[SPARK-32232][ML][PySpark] Make sure ML has the same default solver values between Scala and Python

mllib/src/main/scala/org/apache/spark/ml/regression/FMRegressor.scala

@@ -112,6 +112,10 @@ private[ml] trait FactorizationMachinesParams extends PredictorParams
     "The solver algorithm for optimization. Supported options: " +
       s"${supportedSolvers.mkString(", ")}. (Default adamW)",
     ParamValidators.inArray[String](supportedSolvers))
+
+  setDefault(factorSize -> 8, fitIntercept -> true, fitLinear -> true, regParam -> 0.0,
+    miniBatchFraction -> 1.0, initStd -> 0.01, maxIter -> 100, stepSize -> 1.0, tol -> 1E-6,
+    solver -> AdamW)
 }
 
 private[ml] trait FactorizationMachines extends FactorizationMachinesParams {
@@ -308,7 +312,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setFactorSize(value: Int): this.type = set(factorSize, value)
-  setDefault(factorSize -> 8)
 
   /**
    * Set whether to fit intercept term.
@@ -318,7 +321,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setFitIntercept(value: Boolean): this.type = set(fitIntercept, value)
-  setDefault(fitIntercept -> true)
 
   /**
    * Set whether to fit linear term.
@@ -328,7 +330,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setFitLinear(value: Boolean): this.type = set(fitLinear, value)
-  setDefault(fitLinear -> true)
 
   /**
    * Set the L2 regularization parameter.
@@ -338,7 +339,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setRegParam(value: Double): this.type = set(regParam, value)
-  setDefault(regParam -> 0.0)
 
   /**
    * Set the mini-batch fraction parameter.
@@ -348,7 +348,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setMiniBatchFraction(value: Double): this.type = set(miniBatchFraction, value)
-  setDefault(miniBatchFraction -> 1.0)
 
   /**
    * Set the standard deviation of initial coefficients.
@@ -358,7 +357,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setInitStd(value: Double): this.type = set(initStd, value)
-  setDefault(initStd -> 0.01)
 
   /**
    * Set the maximum number of iterations.
@@ -368,7 +366,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setMaxIter(value: Int): this.type = set(maxIter, value)
-  setDefault(maxIter -> 100)
 
   /**
    * Set the initial step size for the first step (like learning rate).
@@ -378,7 +375,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setStepSize(value: Double): this.type = set(stepSize, value)
-  setDefault(stepSize -> 1.0)
 
   /**
    * Set the convergence tolerance of iterations.
@@ -388,7 +384,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setTol(value: Double): this.type = set(tol, value)
-  setDefault(tol -> 1E-6)
 
   /**
    * Set the solver algorithm used for optimization.
@@ -399,7 +394,6 @@ class FMRegressor @Since("3.0.0") (
    */
   @Since("3.0.0")
   def setSolver(value: String): this.type = set(solver, value)
-  setDefault(solver -> AdamW)
 
   /**
    * Set the random seed for weight initialization.

mllib/src/main/scala/org/apache/spark/ml/regression/GeneralizedLinearRegression.scala

@@ -181,6 +181,9 @@ private[regression] trait GeneralizedLinearRegressionBase extends PredictorParam
       s"${supportedSolvers.mkString(", ")}. (Default irls)",
     ParamValidators.inArray[String](supportedSolvers))
 
+  setDefault(family -> Gaussian.name, variancePower -> 0.0, maxIter -> 25, tol -> 1E-6,
+    regParam -> 0.0, solver -> IRLS)
+
   @Since("2.0.0")
   override def validateAndTransformSchema(
       schema: StructType,
@@ -257,7 +260,6 @@ class GeneralizedLinearRegression @Since("2.0.0") (@Since("2.0.0") override val
    */
   @Since("2.0.0")
   def setFamily(value: String): this.type = set(family, value)
-  setDefault(family -> Gaussian.name)
 
   /**
    * Sets the value of param [[variancePower]].
@@ -268,7 +270,6 @@ class GeneralizedLinearRegression @Since("2.0.0") (@Since("2.0.0") override val
    */
   @Since("2.2.0")
   def setVariancePower(value: Double): this.type = set(variancePower, value)
-  setDefault(variancePower -> 0.0)
 
   /**
    * Sets the value of param [[linkPower]].
@@ -305,7 +306,6 @@ class GeneralizedLinearRegression @Since("2.0.0") (@Since("2.0.0") override val
    */
   @Since("2.0.0")
   def setMaxIter(value: Int): this.type = set(maxIter, value)
-  setDefault(maxIter -> 25)
 
   /**
    * Sets the convergence tolerance of iterations.
@@ -316,7 +316,6 @@ class GeneralizedLinearRegression @Since("2.0.0") (@Since("2.0.0") override val
    */
   @Since("2.0.0")
   def setTol(value: Double): this.type = set(tol, value)
-  setDefault(tol -> 1E-6)
 
