Saving and loading models

Note

The Python API is still in development and the serialisation format may change in the future. It is recommended to only use Pickle for temporary storage and not for long term storage.

KatScript offers one way to easily store a particular optical layout which describes the components, detectors, and analyses to be performed. Whilst this is convenient and can be stored as a simple ASCII file for long term storage, it is limited in how complex a model can be described. The Python API offers full access to make models which use advanced features in Finesse as well as custom user code. In such cases the finesse.model.Model objects much be serialised and deserialised and saved somewhere.

In Python, Pickle is the standard library for serialising objects. Pickling tries to save the object as a binary representation of the object, which can then be loaded back into memory. The benefit of this is that arbitrarily complex objects can be saved and loaded. However, there are several downsides to Pickle:

  • Pickle is not human readable, so it is not suitable for long term storage

  • Pickle is not secure, loading a Pickle file can execute arbitrary code so care should be taken on the origin of the Pickle file

  • Pickle is not guaranteed to be compatible between different versions of Python

  • Pickle is not guaranteed to be compatible between different versions of Finesse

  • Pickle is not guaranteed to be compatible between different versions of the Python API

  • Pickle is not guaranteed to be compatible between different operating systems, such as Windows and Linux, or different architectures, such as x86 and ARM.

With this being said, it is still very useful to be able to save and load models for temporary short term storage, such as when running an optimisation and saving the state of the model at each iteration. Or serialising models to distribute to other processors in a cluster or multi core system.

The interface to load and save models is very simple. The finesse.model.Model.save() and finesse.model.load() methods can be used like so:

import finesse
from finesse import Model
from finesse.components import Laser, Mirror, Space, Beamsplitter
from finesse.detectors import PowerDetector

model = Model()
l1 = model.add(Laser("l1", P=10))
pd1 = model.add(PowerDetector("pd1", l1.p1.o))

model.save("model.pkl")
loaded_model = finesse.model.load('model.pkl')

print(model, loaded_model)

<finesse.model.Model object at 0x7fd1fb16b4d0> <finesse.model.Model object at 0x7fd1fb182990>

Warning

Pickling is not always a fool proof way to save and load models. Sometimes the saving works and the loading does not. This is usually due to internal issues in Finesse that can be fixed. Please report any issues you have to the Gitlab project issue tracker.