Train on rice

[idealwhite]

Hi, I really appreciate your job! Could you give some instructions on how to process and train model on RACE dataset?

[intfloat]

Thank you for your interests.

Both RACE dataset and SemEval-2018 Task 10 dataset are multiple-choice reading comprehension datasets. All you need to do is to preprocess RACE dataset into the same format as required by this repository. You can check out src/preprocess.py preprocess_race_dataset for reference.

Be cautious that RACE dataset is much larger and takes longer time to train.

[idealwhite]

THanks for your reply, it works well!