Skip to content

Latest commit

 

History

History
21 lines (15 loc) · 547 Bytes

ranker.md

File metadata and controls

21 lines (15 loc) · 547 Bytes

Ranking model (Training)

  1. Training ranker model with FairytaleQA
python train.py #--model_file ${model_file}

This produces the model checkpoint at ./model_ckpt

2 Apply ranker model for QA-pairs

Following code generates checkpoint ./model_ckpt/roberta-base_5e-07/5.pth automatically. OR You can download the model checkpoints

python load_generated.py --source ${source} --output_dir ${output_dir}

${source} : pre-constructed QA pairs for ranking

${output_dir} : directory for ranked QA-pairs