Skip to content

Conversation

@Lewuathe
Copy link
Contributor

Copied ML models must have the same parent of original ones

@SparkQA
Copy link

SparkQA commented Jul 16, 2015

Test build #37502 has finished for PR 7447 at commit 6a5b495.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds no public classes.

@SparkQA
Copy link

SparkQA commented Jul 29, 2015

Test build #38842 has finished for PR 7447 at commit f091a6b.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds no public classes.

@jkbradley
Copy link
Member

@Lewuathe Sorry for the delay, but this would actually be important to get into Spark 1.5. Would you have time to fix the merge conflicts, and then I can make a final pass? Thank you!

@Lewuathe
Copy link
Contributor Author

Lewuathe commented Aug 4, 2015

@jkbradley I merged master into this branch. Could you check it? Thank you!

@SparkQA
Copy link

SparkQA commented Aug 4, 2015

Test build #39661 has finished for PR 7447 at commit 5316908.

  • This patch fails SparkR unit tests.
  • This patch merges cleanly.
  • This patch adds no public classes.

Copy link
Member

Choose a reason for hiding this comment

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

fix indentation

Copy link
Member

Choose a reason for hiding this comment

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

organize imports (Please check below for other files too.)

@jkbradley
Copy link
Member

@Lewuathe Thanks for the update. Looking at this again, I've become more convinced that the check should piggy-back on other tests. For most models, it fits naturally with the standard ParamsSuite.checkParams check. For a few models, it's pretty expensive to add a unit test which requires fitting a new model. Could you please put the check into existing unit tests?

@SparkQA
Copy link

SparkQA commented Aug 6, 2015

Test build #40081 has finished for PR 7447 at commit 1e0b610.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds the following public classes (experimental):
    • sealed trait LogisticRegressionTrainingSummary extends LogisticRegressionSummary
    • sealed trait LogisticRegressionSummary extends Serializable
    • class IsotonicRegression(override val uid: String) extends Estimator[IsotonicRegressionModel]
    • final class SpecificMutableRow(val values: Array[MutableValue])
    • final class SpecificRow extends $
    • trait BaseGenericInternalRow extends InternalRow
    • class GenericMutableRow(values: Array[Any]) extends MutableRow with BaseGenericInternalRow

@SparkQA
Copy link

SparkQA commented Aug 7, 2015

Test build #40106 has finished for PR 7447 at commit 4c60f8d.

  • This patch fails SparkR unit tests.
  • This patch merges cleanly.
  • This patch adds the following public classes (experimental):
    • sealed trait LogisticRegressionTrainingSummary extends LogisticRegressionSummary
    • sealed trait LogisticRegressionSummary extends Serializable
    • class IsotonicRegression(override val uid: String) extends Estimator[IsotonicRegressionModel]
    • final class SpecificMutableRow(val values: Array[MutableValue])
    • final class SpecificRow extends $
    • trait BaseGenericInternalRow extends InternalRow
    • class GenericMutableRow(values: Array[Any]) extends MutableRow with BaseGenericInternalRow
    • case class StringTranslate(srcExpr: Expression, matchingExpr: Expression, replaceExpr: Expression)

@Lewuathe
Copy link
Contributor Author

@jkbradley Could you check it? We can keep separate test case because some suite such as RandomForestRegressionSuite has no param check case.

@SparkQA
Copy link

SparkQA commented Aug 10, 2015

Test build #40289 has finished for PR 7447 at commit 5d798e9.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds no public classes.

@jkbradley
Copy link
Member

@Lewuathe I realized my suggestion of including the check within the Params check was not a good idea since those checks create models without parents. (The parent check therefore just ensures that null equals null.) Sorry for asking for yet another update, but could you please update the tests to piggy-back off of other tests which create models via fit()? Those tests will create models with actual parents, so that the checkCopy() method will do a reasonable test. Please let me know if this makes sense to you.

Thank you!

@Lewuathe
Copy link
Contributor Author

@jkbradley That sounds good to me at last. I'll update. Thank you!

@SparkQA
Copy link

SparkQA commented Aug 11, 2015

Test build #40449 has finished for PR 7447 at commit 9445cb7.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds no public classes.

Copy link
Member

Choose a reason for hiding this comment

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

fix indentation

Copy link
Member

Choose a reason for hiding this comment

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

This should be in a different test. The model should be produced via fit(), not constructed from scratch, so that the model has a parent.

@SparkQA
Copy link

SparkQA commented Aug 12, 2015

Test build #40625 has finished for PR 7447 at commit 200dd11.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds no public classes.

Copy link
Member

Choose a reason for hiding this comment

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

I should have been clearer: Can you please check both that the uids match and the equality operator?

assert(copied.parent.uid == model.parent.uid)
assert(copied.parent == model.parent)

@jkbradley
Copy link
Member

Looks good except that 1 item

@SparkQA
Copy link

SparkQA commented Aug 13, 2015

Test build #40759 has finished for PR 7447 at commit c01e93c.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds the following public classes (experimental):
    • trait Identifiable
    • class VectorUDT extends UserDefinedType[Vector]
    • class KMeans(JavaEstimator, HasFeaturesCol, HasPredictionCol, HasMaxIter, HasTol, HasSeed):

@jkbradley
Copy link
Member

LGTM. Merging with master and branch-1.5
Thanks!

asfgit pushed a commit that referenced this pull request Aug 13, 2015
…here is a parent

Copied ML models must have the same parent of original ones

Author: lewuathe <[email protected]>
Author: Lewuathe <[email protected]>

Closes #7447 from Lewuathe/SPARK-9073.

(cherry picked from commit 2932e25)
Signed-off-by: Joseph K. Bradley <[email protected]>
@asfgit asfgit closed this in 2932e25 Aug 13, 2015
CodingCat pushed a commit to CodingCat/spark that referenced this pull request Aug 17, 2015
…here is a parent

Copied ML models must have the same parent of original ones

Author: lewuathe <[email protected]>
Author: Lewuathe <[email protected]>

Closes apache#7447 from Lewuathe/SPARK-9073.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants