Extend Restart to Adaptive samplers

developer_tools/XSDSchemas/Samplers.xsd

@@ -398,6 +398,7 @@
         <xsd:element name="initialSeed"      type="xsd:integer"                minOccurs="0" maxOccurs="1"/>
         <xsd:element name="variable"         type="variableType"             minOccurs="1" maxOccurs="unbounded"/>
         <xsd:element name="constant" type="constantVarType"        minOccurs="0" maxOccurs='unbounded'/>
+        <xsd:element name="Restart"      type="AssemblerObjectType"   minOccurs="0"  maxOccurs="1"/>
       </xsd:sequence>
     <xsd:attribute name="name"      type="xsd:string" use="required"/>
     <xsd:attribute name="verbosity" type="verbosityAttr"/>

framework/JobHandler.py

@@ -377,6 +377,18 @@ def addClientJob(self, args, functionToRun, identifier, metadata=None, uniqueHan
                 forceUseThreads = True, uniqueHandler = uniqueHandler,
                 clientQueue = True)
 
+  def addFinishedJob(self, data):
+    """
+      Takes an already-finished job (for example, a restart realization) and adds it to the finished queue.
+      @ In, data, dict, completed realization
+      @ Out, None
+    """
+    # create a placeholder runner
+    run = Runners.PassthroughRunner(self.messageHandler, data)
+    # place it on the finished queue
+    with self.__queueLock:
+      self.__finished.append(run)
+
   def isFinished(self):
     """
       Method to check if all the runs in the queue are finished

framework/Runners/PassthroughRunner.py

@@ -0,0 +1,86 @@
+# Copyright 2017 Battelle Energy Alliance, LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+  Module for Passthrough Runner class, which skips evaluation. Used particularly
+  for restarting Samplers from existing data currently.
+"""
+import numpy as np
+from .Runner import Runner
+
+class PassthroughRunner(Runner):
+  """
+    A runner for when we already have the answer, but need to go through the mechanics.
+  """
+  def __init__(self, messageHandler, data, metadata=None, uniqueHandler="any", profile=False):
+    """
+      Init method
+      @ In, messageHandler, MessageHandler object, the global RAVEN message
+        handler object
+      @ In, data, dict, fully-evaluated realization
+      @ In, metadata, dict, optional, dictionary of metadata associated with
+        this run
+      @ In, uniqueHandler, string, optional, it is a special keyword attached to
+        this runner. For example, if present, to retrieve this runner using the
+        method jobHandler.getFinished, the uniqueHandler needs to be provided.
+        If uniqueHandler == 'any', every "client" can get this runner
+      @ In, profile, bool, optional, if True then at deconstruction timing statements will be printed
+      @ Out, None
+    """
+    super(PassthroughRunner, self).__init__(messageHandler, metadata=metadata, uniqueHandler=uniqueHandler, profile=profile)
+    self._data = data   # realization with completed data
+    self.returnCode = 0 # passthrough was born successful
+
+  def isDone(self):
+    """
+      Method to check if the calculation associated with this Runner is finished
+      @ In, None
+      @ Out, isDone, bool, is it finished?
+    """
+    return True # passthrough was born done
+
+  def getReturnCode(self):
+    """
+      Returns the return code from "running the code."
+      @ In, None
+      @ Out, returnCode, int,  the return code of this evaluation
+    """
+    return self.returnCode
+
+  def getEvaluation(self):
+    """
+      Return solution.
+      @ In, None
+      @ Out, result, dict, results
+    """
+    result = {}
+    result.update(dict((key, np.atleast_1d(value)) for key, value in self._data['inputs'].items()))
+    result.update(dict((key, np.atleast_1d(value)) for key, value in self._data['outputs'].items()))
+    result.update(dict((key, np.atleast_1d(value)) for key, value in self._data['metadata'].items()))
+    return result
+
+  def start(self):
+    """
+      Method to start the job associated to this Runner
+      @ In, None
+      @ Out, None
+    """
+    pass # passthrough was born done
+
+  def kill(self):
+    """
+      Method to kill the job associated to this Runner
+      @ In, None
+      @ Out, None
+    """
+    pass # passthrough was born done; you can't kill it

framework/Runners/Runner.py

@@ -38,7 +38,7 @@ class Runner(MessageHandler.MessageUser):
     Generic base class for running codes and models in parallel environments
     both internally (shared data) and externally.
   """
-  def __init__(self, messageHandler, identifier = None, metadata = None, uniqueHandler = "any", profile = False):
+  def __init__(self, messageHandler, identifier=None, metadata=None, uniqueHandler="any", profile=False):
     """
       Initialize command variable
       @ In, messageHandler, MessageHandler instance, the global RAVEN message handler instance

framework/Runners/__init__.py

@@ -28,6 +28,7 @@
 from .InternalRunner import InternalRunner
 from .SharedMemoryRunner import SharedMemoryRunner
 from .DistributedMemoryRunner import DistributedMemoryRunner
+from .PassthroughRunner import PassthroughRunner
 from .Error import Error
 
