Co-occurrence Statistics Data of the Pile and BERT's pretraining data
收藏数据链接：
官方服务：
资源简介：
This repository contains part of data for the paper "Why Do Neural Language Models Still Need Commonsense Knowledge?" This includes the co-occurrence statistics data computed from the Pile dataset and the pretraining data of BERT, saved in the co-occurrence matrix and occurrence matrix in the numpy (.npy) format.
提供机构：
Kang, Cheongwoong创建时间：
2024-06-20
