class H2OGridSearchParams(H2OGridSearchRandomDiscreteCriteriaParams,
H2OGridSearchCartesianCriteriaParams,
H2OGridSearchCommonCriteriaParams):
##
# Param definitions
##
algo = Param(Params._dummy(), "algo", "Algo to run grid search on",
H2OTypeConverters.toH2OGridSearchSupportedAlgo())
hyperParameters = Param(Params._dummy(), "hyperParameters",
"Grid Search Hyper Params map",
H2OTypeConverters.toDictionaryWithAnyElements())
selectBestModelBy = Param(
Params._dummy(), "selectBestModelBy",
"Specifies the metric which is used for comparing and sorting the models returned by the grid.",
H2OTypeConverters.toEnumString(
"ai.h2o.sparkling.ml.internals.H2OMetric"))
parallelism = Param(
Params._dummy(), "parallelism",
"""Level of model-building parallelism, the possible values are:
0 -> H2O selects parallelism level based on cluster configuration, such as number of cores
1 -> Sequential model building, no parallelism
n>1 -> n models will be built in parallel if possible""",
H2OTypeConverters.toInt())
##
# Getters
##
def getAlgo(self):
javaAlgo = self._java_obj.getAlgo()
if javaAlgo is None:
return None
algoName = javaAlgo.getClass().getSimpleName()
if algoName.endswith("Classifier"):
import ai.h2o.sparkling.ml.algos.classification
algo = getattr(ai.h2o.sparkling.ml.algos.classification,
algoName)()
elif algoName.endswith("Regressor"):
import ai.h2o.sparkling.ml.algos.regression
algo = getattr(ai.h2o.sparkling.ml.algos.regression, algoName)()
else:
import ai.h2o.sparkling.ml.algos
algo = getattr(ai.h2o.sparkling.ml.algos, algoName)()
algo._resetUid(javaAlgo.uid())
algo._java_obj = javaAlgo
algo._transfer_params_from_java()
return algo
def getHyperParameters(self):
return self.getOrDefault(self.hyperParameters)
def getSelectBestModelBy(self):
return self.getOrDefault(self.selectBestModelBy)
def getParallelism(self):
return self.getOrDefault(self.parallelism)
##
# Setters
##
def setAlgo(self, value):
self._set(algo=value)
self._transfer_params_to_java()
return self
def setHyperParameters(self, value):
return self._set(hyperParameters=value)
def setSelectBestModelBy(self, value):
return self._set(selectBestModelBy=value)
def setParallelism(self, value):
return self._set(parallelism=value)
class H2OGridSearchParams(H2OCommonSupervisedParams):
##
# Param definitions
##
algo = Param(Params._dummy(), "algo", "Algo to run grid search on",
H2OTypeConverters.toH2OGridSearchSupportedAlgo())
hyperParameters = Param(Params._dummy(), "hyperParameters",
"Grid Search Hyper Params map",
H2OTypeConverters.toDictionaryWithAnyElements())
strategy = Param(
Params._dummy(), "strategy", "strategy",
H2OTypeConverters.toEnumString(
"hex.grid.HyperSpaceSearchCriteria$Strategy"))
maxRuntimeSecs = Param(Params._dummy(), "maxRuntimeSecs", "maxRuntimeSecs",
H2OTypeConverters.toFloat())
maxModels = Param(Params._dummy(), "maxModels", "maxModels",
H2OTypeConverters.toInt())
stoppingRounds = Param(Params._dummy(), "stoppingRounds", "stoppingRounds",
H2OTypeConverters.toInt())
stoppingTolerance = Param(Params._dummy(), "stoppingTolerance",
"stoppingTolerance", H2OTypeConverters.toFloat())
stoppingMetric = Param(
Params._dummy(), "stoppingMetric", "stoppingMetric",
H2OTypeConverters.toEnumString("hex.ScoreKeeper$StoppingMetric"))
selectBestModelBy = Param(
Params._dummy(), "selectBestModelBy",
"Specifies the metric which is used for comparing and sorting the models returned by the grid.",
H2OTypeConverters.toEnumString(
"ai.h2o.sparkling.ml.internals.H2OMetric"))
parallelism = Param(
Params._dummy(), "parallelism",
"""Level of model-building parallelism, the possible values are:
0 -> H2O selects parallelism level based on cluster configuration, such as number of cores
1 -> Sequential model building, no parallelism
n>1 -> n models will be built in parallel if possible""",
H2OTypeConverters.toInt())
##
# Getters
##
def getAlgo(self):
javaAlgo = self.getOrDefault(self.algo)
algoName = javaAlgo.parameters().algoName()
if algoName == "GBM":
from ai.h2o.sparkling.ml.algos import H2OGBM
algo = H2OGBM()
elif algoName == "DeepLearning":
from ai.h2o.sparkling.ml.algos import H2ODeepLearning
algo = H2ODeepLearning()
elif algoName == "XGBoost":
from ai.h2o.sparkling.ml.algos import H2OXGBoost
