A pytorch implement of bert baseline for machine comprehension task in 2019 cail(法研杯 阅读理解)
Please visit the official website for the task description and data at http://cail.cipsc.org.cn/instruction.html
This code provides a coarse Bert Joint Baseline. While it can reach a not bad f1 score (~73), further optimizations on data processing, hyper-parameters tuning and candidates strategy are still expected.
python train_start.py
python test.py
If it helps, welcome to star me. Enjoying the match!