You need to download XLM-Roberta before running the code-----https://huggingface.co/xlm-roberta-base/tree/main

1.Place the file（process001_42k.py） in the same folder as the data set
 run code_offensive\process001_42k.py
5 fold stratified sampling is performed

2.run code_offensive\main.sh 
In this way, the model can be trained first, and then the results can be predicted using weights,see the comment line for details