search

timoschick / pet

This repository contains the code for "Exploiting Cloze Questions for Few-Shot Text Classification and Natural Language Inference"
https://arxiv.org/abs/2001.07676
Apache License 2.0
1.62k stars 283 forks source link

how to get the datafiles #44

Closed jackfeinmann5 closed 3 years ago

jackfeinmann5 commented 3 years ago

Hi I am not sure how to download the data for each task, could you kindly upload the files somewhere or include the links from which repo you got the data for each tasks?

Could you also comment how I can get unlabeled.jsonl for each task like boolq or rte?

thanks for your help

timoschick commented 3 years ago

If you're interested in the SuperGLUE tasks, you can find all train/unlabeled examples in the FewGlue repository; the original train/validation/test files are available here. Data for AG's News, Yelp Full and Yahoo Questions is from here. For all summarization tasks, required files are loaded using huggingface's datasets library.