This is an automated email from the ASF dual-hosted git repository.
zhengruifeng pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/spark.git
The following commit(s) were added to refs/heads/master by this push:
new bf3bb385d7b9 [SPARK-57910][ML][PYTHON] Make ALS reuse the shared
HasIntermediateStorageLevel param
bf3bb385d7b9 is described below
commit bf3bb385d7b952f4162452f5d677120111308fb4
Author: Mao Li <[email protected]>
AuthorDate: Tue Jul 7 08:42:16 2026 +0800
[SPARK-57910][ML][PYTHON] Make ALS reuse the shared
HasIntermediateStorageLevel param
### What changes were proposed in this pull request?
This is a follow-up of SPARK-57860
(https://github.com/apache/spark/pull/56949), as suggested by zhengruifeng
during that review: now that the shared param trait
`HasIntermediateStorageLevel` exists, make `ALS` reuse it instead of declaring
its own duplicate `intermediateStorageLevel` param.
- Scala: `ALSParams` now extends `HasIntermediateStorageLevel`; the
locally-declared `intermediateStorageLevel` param, its getter, and its
`setDefault` entry are removed (the shared trait provides all three with
identical semantics: default `"MEMORY_AND_DISK"`, value cannot be `"NONE"`).
- Python: `_ALSParams` now mixes in `HasIntermediateStorageLevel` from
`pyspark.ml.param.shared`; the locally-declared `Param`, its getter, and its
default are removed likewise.
`ALS.setIntermediateStorageLevel` (Scala and Python) is untouched;
`finalStorageLevel` stays local to ALS since it is ALS-specific.
### Why are the changes needed?
Removes a duplicated param definition. The shared trait was deliberately
modeled after ALS's existing param in SPARK-57860, so ALS itself should be its
first reuse besides the new adopters. One definition to maintain,
guaranteed-consistent validation and default.
### Does this PR introduce _any_ user-facing change?
No behavior change. Param name, default (`"MEMORY_AND_DISK"`), validation
(must be a valid `StorageLevel`, not `"NONE"`), getter, and setter are all
unchanged.
The only visible difference is the param's doc string, which now uses the
shared trait's wording:
- before: `StorageLevel for intermediate datasets. Cannot be 'NONE'.`
- after: `StorageLevel for intermediate datasets. Pass in a string
representation of `StorageLevel`. Cannot be 'NONE'.`
### How was this patch tested?
Existing `ALSStorageSuite` already covers the param end-to-end (default
level applied to intermediate RDDs, non-default level honored, `"NONE"`/invalid
values rejected) and passes unchanged (all 4 `ml.recommendation` suites,
38/38). `pyspark.ml.tests.test_param` (Scala/Python param parity via
`test_java_params`) passes locally.
MiMa reports two `FinalMethodProblem`s on `ALS.intermediateStorageLevel` /
`ALS.getIntermediateStorageLevel`, because the members are declared `final` in
the shared trait while ALS's local ones were not. Overriding a param `val` to
change its semantics is not a supported usage, and the param's name, type,
default and validation are unchanged, so this PR adds two narrowly-scoped
`MimaExcludes` entries with an explaining comment.
### Was this patch authored or co-authored using generative AI tooling?
Cooperate with: Claude Code (Fable 5)
Closes #56979 from maoli67660/SPARK-57910.
Authored-by: Mao Li <[email protected]>
Signed-off-by: Ruifeng Zheng <[email protected]>
---
.../org/apache/spark/ml/recommendation/ALS.scala | 17 +--------------
project/MimaExcludes.scala | 9 +++++++-
python/pyspark/ml/recommendation.py | 24 ++++++++--------------
3 files changed, 18 insertions(+), 32 deletions(-)
diff --git a/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala
b/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala
index 1cee915046c0..43fa2c0ed3d2 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala
@@ -139,7 +139,7 @@ private[recommendation] trait ALSModelParams extends Params
with HasPredictionCo
* Common params for ALS.
*/
private[recommendation] trait ALSParams extends ALSModelParams with HasMaxIter
with HasRegParam
- with HasCheckpointInterval with HasSeed {
+ with HasCheckpointInterval with HasSeed with HasIntermediateStorageLevel {
/**
* Param for rank of the matrix factorization (positive).
@@ -215,20 +215,6 @@ private[recommendation] trait ALSParams extends
ALSModelParams with HasMaxIter w
/** @group getParam */
def getNonnegative: Boolean = $(nonnegative)
- /**
- * Param for StorageLevel for intermediate datasets. Pass in a string
representation of
- * `StorageLevel`. Cannot be "NONE".
- * Default: "MEMORY_AND_DISK".
