[
https://issues.apache.org/jira/browse/SPARK-21685?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16123738#comment-16123738
]
Ratan Rai Sur commented on SPARK-21685:
---------------------------------------
The python wrapper is generated so I've pasted it here so you don't have to
build it:
{code:java}
# Copyright (C) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License. See LICENSE in the 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 pyspark.ml.wrapper import JavaTransformer, JavaEstimator, JavaModel
from pyspark.ml.common import inherit_doc
from mmlspark.Utils import *
@inherit_doc
class _CNTKModel(ComplexParamsMixin, JavaMLReadable, JavaMLWritable,
JavaTransformer):
"""
The ``CNTKModel`` evaluates a pre-trained CNTK model in parallel. The
``CNTKModel`` takes a path to a model and automatically loads and
distributes the model to workers for parallel evaluation using CNTK's
java bindings.
The ``CNTKModel`` loads the pretrained model into the ``Function`` class
of CNTK. One can decide which node of the CNTK Function computation
graph to evaluate by passing in the name of the output node with the
output node parameter. Currently the ``CNTKModel`` supports single
input single output models.
The ``CNTKModel`` takes an input column which should be a column of
spark vectors and returns a column of spark vectors representing the
activations of the selected node. By default, the CNTK model defaults
to using the model's first input and first output node.
Args:
inputCol (str): The name of the input column (undefined)
inputNode (int): index of the input node (default: 0)
miniBatchSize (int): size of minibatches (default: 10)
model (object): Array of bytes containing the serialized CNTKModel
(undefined)
outputCol (str): The name of the output column (undefined)
outputNodeIndex (int): index of the output node (default: 0)
outputNodeName (str): name of the output node (undefined)
"""
@keyword_only
def __init__(self, inputCol=None, inputNode=0, miniBatchSize=10,
model=None, outputCol=None, outputNodeIndex=0, outputNodeName=None):
super(_CNTKModel, self).__init__()
self._java_obj = self._new_java_obj("com.microsoft.ml.spark.CNTKModel")
self.inputCol = Param(self, "inputCol", "inputCol: The name of the
input column (undefined)")
self.inputNode = Param(self, "inputNode", "inputNode: index of the
input node (default: 0)")
self._setDefault(inputNode=0)
self.miniBatchSize = Param(self, "miniBatchSize", "miniBatchSize: size
of minibatches (default: 10)")
self._setDefault(miniBatchSize=10)
self.model = Param(self, "model", "model: Array of bytes containing the
serialized CNTKModel (undefined)")
self.outputCol = Param(self, "outputCol", "outputCol: The name of the
output column (undefined)")
self.outputNodeIndex = Param(self, "outputNodeIndex", "outputNodeIndex:
index of the output node (default: 0)")
self._setDefault(outputNodeIndex=0)
self.outputNodeName = Param(self, "outputNodeName", "outputNodeName:
name of the output node (undefined)")
if hasattr(self, "_input_kwargs"):
kwargs = self._input_kwargs
else:
kwargs = self.__init__._input_kwargs
self.setParams(**kwargs)
@keyword_only
def setParams(self, inputCol=None, inputNode=0, miniBatchSize=10,
model=None, outputCol=None, outputNodeIndex=0, outputNodeName=None):
"""
Set the (keyword only) parameters
Args:
inputCol (str): The name of the input column (undefined)
inputNode (int): index of the input node (default: 0)
miniBatchSize (int): size of minibatches (default: 10)
model (object): Array of bytes containing the serialized CNTKModel
(undefined)
outputCol (str): The name of the output column (undefined)
outputNodeIndex (int): index of the output node (default: 0)
outputNodeName (str): name of the output node (undefined)
"""
if hasattr(self, "_input_kwargs"):
kwargs = self._input_kwargs
else:
kwargs = self.__init__._input_kwargs
return self._set(**kwargs)
