pyepo 2.2.3__tar.gz → 2.2.4__tar.gz

{pyepo-2.2.3 → pyepo-2.2.4}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: pyepo
-Version: 2.2.3
-Summary: PyTorch-based End-to-End Predict-then-Optimize Tool
+Version: 2.2.4
+Summary: PyTorch/JAX-based End-to-End Predict-then-Optimize Tool
 Author-email: Bo Tang <bolucas.tang@mail.utoronto.ca>
 License: MIT
 Project-URL: Homepage, https://github.com/khalil-research/PyEPO
@@ -9,7 +9,7 @@ Project-URL: Documentation, https://khalil-research.github.io/PyEPO
 Project-URL: Repository, https://github.com/khalil-research/PyEPO
 Project-URL: Issues, https://github.com/khalil-research/PyEPO/issues
 Project-URL: Paper, https://link.springer.com/article/10.1007/s12532-024-00255-x
-Keywords: predict-then-optimize,end-to-end,decision-focused learning,optimization,deep learning,pytorch,linear programming,integer programming
+Keywords: predict-then-optimize,end-to-end,decision-focused learning,optimization,deep learning,pytorch,jax,linear programming,integer programming
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10

{pyepo-2.2.3 → pyepo-2.2.4}/pyepo/model/bases.py

@@ -16,7 +16,6 @@ from __future__ import annotations
 
 import math
 from collections import defaultdict
-from copy import deepcopy
 from itertools import combinations
 from numbers import Integral, Real
 from typing import TYPE_CHECKING
@@ -105,9 +104,6 @@ class shortestPathBase(optModel):
         self.arcs = _get_grid_arcs(self.grid)
         super().__init__(*args, **kwargs)
 
-    def get_config(self) -> dict:
-        return {**super().get_config(), "grid": self.grid}
-
     @property
     def num_cost(self) -> int:
         return len(self.arcs)
@@ -208,14 +204,6 @@ class portfolioBase(optModel):
             raise ValueError("gamma must be greater than or equal to zero.")
         super().__init__(*args, **kwargs)
 
-    def get_config(self) -> dict:
-        return {
-            **super().get_config(),
-            "num_assets": self.num_assets,
-            "covariance": self.covariance.copy(),
-            "gamma": self.gamma,
-        }
-
     @property
     def num_cost(self) -> int:
         return self.num_assets
@@ -263,9 +251,6 @@ class tspABBase(optModel):
         self._extra_constrs: list = []
         super().__init__(*args, **kwargs)
 
-    def get_config(self) -> dict:
-        return {**super().get_config(), "num_nodes": self.num_nodes}
-
     @property
     def num_cost(self) -> int:
         # use edges; backend's self.x has 2*num_edges directed Vars
@@ -381,15 +366,6 @@ class vrpABBase(optModel):
         self._extra_constrs: list = []
         super().__init__(*args, **kwargs)
 
-    def get_config(self) -> dict:
-        return {
-            **super().get_config(),
-            "num_nodes": self.num_nodes,
-            "demands": deepcopy(self.demands),
-            "capacity": self.capacity,
-            "num_vehicle": self.num_vehicle,
-        }
-
     @property
     def num_cost(self) -> int:
         # one predicted cost per undirected edge

{pyepo-2.2.3 → pyepo-2.2.4}/pyepo/model/grb/grbmodel.py

@@ -171,7 +171,8 @@ class optGrbModel(optModel):
         else:
             # LinExpr(coeffs, vars) builds the affine expression in one C call
             vars_list = new_model._vars_list
-            assert vars_list is not None
+            if vars_list is None:
+                raise RuntimeError("Gurobi variable list is unavailable.")
             expr = gp.LinExpr(coefs_np.tolist(), vars_list) <= rhs
             new_model._model.addConstr(expr)
         # track for replay on relax

