Source code for mmlspark.vw.VowpalWabbitFeaturizer

# Copyright (C) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License. See LICENSE in project root for information.


import sys
if sys.version >= '3':
    basestring = str

from pyspark.ml.param.shared import *
from pyspark import keyword_only
from pyspark.ml.util import JavaMLReadable, JavaMLWritable
from mmlspark.core.serialize.java_params_patch import *
from pyspark.ml.wrapper import JavaTransformer, JavaEstimator, JavaModel
from pyspark.ml.common import inherit_doc
from mmlspark.core.schema.Utils import *

[docs]@inherit_doc class VowpalWabbitFeaturizer(ComplexParamsMixin, JavaMLReadable, JavaMLWritable, JavaTransformer): """ Args: inputCols (list): The names of the input columns (default: [Ljava.lang.String;@5e8e5f44) numBits (int): Number of bits used to mask (default: 30) outputCol (str): The name of the output column (default: features) prefixStringsWithColumnName (bool): Prefix string features with column name (default: true) preserveOrderNumBits (int): Number of bits used to preserve the feature order. This will reduce the hash size. Needs to be large enough to fit count the maximum number of words (default: 0) seed (int): Hash seed (default: 0) stringSplitInputCols (list): Input cols that should be split at word boundaries (default: [Ljava.lang.String;@4d097470) sumCollisions (bool): Sums collisions if true, otherwise removes them (default: true) """ @keyword_only def __init__(self, inputCols=[], numBits=30, outputCol="features", prefixStringsWithColumnName=True, preserveOrderNumBits=0, seed=0, stringSplitInputCols=[], sumCollisions=True): super(VowpalWabbitFeaturizer, self).__init__() self._java_obj = self._new_java_obj("com.microsoft.ml.spark.vw.VowpalWabbitFeaturizer") self.inputCols = Param(self, "inputCols", "inputCols: The names of the input columns (default: [Ljava.lang.String;@5e8e5f44)") self._setDefault(inputCols=[]) self.numBits = Param(self, "numBits", "numBits: Number of bits used to mask (default: 30)") self._setDefault(numBits=30) self.outputCol = Param(self, "outputCol", "outputCol: The name of the output column (default: features)") self._setDefault(outputCol="features") self.prefixStringsWithColumnName = Param(self, "prefixStringsWithColumnName", "prefixStringsWithColumnName: Prefix string features with column name (default: true)") self._setDefault(prefixStringsWithColumnName=True) self.preserveOrderNumBits = Param(self, "preserveOrderNumBits", "preserveOrderNumBits: Number of bits used to preserve the feature order. This will reduce the hash size. Needs to be large enough to fit count the maximum number of words (default: 0)") self._setDefault(preserveOrderNumBits=0) self.seed = Param(self, "seed", "seed: Hash seed (default: 0)") self._setDefault(seed=0) self.stringSplitInputCols = Param(self, "stringSplitInputCols", "stringSplitInputCols: Input cols that should be split at word boundaries (default: [Ljava.lang.String;@4d097470)") self._setDefault(stringSplitInputCols=[]) self.sumCollisions = Param(self, "sumCollisions", "sumCollisions: Sums collisions if true, otherwise removes them (default: true)") self._setDefault(sumCollisions=True) if hasattr(self, "_input_kwargs"): kwargs = self._input_kwargs else: kwargs = self.__init__._input_kwargs self.setParams(**kwargs)
[docs] @keyword_only def setParams(self, inputCols=[], numBits=30, outputCol="features", prefixStringsWithColumnName=True, preserveOrderNumBits=0, seed=0, stringSplitInputCols=[], sumCollisions=True): """ Set the (keyword only) parameters Args: inputCols (list): The names of the input columns (default: [Ljava.lang.String;@5e8e5f44) numBits (int): Number of bits used to mask (default: 30) outputCol (str): The name of the output column (default: features) prefixStringsWithColumnName (bool): Prefix string features with column name (default: true) preserveOrderNumBits (int): Number of bits used to preserve the feature order. This will reduce the hash size. Needs to be large enough to fit count the maximum number of words (default: 0) seed (int): Hash seed (default: 0) stringSplitInputCols (list): Input cols that should be split at word boundaries (default: [Ljava.lang.String;@4d097470) sumCollisions (bool): Sums collisions if true, otherwise removes them (default: true) """ if hasattr(self, "_input_kwargs"): kwargs = self._input_kwargs else: kwargs = self.__init__._input_kwargs return self._set(**kwargs)
