Conversion Guide for the Auto-IVC (IndepVarComp) Feature¶

As of the OpenMDAO 3.2 release, it is no longer necessary to add an IndepVarComp to your model to handle the assignment of unconnected inputs as design variables.

Declaring Design Variables¶

Pre Auto IVC

This is what we used to do

import openmdao.api as om

prob = om.Problem()
indeps = prob.model.add_subsystem('indeps', om.IndepVarComp())
indeps.add_output('x', 3.0)
indeps.add_output('y', -4.0)

prob.model.add_subsystem('paraboloid',
                         om.ExecComp('f = (x-3)**2 + x*y + (y+4)**2 - 3'))

prob.model.connect('indeps.x', 'paraboloid.x')
prob.model.connect('indeps.y', 'paraboloid.y')

# setup the optimization
prob.driver = om.ScipyOptimizeDriver()
prob.driver.options['optimizer'] = 'SLSQP'

prob.model.add_design_var('indeps.x', lower=-50, upper=50)
prob.model.add_design_var('indeps.y', lower=-50, upper=50)
prob.model.add_objective('paraboloid.f')

prob.setup()
prob.run_driver()

With Auto IVC

This is how we handle IVCs now

prob = om.Problem()

prob.model.add_subsystem('paraboloid',
                         om.ExecComp('f = (x-3)**2 + x*y + (y+4)**2 - 3'),
                         promotes_inputs=['x', 'y'])

# setup the optimization
prob.driver = om.ScipyOptimizeDriver()
prob.driver.options['optimizer'] = 'SLSQP'

prob.model.add_design_var('x', lower=-50, upper=50)
prob.model.add_design_var('y', lower=-50, upper=50)
prob.model.add_objective('paraboloid.f')

prob.setup()

prob['x'] = 3.0
prob['y'] = -4.0

prob.run_driver()

Declaring a Multi-Component Input as a Design Variable¶

Pre Auto IVC

import openmdao.api as om

prob = om.Problem()
indeps = prob.model.add_subsystem('indeps', om.IndepVarComp())
indeps.add_output('x', 3.0)
indeps.add_output('y', -4.0)

prob.model.add_subsystem('parab', Paraboloid())

# define the component whose output will be constrained
prob.model.add_subsystem('const', om.ExecComp('g = x + y'))

prob.model.connect('indeps.x', ['parab.x', 'const.x'])
prob.model.connect('indeps.y', ['parab.y', 'const.y'])

# setup the optimization
prob.driver = om.ScipyOptimizeDriver()
prob.driver.options['optimizer'] = 'COBYLA'

prob.model.add_design_var('indeps.x', lower=-50, upper=50)
prob.model.add_design_var('indeps.y', lower=-50, upper=50)
prob.model.add_objective('parab.f_xy')

# to add the constraint to the model
prob.model.add_constraint('const.g', lower=0, upper=10.)
# prob.model.add_constraint('const.g', equals=0.)

prob.setup()
prob.run_driver()

With Auto IVC

prob = om.Problem()
prob.model.add_subsystem('parab', Paraboloid(),
                         promotes_inputs=['x', 'y'])

# define the component whose output will be constrained
prob.model.add_subsystem('const', om.ExecComp('g = x + y'),
                         promotes_inputs=['x', 'y'])

# Design variables 'x' and 'y' span components, so we need to provide a common initial
# value for them.
prob.model.set_input_defaults('x', 3.0)
prob.model.set_input_defaults('y', -4.0)

# setup the optimization
prob.driver = om.ScipyOptimizeDriver()
prob.driver.options['optimizer'] = 'COBYLA'

prob.model.add_design_var('x', lower=-50, upper=50)
prob.model.add_design_var('y', lower=-50, upper=50)
prob.model.add_objective('parab.f_xy')

# to add the constraint to the model
prob.model.add_constraint('const.g', lower=0, upper=10.)

prob.setup()
prob.run_driver()

Declaring a New Name for a Promoted Input¶

Pre Auto IVC

prob = om.Problem()

indeps = prob.model.add_subsystem('indeps', om.IndepVarComp())
indeps.add_output('width', 3.0)
indeps.add_output('length', -4.0)

prob.model.add_subsystem('paraboloid',
                       om.ExecComp('f = (x-3)**2 + x*y + (y+4)**2 - 3'))

prob.model.connect('indeps.width', 'paraboloid.x')
prob.model.connect('indeps.length', 'paraboloid.y')

prob.setup()

With Auto IVC

prob = om.Problem()

prob.model.add_subsystem('paraboloid',
                         om.ExecComp('f = (x-3)**2 + x*y + (y+4)**2 - 3'),
                         promotes_inputs=[('x', 'width'), ('y', 'length')])

# Could also set these after setup.
prob.model.set_input_defaults('width', 3.0)
prob.model.set_input_defaults('length', -4.0)

prob.setup()

Declare an Input Defined with Source Indices as a Design Variable¶

Pre Auto IVC

class MyComp1(om.ExplicitComponent):
    def setup(self):
        # this input will connect to entries 0, 1, and 2 of its source
        self.add_input('x', np.ones(3), src_indices=[0, 1, 2])
        self.add_output('y', 1.0)

    def compute(self, inputs, outputs):
        outputs['y'] = np.sum(inputs['x'])*2.0

class MyComp2(om.ExplicitComponent):
    def setup(self):
        # this input will connect to entries 3 and 4 of its source
        self.add_input('x', np.ones(2), src_indices=[3, 4])
        self.add_output('y', 1.0)

    def compute(self, inputs, outputs):
        outputs['y'] = np.sum(inputs['x'])*4.0

p = om.Problem()

p.model.add_subsystem('indep', om.IndepVarComp('x', np.ones(5)),
                    promotes_outputs=['x'])
p.model.add_subsystem('C1', MyComp1(), promotes_inputs=['x'])
p.model.add_subsystem('C2', MyComp2(), promotes_inputs=['x'])

p.model.add_design_var('x')
p.setup()
p.run_model()

