openscvx 0.5.3.dev23__tar.gz → 0.5.3.dev25__tar.gz

{openscvx-0.5.3.dev23/openscvx.egg-info → openscvx-0.5.3.dev25}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: openscvx
-Version: 0.5.3.dev23
+Version: 0.5.3.dev25
 Summary: A general Python-based successive convexification implementation which uses a JAX backend.
 Author-email: Chris Hayner and Griffin Norris <haynec@uw.edu>
 License: Apache Software License

{openscvx-0.5.3.dev23 → openscvx-0.5.3.dev25}/docs/UnderTheHood/batching_jit_grad.md

@@ -60,7 +60,7 @@ that one difference is the whole point.
   no outer `vmap` rule (`call_exported` cannot be batched), so an exported
   artifact can never be re-vmapped, and every fresh process re-traces and
   re-compiles the entire loop.
-* `solve_batched(x0_stack, xf_stack)` — *the library* owns the axis. The vmap
+* `solve_batched(x_initial=x0_batch)` — *the library* owns the axis. The vmap
   is applied **inside** the method, before any export, so the whole loop lowers
   to ordinary batched XLA ops and serializes cleanly. Under
   `save_compiled=True` it is written to the solver cache on the first run and
@@ -74,10 +74,131 @@ problem.initialize()
 