{pyepo-2.2.3 → pyepo-2.2.4}/pyepo/model/grb/tsp.py

@@ -178,9 +178,6 @@ class tspDFJModel(tspABModel):
         self._recycled_keys: set = set()
         super().__init__(num_nodes, *args, **kwargs)
 
-    def get_config(self) -> dict:
-        return {**super().get_config(), "recycle_cuts": self.recycle_cuts}
-
     def _getModel(self) -> tuple:
         """
         A method to build Gurobi model

{pyepo-2.2.3 → pyepo-2.2.4}/pyepo/model/grb/vrp.py

@@ -70,9 +70,6 @@ class vrpRCIModel(vrpABModel):
         self._recycled_keys: set = set()
         super().__init__(num_nodes, demands, capacity, num_vehicle)
 
-    def get_config(self) -> dict:
-        return {**super().get_config(), "recycle_cuts": self.recycle_cuts}
-
     def _getModel(self) -> tuple:
         """
         A method to build Gurobi model

{pyepo-2.2.3 → pyepo-2.2.4}/pyepo/model/opt.py

@@ -5,6 +5,8 @@ Abstract optimization model
 
 from __future__ import annotations
 
+import functools
+import inspect
 from abc import ABC, abstractmethod
 from copy import deepcopy
 from dataclasses import dataclass
@@ -41,6 +43,27 @@ class ModelSpec:
         return self.model_type.from_config(self._config)
 
 
+def _capture_init_config(init, args, kwargs) -> dict:
+    """Flatten a constructor call into keyword arguments that rebuild the model."""
+    sig = inspect.signature(init)
+    bound = sig.bind(None, *args, **kwargs)
+    config = {}
+    for i, (name, value) in enumerate(bound.arguments.items()):
+        # the first bound argument is self
+        if i == 0:
+            continue
+        kind = sig.parameters[name].kind
+        # **kwargs: merge captured keywords in directly
+        if kind is inspect.Parameter.VAR_KEYWORD:
+            config.update(deepcopy(value))
+        # nameless positionals cannot replay by keyword; override get_config to keep them
+        elif kind in (inspect.Parameter.VAR_POSITIONAL, inspect.Parameter.POSITIONAL_ONLY):
+            continue
+        else:
+            config[name] = deepcopy(value)
+    return config
+
+
 class optModel(ABC):
     """
     Abstract base class for predict-then-optimize models.
@@ -53,11 +76,6 @@ class optModel(ABC):
     and MPAX (``optMpaxModel``); subclass ``optModel`` directly to integrate
     any other solver or algorithm.
 
-    Models that take constructor arguments should override ``get_config`` and
-    cooperatively merge ``super().get_config()``. The resulting configuration
-    powers ``rebuild()``, multiprocessing workers, and sklearn scorers without
-    inspecting constructor signatures or runtime solver state.
-
     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
@@ -73,6 +91,22 @@ class optModel(ABC):
     arcs: list
     _cost_vars: list
 
+    def __init_subclass__(cls, **kwargs) -> None:
+        super().__init_subclass__(**kwargs)
+        # only wrap a subclass that defines its own __init__
+        if "__init__" not in cls.__dict__:
+            return
+        user_init = cls.__init__
+
+        @functools.wraps(user_init)
+        def _init_capturing(self, *args, **kwargs):
+            # record only the outermost call; nested super().__init__ leaves it intact
+            if "_init_config" not in self.__dict__:
+                self._init_config = _capture_init_config(user_init, args, kwargs)
+            user_init(self, *args, **kwargs)
+
+        cls.__init__ = _init_capturing
+
     def __init__(self) -> None:
         # Cache for models whose solver variables do not map one-to-one to
         # predicted costs (for example directed TSP/VRP formulations).
@@ -86,8 +120,8 @@ class optModel(ABC):
         return "optModel " + self.__class__.__name__
 
     def get_config(self) -> dict:
-        """Return the explicit constructor configuration for this model."""
-        return {}
+        """Return the constructor configuration for this model."""
+        return deepcopy(self.__dict__.get("_init_config", {}))
 
