Skip to content

[SPARK-28736][SPARK-28735][PYTHON][ML][TESTS] Fix PySpark ML tests to pass in JDK 11 #25475

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
HyukjinKwon wants to merge 1 commit into from
Closed

[SPARK-28736][SPARK-28735][PYTHON][ML][TESTS] Fix PySpark ML tests to pass in JDK 11 #25475

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion python/pyspark/ml/tests/test_algorithms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,7 +86,7 @@ def test_raw_and_probability_prediction(self):
expected_rawPrediction = [-11.6081922998, -8.15827998691, 22.17757045]
self.assertTrue(result.prediction, expected_prediction)
self.assertTrue(np.allclose(result.probability, expected_probability, atol=1E-4))
self.assertTrue(np.allclose(result.rawPrediction, expected_rawPrediction, atol=1E-4))
self.assertTrue(np.allclose(result.rawPrediction, expected_rawPrediction, atol=1))
Copy link
Member

@dongjoon-hyun dongjoon-hyun Aug 16, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is 1 the minimum difference?

Copy link
Member Author

@HyukjinKwon HyukjinKwon Aug 16, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yup ..

JDK 8:

[-11.19194106875243,-7.677866573997363,21.280214474039443]

JDK 11:

[-11.608192299802019,-8.158279986906651,22.177570449962918]

Seems multiple floats affects the results while they are roughly correct.

Copy link
Member

@srowen srowen Aug 16, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure where the difference comes from, but it could be subtle differences in randomization or something across the JDKs. If these two tests are the only ones that vary, I think we're OK. I agree with loosening the bound here as these are log-odds, and I suspect the test values were picked just because it's what some previous run spit out (that is, it's too specific)



class OneVsRestTests(SparkSessionTestCase):
Expand Down
6 changes: 3 additions & 3 deletions python/pyspark/mllib/clustering.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -383,11 +383,11 @@ class GaussianMixtureModel(JavaModelWrapper, JavaSaveable, JavaLoader):
>>> model.predict([-0.1,-0.05])
0
>>> softPredicted = model.predictSoft([-0.1,-0.05])
Copy link
Member Author

@HyukjinKwon HyukjinKwon Aug 16, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For instance, weights within Gaussian mixture model:

JDK 8

weights: WrappedArray(0.49520257460263445, 0.33813075873069875, 0.16666666666666685)

JDK 11

weights: WrappedArray(0.5000000000000001, 0.33333333333333326, 0.16666666666666666)

Copy link
Member

@srowen srowen Aug 16, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also probably OK for the same reason. The test was too specific.

>>> abs(softPredicted[0] - 1.0) < 0.001
>>> abs(softPredicted[0] - 1.0) < 0.03
True
>>> abs(softPredicted[1] - 0.0) < 0.001
>>> abs(softPredicted[1] - 0.0) < 0.03
True
>>> abs(softPredicted[2] - 0.0) < 0.001
>>> abs(softPredicted[2] - 0.0) < 0.03
True

>>> path = tempfile.mkdtemp()
Expand Down