hotpotqa / hotpot

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

A question about evaluate.

tomtang110 opened this issue · comments

Hi
I run my results in the eval() in hotpot_evaluate_v1.py, however, the result may be not the same with your scores in leaderboard. Could you tell me the correct function to evaluate?
屏幕快照 2019-07-07 14 35 18

Hi, are you sure you're using our script on a complete output file (an example of predictions on the dev set can be found on the website)? Our script should print out a JSON object containing various metrics, and your output looks very different from it.