 # First process: traces, exports, writes <cache>/compiled_solve_batched_<hash>.jax
 # Later processes: deserialize-and-.call, no compile.
-batched = problem.solve_batched(x0_stack, xf_stack, params)
+batched = problem.solve_batched(x_initial=x0_batch, parameters=params)
 # batched.x.shape == (B, N, n_states) — same pytree jax.vmap(solve_jax) returns.
 ```
 
+### One rule: add a leading batch axis to whatever varies
+
+`solve_batched` takes exactly `solve_jax`'s arguments. Every batchable input
+has a *declared* (unbatched) shape the library knows — `(n_x,)` for the
+boundary pins, `(N, n_x)` / `(N, n_u)` for the trajectory guesses, the
+parameter's stored shape for each parameter, and `()` for the SCP
+hyperparameters. An argument that matches its declared shape is **shared** by
+every batch element; one with exactly one extra leading axis is **batched**
+along it. Rank against the declared shape decides — never the leading-axis
+value — so a shared `(B, n)` matrix parameter is never misread as batched.
+(`B` in these shape patterns is plain notation for the batch size, not
+syntax: it is never passed anywhere — like `jax.vmap`, the library reads it
+off the leading axis of whatever you batched and cross-checks that all
+batched inputs agree.)
+
+```python
+results = problem.solve_batched(
+    x_initial=x0_batch,                  # (B, n_x): rank = declared+1 -> batched
+    parameters={
+        "gate_center": centers,          # (B, 3) vs declared (3,)  -> batched
+        "gate_radius": 0.5,              # declared shape           -> shared
+    },
+)
+```
+
+For the rare entry the rank rule cannot express — e.g. batching a declared
+rank-1 parameter element-wise — pass an explicit `in_axes` in `jax.vmap`'s
+vocabulary, including its prefix semantics (`0` = batched along the leading
+axis, `None` = shared; a prefix applies to everything below it; partial specs
+fall back to the rank rule):
+
+```python
+results = problem.solve_batched(
+    parameters={"weights": w},           # (B,) is also w's declared shape...
+    in_axes={"parameters": {"weights": 0}},  # ...so say "batched" explicitly
+)
+
+results = problem.solve_batched(parameters=params, in_axes={"parameters": 0})
+# prefix: every passed parameter is batched, exactly like jax.vmap's
+# pytree-prefix in_axes; a bare in_axes=0 batches every passed input.
+```
+
+One deliberate divergence from `jax.vmap`: the *default* is "infer by rank",
+not `0`-everything — an all-shared batched solve has no batch axis to find,
+and inference is what makes mixed shared/batched dicts ergonomic.
+
+### Hyperparameter sweeps: the `algorithm` dict
+
+Solve-time mirrors construction-time: the `Problem` constructor configures
+the SCP algorithm with a dict (`algorithm={"ep_tr": 1e-5, "lam_prox": 1.0}`),
+and the solve methods take the same box with the same names as per-solve,
+batchable inputs. Every `AlgorithmState` field is a runtime input riding the
+state pytree, and the `algorithm` dict reaches any of them by name —
+convergence thresholds (`ep_tr` / `ep_vb` / `ep_vc`), penalty weights
+(`lam_prox` / `lam_vc` / `lam_cost`), the seed iterate, autotuner
+hyperparameters. There is no whitelist: the field declaration is the
+registration, and the `AlgorithmState` attribute docs are the reference.
+(`max_iters` survives as a named kwarg — sugar for `algorithm={"k_max":
+...}`; passing both raises, as does an entry that shadows any other named
+kwarg's field. Construction-only keys of the constructor dict — `autotuner`,
+`t_max` — raise pointing back at the constructor.)
+
+* On `solve_jax`, each value matches the field's shape, or is a scalar
+  broadcast to it: `solve_jax(algorithm={"ep_tr": 1e-6, "lam_prox": 2.0})`
+  retraces nothing.
+* On `solve_batched`, each value follows the rank rule against the field's
+  shape, plus two **fill** forms for array-shaped fields: a scalar (shared)
+  or a `(B,)` vector (one scalar per element, broadcast to the field shape).
+  Exact shapes win when both parse; `in_axes={"algorithm": {name: 0}}`
+  forces the per-element reading. All against one cached artifact:
+
+```python
+B = 8  # batch size. Just an int for building the arrays below — there is no
+       # B argument; solve_batched infers it from the leading-axis lengths.
+sweep = problem.solve_batched(algorithm={"ep_tr": jnp.logspace(-6, -3, B)})
+weights = problem.solve_batched(algorithm={"lam_prox": jnp.array([0.5, 1.0, 4.0])})
+budgets = problem.solve_batched(max_iters=jnp.array([5, 10, 20]))  # per-element caps
+```
+
+The batch runs until its slowest element converges or exhausts its own cap;
+finished elements are frozen, not re-iterated, so each element matches the
+corresponding single `solve_jax` run.
+
+**Every built-in autotuner knob batches for free.** All numeric knobs of the
+built-in autotuners are declared hyperparameters — the trust-region factors
+(`gamma_1` / `gamma_2`), the acceptance thresholds (`eta_0` / `eta_1` /
+`eta_2`), the weight clips, the cost-relaxation schedule — so any of them is a
+`solve_jax` override and a `solve_batched` sweep target by name:
+
+```python
+# sweep the trust-region growth factor across a batch, one cached artifact
+results = problem.solve_batched(algorithm={"gamma_1": jnp.linspace(1.5, 3.0, B)})
+```
+
+**New autotuners batch the same way.** A custom autotuner's tunable knobs
+follow the same derivation: declare them as fields on a `HyperParams` subclass
+— bare annotations, no decorator; subclassing applies the frozen-dataclass
+transform and the JAX pytree registration, and the annotation picks the dtype
+(`int` gets the iteration counter's, `float` the problem's) — assign an
+instance to `self.hyper`, and read `state.hyper.<field>` inside
+`update_weights` (see the `AutotuningBase` docstring). The declaration is the
+registration — each field immediately works as a `solve_jax` override, a
+`solve_batched` sweep target, and a runtime input of the exported artifact,
+with zero edits anywhere else:
+
+```python
+from openscvx.algorithms import AutotuningBase, HyperParams
+
+class MyHyper(HyperParams):
+    ramp: float = 2.0
+
+class MyAutotuner(AutotuningBase):
+    def __init__(self, ramp: float = 2.0):
+        self.hyper = MyHyper(ramp=ramp)
+
+    def update_weights(self, state, ...):
+        ...state.hyper.ramp...
+
+results = problem.solve_batched(algorithm={"ramp": jnp.linspace(1.5, 3.0, B)})
+```
+
 Pick `solve_batched` when **cross-process cold-start dominates** — short-lived
 processes (CI sweeps, notebook restarts, deployed workers) that otherwise pay
 the compile every launch. Pick `jax.vmap(solve_jax)` for everything in-program,
@@ -111,15 +232,21 @@ object contributing through its own `_hash_into` (the same protocol the symbolic
 layer uses):
 
 * the convex backend class + its `solver_args`,
-* the algorithm + autotuner + convergence thresholds + initial weights,
+* the algorithm + autotuner + initial weights,
 * the discretizer scheme (class + hold type + ODE solver),
 * the state/control scaling matrices `inv_S_x` / `inv_S_u`,
-* the fixed batch size `B` (the artifact is exported at one `B`), and
+* the fixed batch size `B` (the artifact is exported at one `B`),
+* the per-parameter shared/batched split (baked into the `jax.vmap` program),
+  and
 * the JAX version (exported artifacts are not guaranteed to survive an
   incompatible jax bump).
 
-Change any of these — backend, a tolerance, `k_max`, a state bound — and you get
-a different cache path, so a stale artifact is never reused.
+Change any of these — backend, a penalty weight, a state bound, which
+parameters are batched — and you get a different cache path, so a stale
+artifact is never reused. Everything the `algorithm` dict can reach — convergence
+thresholds, `max_iters`, weights, autotuner hyperparameters — is deliberately
+*not* in the key: those are runtime inputs on the state pytree, so one
+artifact serves every setting of them.
 
 `examples/abstract/brachistochrone_batched.py` shows both paths side by side;
 re-running it is itself the cross-process demo.
@@ -158,10 +285,15 @@ problem.solve_jax(
     parameters=None,     # parameters dict for this solve; falls back to
                          #     ``self._parameters``
     *,
-    max_iters=None,      # SCP iteration cap; non-default rebuilds the
-                         #     cached ``lax.while_loop`` closure (one extra
-                         #     trace; subsequent calls at the same cap hit
-                         #     the cache)
+    x_guess=None,        # (N, n_x) state trajectory warm-start
+    u_guess=None,        # (N, n_u) control trajectory warm-start
+    max_iters=None,      # SCP iteration cap; sugar for algorithm={"k_max": ...};
+                         #     falls back to the configured k_max
+    algorithm=None,      # any AlgorithmState field by name (ep_tr, lam_prox,
+                         #     ...) — the constructor's algorithm-dict names;
+                         #     values match the field shape or fill it from a
+                         #     scalar. Runtime inputs on the state pytree — no
+                         #     value forces a retrace
 )
 ```
 

