Skip to content

GitLab

Switch branch/tag
Find file BlameHistoryPermalink
  • Owen Wang's avatar
    make prediction count evaluation stable with DDP · cbdd46ab
    Owen Wang authored 
    Summary: Prediction count evaluator needs to gather it's state before computing metrics, otherwise when parallelized across N GPUs, we only get metrics computed from 1/N of the dataset, increasing our eval signal's variance.

Reviewed By: wat3rBro

Differential Revision: D27416864

fbshipit-source-id: b2c5334cd5a38bebcd06c6ace1627a6b71645fdd
    cbdd46ab
test_prediction_count_evaluation.py 2.48 KB