Return transformed target in ComponentGraph.fit_features #2780
Conversation
Codecov Report
@@ Coverage Diff @@
## main #2780 +/- ##
=======================================
+ Coverage 99.8% 99.8% +0.1%
=======================================
Files 298 298
Lines 27604 27631 +27
=======================================
+ Hits 27536 27563 +27
Misses 68 68
Continue to review full report at Codecov.
|
freddyaboulton
force-pushed
the
2703-compute-estimator-features-transformed-target
branch
from
25330ac to
d961cf6
Compare
freddyaboulton
requested review from
ParthivNaresh,
angela97lin,
bchen1116,
christopherbunn,
dsherry,
eccabay and
jeremyliweishih
ParthivNaresh
left a comment
ParthivNaresh
left a comment
There was a problem hiding this comment.
Looks good! Does this require performance testing and/or pre and post comparison for permutation importance?
| }, | ||
| ) | ||
| pipeline.fit(X, y) | ||
| pd.testing.assert_series_equal(mock_to_check.call_args[0][1], y + 2) |
There was a problem hiding this comment.
Interesting how this fails with Woodwork 0.6.0 because you're trying to add an int to a categorical but passes with 0.7.1 :)
There was a problem hiding this comment.
Then how come min-dependencies git-test-other passes? That installs ww 0.6.0? 🤔
There was a problem hiding this comment.
(I have woodwork 0.6.0 locally and this passes 🤔 )
evalml/pipelines/component_graph.py
Outdated
| Returns: | ||
| pd.DataFrame: Transformed values. | ||
| Tuple: pd.DataFrame, pd.Series: Transformed features and target. |
There was a problem hiding this comment.
Thank you 😂
I'm not sure if this intended but it looks like sphinx / napoleon separates the return type only on the first colon: https://feature-labs-inc-evalml--2780.com.readthedocs.build/en/2780/autoapi/evalml/pipelines/index.html#evalml.pipelines.ComponentGraph.fit_features
If we do Tuple(pd.DataFrame, pd.Series) (unclear of exact syntax, might be square brackets heh), we could get it all on the return type line :)
There was a problem hiding this comment.
Thank you for letting me know!
| }, | ||
| ) | ||
| pipeline.fit(X, y) | ||
| pd.testing.assert_series_equal(mock_to_check.call_args[0][1], y + 2) |
There was a problem hiding this comment.
(I have woodwork 0.6.0 locally and this passes 🤔 )
freddyaboulton
force-pushed
the
2703-compute-estimator-features-transformed-target
branch
from
3e92131 to
dd83f62
Compare
freddyaboulton
deleted the
2703-compute-estimator-features-transformed-target
branch
|
@ParthivNaresh Forgot to reply to your comment. I don't think the perf tests are necessary in this case. This really only impacts time series pipelines with a detrender or log transformer and those are not added to the perf tests pipelines (because they don't meet the log normal check and because we don't automatically add detrenders to pipelines). |
Pull Request Description
Fixes #2703
I realized it's easier to update
fit_featuresas opposed tocompute_estimator_features. It's also a smaller change this way.This can go in the release in two weeks. Just opening up for review.
After creating the pull request: in order to pass the release_notes_updated check you will need to update the "Future Release" section of
docs/source/release_notes.rstto include this pull request by adding :pr:123.