Add FMU Exception handling and unit test

modestpy/estim/ga_parallel/ga_parallel.py

@@ -6,11 +6,11 @@
 """
 import logging
 import os
-import pandas as pd
-import numpy as np
 from random import random
-from multiprocessing import Manager
-from multiprocessing.managers import BaseManager
+
+import numpy as np
+import pandas as pd
+
 from modestga import minimize
 from modestpy.fmi.model import Model
 from modestpy.estim.estpar import EstPar
@@ -21,9 +21,20 @@
 
 
 class ObjectiveFun:
+    """Objective function for paralell GA.
+
+    This class is callable and it is a wrapper
+    containing FMU initialization and simulation.
+
+    Note:
+        If FMU simulation fails, the returned error is 1e8
+        (attribute `err_sim_failed`) and the FMU is
+        reinitialized before next simulation.
+    """
     def __init__(self, fmu_path, inp, known, est, ideal, ftype='RMSE'):
         self.logger = logging.getLogger(type(self).__name__)
         self.model = None
+        self.sim_failed = False
         self.fmu_path = fmu_path
         self.inp = inp
 
@@ -41,6 +52,7 @@ def __init__(self, fmu_path, inp, known, est, ideal, ftype='RMSE'):
         self.output_names = [var for var in ideal]
         self.ftype = ftype
         self.best_err = 1e7
+        self.err_sim_failed = 1e8
         self.res = pd.DataFrame()
 
         self.logger.debug(f"fmu_path = {fmu_path}")
@@ -54,12 +66,17 @@ def rescale(self, v, lo, hi):
         return lo + v * (hi - lo)
 
     def __call__(self, x, *args):
-        # Instantiate the model
+
         self.logger.debug(f"x = {x}")
-        if self.model is None:
+
+        # Instantiate the model if
+        # - there's no model yet
+        # - previous simulation failed
+        if (self.model is None) or (self.sim_failed is True):
             self.model = self._get_model_instance(
                 self.fmu_path, self.inp, self.known, self.est, self.output_names
             )
+            self.sim_failed = False
             logging.debug(f"Model instance returned: {self.model}")
 
         # Updated parameters are stored in x. Need to update the model.
@@ -69,15 +86,25 @@ def __call__(self, x, *args):
                 parameters[ep.name] = self.rescale(v, ep.lo, ep.hi)
         except TypeError as e:
             raise e
+
         self.logger.debug(f"parameters = {parameters}")
+
         self.model.parameters_from_df(parameters)
+
         self.logger.debug(f"est: {self.est}")
         self.logger.debug(f"parameters: {parameters}")
         self.logger.debug(f"model: {self.model}")
         self.logger.debug("Calling simulation...")
-        result = self.model.simulate()
-        self.logger.debug(f"result: {result}")
-        err = calc_err(result, self.ideal, ftype=self.ftype)['tot']
+
+        try:
+            result = self.model.simulate()
+            self.logger.debug(f"result: {result}")
+            err = calc_err(result, self.ideal, ftype=self.ftype)['tot']
+        except Exception as e:
+            self.model_failed = True
+            err = self.err_sim_failed
+            self.logger.error(str(e))
+
         # Update best error and result
         if err < self.best_err:
             self.best_err = err
@@ -92,19 +119,31 @@ def _get_model_instance(self, fmu_path, inputs, known_pars, est, output_names):
         self.logger.debug(f"est = {est}")
         self.logger.debug(f"estpars_2_df(est) = {estpars_2_df(est)}")
         self.logger.debug(f"output_names = {output_names}")
-        model = Model(fmu_path)
-        model.inputs_from_df(inputs)
-        model.parameters_from_df(known_pars)
-        model.parameters_from_df(estpars_2_df(est))
-        model.specify_outputs(output_names)
+
+        try:
+            model = Model(fmu_path)
+            model.inputs_from_df(inputs)
+            model.parameters_from_df(known_pars)
+            model.parameters_from_df(estpars_2_df(est))
+            model.specify_outputs(output_names)
+            res = model.simulate()
+        except Exception as e:
+            self.logger.error(str(e))
+            msg = "Can't initialize FMU with given parameters. Will try with defaults."
+            self.logger.error(msg)
+            model = Model(fmu_path)
+            model.inputs_from_df(inputs)
+            model.specify_outputs(output_names)
+            res = model.simulate()
+
         self.logger.debug(f"Model instance initialized: {model}")
         self.logger.debug(f"Model instance initialized: {model.model}")
-        res = model.simulate()
         self.logger.debug(f"test result: {res}")
+
         return model
 
 
-class MODESTGA(object):
+class MODESTGA:
     """
     Parallel Genetic Algorithm based on modestga
     (https://github.com/krzysztofarendt/modestga).

modestpy/test/run.py

@@ -7,6 +7,7 @@
 import unittest
 from modestpy.test import test_fmpy
 from modestpy.test import test_ga
+from modestpy.test import test_ga_parallel
 from modestpy.test import test_ps
 from modestpy.test import test_scipy
 from modestpy.test import test_estimation

modestpy/test/test_ga_parallel.py

@@ -70,6 +70,18 @@ def test_modestga_default(self):
         par_df = ga.estimate()
         assert type(par_df) is pd.DataFrame
 
+    def test_modestga_simulation_fail(self):
+        est_fail = {
+            "R1": (-99, -100., 0.),
+            "R2": (-1e6, -1e7, -1e5),
+            "C": (0., -1e-10, 1e-10)
+        }
+        gen = 3
+        ga = MODESTGA(self.fmu_path, self.inp, self.known, est_fail, self.ideal,
+                      generations=gen)
+        par_df = ga.estimate()
+        assert type(par_df) is pd.DataFrame
+
     def test_modestga_1_worker(self):
         ga = MODESTGA(self.fmu_path, self.inp, self.known, self.est, self.ideal,
                       generations=self.gen,
@@ -107,6 +119,7 @@ def test_modestga_2_workers_large_pop(self):
 def suite():
     suite = unittest.TestSuite()
     suite.addTest(TestMODESTGA('test_modestga_default'))
+    suite.addTest(TestMODESTGA('test_modestga_simulation_fail'))
     suite.addTest(TestMODESTGA('test_modestga_1_worker'))
     suite.addTest(TestMODESTGA('test_modestga_2_workers_small_pop'))
     suite.addTest(TestMODESTGA('test_modestga_2_workers_large_pop'))
@@ -116,4 +129,3 @@ def suite():
 if __name__ == '__main__':
     config_logger(filename='unit_tests.log', level='DEBUG')
     unittest.main()
-

setup.py

@@ -2,7 +2,7 @@
 
 setup(
     name='modestpy',
-    version='0.1',
+    version='0.1.1',
     description='FMI-compliant model identification package',
     url='https://github.com/sdu-cfei/modest-py',
     keywords='fmi fmu optimization model identification estimation',