{openscvx-0.5.3.dev23 → openscvx-0.5.3.dev25}/examples/abstract/brachistochrone_batched.py

@@ -6,7 +6,7 @@ starting x-coordinate — and contrasts the two batching entry points:
 * ``jax.vmap(problem.solve_jax)`` — the caller owns the batch axis. Maximally
   idiomatic, composes with ``grad`` / ``scan``, but in-process JIT only: every
   fresh process re-traces and re-compiles the whole SCP loop.
-* ``problem.solve_batched(x0_stack, xf_stack)`` — the library owns the batch
+* ``problem.solve_batched(x_initial=...)`` — the library owns the batch
   axis, so the whole vmapped loop is a single function that ``jax.export`` can
   serialize. Under ``save_compiled=True`` it is written to the solver cache on
   the first run and deserialized on later runs, skipping that compile. Reach
@@ -140,16 +140,28 @@ if __name__ == "__main__":
     # The same batched solve, but ``solve_batched`` owns the vmap so the whole
     # loop is one exportable artifact. Under ``save_compiled=True`` it lands in
     # the solver cache on the first run; re-run this script (a fresh process) to
-    # see it deserialized instead of recompiled. ``xf`` is shared across the
-    # batch, so broadcast the default terminal pin to the same leading axis.
+    # see it deserialized instead of recompiled. One rule: ``x_initial`` is
+    # ``(B, n_x)`` — one extra leading axis — so it is batched; the terminal
+    # pin is omitted, so every element shares the default.
     export_problem = build_problem()
     export_problem.settings.sim.save_compiled = True
     export_problem.initialize()
 