[docs] def setInputCols(self, value): """ Args: inputCols (list): The names of the input columns (default: [Ljava.lang.String;@5e8e5f44) """ self._set(inputCols=value) return self
[docs] def getInputCols(self): """ Returns: list: The names of the input columns (default: [Ljava.lang.String;@5e8e5f44) """ return self.getOrDefault(self.inputCols)
[docs] def setNumBits(self, value): """ Args: numBits (int): Number of bits used to mask (default: 30) """ self._set(numBits=value) return self
[docs] def getNumBits(self): """ Returns: int: Number of bits used to mask (default: 30) """ return self.getOrDefault(self.numBits)
[docs] def setOutputCol(self, value): """ Args: outputCol (str): The name of the output column (default: features) """ self._set(outputCol=value) return self
[docs] def getOutputCol(self): """ Returns: str: The name of the output column (default: features) """ return self.getOrDefault(self.outputCol)
[docs] def setPrefixStringsWithColumnName(self, value): """ Args: prefixStringsWithColumnName (bool): Prefix string features with column name (default: true) """ self._set(prefixStringsWithColumnName=value) return self
[docs] def getPrefixStringsWithColumnName(self): """ Returns: bool: Prefix string features with column name (default: true) """ return self.getOrDefault(self.prefixStringsWithColumnName)
[docs] def setPreserveOrderNumBits(self, value): """ Args: preserveOrderNumBits (int): Number of bits used to preserve the feature order. This will reduce the hash size. Needs to be large enough to fit count the maximum number of words (default: 0) """ self._set(preserveOrderNumBits=value) return self
[docs] def getPreserveOrderNumBits(self): """ Returns: int: Number of bits used to preserve the feature order. This will reduce the hash size. Needs to be large enough to fit count the maximum number of words (default: 0) """ return self.getOrDefault(self.preserveOrderNumBits)
[docs] def setSeed(self, value): """ Args: seed (int): Hash seed (default: 0) """ self._set(seed=value) return self
[docs] def getSeed(self): """ Returns: int: Hash seed (default: 0) """ return self.getOrDefault(self.seed)
[docs] def setStringSplitInputCols(self, value): """ Args: stringSplitInputCols (list): Input cols that should be split at word boundaries (default: [Ljava.lang.String;@4d097470) """ self._set(stringSplitInputCols=value) return self
[docs] def getStringSplitInputCols(self): """ Returns: list: Input cols that should be split at word boundaries (default: [Ljava.lang.String;@4d097470) """ return self.getOrDefault(self.stringSplitInputCols)
[docs] def setSumCollisions(self, value): """ Args: sumCollisions (bool): Sums collisions if true, otherwise removes them (default: true) """ self._set(sumCollisions=value) return self
[docs] def getSumCollisions(self): """ Returns: bool: Sums collisions if true, otherwise removes them (default: true) """ return self.getOrDefault(self.sumCollisions)
[docs] @classmethod def read(cls): """ Returns an MLReader instance for this class. """ return JavaMMLReader(cls)
[docs] @staticmethod def getJavaPackage(): """ Returns package name String. """ return "com.microsoft.ml.spark.vw.VowpalWabbitFeaturizer"
@staticmethod def _from_java(java_stage): module_name=VowpalWabbitFeaturizer.__module__ module_name=module_name.rsplit(".", 1)[0] + ".VowpalWabbitFeaturizer" return from_java(java_stage, module_name)