Skip to content

some classic models+BERT implements in RACE and SQuAD dataset using Pytorch

Notifications You must be signed in to change notification settings

yuansky/MRC-Pytorch

 
 

Repository files navigation

简介

该仓库实现了一些阅读理解的经典模型,为了对让同学对阅读理解有全面的认识,我个人对多种类型的数据集都进行了相关的模型实现。具体可以参见 docs 文件夹下的文档。

Dataset

本仓库采用了多个类型的数据集,选择的都是比较有代表性,比较火热的数据集,包括:

  • 英文:CLOTH, RACE, SQuAD 2.0,TriviaQA, CoQA, MS MARCO,
  • 中文:DuReader, CMRC 2018, CMRC 2019

运行


Results

模型 loss acc
GAReader 1.30194 38.852

Bert Results

模型 loss acc
BertOrigin 0.944 64.471

About

some classic models+BERT implements in RACE and SQuAD dataset using Pytorch

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%