With Auto IVC

class MyComp1(om.ExplicitComponent):
    def setup(self):
        # this input will connect to entries 0, 1, and 2 of its source
        self.add_input('x', np.ones(3), src_indices=[0, 1, 2])
        self.add_output('y', 1.0)

    def compute(self, inputs, outputs):
        outputs['y'] = np.sum(inputs['x'])*2.0

class MyComp2(om.ExplicitComponent):
    def setup(self):
        # this input will connect to entries 3 and 4 of its source
        self.add_input('x', np.ones(2), src_indices=[3, 4])
        self.add_output('y', 1.0)

    def compute(self, inputs, outputs):
        outputs['y'] = np.sum(inputs['x'])*4.0

p = om.Problem()

# IndepVarComp is required to define the full size of the source vector.
p.model.add_subsystem('indep', om.IndepVarComp('x', np.ones(5)),
                      promotes_outputs=['x'])
p.model.add_subsystem('C1', MyComp1(), promotes_inputs=['x'])
p.model.add_subsystem('C2', MyComp2(), promotes_inputs=['x'])

p.model.add_design_var('x')
p.setup()
p.run_model()

Setting Default Units for an Input¶

Pre Auto IVC

prob = om.Problem()

ivc = om.IndepVarComp()
ivc.add_output('x2', 100.0, units='degC')
prob.model.add_subsystem('T1', ivc,
                       promotes_outputs=['x2'])

# Input units in degF
prob.model.add_subsystem('tgtF', TgtCompF(),
                       promotes_inputs=['x2'])

# Input units in degC
prob.model.add_subsystem('tgtC', TgtCompC(),
                       promotes_inputs=['x2'])

# Input units in deg
prob.model.add_subsystem('tgtK', TgtCompK(),
                       promotes_inputs=['x2'])

prob.setup()

With Auto IVC

prob = om.Problem()

# Input units in degF
prob.model.add_subsystem('tgtF', TgtCompF(),
                         promotes_inputs=['x2'])

# Input units in degC
prob.model.add_subsystem('tgtC', TgtCompC(),
                         promotes_inputs=['x2'])

# Input units in degK
prob.model.add_subsystem('tgtK', TgtCompK(),
                         promotes_inputs=['x2'])

prob.model.set_input_defaults('x2', 100.0, units='degC')

prob.setup()

Creating a Distributed Component with Unconnected Inputs¶

Pre Auto IVC

size = 4

prob = om.Problem()

prob.model.add_subsystem("C1", DistribNoncontiguousComp(arr_size=size),
                       promotes=['invec', 'outvec'])

prob.setup()

rank = prob.model.comm.rank
if rank == 0:
    prob.set_val('invec', np.array([1.0, 3.0]))
else:
    prob.set_val('invec', np.array([5.0, 7.0]))

prob.run_model()

With Auto IVC

size = 4

prob = om.Problem()

# An IndepVarComp is required on all unconnected distributed inputs.
ivc = om.IndepVarComp()
ivc.add_output('invec', np.ones(size), distributed=True)
prob.model.add_subsystem('P', ivc,
                         promotes_outputs=['invec'])

prob.model.add_subsystem("C1", DistribNoncontiguousComp(arr_size=size),
                         promotes=['invec', 'outvec'])

prob.setup()

prob.set_val('P.invec', np.array([1.0, 3.0, 5.0, 7.0]))

prob.run_model()

%%px

import openmdao.api as om

from openmdao.utils.array_utils import take_nth

class DistribNoncontiguousComp(om.ExplicitComponent):
    """Uses 4 procs and takes non-contiguous input var slices and has output
    var slices as well
    """

    def initialize(self):
        self.options.declare('arr_size', types=int, default=11,
                             desc="Size of input and output vectors.")

    def compute(self, inputs, outputs):
        outputs['outvec'] = inputs['invec']*2.0

    def setup(self):

        comm = self.comm
        rank = comm.rank
        arr_size = self.options['arr_size']
        idxs = list(take_nth(rank, comm.size, range(arr_size)))

        self.add_input('invec', np.ones(len(idxs), float), distributed=True)
        self.add_output('outvec', np.ones(len(idxs), float), distributed=True)

Setting and Getting Inputs¶

Pre Auto IVC

prob = om.Problem()
indeps = prob.model.add_subsystem('indeps', om.IndepVarComp())
indeps.add_output('x', 3.0)
indeps.add_output('y', -4.0)

prob.model.add_subsystem('paraboloid',
                         om.ExecComp('f = (x-3)**2 + x*y + (y+4)**2 - 3'))

prob.model.connect('indeps.x', 'paraboloid.x')
prob.model.connect('indeps.y', 'paraboloid.y')

prob.setup()

x = prob.get_val('indeps.x')
prob.set_val('indeps.y', 15.0)

With Auto IVC

prob = om.Problem()

prob.model.add_subsystem('paraboloid',
                         om.ExecComp('f = (x-3)**2 + x*y + (y+4)**2 - 3'),
                         promotes_inputs=['x', 'y'])

prob.setup()

x = prob.get_val('x')
prob.set_val('y', 15.0)