_supported) for the corresponding GPUs of each Nvidia architecture.
 + CUDA 11.4 and above.
 + PyTorch 1.12 and above.
-+ **ModelScope**：Install FlashAttention by executing `pip install modelscope`.
 + Other dependencies as required in [requirements.txt](requirements.txt).
++ **ModelScope**：Install FlashAttention by executing `pip install modelscope`.
++ In order to cooperate with the use of **ModelScope**, the version of **transformers** is best controlled between (4.10.0, 4.30.2).
 
 ## Use it in Chinese-CLIP!
 It is not complicated to apply knowledge distillation to the image side in Chinese-CLIP finetune. Just add the `--distllation` configuration item to the sh script of finetune.