optimization.variable.conic#
Conic Constant Variable Module
This module contains the class for a conic constant variable in an optic system. The ConicVariable class is a subclass of the VariableBehavior class that represents a variable for the conic constant of a surface in an optic system. It is used in the optimization process for conic surfaces.
Kramer Harrison, 2024
Classes
|
Represents a variable for the conic constant of a surface in an optic. |
- class ConicVariable(optic, surface_number, scaler=None, **kwargs)[source]#
Represents a variable for the conic constant of a surface in an optic.
- Parameters:
- Variables:
- Methods:
get_value – Returns the current conic constant of the surface.
update_value – Updates the conic value of the surface.
- get_value()[source]#
Returns the current conic constant of the surface.
- Returns:
The conic constant of the surface.
- Return type:
- inverse_scale(scaled_value)#
Inverse scale the value of the variable.
- Parameters:
scaled_value – The scaled value to inverse scale
- release_graph()#
Drop the autograd graph that the last update left in the optic.
A Torch optimizer calls this for all variables before it writes the values of a new step (
VariableManager.release_graphs). A variable that writes one entry of an array, which other variables can share (a coefficient array), must start each step from plain values. Otherwise the array still holds the graph of the last step: the next backward pass then runs into that freed graph and fails, or the graph grows with each step.The default does nothing: a variable that replaces a whole attribute leaves no old graph behind.
- scale(value)#
Scale the value of the variable for improved optimization performance.
- Parameters:
value – The value to scale