- *
- * @group expertParam
- */
- val intermediateStorageLevel = new Param[String](this,
"intermediateStorageLevel",
- "StorageLevel for intermediate datasets. Cannot be 'NONE'.",
- (s: String) => Try(StorageLevel.fromString(s)).isSuccess && s != "NONE")
-
- /** @group expertGetParam */
- def getIntermediateStorageLevel: String = $(intermediateStorageLevel)
-
/**
* Param for StorageLevel for ALS model factors. Pass in a string
representation of
* `StorageLevel`.
@@ -246,7 +232,6 @@ private[recommendation] trait ALSParams extends
ALSModelParams with HasMaxIter w
setDefault(rank -> 10, maxIter -> 10, regParam -> 0.1, numUserBlocks -> 10,
numItemBlocks -> 10,
implicitPrefs -> false, alpha -> 1.0, userCol -> "user", itemCol -> "item",
ratingCol -> "rating", nonnegative -> false, checkpointInterval -> 10,
- intermediateStorageLevel -> StorageLevelMapper.MEMORY_AND_DISK.name(),
finalStorageLevel -> StorageLevelMapper.MEMORY_AND_DISK.name(),
coldStartStrategy -> "nan")
/**
diff --git a/project/MimaExcludes.scala b/project/MimaExcludes.scala
index df2b51f3b732..442912429aa4 100644
--- a/project/MimaExcludes.scala
+++ b/project/MimaExcludes.scala
@@ -37,7 +37,14 @@ object MimaExcludes {
lazy val v50excludes: Seq[Problem => Boolean] = v43excludes
// Exclude rules for 4.3.x from 4.2.0 (add 4.3-specific filters below as
needed).
- lazy val v43excludes: Seq[Problem => Boolean] = v42excludes
+ lazy val v43excludes: Seq[Problem => Boolean] = v42excludes ++ Seq(
+ // [SPARK-57910] ALS reuses the shared HasIntermediateStorageLevel param;
the param and its
+ // getter are declared final in the shared trait (same name, type, default
and validation).
+ ProblemFilters.exclude[FinalMethodProblem](
+ "org.apache.spark.ml.recommendation.ALS.intermediateStorageLevel"),
+ ProblemFilters.exclude[FinalMethodProblem](
+ "org.apache.spark.ml.recommendation.ALS.getIntermediateStorageLevel")
+ )
// Exclude rules for 4.2.x from 4.1.0
lazy val v42excludes = v41excludes ++ Seq(
diff --git a/python/pyspark/ml/recommendation.py
b/python/pyspark/ml/recommendation.py
index 97a638ed6743..ee4a0c8d2e96 100644
--- a/python/pyspark/ml/recommendation.py
+++ b/python/pyspark/ml/recommendation.py
@@ -26,6 +26,7 @@ from pyspark.ml.param.shared import (
HasRegParam,
HasCheckpointInterval,
HasSeed,
+ HasIntermediateStorageLevel,
)
from pyspark.ml.wrapper import JavaEstimator, JavaModel
from pyspark.ml.common import inherit_doc
@@ -98,7 +99,14 @@ class _ALSModelParams(HasPredictionCol, HasBlockSize):
@inherit_doc
-class _ALSParams(_ALSModelParams, HasMaxIter, HasRegParam,
HasCheckpointInterval, HasSeed):
+class _ALSParams(
+ _ALSModelParams,
+ HasMaxIter,
+ HasRegParam,
+ HasCheckpointInterval,
+ HasSeed,
+ HasIntermediateStorageLevel,
+):
"""
Params for :py:class:`ALS`.
@@ -145,12 +153,6 @@ class _ALSParams(_ALSModelParams, HasMaxIter, HasRegParam,
HasCheckpointInterval
"whether to use nonnegative constraint for least squares",
typeConverter=TypeConverters.toBoolean,
)
- intermediateStorageLevel: Param[str] = Param(
- Params._dummy(),
- "intermediateStorageLevel",
- "StorageLevel for intermediate datasets. Cannot be 'NONE'.",
- typeConverter=TypeConverters.toString,
- )
finalStorageLevel: Param[str] = Param(
Params._dummy(),
"finalStorageLevel",
@@ -173,7 +175,6 @@ class _ALSParams(_ALSModelParams, HasMaxIter, HasRegParam,
HasCheckpointInterval
ratingCol="rating",
nonnegative=False,
checkpointInterval=10,
- intermediateStorageLevel="MEMORY_AND_DISK",
finalStorageLevel="MEMORY_AND_DISK",
coldStartStrategy="nan",
)
@@ -227,13 +228,6 @@ class _ALSParams(_ALSModelParams, HasMaxIter, HasRegParam,
HasCheckpointInterval
"""
return self.getOrDefault(self.nonnegative)
- @since("2.0.0")
- def getIntermediateStorageLevel(self) -> str:
- """
- Gets the value of intermediateStorageLevel or its default value.
- """
- return self.getOrDefault(self.intermediateStorageLevel)
-
@since("2.0.0")
def getFinalStorageLevel(self) -> str:
"""
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