   /**
    * Sets the regularization parameter for L2 regularization.
@@ -332,7 +331,6 @@ class GeneralizedLinearRegression @Since("2.0.0") (@Since("2.0.0") override val
    */
   @Since("2.0.0")
   def setRegParam(value: Double): this.type = set(regParam, value)
-  setDefault(regParam -> 0.0)
 
   /**
    * Sets the value of param [[weightCol]].
@@ -364,7 +362,6 @@ class GeneralizedLinearRegression @Since("2.0.0") (@Since("2.0.0") override val
    */
   @Since("2.0.0")
   def setSolver(value: String): this.type = set(solver, value)
-  setDefault(solver -> IRLS)
 
   /**
    * Sets the link prediction (linear predictor) column name.

python/pyspark/ml/classification.py

@@ -2421,6 +2421,10 @@ class _MultilayerPerceptronParams(_ProbabilisticClassifierParams, HasSeed, HasMa
     initialWeights = Param(Params._dummy(), "initialWeights", "The initial weights of the model.",
                            typeConverter=TypeConverters.toVector)
 
+    def __init__(self):
+        super(_MultilayerPerceptronParams, self).__init__()
+        self._setDefault(maxIter=100, tol=1E-6, blockSize=128, stepSize=0.03, solver="l-bfgs")
+
     @since("1.6.0")
     def getLayers(self):
         """
@@ -2524,7 +2528,6 @@ def __init__(self, featuresCol="features", labelCol="label", predictionCol="pred
         super(MultilayerPerceptronClassifier, self).__init__()
         self._java_obj = self._new_java_obj(
             "org.apache.spark.ml.classification.MultilayerPerceptronClassifier", self.uid)
-        self._setDefault(maxIter=100, tol=1E-6, blockSize=128, stepSize=0.03, solver="l-bfgs")
         kwargs = self._input_kwargs
         self.setParams(**kwargs)
 
@@ -3120,9 +3123,6 @@ def __init__(self, featuresCol="features", labelCol="label", predictionCol="pred
         super(FMClassifier, self).__init__()
         self._java_obj = self._new_java_obj(
             "org.apache.spark.ml.classification.FMClassifier", self.uid)
-        self._setDefault(factorSize=8, fitIntercept=True, fitLinear=True, regParam=0.0,
-                         miniBatchFraction=1.0, initStd=0.01, maxIter=100, stepSize=1.0,
-                         tol=1e-6, solver="adamW")
         kwargs = self._input_kwargs
         self.setParams(**kwargs)
 

python/pyspark/ml/regression.py

@@ -1891,6 +1891,11 @@ class _GeneralizedLinearRegressionParams(_PredictorParams, HasFitIntercept, HasM
                       "or empty, we treat all instance offsets as 0.0",
                       typeConverter=TypeConverters.toString)
 
+    def __init__(self):
+        super(_GeneralizedLinearRegressionParams, self).__init__()
+        self._setDefault(family="gaussian", maxIter=25, tol=1e-6, regParam=0.0, solver="irls",
+                         variancePower=0.0, aggregationDepth=2)
+
     @since("2.0.0")
     def getFamily(self):
         """
@@ -2023,8 +2028,6 @@ def __init__(self, labelCol="label", featuresCol="features", predictionCol="pred
         super(GeneralizedLinearRegression, self).__init__()
         self._java_obj = self._new_java_obj(
             "org.apache.spark.ml.regression.GeneralizedLinearRegression", self.uid)
-        self._setDefault(family="gaussian", maxIter=25, tol=1e-6, regParam=0.0, solver="irls",
-                         variancePower=0.0, aggregationDepth=2)
         kwargs = self._input_kwargs
 
         self.setParams(**kwargs)
@@ -2398,6 +2401,12 @@ class _FactorizationMachinesParams(_PredictorParams, HasMaxIter, HasStepSize, Ha
     solver = Param(Params._dummy(), "solver", "The solver algorithm for optimization. Supported " +
                    "options: gd, adamW. (Default adamW)", typeConverter=TypeConverters.toString)
 