 # from .Factory import knownTypes

framework/Steps.py

@@ -626,9 +626,10 @@ def _localInitializeStep(self,inDictionary):
     for inputIndex in range(inDictionary['jobHandler'].runInfoDict['batchSize']):
       if inDictionary[self.samplerType].amIreadyToProvideAnInput():
         try:
-          newInput = self._findANewInputToRun(inDictionary[self.samplerType], inDictionary['Model'], inDictionary['Input'], inDictionary['Output'])
-          inDictionary["Model"].submit(newInput, inDictionary[self.samplerType].type, inDictionary['jobHandler'], **copy.deepcopy(inDictionary[self.samplerType].inputInfo))
-          self.raiseADebug('Submitted input '+str(inputIndex+1))
+          newInput = self._findANewInputToRun(inDictionary[self.samplerType], inDictionary['Model'], inDictionary['Input'], inDictionary['Output'], inDictionary['jobHandler'])
+          if newInput is not None:
+            inDictionary["Model"].submit(newInput, inDictionary[self.samplerType].type, inDictionary['jobHandler'], **copy.deepcopy(inDictionary[self.samplerType].inputInfo))
+            self.raiseADebug('Submitted input '+str(inputIndex+1))
         except utils.NoMoreSamplesNeeded:
           self.raiseAMessage('Sampler returned "NoMoreSamplesNeeded".  Continuing...')
 
@@ -710,8 +711,9 @@ def _localTakeAstepRun(self,inDictionary):
 
           if sampler.amIreadyToProvideAnInput():
             try:
-              newInput = self._findANewInputToRun(sampler, model, inputs, outputs)
-              model.submit(newInput, inDictionary[self.samplerType].type, jobHandler, **copy.deepcopy(sampler.inputInfo))
+              newInput = self._findANewInputToRun(sampler, model, inputs, outputs, jobHandler)
+              if newInput is not None:
+                model.submit(newInput, inDictionary[self.samplerType].type, jobHandler, **copy.deepcopy(sampler.inputInfo))
             except utils.NoMoreSamplesNeeded:
               self.raiseAMessage(' ... Sampler returned "NoMoreSamplesNeeded".  Continuing...')
               break
@@ -728,7 +730,7 @@ def _localTakeAstepRun(self,inDictionary):
     # if any collected runs failed, let the sampler treat them appropriately, and any other closing-out actions
     sampler.finalizeSampler(self.failedRuns)
 
-  def _findANewInputToRun(self, sampler, model, inputs, outputs):
+  def _findANewInputToRun(self, sampler, model, inputs, outputs, jobHandler):
     """
       Repeatedly calls Sampler until a new run is found or "NoMoreSamplesNeeded" is raised.
       @ In, sampler, Sampler, the sampler in charge of generating the sample
@@ -741,18 +743,20 @@ def _findANewInputToRun(self, sampler, model, inputs, outputs):
         (i.e., a DataObject, File, or HDF5, I guess? Maybe these should all
         inherit from some base "Data" so that we can ensure a consistent
         interface for these?)
-      @ Out, newInp, list, list containing the new inputs
+      @ In, jobHandler, object, the raven object used to handle jobs
+      @ Out, newInp, list, list containing the new inputs (or None if a restart)
     """
     #The value of "found" determines what the Sampler is ready to provide.
     #  case 0: a new sample has been discovered and can be run, and newInp is a new input list.
     #  case 1: found the input in restart, and newInp is a realization dicitonary of data to use
-    found = None
-    while found != 0:
-      found, newInp = sampler.generateInput(model,inputs)
-      if found == 1:
-        # loop over the outputs for this step and collect the data for each
-        for collector, outIndex in self._outputDictCollectionLambda:
-          collector([newInp,outputs[outIndex]])
+    found, newInp = sampler.generateInput(model,inputs)
+    if found == 1:
+      # "submit" the finished run
+      jobHandler.addFinishedJob(newInp)
+      return None
+      # NOTE: we return None here only because the Sampler's "counter" is not correctly passed
+      # through if we add several samples at once through the restart. If we actually returned
+      # a Realization object from the Sampler, this would not be a problem. - talbpaul
     return newInp
 #
 #

