-
Notifications
You must be signed in to change notification settings - Fork 819
/
PySparkWrapperTest.scala
341 lines (306 loc) · 15.8 KB
/
PySparkWrapperTest.scala
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
// Copyright (C) Microsoft Corporation. All rights reserved.
// Licensed under the MIT License. See LICENSE in project root for information.
package com.microsoft.ml.spark.codegen
import java.io.File
import com.microsoft.ml.spark.core.env.FileUtilities._
import com.microsoft.ml.spark.core.serialize.ComplexParam
import Config._
import org.apache.commons.lang3.StringUtils
import org.apache.spark.ml.evaluation.Evaluator
import org.apache.spark.ml.param.{Param, Params, ServiceParam}
import org.apache.spark.ml.{Estimator, PipelineStage, Transformer}
/** :: DeveloperApi ::
* Abstraction for PySpark wrapper generators.
*/
abstract class PySparkWrapperParamsTest(entryPoint: Params,
entryPointName: String,
entryPointQualifiedName: String) extends WritableWrapper {
// general classes are imported from the mmlspark directy;
// internal classes have to be imported from their packages
private def importClass(entryPointName:String):String = {
val packageString = if (subPackages.isEmpty) {
"mmlspark"
}else{
"mmlspark." + subPackages.mkString(".")
}
if (entryPointName startsWith internalPrefix) {
s"from $packageString.$entryPointName import $entryPointName"
} else {
s"from $packageString import $entryPointName"
}
}
protected def classTemplate(classParams: String, paramGettersAndSetters: String) =
s"""|import unittest
|import pandas as pd
|import numpy as np
|import pyspark.ml, pyspark.ml.feature
|from pyspark import SparkContext
|from pyspark.sql import SQLContext, SparkSession
|from pyspark.ml.classification import LogisticRegression
|from pyspark.ml.regression import LinearRegression
|${importClass(entryPointName)}
|from pyspark.ml.feature import Tokenizer
|from mmlspark.train import TrainClassifier
|from mmlspark.featurize import ValueIndexer
|
|spark = SparkSession.builder \\
| .master("local[*]") \\
| .appName("$entryPointName") \\
| .config("spark.jars.packages", "com.microsoft.ml.spark:${packageName}_2.11:$version") \\
| .config("spark.executor.heartbeatInterval", "60s") \\
| .getOrCreate()
|
|sc = spark.sparkContext
|
|class ${entryPointName}Test(unittest.TestCase):
| def test_placeholder(self):
| True
|
|$paramGettersAndSetters
|
|""".stripMargin
protected val unittestString =
s"""|
|import os, xmlrunner
|if __name__ == "__main__":
| result = unittest.main(testRunner=xmlrunner.XMLTestRunner(output=os.getenv("TEST_RESULTS","TestResults")),
| failfast=False, buffer=False, catchbreak=False)
|""".stripMargin
protected def setAndGetTemplate(paramName: String, value: String) =
s"""| def test_set$paramName(self):
| my$entryPointName = $entryPointName()
| val = $value
| my$entryPointName.set$paramName(val)
| retVal = my$entryPointName.get$paramName()
| self.assertEqual(val, retVal)
|""".stripMargin
protected def tryFitSetupTemplate(entryPointName: String) =
s"""| def test_$entryPointName(self):
| dog = "dog"
| cat = "cat"
| bird = "bird"
| tmp1 = {
| "col1": [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1],
| "col2": [2, 3, 4, 5, 1, 3, 3, 4, 0, 2, 3, 4],
| "col3": [0.50, 0.40, 0.78, 0.12, 0.50, 0.40, 0.78, 0.12, 0.50, 0.40, 0.78, 0.12],
