Update README.md

README.md

@@ -17,9 +17,9 @@ inference: False
 
 A deep VAE model pretrained on Wudao dataset. Both encoder and decoder are based on GPT-2 architecture. Such model is particularly suitable for paraphrasing, semantic updating and fine-grained attributes control.
 
-**请注意本模型是在通用语料上进行的预训练。这增加了模型的泛化能力使其能够在微调时快速适应到下游特定领域上，但同时也弱化了其对通用文本的重构能力。如要获得最佳效果请在特定领域微调后使用，并参考本系列开源的CVAE的做法与效果 [Randeng-DELLA-226M-CVAE-NER-Chinese](https://huggingface.co/IDEA-CCNL/Randeng-DELLA-226M-CVAE-NER-Chinese)。
+**请注意本模型是在通用语料上进行的预训练。这增加了模型的泛化能力使其能够在微调时快速适应到下游特定领域上，但同时也弱化了其对通用文本的重构能力。如要获得最佳效果请在特定领域微调后使用，并参考本系列开源的CVAE的做法与效果 [Randeng-DELLA-226M-CVAE-NER-Chinese](https://huggingface.co/IDEA-CCNL/Randeng-DELLA-226M-CVAE-NER-Chinese)。**
 
-Please bear in mind that this model is pre-trained in open domian dataset. Such pretraining enhanced its generalizability and made it capable of adapting to specific domain easily, however it also lessened its strength to reconstruct given texts. To get the maximum effect of this model, consider finetuning it in your desired task domain. You can find such example in [Randeng-DELLA-226M-CVAE-NER-Chinese](https://huggingface.co/IDEA-CCNL/Randeng-DELLA-226M-CVAE-NER-Chinese)**
+**Please bear in mind that this model is pre-trained in open domian dataset. Such pretraining enhanced its generalizability and made it capable of adapting to specific domain easily, however it also lessened its strength to reconstruct given texts. To get the maximum effect of this model, consider finetuning it in your desired task domain. You can find such example in [Randeng-DELLA-226M-CVAE-NER-Chinese](https://huggingface.co/IDEA-CCNL/Randeng-DELLA-226M-CVAE-NER-Chinese)**
 
 ## 模型分类 Model Taxonomy