tests/framework/Samplers/Restart/adaptive_mc.xml

@@ -0,0 +1,197 @@
+<?xml version="1.0" ?>
+<Simulation verbosity="debug">
+  <TestInfo>
+    <name>framework/Samplers/Restart.AMC</name>
+    <author>talbpaul</author>
+    <created>2020-07-16</created>
+    <classesTested>Samplers.AdaptiveMonteCarlo</classesTested>
+    <description>
+      Tests restarting an Adaptive Monte Carlo sampling from restart.  \texttt{makeCoarse} samples initial data, then \texttt{makeRestart}
+      makes additional samples, restarting from the first set of samples.  \texttt{makeFine} does all the samples without restart
+      for comparison.  The model for \texttt{coarse} always returns a value of 1, while the model for \texttt{restart} returns a value of 2, so
+      you can tell which samples came from which sampling strategy. Further, the solution export for
+      the restart should contain all the sample points, not merely the newly-sampled points.
+    </description>
+  </TestInfo>
+
+  <RunInfo>
+    <WorkingDir>AMC</WorkingDir>
+    <Sequence>makeCoarse,makeRestart,makeFine,print</Sequence>
+  </RunInfo>
+
+  <Steps>
+    <MultiRun name="makeCoarse">
+      <Input class="DataObjects" type="PointSet">dummyIN</Input>
+      <Model class="Models" type="ExternalModel">coarsemod</Model>
+      <Sampler class="Samplers" type="AdaptiveMonteCarlo">coarse</Sampler>
+      <SolutionExport class="DataObjects" type="PointSet">exportCoarse</SolutionExport>
+      <Output class="DataObjects" type="PointSet">solns</Output>
+    </MultiRun>
+    <MultiRun name="makeFine">
+      <Input class="DataObjects" type="PointSet">dummyIN</Input>
+      <Model class="Models" type="ExternalModel">finemod</Model>
+      <Sampler class="Samplers" type="AdaptiveMonteCarlo">fine</Sampler>
+      <SolutionExport class="DataObjects" type="PointSet">exportFine</SolutionExport>
+      <Output class="DataObjects" type="PointSet">solnsFine</Output>
+    </MultiRun>
+    <MultiRun name="makeRestart">
+      <Input class="DataObjects" type="PointSet">solns</Input>
+      <Model class="Models" type="ExternalModel">finemod</Model>
+      <Sampler class="Samplers" type="AdaptiveMonteCarlo">restart</Sampler>
+      <SolutionExport class="DataObjects" type="PointSet">exportRestart</SolutionExport>
+      <Output class="DataObjects" type="PointSet">solnsRestart</Output>
+    </MultiRun>
+    <IOStep name="print">
+      <Input class="DataObjects" type="PointSet">solns</Input>
+      <Input class="DataObjects" type="PointSet">solnsFine</Input>
+      <Input class="DataObjects" type="PointSet">solnsRestart</Input>
+      <Input class="DataObjects" type="PointSet">exportCoarse</Input>
+      <Input class="DataObjects" type="PointSet">exportFine</Input>
+      <Input class="DataObjects" type="PointSet">exportRestart</Input>
+      <Output class="OutStreams" type="Print">coarse</Output>
+      <Output class="OutStreams" type="Print">fine</Output>
+      <Output class="OutStreams" type="Print">restart</Output>
+      <Output class="OutStreams" type="Print">exp_coarse</Output>
+      <Output class="OutStreams" type="Print">exp_fine</Output>
+      <Output class="OutStreams" type="Print">exp_restart</Output>
+    </IOStep>
+  </Steps>
+
+  <Distributions>
+    <Uniform name="u1">
+      <lowerBound>1</lowerBound>
+      <upperBound>2</upperBound>
+    </Uniform>
+    <Uniform name="u2">
+      <lowerBound>2</lowerBound>
+      <upperBound>3</upperBound>
+    </Uniform>
+  </Distributions>
+
+  <Samplers>
+    <AdaptiveMonteCarlo name="coarse">
+      <TargetEvaluation class="DataObjects" type="PointSet">solns</TargetEvaluation>
+      <Convergence>
+        <limit>50</limit>
+        <persistence>5</persistence>
