kernel/write_an_optimizer_state.py

While we do not recommend doing serious work using optimizer states written in Python, it is often useful when prototyping or testing code. Copy this example and modify as needed.

## \example kernel/write_an_optimizer_state.py
# While we do not recommend doing serious work using optimizer states
# written in Python, it is often useful when prototyping or testing code.
# Copy this example and modify as needed.

import IMP
import sys

IMP.setup_from_argv(sys.argv, "Optimizer state")


class MyOptimizerState(IMP.OptimizerState):
    "An optimizer state which prints out the last scores of some restraints"

    def __init__(self, rs):
        super().__init__(rs[0].get_model(), "MyOptimizerState%1%")
        self.rs = rs

    def update(self):
        for r in self.rs:
            print(r.get_name(), r.get_last_score())


# some code to create and evaluate it
k = IMP.FloatKey("a key")
m = IMP.Model()
# we don't have any real restraints in the kernel
r0 = IMP._ConstRestraint(m, [], 1)
r0.set_name("restraint 0")

r1 = IMP._ConstRestraint(m, [], 2)
r1.set_name("restraint 1")

rs = IMP.RestraintSet([r0, r1], 1.0)
sf = rs.create_scoring_function()

os = MyOptimizerState([r0, r1])
os.set_name("Python optimizer state")
# we don't have any optimizers either
co = IMP._ConstOptimizer(m)
co.set_scoring_function(sf)
co.add_optimizer_state(os)
print(co.optimize(100))

# this is needed to clean up memory properly for some reason
co.remove_optimizer_state(os)
del os
del m