Closed robinholzi closed 5 days ago
( % to main) ( % to main)
Lmk when i should review this
I think it should be called "batching requests" because we do not actually parallelize them?
I think it should be called "batching requests" because we do not actually parallelize them?
Ah right, the evaluator doesn't do parallelization (yet)
Attention: Patch coverage is 96.22642%
with 2 lines
in your changes missing coverage. Please review.
Project coverage is 82.93%. Comparing base (
c6622d3
) to head (8494963
).
Files | Patch % | Lines |
---|---|---|
...yn/supervisor/internal/pipeline_executor/models.py | 81.81% | 2 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
I'll merge this now with the rho-loss unit test failing
the failed one is a slightly flaky test. no problem at all
Motivation
We want to utilize the batching logic introduced in https://github.com/eth-easl/modyn/pull/554 by grouping requests in the supervisor.
This will break the analytics tool because of the adjusted logging. Once we have a completed run I'll open a followup PR.