-    x_term_stack = jnp.broadcast_to(export_problem.state.x_term_pin, x_initial_stack.shape)
-    batched = export_problem.solve_batched(x_initial_stack, x_term_stack)
+    batched = export_problem.solve_batched(x_initial=x_initial_stack)
 
     print(f"\nsolve_batched (save_compiled=True) over {x_initial_stack.shape[0]} ICs:")
     print(f"  result.x.shape:   {batched.x.shape}")
     print(f"  result.t_final:   {np.asarray(batched.t_final).reshape(-1)}")
     print(f"  result.converged: {np.asarray(batched.converged)}")
+
+    # Algorithm knobs batch by the same rule, through the `algorithm` dict —
+    # the same names as the Problem constructor's algorithm config. ep_tr is
+    # a scalar field, so a (B,) vector sweeps the convergence tolerance per
+    # element — same artifact, no recompile, since the knobs are runtime
+    # inputs on the state pytree. The batch size is never passed: it is read
+    # off the array's leading axis (4 tolerances -> B = 4).
+    sweep = export_problem.solve_batched(
+        algorithm={"ep_tr": jnp.logspace(-4, -1, 4), "ep_vc": 1e-7}
+    )
+    print(f"  ep_tr sweep t_final:   {np.asarray(sweep.t_final).reshape(-1)}")
+    print(f"  ep_tr sweep converged: {np.asarray(sweep.converged)}")

{openscvx-0.5.3.dev23 → openscvx-0.5.3.dev25}/examples/rocket/6DoF_pdg_batched_ic.py

@@ -2,7 +2,7 @@
 
 Runs ``N_BATCH`` SCP solves in parallel, each starting from a different random
 initial position drawn uniformly within the position bounds.  Per-batch ICs are
-passed through ``x0_stack``: position is declared ``Fix`` at the symbolic level
+passed through ``x_initial``: position is declared ``Fix`` at the symbolic level
 (so the solver consumes ``x_init_pin``), then each batch row overwrites the
 position slice with the sampled values.  This avoids the per-iteration
 ``(position == parameter).convex()`` constraint while still enforcing the full
@@ -110,7 +110,7 @@ mass.final = [ox.Maximize(1.5)]
 position = ox.State("position", shape=(3,))
 position.max = [10.0, 10.0, 10.0]
 position.min = [-10.0, -10.0, -10.0]
-# Fix so x0_stack pins are consumed; per-batch values overwrite these defaults.
+# Fix so x_initial pins are consumed; per-batch values overwrite these defaults.
 position.initial = [7.5, 4.5, 2.5]
 position.final = [0.0, ox.Free(0.0), ox.Free(0.0)]
 
@@ -193,7 +193,7 @@ constraint_exprs = []
 for st in states:
     constraint_exprs.extend([ox.ctcs(st <= st.max), ox.ctcs(st.min <= st)])
 
-# Initial position via x0_stack (Fix BC); terminal lateral via convex constraint.
+# Initial position via x_initial pins (Fix BC); terminal lateral via convex constraint.
 constraint_exprs.append((position[1:3] == final_position).convex().at([N - 1]))
 
 constraint_exprs.append(ox.ctcs(1.0 * (mass - m_dry) >= 0))
@@ -414,15 +414,14 @@ if __name__ == "__main__":
     print(f"Sampled {N_BATCH} initial positions  (alt ∈ [3, 9] m, lat ∈ [-7, 7] m)")
 
     # ── Compile + solve ───────────────────────────────────────────────────────
+    # x_initial is (B, n_x) — one extra leading axis — so it is batched; the
+    # terminal pin is omitted, so every element shares the default.
     default_init = np.asarray(problem.state.x_init_pin)
-    x0_stack = np.broadcast_to(default_init, (N_BATCH, default_init.shape[0])).copy()
-    x0_stack[:, POSITION_SLICE] = ic_batch
-    x0_stack = jnp.asarray(x0_stack)
-    xf_pin = np.asarray(problem.state.x_term_pin)
-    xf_stack = jnp.broadcast_to(xf_pin, (N_BATCH, xf_pin.shape[0]))
+    x_initial = np.broadcast_to(default_init, (N_BATCH, default_init.shape[0])).copy()
+    x_initial[:, POSITION_SLICE] = ic_batch
 
     print("Compiling and running solve_batched …")
-    results = problem.solve_batched(x0_stack=x0_stack, xf_stack=xf_stack)
+    results = problem.solve_batched(x_initial=jnp.asarray(x_initial))
 
     # ── Post-process: propagate all B solutions through nonlinear dynamics ────
     print("Post-processing (nonlinear propagation) …")