| "col4": [0.60, 0.50, 0.99, 0.34, 0.60, 0.50, 0.99, 0.34, 0.60, 0.50, 0.99, 0.34],
| "col5": [dog, cat, dog, cat, dog, bird, dog, cat, dog, bird, dog, cat],
| "col6": [cat, dog, bird, dog, bird, dog, cat, dog, cat, dog, bird, dog],
| "image": [cat, dog, bird, dog, bird, dog, cat, dog, cat, dog, bird, dog]
| }
| sqlC = SQLContext(sc)
| pddf = pd.DataFrame(tmp1)
| pddf["col1"] = pddf["col1"].astype(np.float64)
| pddf["col2"] = pddf["col2"].astype(np.int32)
| data = sqlC.createDataFrame(pddf)
|""".stripMargin
protected def tryTransformTemplate(entryPointName: String, param: String) =
s"""| my$entryPointName = $entryPointName($param)
| prediction = my$entryPointName.transform(data)
| self.assertNotEqual(prediction, None)
|""".stripMargin
protected def tryFitTemplate(entryPointName: String, model: String) =
s"""| my$entryPointName = $entryPointName(model=$model, labelCol="col1", numFeatures=5)
| model = my$entryPointName.fit(data)
| self.assertNotEqual(model, None)""".stripMargin
protected def tryMultiColumnFitTemplate(entryPointName: String, model: String) =
s"""| my$entryPointName = $entryPointName(baseStage=$model, inputCols=["col1"], outputCols=["out"])
| model = my$entryPointName.fit(data)
| self.assertNotEqual(model, None)""".stripMargin
private def evaluateSetupTemplate(entryPointName: String) =
s"""| def test_$entryPointName(self):
| data = {
| "labelColumn": [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1],
| "col1": [2, 3, 4, 5, 1, 3, 3, 4, 0, 2, 3, 4],
| "col2": [0.50, 0.40, 0.78, 0.12, 0.50, 0.40, 0.78, 0.12, 0.50, 0.40, 0.78, 0.12],
| "col3": [0.60, 0.50, 0.99, 0.34, 0.60, 0.50, 0.99, 0.34, 0.60, 0.50, 0.99, 0.34],
| "col4": [0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3]
| }
| sqlC = SQLContext(sc)
| pddf = pd.DataFrame(data)
| data = sqlC.createDataFrame(pddf)
| model = TrainClassifier(model=LogisticRegression(), labelCol="labelColumn",
| numFeatures=256).fit(data)
|""".stripMargin
protected def computeStatisticsTemplate(entryPointName: String) =
s"""|${evaluateSetupTemplate(entryPointName)}
| scoredData = model.transform(data)
| scoredData.limit(10).toPandas()
| evaluatedData = $entryPointName().transform(scoredData)
| self.assertNotEqual(evaluatedData, None)
|""".stripMargin
protected def valueIndexerModelTemplate(entryPointName: String) =
s"""|${tryFitSetupTemplate(entryPointName)}
| valueData = $entryPointName(inputCol="col5", outputCol="catOutput",
| dataType="string", levels=["dog", "cat", "bird"]).transform(data)
| self.assertNotEqual(valueData, None)
|""".stripMargin
protected def indexToValueTemplate(entryPointName: String) =
s"""|${tryFitSetupTemplate(entryPointName)}
| indexModel = ValueIndexer(inputCol="col5", outputCol="catOutput").fit(data)
| indexedData = indexModel.transform(data)
| valueData = $entryPointName(inputCol="catOutput", outputCol="origDomain").transform(indexedData)
| self.assertNotEqual(valueData, None)
|""".stripMargin
protected def evaluateTemplate(entryPointName: String) =
s"""|${evaluateSetupTemplate(entryPointName)}
| model = TrainClassifier(model=LogisticRegression(), labelCol="labelColumn",
| numFeatures=256).fit(data)
| evaluateModels = FindBestModel(models=[model, model]).fit(data)
| bestModel = evaluateModels.transform(data)
| self.assertNotEqual(bestModel, None)
|""".stripMargin
// These com.microsoft.ml.spark.core.serialize.params are need custom handling.