+    def __init__(self):
+        super(_FactorizationMachinesParams, self).__init__()
+        self._setDefault(factorSize=8, fitIntercept=True, fitLinear=True, regParam=0.0,
+                         miniBatchFraction=1.0, initStd=0.01, maxIter=100, stepSize=1.0,
+                         tol=1e-6, solver="adamW")
+
     @since("3.0.0")
     def getFactorSize(self):
         """
@@ -2489,9 +2498,6 @@ def __init__(self, featuresCol="features", labelCol="label", predictionCol="pred
         super(FMRegressor, self).__init__()
         self._java_obj = self._new_java_obj(
             "org.apache.spark.ml.regression.FMRegressor", self.uid)
-        self._setDefault(factorSize=8, fitIntercept=True, fitLinear=True, regParam=0.0,
-                         miniBatchFraction=1.0, initStd=0.01, maxIter=100, stepSize=1.0,
-                         tol=1e-6, solver="adamW")
         kwargs = self._input_kwargs
         self.setParams(**kwargs)
 

python/pyspark/ml/tests/test_persistence.py

@@ -21,19 +21,78 @@
 import unittest
 
 from pyspark.ml import Transformer
-from pyspark.ml.classification import DecisionTreeClassifier, LogisticRegression, OneVsRest, \
-    OneVsRestModel
+from pyspark.ml.classification import DecisionTreeClassifier, FMClassifier, \
+    FMClassificationModel, LogisticRegression, MultilayerPerceptronClassifier, \
+    MultilayerPerceptronClassificationModel, OneVsRest, OneVsRestModel
 from pyspark.ml.clustering import KMeans
 from pyspark.ml.feature import Binarizer, HashingTF, PCA
 from pyspark.ml.linalg import Vectors
 from pyspark.ml.param import Params
 from pyspark.ml.pipeline import Pipeline, PipelineModel
-from pyspark.ml.regression import DecisionTreeRegressor, LinearRegression
+from pyspark.ml.regression import DecisionTreeRegressor, GeneralizedLinearRegression, \
+    GeneralizedLinearRegressionModel, \
+    LinearRegression
 from pyspark.ml.util import DefaultParamsReadable, DefaultParamsWriter
 from pyspark.ml.wrapper import JavaParams
 from pyspark.testing.mlutils import MockUnaryTransformer, SparkSessionTestCase
 
 
+class TestDefaultSolver(SparkSessionTestCase):
+
+    def test_multilayer_load(self):
+        df = self.spark.createDataFrame([(0.0, Vectors.dense([0.0, 0.0])),
+                                         (1.0, Vectors.dense([0.0, 1.0])),
+                                         (1.0, Vectors.dense([1.0, 0.0])),
+                                         (0.0, Vectors.dense([1.0, 1.0]))],
+                                        ["label",  "features"])
+
+        mlp = MultilayerPerceptronClassifier(layers=[2, 2, 2], seed=123)
+        model = mlp.fit(df)
+        self.assertEqual(model.getSolver(), "l-bfgs")
+        transformed1 = model.transform(df)
+        path = tempfile.mkdtemp()
+        model_path = path + "/mlp"
+        model.save(model_path)
+        model2 = MultilayerPerceptronClassificationModel.load(model_path)
+        self.assertEqual(model2.getSolver(), "l-bfgs")
+        transformed2 = model2.transform(df)
+        self.assertEqual(transformed1.take(4), transformed2.take(4))
+
+    def test_fm_load(self):
+        df = self.spark.createDataFrame([(1.0, Vectors.dense(1.0)),
+                                         (0.0, Vectors.sparse(1, [], []))],
+                                        ["label",  "features"])
+        fm = FMClassifier(factorSize=2, maxIter=50, stepSize=2.0)
+        model = fm.fit(df)
+        self.assertEqual(model.getSolver(), "adamW")
+        transformed1 = model.transform(df)
+        path = tempfile.mkdtemp()
+        model_path = path + "/fm"
+        model.save(model_path)
+        model2 = FMClassificationModel.load(model_path)
+        self.assertEqual(model2.getSolver(), "adamW")
+        transformed2 = model2.transform(df)
+        self.assertEqual(transformed1.take(2), transformed2.take(2))
+
+    def test_glr_load(self):
+        df = self.spark.createDataFrame([(1.0, Vectors.dense(0.0, 0.0)),
+                                         (1.0, Vectors.dense(1.0, 2.0)),
+                                         (2.0, Vectors.dense(0.0, 0.0)),
+                                         (2.0, Vectors.dense(1.0, 1.0))],
+                                        ["label",  "features"])
+        glr = GeneralizedLinearRegression(family="gaussian", link="identity", linkPredictionCol="p")
+        model = glr.fit(df)
+        self.assertEqual(model.getSolver(), "irls")
+        transformed1 = model.transform(df)
+        path = tempfile.mkdtemp()
+        model_path = path + "/glr"
+        model.save(model_path)
+        model2 = GeneralizedLinearRegressionModel.load(model_path)
+        self.assertEqual(model2.getSolver(), "irls")
+        transformed2 = model2.transform(df)
+        self.assertEqual(transformed1.take(4), transformed2.take(4))
+
+
 class PersistenceTest(SparkSessionTestCase):
 
     def test_linear_regression(self):