comments about bugs in metric calculation
#20
by tytskiy - opened
These bugs could substantially distort absolute metric values (especially NDCG), but as far as we can judge it does not change the ranking of the models reported in the paper.
P.S. Thanks to Kirill Khrylchenko for identifying these issues
tytskiy changed pull request title from Update benchmarks/yambda/evaluation/metrics.py to comments about bugs in metric calculation
tytskiy changed pull request status to closed