     @classmethod
     def from_config(cls, config: dict) -> Self:

{pyepo-2.2.3 → pyepo-2.2.4}/pyepo.egg-info/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: pyepo
-Version: 2.2.3
-Summary: PyTorch-based End-to-End Predict-then-Optimize Tool
+Version: 2.2.4
+Summary: PyTorch/JAX-based End-to-End Predict-then-Optimize Tool
 Author-email: Bo Tang <bolucas.tang@mail.utoronto.ca>
 License: MIT
 Project-URL: Homepage, https://github.com/khalil-research/PyEPO
@@ -9,7 +9,7 @@ Project-URL: Documentation, https://khalil-research.github.io/PyEPO
 Project-URL: Repository, https://github.com/khalil-research/PyEPO
 Project-URL: Issues, https://github.com/khalil-research/PyEPO/issues
 Project-URL: Paper, https://link.springer.com/article/10.1007/s12532-024-00255-x
-Keywords: predict-then-optimize,end-to-end,decision-focused learning,optimization,deep learning,pytorch,linear programming,integer programming
+Keywords: predict-then-optimize,end-to-end,decision-focused learning,optimization,deep learning,pytorch,jax,linear programming,integer programming
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10

{pyepo-2.2.3 → pyepo-2.2.4}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pyepo"
-version = "2.2.3"
-description = "PyTorch-based End-to-End Predict-then-Optimize Tool"
+version = "2.2.4"
+description = "PyTorch/JAX-based End-to-End Predict-then-Optimize Tool"
 readme = { file = "README.md", content-type = "text/markdown" }
 license = { text = "MIT" }
 authors = [{ name = "Bo Tang", email = "bolucas.tang@mail.utoronto.ca" }]
@@ -17,6 +17,7 @@ keywords = [
     "optimization",
     "deep learning",
     "pytorch",
+    "jax",
     "linear programming",
     "integer programming",
 ]

{pyepo-2.2.3 → pyepo-2.2.4}/tests/test_10_utils.py

@@ -43,6 +43,24 @@ class ConfigModel(optModel):
         return np.zeros(len(self.values)), 0.0
 
 
+class AutoConfigModel(optModel):
+    """Solver-free model that relies on optModel's automatic config capture."""
+
+    def __init__(self, values, **kwargs):
+        self.values = values
+        self.kwargs = kwargs
+        super().__init__()
+
+    def _getModel(self):
+        return None, list(range(len(self.values)))
+
+    def setObj(self, c):
+        self.cost = np.asarray(c)
+
+    def solve(self):
+        return np.zeros(len(self.values)), 0.0
+
+
 # ============================================================
 # unionFind (pure)
 # ============================================================
@@ -222,6 +240,28 @@ class TestModelSpec:
         np.testing.assert_array_equal(sol, [0.0, 0.0, 0.0])
         assert obj == 0.0
 
+    def test_auto_config_snapshots_constructor_inputs(self):
+        values = [1, 2, 3]
+        nested = {"tag": ["x"]}
+        model = AutoConfigModel(values, nested=nested)
+        values[0] = 99
+        nested["tag"][0] = "changed"
+
+        rebuilt = model.rebuild()
+
+        assert rebuilt.values == [1, 2, 3]
+        assert rebuilt.kwargs == {"nested": {"tag": ["x"]}}
+
+    def test_auto_config_export_is_independent(self):
+        model = AutoConfigModel([1, 2, 3], nested={"tag": ["x"]})
+        config = model.get_config()
+        config["values"][0] = 99
+        config["nested"]["tag"][0] = "changed"
+
+        assert model.get_config()["values"] == [1, 2, 3]
+        assert model.get_config()["nested"] == {"tag": ["x"]}
+        assert model.rebuild().kwargs == {"nested": {"tag": ["x"]}}
+
 
 # ============================================================
 # getArgs (needs a real optModel)