{openscvx-0.5.3.dev23 → openscvx-0.5.3.dev25}/openscvx/_version.py

@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
 commit_id: str | None
 __commit_id__: str | None
 
-__version__ = version = '0.5.3.dev23'
-__version_tuple__ = version_tuple = (0, 5, 3, 'dev23')
+__version__ = version = '0.5.3.dev25'
+__version_tuple__ = version_tuple = (0, 5, 3, 'dev25')
 
-__commit_id__ = commit_id = 'g107a1d50b'
+__commit_id__ = commit_id = 'ge4b992a50'

{openscvx-0.5.3.dev23 → openscvx-0.5.3.dev25}/openscvx/algorithms/__init__.py

@@ -37,6 +37,8 @@ from typing import Annotated, Any, Dict, List, Optional, Union
 from pydantic import BaseModel, ConfigDict, Field, field_validator
 
 from .autotuner import (
+    AcceptanceRatioAutotuner,
+    AcceptanceRatioHyper,
     AdaptiveProximalWeight,
     AdaptiveProximalWeightSpec,
     AugmentedLagrangian,
@@ -52,7 +54,9 @@ from .base import (
     AlgorithmHistory,
     AlgorithmState,
     AutotuningBase,
+    CandidateIterate,
     DiscretizationResult,
+    HyperParams,
     adaptive_state_code_to_str,
 )
 from .optimization_results import OptimizationResults
@@ -137,6 +141,7 @@ __all__ = [
     "AlgorithmHistory",
     "AlgorithmState",
     "AdaptiveStateCode",
+    "CandidateIterate",
     "adaptive_state_code_to_str",
     "DiscretizationResult",
     "Weights",
@@ -145,6 +150,9 @@ __all__ = [
     # PTR algorithm
     "PenalizedTrustRegion",
     "AutotuningBase",
+    "HyperParams",
+    "AcceptanceRatioAutotuner",
+    "AcceptanceRatioHyper",
     "AugmentedLagrangian",
     "AdaptiveProximalWeight",
     "ConstantProximalWeight",

{openscvx-0.5.3.dev23 → openscvx-0.5.3.dev25}/openscvx/algorithms/autotuner/__init__.py

@@ -1,11 +1,14 @@
 """SCP weight autotuning strategies."""
 
+from .acceptance_ratio import AcceptanceRatioAutotuner, AcceptanceRatioHyper
 from .adaptive_proximal_weight import AdaptiveProximalWeight, AdaptiveProximalWeightSpec
 from .augmented_lagrangian import AugmentedLagrangian, AugmentedLagrangianSpec
 from .constant_proximal_weight import ConstantProximalWeight, ConstantProximalWeightSpec
 from .ramp_proximal_weight import RampProximalWeight, RampProximalWeightSpec
 
 __all__ = [
+    "AcceptanceRatioAutotuner",
+    "AcceptanceRatioHyper",
     "AdaptiveProximalWeight",
     "AdaptiveProximalWeightSpec",
     "AugmentedLagrangian",

