search

eth-easl / modyn

Modyn is a research-platform for training ML models on dynamic datasets.
MIT License
22 stars 3 forks source link

Fix batch number #533

Closed XianzheMa closed 2 weeks ago

XianzheMa commented 2 weeks ago

Previously, we didn't record the number of passed batches correctly: We use a batch_number which is generated purely from enumeration in dataloader. Therefore this number is irrelevant to the number of epochs (only shows how many batches there are in one epoch). A similar issue exists on the iteration on StB when we calculate scores class by class. The number of batches passed in the previous class is not correctly accumulated on the current class.

This PR fixes it.

github-actions[bot] commented 2 weeks ago

Line Coverage: -% ( % to main) Branch Coverage: -% ( % to main)

codecov[bot] commented 2 weeks ago

Codecov Report

Attention: Patch coverage is 83.33333% with 3 lines in your changes missing coverage. Please review.

Project coverage is 82.81%. Comparing base (bf96bfa) to head (9b2a8c9). Report is 1 commits behind head on main.

:exclamation: Current head 9b2a8c9 differs from pull request most recent head 07ebc07

Please upload reports for the commit 07ebc07 to get more accurate results.

Files Patch % Lines
...trainer_server/internal/trainer/pytorch_trainer.py 82.35% 3 Missing :warning:
Additional details and impacted files ```diff @@ Coverage Diff @@ ## main #533 +/- ## ========================================== + Coverage 82.31% 82.81% +0.50% ========================================== Files 217 217 Lines 10125 10124 -1 ========================================== + Hits 8334 8384 +50 + Misses 1791 1740 -51 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.