// For now, just skip them so we have tests that pass.
private lazy val skippedParams = Set[String]("models", "model", "cntkModel", "stage")
protected def isSkippedParam(paramName: String): Boolean = skippedParams.contains(paramName)
protected def isModel(paramName: String): Boolean = paramName.toLowerCase() == "model"
protected def isBaseTransformer(paramName: String): Boolean = paramName.toLowerCase() == "basetransformer"
protected def tryFitString(entryPointName: String): String =
if (entryPointName.contains("Regressor") && !entryPointName.contains("LightGBM"))
tryFitTemplate(entryPointName, "LinearRegression(solver=\"l-bfgs\")")
else if (entryPointName.contains("Classifier") && !entryPointName.contains("LightGBM"))
tryFitTemplate(entryPointName, "LogisticRegression()")
else if (entryPointName.contains("MultiColumnAdapter"))
tryMultiColumnFitTemplate(entryPointName, "ValueIndexer()")
else ""
protected def computeStatisticsString(entryPointName: String): String = computeStatisticsTemplate(entryPointName)
protected def evaluateString(entryPointName: String): String = evaluateTemplate(entryPointName)
protected def indexToValueString(entryPointName: String): String = indexToValueTemplate(entryPointName)
protected def valueIndexerModelString(entryPointName: String): String = valueIndexerModelTemplate(entryPointName)
protected def tryTransformString(entryPointName: String): String = {
val param: String =
entryPointName match {
case "_CNTKModel" | "MultiTokenizer" | "NltTokenizeTransform" | "TextTransform"
| "TextNormalizerTransform" | "WordTokenizeTransform" => "inputCol=\"col5\""
case "DataConversion" => "cols=[\"col1\"], convertTo=\"double\""
case "DropColumns" => "cols=[\"col1\"]"
case "EnsembleByKey" => "keys=[\"col1\"], cols=[\"col3\"]"
case "FastVectorAssembler" => "inputCols=\"col1\""
case "IndexToValue" => "inputCol=\"catOutput\""
case "MultiNGram" => "inputColumns=np.array([ \"col5\", \"col6\" ])"
case "RenameColumn" => "inputCol=\"col5\", outputCol=\"catOutput1\""
case "Repartition" => "n=2"
case "SelectColumns" => "cols=[\"col1\"]"
case "TextPreprocessor" => "inputCol=\"col5\", outputCol=\"catOutput1\", normFunc=\"identity\""
case "ValueIndexerModel" => "inputCol=\"col5\", outputCol=\"catOutput\", " +
"dataType=\"string\", levels=[\"dog\", \"cat\", \"bird\"]"
case "WriteBlob" => "blobPath=\"file:///tmp/" + java.util.UUID.randomUUID + ".tsv\""
case _ => ""
}
tryTransformTemplate(entryPointName, param)
}
protected def getPythonizedDefault(paramDefault: String, paramType: String,
defaultStringIsParsable: Boolean): String =
paramType match {
case "BooleanParam" =>
StringUtils.capitalize(paramDefault)
case "DoubleParam" | "FloatParam" | "IntParam" | "LongParam" =>
paramDefault
case x if x == "Param" || defaultStringIsParsable =>
"\"" + paramDefault + "\""
case _ =>
"None"
}
protected def getParamDefault(param: Param[_]): (String, String) = {
if (!entryPoint.hasDefault(param)) ("None", null)
else {
val paramParent: String = param.parent
val paramDefault = entryPoint.getDefault(param).get.toString
if (paramDefault.toLowerCase.contains(paramParent.toLowerCase))
("None",
paramDefault.substring(paramDefault.lastIndexOf(paramParent) + paramParent.length))
else {
val defaultStringIsParsable: Boolean =
try {
entryPoint.getParam(param.name).w(paramDefault)
true
} catch {
case e: Exception => false
}
(getPythonizedDefault(paramDefault, param.getClass.getSimpleName, defaultStringIsParsable),