+        <forceIteration>False</forceIteration>
+        <expectedValue prefix="mean" tol="5e-1">x1</expectedValue>
+      </Convergence>
+      <initialSeed>42</initialSeed>
+      <variable name="x1">
+        <distribution>u1</distribution>
+      </variable>
+      <variable name="x2">
+        <distribution>u2</distribution>
+      </variable>
+    </AdaptiveMonteCarlo>
+    <AdaptiveMonteCarlo name="restart">
+      <TargetEvaluation class="DataObjects" type="PointSet">solnsRestart</TargetEvaluation>
+      <Convergence>
+        <limit>50</limit>
+        <persistence>5</persistence>
+        <forceIteration>False</forceIteration>
+        <expectedValue prefix="mean" tol="1e-1">x1</expectedValue>
+      </Convergence>
+      <initialSeed>42</initialSeed>
+      <variable name="x1">
+        <distribution>u1</distribution>
+      </variable>
+      <variable name="x2">
+        <distribution>u2</distribution>
+      </variable>
+      <Restart class="DataObjects" type="PointSet">solns</Restart>
+    </AdaptiveMonteCarlo>
+    <AdaptiveMonteCarlo name="fine">
+      <TargetEvaluation class="DataObjects" type="PointSet">solnsFine</TargetEvaluation>
+      <Convergence>
+        <limit>50</limit>
+        <persistence>5</persistence>
+        <forceIteration>False</forceIteration>
+        <expectedValue prefix="mean" tol="1e-1">x1</expectedValue>
+      </Convergence>
+      <initialSeed>42</initialSeed>
+      <variable name="x1">
+        <distribution>u1</distribution>
+      </variable>
+      <variable name="x2">
+        <distribution>u2</distribution>
+      </variable>
+    </AdaptiveMonteCarlo>
+  </Samplers>
+
+  <Models>
+    <Dummy name="MyDummy" subType=""/>
+    <ExternalModel ModuleToLoad="../coarse" name="coarsemod" subType="">
+      <variables>x1,x2,ans</variables>
+    </ExternalModel>
+    <ExternalModel ModuleToLoad="../fine" name="finemod" subType="">
+      <variables>x1,x2,ans</variables>
+    </ExternalModel>
+  </Models>
+
+  <DataObjects>
+    <PointSet name="dummyIN">
+      <Input>x1,x2</Input>
+      <Output>OutputPlaceHolder</Output>
+    </PointSet>
+    <PointSet name="solns">
+      <Input>x1,x2</Input>
+      <Output>ans</Output>
+    </PointSet>
+    <PointSet name="solnsFine">
+      <Input>x1,x2</Input>
+      <Output>ans</Output>
+    </PointSet>
+    <PointSet name="solnsRestart">
+      <Input>x1,x2</Input>
+      <Output>ans</Output>
+    </PointSet>
+    <PointSet name="exportCoarse">
+      <Input>solutionUpdate</Input>
+      <Output>mean_x1,mean_ste_x1</Output>
+    </PointSet>
+    <PointSet name="exportFine">
+      <Input>solutionUpdate</Input>
+      <Output>mean_x1,mean_ste_x1</Output>
+    </PointSet>
+    <PointSet name="exportRestart">
+      <Input>solutionUpdate</Input>
+      <Output>mean_x1,mean_ste_x1</Output>
+    </PointSet>
+  </DataObjects>
+
+  <OutStreams>
+    <Print name="coarse">
+      <type>csv</type>
+      <source>solns</source>
+      <what>input,output</what>
+    </Print>
+    <Print name="fine">
+      <type>csv</type>
+      <source>solnsFine</source>
+      <what>input,output</what>
+    </Print>
+    <Print name="restart">
+      <type>csv</type>
+      <source>solnsRestart</source>
+      <what>input,output</what>
+    </Print>
+    <Print name="exp_coarse">
+      <type>csv</type>
+      <source>exportCoarse</source>
+      <what>input,output</what>
+    </Print>
+    <Print name="exp_fine">
+      <type>csv</type>
+      <source>exportFine</source>
+      <what>input,output</what>
+    </Print>
+    <Print name="exp_restart">
+      <type>csv</type>
+      <source>exportRestart</source>
+      <what>input,output</what>
+    </Print>
+  </OutStreams>
+
+</Simulation>