Update RMSE, MSLE to handle missing values #542
Conversation
# Conflicts: # etna/metrics/__init__.py # etna/metrics/functional_metrics.py # tests/test_metrics/test_metrics.py
|
🚀 Deployed on https://deploy-preview-542--etna-docs.netlify.app |
etna/metrics/functional_metrics.py
Outdated
| A non-negative floating point value (the best value is 0.0), or an array of floating point values, | ||
| one for each individual target. | ||
| """ | ||
| mse_result = np.asarray(mse(y_true=y_true, y_pred=y_pred, multioutput=multioutput)) |
There was a problem hiding this comment.
Why do you use np.asarray here?
There was a problem hiding this comment.
To avoid typing errors from mypy.
There was a problem hiding this comment.
I'm not sure if it is ok. In multioutput="joint" the result should be float, but np.asarray makes it smth like np.array([value]) instead of value.
I thought that we have tests that check that.
There was a problem hiding this comment.
Yes, but np.sqrt makes from np.array(value) to np.float64.
There was a problem hiding this comment.
- We tested that current behavior is ok.
- It could be not so obvious for other readers why do we use
np.asarray. - Where exactly mypy is raising error? In some other functional metrics we just put
type: ignoreat the last line to ignore the mypy.
There was a problem hiding this comment.
error: Incompatible return value type (got "ndarray[Any, dtype[Any]]", expected "Union[float, Sequence[float], Sequence[Sequence[float]]]") [return-value]
Got it, I'll try to use type: ignore
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #542 +/- ##
==========================================
- Coverage 90.49% 90.48% -0.01%
==========================================
Files 256 256
Lines 17364 17373 +9
==========================================
+ Hits 15713 15720 +7
- Misses 1651 1653 +2 ☔ View full report in Codecov by Sentry. |
# Conflicts: # CHANGELOG.md
Before submitting (must do checklist)
Proposed Changes
See #526
Closing issues
Closes #526.