openscvx-0.5.3.dev25/openscvx/algorithms/autotuner/acceptance_ratio.py

@@ -0,0 +1,245 @@
+"""Acceptance-ratio autotuner family.
+
+The acceptance-ratio autotuners share one decision rule: an acceptance ratio
+:math:`\\rho` between the predicted and actual reduction of the nonlinear
+objective drives a four-bucket update of the trust-region weight ``lam_prox``
+(``REJECT`` / ``ACCEPT_HIGHER`` / ``ACCEPT_CONSTANT`` / ``ACCEPT_LOWER``). What
+differs between members is the *multiplier* rule — how the virtual-control and
+virtual-buffer weights (``lam_vc`` / ``lam_vb_*``) respond once an iterate is
+accepted.
+
+:class:`AcceptanceRatioAutotuner` owns the shared decision body and exposes a
+single :meth:`AcceptanceRatioAutotuner._update_multipliers` hook for the
+multiplier rule; subclasses supply only that hook.
+:class:`AdaptiveProximalWeight` is the hook-default (multipliers held
+constant); :class:`AugmentedLagrangian` overrides it with the
+constraint-violation update.
+"""
+
+from typing import TYPE_CHECKING, List
+
+import jax
+import jax.numpy as jnp
+
+from openscvx.config import Config
+from openscvx.utils.printing import (
+    Column,
+    Verbosity,
+    color_acceptance_ratio,
+    color_adaptive_state,
+)
+
+from ..base import AdaptiveStateCode, AutotuningBase, HyperParams
+
+if TYPE_CHECKING:
+    from openscvx.lowered import LoweredJaxConstraints
+
+    from ..base import AlgorithmState, CandidateIterate
+
+
+class AcceptanceRatioHyper(HyperParams):
+    """Knobs shared by every acceptance-ratio autotuner.
+
+    ``lam_cost_drop`` is the iteration after which ``lam_cost`` relaxation
+    applies (``state.k > lam_cost_drop``): ``-1`` relaxes from the first
+    iteration, and the default ``lam_cost_relax=1.0`` makes that a no-op.
+    """
+
+    gamma_1: float = 2.0
+    gamma_2: float = 0.5
+    eta_0: float = 1e-2
+    eta_1: float = 1e-1
+    eta_2: float = 0.8
+    lam_prox_min: float = 1e-3
+    lam_prox_max: float = 1e4
+    lam_cost_drop: int = -1
+    lam_cost_relax: float = 1.0
+
+
+class AcceptanceRatioAutotuner(AutotuningBase):
+    """Four-bucket acceptance-ratio update on ``lam_prox``.
+
+    The acceptance ratio :math:`\\rho` (actual over predicted nonlinear
+    reduction) selects one of four buckets — ``REJECT`` / ``ACCEPT_HIGHER`` /
+    ``ACCEPT_CONSTANT`` / ``ACCEPT_LOWER`` — via a ``jnp.where`` cascade so the
+    whole update traces under ``jax.jit``. Subclasses supply the multiplier
+    rule (how ``lam_vc`` / ``lam_vb_*`` respond on acceptance) through
+    :meth:`_update_multipliers`, and must assign an
+    :class:`AcceptanceRatioHyper` (or subclass) instance to ``self.hyper`` in
+    ``__init__`` — the shared body reads its knobs off ``state.hyper``.
+
+    ``update_weights`` is a pure functional update on the
+    :class:`AlgorithmState` pytree; see the base-class contract.
+    """
+
+    COLUMNS: List[Column] = [
+        Column("J_nonlin", "J_nonlin", 8, "{: .1e}", None, Verbosity.STANDARD),
+        Column("J_lin", "J_lin", 8, "{: .1e}", None, Verbosity.STANDARD),
+        Column("pred_reduction", "pred_red", 9, "{: .1e}", min_verbosity=Verbosity.FULL),
+        Column("actual_reduction", "act_red", 9, "{: .1e}", min_verbosity=Verbosity.FULL),
+        Column(
+            "acceptance_ratio",
+            "acc_ratio",
+            9,
+            "{: .2e}",
+            color_acceptance_ratio,
+            Verbosity.STANDARD,
+        ),
+        Column("lam_prox", "lam_prox", 8, "{: .1e}", min_verbosity=Verbosity.FULL),
+        Column("adaptive_state", "Adaptive", 16, "{}", color_adaptive_state, Verbosity.FULL),
+    ]
+
+    def _update_multipliers(
+        self,
+        state: "AlgorithmState",
+        candidate: "CandidateIterate",
+        nodal_constraints: "LoweredJaxConstraints",
+        params: dict,
+        settings: Config,
+        new_lam_prox: jnp.ndarray,
+    ) -> dict:
+        """Return the multiplier-weight updates to apply on an accepted iterate.
+
+        Each returned entry maps an :class:`AlgorithmState` field name
+        (``lam_vc`` / ``lam_vb_nodal`` / ``lam_vb_cross``) to its updated
+        value, computed against the *candidate* trajectory; the shared body
+        gates each by acceptance (a rejected iterate keeps the previous value).
+        The default holds every multiplier constant — return ``{}``.
+        """
+        return {}
+
+    def update_weights(
+        self,
+        state: "AlgorithmState",
+        candidate: "CandidateIterate",
+        nodal_constraints: "LoweredJaxConstraints",
+        settings: Config,
+        params: dict,
+    ) -> "AlgorithmState":
+        """Return the next-iterate state per the acceptance-ratio rules.
+
+        Pure functional update — see class docstring.
+        """
+        candidate_x_prop = candidate.x_prop_plus[1:]
+        nonlin_cost, nonlin_pen, nodal_pen = self.calculate_nonlinear_penalty(
+            candidate_x_prop,
+            candidate.x,
+            candidate.u,
+            state.lam_vc,
+            state.lam_vb_nodal,
+            state.lam_vb_cross,
