You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Every PR runs a job called pandas-tests that run the Pandas unit tests using cudf.pandas. Currently, there are two issues with this job:
the results of those tests aren't reported in the job summary or anywhere convenient to look at
the job always passes, regardless of how many tests were passed or failed
Ideally, we should post the results as a job summary and fail the job if the number of tests passed relative to the development branch falls below a threshold (say, 0.1%).
The text was updated successfully, but these errors were encountered:
The reporting piece of this was addressed in #15369 and follow-ups. @galipremsagar do we want to do the other part and add a failure condition as well?
We currently use nightly reports as baseline, so we can't have a failure condition based on a threshold as it can swing widely based on the number of PRs being merged on a given day. We can revisit the second part when we have a more accurate(upto-date) branch based baselines.
Every PR runs a job called
pandas-tests
that run the Pandas unit tests usingcudf.pandas
. Currently, there are two issues with this job:Ideally, we should post the results as a job summary and fail the job if the number of tests passed relative to the development branch falls below a threshold (say, 0.1%).
The text was updated successfully, but these errors were encountered: