constant_proximal_weight

Constant proximal-weight autotuner.

Keeps lam_prox fixed across iterations while still relaxing lam_cost after the configured lam_cost_drop iteration.

ConstantProximalWeight

Bases: AutotuningBase

Hold lam_prox constant; relax lam_cost after lam_cost_drop.

Useful when you want a fixed trust-region size without adaptation.

update_weights is a pure functional update on the :class:AlgorithmState pytree; see the base-class contract.

Source code in openscvx/algorithms/autotuner/constant_proximal_weight.py

class ConstantProximalWeight(AutotuningBase):
    """Hold ``lam_prox`` constant; relax ``lam_cost`` after ``lam_cost_drop``.

    Useful when you want a fixed trust-region size without adaptation.

    ``update_weights`` is a pure functional update on the
    :class:`AlgorithmState` pytree; see the base-class contract.
    """

    COMPUTES_ACCEPTANCE_METRICS = False

    def __init__(
        self,
        lam_cost_drop: int = -1,
        lam_cost_relax: float = 1.0,
    ):
        self.hyper = ConstantProximalWeightHyper(
            lam_cost_drop=lam_cost_drop,
            lam_cost_relax=lam_cost_relax,
        )

    def update_weights(
        self,
        state: "AlgorithmState",
        candidate: "CandidateIterate",
        nodal_constraints: "LoweredJaxConstraints",
        settings: Config,
        params: dict,
    ) -> "AlgorithmState":
        """Return the next-iterate state.

        Pure functional update — see class docstring.
        """
        lam_cost_next = self._relaxed_lam_cost(state)

        return state.replace(
            x=candidate.x,
            u=candidate.u,
            x_prop=candidate.x_prop,
            x_prop_plus=candidate.x_prop_plus,
            lam_cost=lam_cost_next,
            adaptive_state_code=jnp.asarray(
                int(AdaptiveStateCode.ACCEPT_CONSTANT), dtype=jnp.int32
            ),
        )

update_weights(state: AlgorithmState, candidate: CandidateIterate, nodal_constraints: LoweredJaxConstraints, settings: Config, params: dict) -> AlgorithmState

Return the next-iterate state.

Pure functional update — see class docstring.

Source code in openscvx/algorithms/autotuner/constant_proximal_weight.py

def update_weights(
    self,
    state: "AlgorithmState",
    candidate: "CandidateIterate",
    nodal_constraints: "LoweredJaxConstraints",
    settings: Config,
    params: dict,
) -> "AlgorithmState":
    """Return the next-iterate state.

    Pure functional update — see class docstring.
    """
    lam_cost_next = self._relaxed_lam_cost(state)

    return state.replace(
        x=candidate.x,
        u=candidate.u,
        x_prop=candidate.x_prop,
        x_prop_plus=candidate.x_prop_plus,
        lam_cost=lam_cost_next,
        adaptive_state_code=jnp.asarray(
            int(AdaptiveStateCode.ACCEPT_CONSTANT), dtype=jnp.int32
        ),
    )

ConstantProximalWeightHyper

Bases: LamCostRelaxHyper

Declared hyperparameters for :class:ConstantProximalWeight.

The lam_cost relaxation knobs are this autotuner's only tunables, so it declares no fields beyond the shared :class:LamCostRelaxHyper ones.

Source code in openscvx/algorithms/autotuner/constant_proximal_weight.py

class ConstantProximalWeightHyper(LamCostRelaxHyper):
    """Declared hyperparameters for :class:`ConstantProximalWeight`.

    The ``lam_cost`` relaxation knobs are this autotuner's only tunables, so it
    declares no fields beyond the shared :class:`LamCostRelaxHyper` ones.
    """

ConstantProximalWeightSpec

Bases: BaseModel

Validates ConstantProximalWeight configuration from dict/YAML input.

Source code in openscvx/algorithms/autotuner/constant_proximal_weight.py

class ConstantProximalWeightSpec(BaseModel):
    """Validates ConstantProximalWeight configuration from dict/YAML input."""

    type: Literal["ConstantProximalWeight"] = "ConstantProximalWeight"
    lam_cost_drop: int = -1
    lam_cost_relax: float = 1.0

    model_config = ConfigDict(extra="forbid")

    def build(self) -> ConstantProximalWeight:
        return ConstantProximalWeight(**self.model_dump(exclude={"type"}, exclude_unset=True))