+            state.lam_cost,
+            nodal_constraints,
+            params,
+            settings,
+        )
+        J_nonlin = nonlin_cost + nonlin_pen + nodal_pen
+
+        # Cost relaxation: when state.k > hyper.lam_cost_drop, scale
+        # state.lam_cost; otherwise reset to the algorithm's initial weight
+        # (carried on the pytree as state.lam_cost_init, broadcast at
+        # from_settings()). Both constants ride the pytree so per-solve
+        # overrides and vmap sweeps reach the traced body. Scalar
+        # lam_cost_relax preserves the user-specified per-state weight ratios.
+        lam_cost_next = jnp.where(
+            state.k > state.hyper.lam_cost_drop,
+            state.lam_cost * state.hyper.lam_cost_relax,
+            state.lam_cost_init,
+        )
+
+        def first_iter(state):
+            # Iter 1: accept unconditionally, leave weights at their init values,
+            # only refresh trajectory + propagation fields.
+            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,
+                J_nonlin=J_nonlin,
+                adaptive_state_code=jnp.asarray(int(AdaptiveStateCode.INITIAL), dtype=jnp.int32),
+            )
+
+        def later_iter(state):
+            # Recompute the previous iterate's J_nonlin from the pytree fields
+            # (state.x/state.u were the *previous* accepted iterate).
+            prev_x_prop = state.x_prop_plus[1:]
+            prev_cost, prev_pen, prev_nodal_pen = self.calculate_nonlinear_penalty(
+                prev_x_prop,
+                state.x,
+                state.u,
+                state.lam_vc,
+                state.lam_vb_nodal,
+                state.lam_vb_cross,
+                state.lam_cost,
+                nodal_constraints,
+                params,
+                settings,
+            )
+            prev_J_nonlin = prev_cost + prev_pen + prev_nodal_pen
+
+            actual = prev_J_nonlin - J_nonlin
+            predicted = prev_J_nonlin - candidate.J_lin
+            # If predicted reduction is exactly zero, force the reject bucket
+            # (rho = -inf) deterministically instead of raising.
+            safe_pred = jnp.where(predicted == 0.0, 1.0, predicted)
+            rho = jnp.where(predicted == 0.0, -jnp.inf, actual / safe_pred)
+
+            is_reject = rho < state.hyper.eta_0
+            is_accept_higher = (rho >= state.hyper.eta_0) & (rho < state.hyper.eta_1)
+            is_accept_constant = (rho >= state.hyper.eta_1) & (rho < state.hyper.eta_2)
+            # is_accept_lower implicit (else)
+            accepted = ~is_reject
+
+            # Compute both lam_prox candidates and gate.
+            lp_higher = jnp.minimum(state.hyper.lam_prox_max, state.hyper.gamma_1 * state.lam_prox)
+            lp_lower = jnp.maximum(state.hyper.lam_prox_min, state.hyper.gamma_2 * state.lam_prox)
+            new_lam_prox = jnp.where(
+                is_reject | is_accept_higher,
+                lp_higher,
+                jnp.where(is_accept_constant, state.lam_prox, lp_lower),
+            )
+
+            # Multiplier (virtual-control / virtual-buffer) updates: computed
+            # against the *candidate* trajectory and gated by `accepted`. Reject
+            # keeps the previous values (so the next subproblem doesn't see a
+            # bumped penalty that wasn't earned). The default hook returns {} —
+            # multipliers held constant.
+            multiplier_updates = self._update_multipliers(
+                state, candidate, nodal_constraints, params, settings, new_lam_prox
+            )
+            gated_multipliers = {
+                name: jnp.where(accepted, value, getattr(state, name))
+                for name, value in multiplier_updates.items()
+            }
+
+            code = jnp.where(
+                is_reject,
+                jnp.int32(AdaptiveStateCode.REJECT),
+                jnp.where(
+                    is_accept_higher,
+                    jnp.int32(AdaptiveStateCode.ACCEPT_HIGHER),
+                    jnp.where(
+                        is_accept_constant,
+                        jnp.int32(AdaptiveStateCode.ACCEPT_CONSTANT),
+                        jnp.int32(AdaptiveStateCode.ACCEPT_LOWER),
+                    ),
+                ),
+            )
+
+            return state.replace(
+                x=jnp.where(accepted, candidate.x, state.x),
+                u=jnp.where(accepted, candidate.u, state.u),
+                x_prop=jnp.where(accepted, candidate.x_prop, state.x_prop),
+                x_prop_plus=jnp.where(accepted, candidate.x_prop_plus, state.x_prop_plus),
+                lam_prox=new_lam_prox,
+                lam_cost=lam_cost_next,
+                J_nonlin=jnp.where(accepted, J_nonlin, state.J_nonlin),
+                predicted_reduction=predicted,
+                actual_reduction=actual,
+                acceptance_ratio=rho,
+                adaptive_state_code=code,
+                **gated_multipliers,
+            )
+
+        return jax.lax.cond(state.k == 1, first_iter, later_iter, state)

