Revert "Add custom name setter and getter for proxy objects in cudf.pandas"
#16267
Conversation
…pandas` …" This reverts commit 2b2058d.
github-actions
bot
added
Python
galipremsagar
added
improvement
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## branch-24.08 #16267 +/- ##
===============================================
Coverage ? 83.44%
===============================================
Files ? 184
Lines ? 28142
Branches ? 0
===============================================
Hits ? 23482
Misses ? 4660
Partials ? 0 ☔ View full report in Codecov by Sentry. |
|
Hey @galipremsagar, do you know where I can see the failures caused by #16234? |
| additional_attributes={ | ||
| "__init__": _DELETE, | ||
| "__setattr__": Index__setattr__, | ||
| "name": property(name), |
There was a problem hiding this comment.
Oh maybe this should've been a _FastSlowAttribute?
| return self._fsproxy_wrapped._name | ||
|
|
||
|
|
||
| def Index__setattr__(self, name, value): |
There was a problem hiding this comment.
Just a note: This is the same __setattr__ method from _FastSlowProxy except customized to set the _name attr(_names for MultiIndex).
edit: add a link
They can be seen in this job summary: https://github.com/rapidsai/cudf/actions/runs/9877265093/attempts/2#summary-27293125763 Since this PR reverts the changes that were causing the failures, you can see the same job summary of this PR here: https://github.com/rapidsai/cudf/actions/runs/9901041027/attempts/2#summary-27373925178 |
Thanks for catching this! I wasn't aware of these tables. Its interesting that the drop is so large, so I'll look into some of the pandas tests with large drops in the table. |
|
I wonder how "volatile" the pandas pass rate is. What's the threshold for large decrease or increase? |
0.01 to 0.03% changes can be safely considered noise. |
Any reason why we don't make the pandas test check fail for changes >0.05% (say)? |
We currently compare against the nightly baseline, which means we will have a delta that is not being caused by the PR alone, but the diff between the nightly and the PR. To fail the job if the changes >0.05% we will to have a baseline that is run on |
|
Resolved by #16270 |
Reverts #16234
This is an investigation PR into sudden spike in failures.