Update README.md

README.md

@@ -44,7 +44,8 @@ When the parameter skip_special_tokens is False:
 
 ## Training data
 
-“Siku Quanshu” full-text corpus was used as Training Data which is same as the project of [SikuBERT](https://huggingface.co/SIKU-BERT/sikubert).
+“Siku Quanshu” full-text corpus was used as Training Data which is same as the project of [SikuBERT](https://huggingface.co/SIKU-BERT/sikubert) to train SikuGPT2.
+[Chinese-poetry](https://github.com/chinese-poetry/chinese-poetry)was used as Training Data to train SikuGPT2-poem based on SikuGPT2.
 
 ## Training procedure