openscvx-0.5.3.dev25/openscvx/algorithms/autotuner/adaptive_proximal_weight.py

@@ -0,0 +1,74 @@
+"""Adaptive proximal-weight autotuner.
+
+Same acceptance-ratio decision logic as the rest of the family for
+``lam_prox``, but ``lam_vc`` and ``lam_vb_*`` are held constant at their
+current values — the default multiplier hook.
+"""
+
+from typing import Literal
+
+from pydantic import BaseModel, ConfigDict
+
+from .acceptance_ratio import AcceptanceRatioAutotuner, AcceptanceRatioHyper
+
+
+class AdaptiveProximalWeight(AcceptanceRatioAutotuner):
+    """PTR-style proximal adaptation with fixed virtual-penalty weights.
+
+    Inherits the four-bucket acceptance-ratio update on ``lam_prox`` from
+    :class:`~openscvx.algorithms.autotuner.acceptance_ratio.AcceptanceRatioAutotuner`
+    and keeps the default multiplier hook, so the virtual-control and
+    virtual-buffer weights are carried unchanged.
+
+    ``update_weights`` is a pure functional update on the
+    :class:`AlgorithmState` pytree; see the base-class contract.
+    """
+
+    def __init__(
+        self,
+        gamma_1: float = 2.0,
+        gamma_2: float = 0.5,
+        eta_0: float = 1e-2,
+        eta_1: float = 1e-1,
+        eta_2: float = 0.8,
+        lam_prox_min: float = 1e-3,
+        lam_prox_max: float = 1e4,
+        lam_cost_drop: int = -1,
+        lam_cost_relax: float = 1.0,
+    ):
+        self.hyper = AcceptanceRatioHyper(
+            gamma_1=gamma_1,
+            gamma_2=gamma_2,
+            eta_0=eta_0,
+            eta_1=eta_1,
+            eta_2=eta_2,
+            lam_prox_min=lam_prox_min,
+            lam_prox_max=lam_prox_max,
+            lam_cost_drop=lam_cost_drop,
+            lam_cost_relax=lam_cost_relax,
+        )
+
+
+# =============================================================================
+# Pydantic spec for dict / YAML validation
+# =============================================================================
+
+
+class AdaptiveProximalWeightSpec(BaseModel):
+    """Validates AdaptiveProximalWeight configuration from dict/YAML input."""
+
+    type: Literal["AdaptiveProximalWeight"] = "AdaptiveProximalWeight"
+    gamma_1: float = 2.0
+    gamma_2: float = 0.5
+    eta_0: float = 1e-2
+    eta_1: float = 1e-1
+    eta_2: float = 0.8
+    lam_prox_min: float = 1e-3
+    lam_prox_max: float = 1e4
+    lam_cost_drop: int = -1
+    lam_cost_relax: float = 1.0
+
+    model_config = ConfigDict(extra="forbid")
+
+    def build(self) -> AdaptiveProximalWeight:
+        return AdaptiveProximalWeight(**self.model_dump(exclude={"type"}, exclude_unset=True))