This repo contains codes and pre-trained models for our paper
He Bai, Peng Shi, Jimmy Lin, Luchen Tan, Kun Xiong, Wen Gao, Jie Liu, Ming Li
ACL SRW 2021
./examples/passage_generation/bash_run_essay.sh $1 $2 $3
This scripts support train, eval, test, and generation 4 modes by changing the value of TRAIN EVAL TEST and GENERATE in this bash script.
The position arguments $1 is the input type, and could be paragraph, and passage. paragraph means add paragraph breaker, while passage means without paragraph breaker.
The position arguments $2 the paragraph breaker type, True means paragraph seperator(SEP), False is the end of paragraph(EOP).
The position arguments $3 is the visible cuda devices.
For writingprompts, we replace all " " into paragraph seperator, which can be find in preprocess.ipynb file.
./examples/passage_generation/bash_run_wp.sh $1 $2 $3
This scripts support train, eval, test, and generation 4 modes by changing the value of TRAIN EVAL TEST and GENERATE in this bash script.
The position arguments $1 is the paragraph breaker type, and could be none, newline, and eos. none means no paragraph breaker, newline is \n, eos is the DIY paragraph breaker.
The position arguments $2 the paragraph breaker type, True means paragraph seperator(SEP), False is the end of paragraph(EOP).
The position arguments $3 is the visible cuda devices.