cshaitao / sailer Goto Github PK

I am trying to use the SAILER_en model from the Hugging Face model hub to generate embeddings for my legal documents. However, I am having trouble generating proper embeddings using this model. The embeddings for each token in my document are the same, regardless of the content of the document. I am simply loading the model and trying to fetch the embedding from last hidden state while passing the input tokens.

I noticed that the sample code provided by Hugging Face for using the SAILER_en model with the transformers library references a BertForCotMAE class that does not exist in the library. I am not sure if this is related to the issue I am experiencing.

Can you provide any guidance on how to use the SAILER_en model correctly to generate embeddings for legal documents? Any help would be greatly appreciated.

Thank you

感谢分享代码~

请问如果文本超过最大长度512您的代码可以处理么？
如果不能处理请问您怎么解决的呢？

请问这套代码改改可以用在cail2023的类案检索赛道吗

请问您是用dense作为下游任务的模型么？
我使用您的模型并使用dense.driver.encode 编码后
使用 dense.faiss_retriever 检索
在LeCaRD任务上的NDCG30只能得到 0.0823
请问有可能是什么问题呢？
感谢

你好，论文里提到的CAIL2022-LCR数据集是cail2022类案检索的任务数据集，但我打开赛事官网发现它所采用的数据集链接就是LeCaRD，请问还有CAIL2022-LCR能够下载的连接嘛，十分感谢！