null)
}
}
}
protected def getPysparkWrapperTestBase: String = {
// Iterate over the com.microsoft.ml.spark.core.serialize.params to build strings
val paramGettersAndSettersString =
entryPoint.params.filter { param => !isSkippedParam(param.name)
}.flatMap { param =>
val value = if (isModel(param.name)) "LogisticRegression()"
else if (isBaseTransformer(param.name)) "Tokenizer()"
else getParamDefault(param)._1
param match {
case p: ServiceParam[_] => None
case p: ComplexParam[_] => None
case _ => Some(setAndGetTemplate(StringUtils.capitalize(param.name), value))
}
}.mkString("\n")
val classParamsString =
entryPoint.params.map(param => param.name + "=" + getParamDefault(param)._1).mkString(", ")
classTemplate(classParamsString, paramGettersAndSettersString)
}
def pysparkWrapperTestBuilder(): String = {
copyrightLines + getPysparkWrapperTestBase
}
private val subPackages = entryPointQualifiedName
.replace("com.microsoft.ml.spark.","")
.split(".".toCharArray.head).dropRight(1)
def writeWrapperToFile(dir: File): Unit = {
val packageDir = subPackages.foldLeft(dir){ case (base, folder) => new File(base, folder)}
packageDir.mkdirs()
new File(packageDir, "__init__.py").createNewFile()
writeFile(new File(packageDir,"test_" + entryPointName + ".py"), pysparkWrapperTestBuilder())
}
}
abstract class PySparkWrapperTest(entryPoint: PipelineStage,
entryPointName: String,
entryPointQualifiedName: String)
extends PySparkWrapperParamsTest(entryPoint, entryPointName, entryPointQualifiedName)
class PySparkEvaluatorTestWrapper(entryPoint: Evaluator,
entryPointName: String,
entryPointQualifiedName: String)
extends PySparkWrapperParamsTest(entryPoint, entryPointName, entryPointQualifiedName)
class PySparkTransformerWrapperTest(entryPoint: Transformer,
entryPointName: String,
entryPointQualifiedName: String)
extends PySparkWrapperTest(entryPoint,
entryPointName,
entryPointQualifiedName) {
// The transformer tests for FastVectorAssembler ... UnrollImage are disabled for the moment.
override def pysparkWrapperTestBuilder(): String = {
val transformTest =
entryPointName match {
case "ComputeModelStatistics" => computeStatisticsString(entryPointName)
case "ComputePerInstanceStatistics" => computeStatisticsString(entryPointName)
case "IndexToValue" => indexToValueString(entryPointName)
case "ValueIndexerModel" => valueIndexerModelString(entryPointName)
case "CheckpointData" | "DataConversion" | "EnsembleByKey" |
"DynamicMiniBatchTransformer" | "FixedMiniBatchTransformer" |
"PartitionConsolidator" | "TimeIntervalMiniBatchTransformer" |
"PartitionSample" | "Cacher" | "DropColumns" | "RenameColumn" |
"Repartition" | "SelectColumns" | "TextPreprocessor" |
"SummarizeData" =>
tryFitSetupTemplate(entryPointName) + tryTransformString(entryPointName)
case _ => ""
}
super.pysparkWrapperTestBuilder + transformTest + unittestString
}
}
class PySparkEstimatorWrapperTest(entryPoint: Estimator[_],
entryPointName: String,
entryPointQualifiedName: String,
companionModelName: String,
companionModelQualifiedName: String)
extends PySparkWrapperTest(entryPoint, entryPointName, entryPointQualifiedName) {
private val modelName = entryPointName + "Model"
override def pysparkWrapperTestBuilder(): String = {
val testString =
if (entryPointName == "FindBestModel")
evaluateString(entryPointName)
else
tryFitSetupTemplate(entryPointName) + tryFitString(entryPointName)
super.pysparkWrapperTestBuilder + testString + unittestString
}
}