fix a link

README.md

@@ -43,7 +43,7 @@ The training data comes this [Kaggle competition](https://www.kaggle.com/c/jigsa
 
 ## Training procedure
 
-You can see [this documentation and codes](https://www.kaggle.com/c/jigsaw-unintended-bias-in-toxicity-classification/data) for how we train the model. It takes about 3 hours in a P-100 GPU.
+You can see [this documentation and codes](https://github.com/MSIA/wenyang_pan_nlp_project_2021) for how we train the model. It takes about 3 hours in a P-100 GPU.
 
 ## Evaluation results