Closed robinholzi closed 2 months ago
( % to main)
( % to main)
@MaxiBoether Thanks for skimming through the PR! I guess skipping the thorough review is reasonable as long as we keep don't completely forget about the fact, that there could be issues! Reviewing would be hard anyways, as the dataframes have no annotations for columns etc. unfortunately.
When building plots for a paper/thesis we have to reinspect the logic anyways. Then it's much easier to validate on a specific case with specific values for fields like 'metric', 'dataset_id', 'eval_handler'!
Attention: Patch coverage is 90.00000%
with 1 line
in your changes missing coverage. Please review.
Project coverage is 82.50%. Comparing base (
f536903
) to head (ff9a424
).
Files | Patch % | Lines |
---|---|---|
...yn/supervisor/internal/pipeline_executor/models.py | 83.33% | 1 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
dependent on #451 and #463
Motivation
It is useful to have a tool that allows us to analyze and compare pipeline runs fast. This enables efficient experimentation and discovering of patterns in the log data. The findings from using this tool enable us to create plots for papers with almost no further data processing.
Changes
Most of the changes are in
/analytics
and are completely greenfield. Some code duplication was needed for the plot components as the UI hooks have to be defined at the server start time (thus not enabling us to generate every component dynamically with all its UI)How to Review
It's more important to review the dataframe wrangling and aggregation compared to the UI boilerplate.