QZ

The reported results of paper Convolutional Neural Networks for Sentence Classification is 81.5 what's the best performance of this code on MR dataset? Indeed, I have tested several repo in...

@attardi @gojomo @xiaoling Thanks for your work! I want to extract the corpus in the format: title, title's description, all mentions/anchors, and their positions and associated entities. So, could I...

the test set and dev set have same results .

1

### Description ### Other Comments when the type is large, the test set and dev set have same results . why ?

没有merge.txt和vocab.json

3

提供下载的预训练模型压缩包中没有vocab.json和merge.txt，那么Roberta模型的tokenizer怎么加载？使用bert的tokenizer么？但是Roberta和bert的tokenizer不一样啊？

does anyone know why the last one is excluded ? Welcome to join this Wechat Group for discussion. 

what does the variable n_ctx mean ? could you explain why it is computed like that ? `n_ctx = min(max(max(len(x[:max_len]) for x in X) for X in [trX, vaX]) +...

what is the index_path? which file generates it ? could it be generated by pyserini?

ERNIE.py与bert.py

3

您好，请问ERNIE.py体现ERNIE模型的地方在哪里呢？为什么我看不出来和bert.py有什么大的区别？

will the corpus be packed and provided?