kp-forks / grobid / 28729242627

push

github

Parallelize evaluation matching in end-to-end evaluation (#1487)

End2End evaluation is already parallelized using Grobid, but not when computing scores. This PR exploit all available CPUs to speed up the evaluation (from ~20 minutes to ~2 minutes in average). 
Also there is a python evaluation script that was updated, now it optionally makes sure that the models and consolidation configuration are correct (and glutton is reachable).

8724 of 25074 branches covered (34.79%)

Branch coverage included in aggregate %.

18673 of 45615 relevant lines covered (40.94%)

1.66 hits per line