pyepo.model.opt¶

Abstract optimization model

Classes¶

`ModelSpec`	Serializable recipe for building a fresh optimization model.
`optModel`	Abstract base class for predict-then-optimize models.

Module Contents¶

class pyepo.model.opt.ModelSpec(model_type: type[optModel], config: dict, args: tuple = ())¶

Serializable recipe for building a fresh optimization model.

model_type: type[optModel]¶

property args: tuple¶: Return an independent copy of positional constructor arguments.

property config: dict¶: Return an independent copy of keyword constructor arguments.

build() → optModel¶: Build a fresh model without sharing mutable configuration values.

class pyepo.model.opt.optModel¶

Bases: abc.ABC

Abstract base class for predict-then-optimize models.

Subclasses wrap an optimization solver or algorithm with a unified _getModel / setObj / solve / num_cost interface that pyepo.func modules call during training. Concrete backends are provided for GurobiPy (optGrbModel), Pyomo (optOmoModel), COPT (optCoptModel), OR-Tools (optOrtModel / optOrtCpModel), and MPAX (optMpaxModel); subclass optModel directly to integrate any other solver or algorithm.

The default objective sense is minimization; set self.modelSense = EPO.MAXIMIZE in _getModel or __init__ for maximization problems (some backends, e.g. Gurobi and COPT, detect this automatically from the underlying solver model).

Variables:

_model (optimization model) – underlying solver model object
modelSense (ModelSense) – EPO.MINIMIZE (default) or EPO.MAXIMIZE

modelSense: pyepo.EPO.ModelSense¶

arcs: list¶

get_config() → dict¶: Return the constructor configuration for this model.

classmethod from_config(config: dict, args: tuple = ()) → Self¶: Build a model from a configuration produced by get_config.

to_spec() → ModelSpec¶: Return a serializable, immutable-snapshot rebuild recipe.

rebuild() → Self¶: Build a structurally equivalent model with clean runtime state.

property num_cost: int¶: number of costs to be predicted

abstractmethod setObj(c: numpy.ndarray | torch.Tensor | list) → None¶

An abstract method to set the objective function

Parameters:: c – cost of objective function

abstractmethod solve() → tuple[numpy.ndarray | torch.Tensor | list, float]¶

An abstract method to solve the model

Returns:: optimal solution (list) and objective value (float)
Return type:: tuple

property c_pred_index: numpy.ndarray | None¶: Variable positions the predicted cost lands on, or None when every variable is predicted (the default).

copy() → Self¶

A method to copy the model

Returns:: new copied model
Return type:: optModel

abstractmethod addConstr(coefs: numpy.ndarray | torch.Tensor | list, rhs: float) → Self¶

A method to add a new constraint. Subclasses should override.

Parameters:

coefs – coefficients of the new constraint
rhs – right-hand side of new constraint

Returns:

new model with the added constraint

Return type:

optModel

abstractmethod relax() → optModel¶: A method to relax the MIP model. Subclasses should override.