algo = H2OXGBoost()
elif algoName == "GLM":
from ai.h2o.sparkling.ml.algos import H2OGLM
algo = H2OGLM()
elif algoName == "DRF":
from ai.h2o.sparkling.ml.algos import H2ODRF
algo = H2ODRF()
else:
raise ValueError('Unsupported algorithm for H2OGridSearch')
algo._resetUid(javaAlgo.uid())
algo._java_obj = javaAlgo
algo._transfer_params_from_java()
return algo
def getHyperParameters(self):
return self.getOrDefault(self.hyperParameters)
def getStrategy(self):
return self.getOrDefault(self.strategy)
def getMaxRuntimeSecs(self):
return self.getOrDefault(self.maxRuntimeSecs)
def getMaxModels(self):
return self.getOrDefault(self.maxModels)
def getStoppingRounds(self):
return self.getOrDefault(self.stoppingRounds)
def getStoppingTolerance(self):
return self.getOrDefault(self.stoppingTolerance)
def getStoppingMetric(self):
return self.getOrDefault(self.stoppingMetric)
def getSelectBestModelBy(self):
return self.getOrDefault(self.selectBestModelBy)
def getParallelism(self):
return self.getOrDefault(self.parallelism)
##
# Setters
##
def setAlgo(self, value):
return self._set(algo=value)
def setHyperParameters(self, value):
return self._set(hyperParameters=value)
def setStrategy(self, value):
return self._set(strategy=value)
def setMaxRuntimeSecs(self, value):
return self._set(maxRuntimeSecs=value)
def setMaxModels(self, value):
return self._set(maxModels=value)
def setStoppingRounds(self, value):
return self._set(stoppingRounds=value)
def setStoppingTolerance(self, value):
return self._set(stoppingTolerance=value)
def setStoppingMetric(self, value):
return self._set(stoppingMetric=value)
def setSelectBestModelBy(self, value):
return self._set(selectBestModelBy=value)
def setParallelism(self, value):
return self._set(parallelism=value)
class H2OGridSearchParams(H2OGridSearchRandomDiscreteCriteriaParams,
H2OGridSearchCartesianCriteriaParams,
H2OGridSearchCommonCriteriaParams):
##
# Param definitions
##
algo = Param(Params._dummy(), "algo", "Algo to run grid search on",
H2OTypeConverters.toH2OGridSearchSupportedAlgo())
hyperParameters = Param(Params._dummy(), "hyperParameters",
"Grid Search Hyper Params map",
H2OTypeConverters.toDictionaryWithAnyElements())
selectBestModelBy = Param(
Params._dummy(), "selectBestModelBy",
"Specifies the metric which is used for comparing and sorting the models returned by the grid.",
H2OTypeConverters.toEnumString(
"ai.h2o.sparkling.ml.internals.H2OMetric"))
parallelism = Param(
Params._dummy(), "parallelism",
"""Level of model-building parallelism, the possible values are:
0 -> H2O selects parallelism level based on cluster configuration, such as number of cores
1 -> Sequential model building, no parallelism
n>1 -> n models will be built in parallel if possible""",
H2OTypeConverters.toInt())
##
# Getters
##
def getAlgo(self):
javaAlgo = self._java_obj.getAlgo()
if javaAlgo is None:
return None
algoName = javaAlgo.parameters().algoName()
if algoName == "GBM":
from ai.h2o.sparkling.ml.algos import H2OGBM
algo = H2OGBM()
elif algoName == "DeepLearning":
from ai.h2o.sparkling.ml.algos import H2ODeepLearning
algo = H2ODeepLearning()
elif algoName == "XGBoost":
from ai.h2o.sparkling.ml.algos import H2OXGBoost
algo = H2OXGBoost()
elif algoName == "GLM":
from ai.h2o.sparkling.ml.algos import H2OGLM
algo = H2OGLM()
elif algoName == "DRF":
from ai.h2o.sparkling.ml.algos import H2ODRF
algo = H2ODRF()
elif algoName == "KMeans":
from ai.h2o.sparkling.ml.algos import H2OKMeans
algo = H2OKMeans()
else:
raise ValueError('Unsupported algorithm for H2OGridSearch')
algo._resetUid(javaAlgo.uid())
algo._java_obj = javaAlgo
algo._transfer_params_from_java()
return algo
def getHyperParameters(self):
return self.getOrDefault(self.hyperParameters)
def getSelectBestModelBy(self):
return self.getOrDefault(self.selectBestModelBy)
def getParallelism(self):
return self.getOrDefault(self.parallelism)
##
# Setters
##
def setAlgo(self, value):
self._set(algo=value)
self._transfer_params_to_java()
return self
def setHyperParameters(self, value):
return self._set(hyperParameters=value)
def setSelectBestModelBy(self, value):
return self._set(selectBestModelBy=value)
def setParallelism(self, value):
return self._set(parallelism=value)