causing 2.4.3__tar.gz → 2.4.5__tar.gz

{causing-2.4.3 → causing-2.4.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: causing
-Version: 2.4.3
+Version: 2.4.5
 Summary: Causing: CAUSal INterpretation using Graphs
 Home-page: https://github.com/realrate/Causing
 Author: Dr. Holger Bartel

{causing-2.4.3 → causing-2.4.5}/causing/graph.py

@@ -193,6 +193,10 @@ def create_graphs(graphs: Iterable[networkx.DiGraph], output_dir: Path, **kwargs
         print("Create", filename)
         dot_str = graph_to_dot(g, **kwargs)
         save_graph(output_dir / filename, dot_str)
+        # save the dot string
+        dot_path = output_dir / f"IME_{g.graph['id']}.dot"
+        with open(dot_path, "w") as f:
+            f.write(dot_str)
 
 
 def remove_node_keep_edges(graph, node):

{causing-2.4.3 → causing-2.4.5}/causing/model.py

@@ -51,65 +51,56 @@ class Model:
     def compute(
         self,
         xdat: np.array,
-        # fix a yval
         fixed_yval: np.array = None,
         fixed_yind: int = None,
-        # fix an arbitrary node going into a yval
         fixed_from_ind: int = None,
         fixed_to_yind: int = None,
         fixed_vals: list = None,
-        # override default parameter values
         parameters: dict[str, float] = {},
     ) -> np.array:
-        """Compute y values for given x values
-
+        """Compute y values for given x values (Optimized Version)
         xdat: m rows, tau columns
         returns: n rows, tau columns
         """
+
         assert xdat.ndim == 2, f"xdat must be m*tau (is {xdat.ndim}-dimensional)"
         assert xdat.shape[0] == self.mdim, f"xdat must be m*tau (is {xdat.shape})"
         tau = xdat.shape[1]
         parameters = self.parameters | parameters
 
-        yhat = np.array([[float("nan")] * tau] * len(self.yvars))
+        # Use np.full for clarity and potential small performance gain
+        yhat = np.full((self.ndim, tau), np.nan)
+
         for i, eq in enumerate(self._model_lam):
+
             if fixed_yind == i:
                 yhat[i, :] = fixed_yval
             else:
-                eq_inputs = np.array(
-                    [[*xval, *yval] for xval, yval in zip(xdat.T, yhat.T)]
-                )
+                # 1. Build inputs (Vectorized)
+                # Use np.vstack for efficient vertical stacking.
+                # `yhat` will have NaNs for unsolved variables, which is correct.
+                eq_inputs = np.vstack([xdat, yhat])
+
+                # 2. Apply fixed values if needed (Vectorized)
                 if fixed_to_yind == i:
-                    eq_inputs[:, fixed_from_ind] = fixed_vals
+                    # Directly modify the correct "row" in the input matrix.
+                    # This is much cleaner and faster.
+                    eq_inputs[fixed_from_ind, :] = fixed_vals
 
+                # 3. Evaluate equation (Vectorized)
                 try:
-                    # print(f"Comuting variable: {self.yvars[i]}")
-                    # yhat[i] = np.array(
-                    #     [eq(*eq_in, *parameters.values()) for eq_in in eq_inputs],
-                    #     dtype=np.float64,
-                    # )
                     np.seterr(under="ignore")
-                    computed_yvars = []
-                    for eq_in in eq_inputs:
-                        computed_yvars.append(eq(*eq_in, *parameters.values()))
-
-                    yhat[i] = np.array(
-                        computed_yvars,
-                        dtype=np.float64,
-                    )
+
+                    # This is the core optimization. We unpack the rows of `eq_inputs`
+                    # as separate arguments into the lambdified function. NumPy will
+                    # then compute the results for all `tau` columns at once.
+                    yhat[i, :] = eq(*eq_inputs, *parameters.values())
+
                 except Exception as e:
-                    # for eq_in in eq_inputs:
-                    #     print("--", self.yvars[i])
-                    #     for var, val in zip(
-                    #         self.vars + list(parameters.keys()),
-                    #         list(eq_in) + list(parameters.values()),
-                    #     ):
-                    #         print(var, "=", val)
-                    #     eq(*eq_in, *parameters.values())
                     raise NumericModelError(
                         f"Failed to compute model value for yvar {self.yvars[i]}: {e}"
                     ) from e
-        assert yhat.shape == (self.ndim, tau)
+
         return yhat
 
     def calc_effects(self, xdat: np.array, xdat_mean=None, yhat_mean=None):
@@ -198,10 +189,25 @@ class Model:
 
     @cached_property
     def _model_lam(self) -> Iterable[Callable]:
-        return [
-            sympy.lambdify(self.vars + list(self.parameters), eq)
-            for eq in self.equations
-        ]
+        """Create lambdified equations with NumPy-compatible functions."""
+        lambdas = []
+        ordered_vars = self.vars + list(self.parameters.keys())
+
+        # Define placeholder for vectorized max function
+        vectorized_max = sympy.Function("vectorized_max")
+
+        # Define custom translation mapping
+        custom_modules = [{"vectorized_max": np.maximum}, "numpy"]
+
+        for i, eq in enumerate(self.equations):
+            # Replace sympy.Max with our placeholder
+            fixed_eq = eq.subs(sympy.Max, vectorized_max)
+
+            # Lambdify with custom NumPy mapping
+            lam = sympy.lambdify(ordered_vars, fixed_eq, modules=custom_modules)
+            lambdas.append(lam)
+
+        return lambdas
 
     @cached_property
     def final_ind(self):

{causing-2.4.3 → causing-2.4.5}/causing.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: causing
-Version: 2.4.3
+Version: 2.4.5
 Summary: Causing: CAUSal INterpretation using Graphs
 Home-page: https://github.com/realrate/Causing
 Author: Dr. Holger Bartel

{causing-2.4.3 → causing-2.4.5}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r", encoding="utf-8") as fh:
 
 setuptools.setup(
     name="causing",
-    version="2.4.3",
+    version="2.4.5",
     author="Dr. Holger Bartel",
     author_email="holger.bartel@realrate.ai",
     description="Causing: CAUSal INterpretation using Graphs",