def setInputCol(self, value):
"""
Args:
inputCol (str): The name of the input column (undefined)
"""
self._set(inputCol=value)
return self
def getInputCol(self):
"""
Returns:
str: The name of the input column (undefined)
"""
return self.getOrDefault(self.inputCol)
def setInputNode(self, value):
"""
Args:
inputNode (int): index of the input node (default: 0)
"""
self._set(inputNode=value)
return self
def getInputNode(self):
"""
Returns:
int: index of the input node (default: 0)
"""
return self.getOrDefault(self.inputNode)
def setMiniBatchSize(self, value):
"""
Args:
miniBatchSize (int): size of minibatches (default: 10)
"""
self._set(miniBatchSize=value)
return self
def getMiniBatchSize(self):
"""
Returns:
int: size of minibatches (default: 10)
"""
return self.getOrDefault(self.miniBatchSize)
def setModel(self, value):
"""
Args:
model (object): Array of bytes containing the serialized CNTKModel
(undefined)
"""
self._set(model=value)
return self
def getModel(self):
"""
Returns:
object: Array of bytes containing the serialized CNTKModel
(undefined)
"""
return self.getOrDefault(self.model)
def setOutputCol(self, value):
"""
Args:
outputCol (str): The name of the output column (undefined)
"""
self._set(outputCol=value)
return self
def getOutputCol(self):
"""
Returns:
str: The name of the output column (undefined)
"""
return self.getOrDefault(self.outputCol)
def setOutputNodeIndex(self, value):
"""
Args:
outputNodeIndex (int): index of the output node (default: 0)
"""
self._set(outputNodeIndex=value)
return self
def getOutputNodeIndex(self):
"""
Returns:
int: index of the output node (default: 0)
"""
return self.getOrDefault(self.outputNodeIndex)
def setOutputNodeName(self, value):
"""
Args:
outputNodeName (str): name of the output node (undefined)
"""
self._set(outputNodeName=value)
return self
def getOutputNodeName(self):
"""
Returns:
str: name of the output node (undefined)
"""
return self.getOrDefault(self.outputNodeName)
@classmethod
def read(cls):
""" Returns an MLReader instance for this class. """
return JavaMMLReader(cls)
@staticmethod
def getJavaPackage():
""" Returns package name String. """
return "com.microsoft.ml.spark.CNTKModel"
@staticmethod
def _from_java(java_stage):
module_name=_CNTKModel.__module__
module_name=module_name.rsplit(".", 1)[0] + ".CNTKModel"
return from_java(java_stage, module_name)
{code}
> Params isSet in scala Transformer triggered by _setDefault in pyspark
> ---------------------------------------------------------------------
>
> Key: SPARK-21685
> URL: https://issues.apache.org/jira/browse/SPARK-21685
> Project: Spark
> Issue Type: Bug
> Components: PySpark
> Affects Versions: 2.1.0
> Reporter: Ratan Rai Sur
>
> I'm trying to write a PySpark wrapper for a Transformer whose transform
> method includes the line
> {code:java}
> require(!(isSet(outputNodeName) && isSet(outputNodeIndex)), "Can't set both
> outputNodeName and outputNodeIndex")
> {code}
> This should only throw an exception when both of these parameters are
> explicitly set.
> In the PySpark wrapper for the Transformer, there is this line in ___init___
> {code:java}
> self._setDefault(outputNodeIndex=0)
> {code}
> Here is the line in the main python script showing how it is being configured
> {code:java}
> cntkModel =
> CNTKModel().setInputCol("images").setOutputCol("output").setModelLocation(spark,
> model.uri).setOutputNodeName("z")
> {code}
> As you can see, only setOutputNodeName is being explicitly set but the
> exception is still being thrown.
> If you need more context,
> https://github.com/RatanRSur/mmlspark/tree/default-cntkmodel-output is the
> branch with the code, the files I'm referring to here that are tracked are
> the following:
> src/cntk-model/src/main/scala/CNTKModel.scala
> notebooks/tests/301 - CIFAR10 CNTK CNN Evaluation.ipynb
> The pyspark wrapper code is autogenerated
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
