added python bindings for edge features

Author: Simon Klix

plugins/machine_learning/python/python_bindings.cpp

@@ -30,6 +30,7 @@ namespace hal
 #endif    // ifdef PYBIND11_MODULE
 
         // Define submodules for namespaces
+        py::module py_edge_feature      = m.def_submodule("edge_feature");
         py::module py_gate_feature      = m.def_submodule("gate_feature");
         py::module py_gate_pair_feature = m.def_submodule("gate_pair_feature");
         py::module py_gate_pair_label   = m.def_submodule("gate_pair_label");
@@ -280,6 +281,12 @@ Get the gates of the context.
             :type: tuple[list[int], list[int]]
         )");
 
+        py_netlist_graph.def_readwrite("edge_features", &machine_learning::NetlistGraph::edge_features, R"(
+            List of features corresponding to the edge list
+
+            :type: list[list[FEATURE_TYPE]]
+        )");
+
         py_netlist_graph.def_readwrite("direction", &machine_learning::NetlistGraph::direction, R"(
             Direction of the graph.
 
@@ -2384,6 +2391,173 @@ Convert the NetNameKeyWord labeler to a string.
 
 :returns: The string representation.
 :rtype: str
+)");
+
+        py::class_<machine_learning::edge_feature::EdgeFeature, RawPtrWrapper<machine_learning::edge_feature::EdgeFeature>> py_edge_feature_class(py_edge_feature,
+                                                                                                                                                  "EdgeFeature",
+                                                                                                                                                  R"(
+Abstract base class representing a feature for an edge between two endpoints.
+)");
+
+        py_edge_feature_class.def(
+            "calculate_feature",
+            [](machine_learning::edge_feature::EdgeFeature& self, machine_learning::Context& ctx, const Endpoint* source, const Endpoint* destination) -> std::optional<std::vector<FEATURE_TYPE>> {
+                auto res = self.calculate_feature(ctx, source, destination);
+                if (res.is_ok())
+                {
+                    return res.get();
+                }
+                else
+                {
+                    log_error("python_context", "Error in EdgeFeature::calculate_feature:\n{}", res.get_error().get());
+                    return std::nullopt;
+                }
+            },
+            py::arg("ctx"),
+            py::arg("source"),
+            py::arg("destination"),
+            R"(
+Calculate the feature vector for the given edge.
+
+:param hal_py.Context ctx: The context.
+:param hal_py.Endpoint source: The source endpoint.
+:param hal_py.Endpoint destination: The destination endpoint.
+:returns: The feature vector or None.
+:rtype: list[hal_py.FEATURE_TYPE] or None
+)");
+
+        py_edge_feature_class.def("to_string",
+                                  &machine_learning::edge_feature::EdgeFeature::to_string,
+                                  R"(
+Get a string representation of this feature.
+
+:returns: A string.
+:rtype: str
+)");
+
+        py::class_<machine_learning::edge_feature::PinTypesOnehot, RawPtrWrapper<machine_learning::edge_feature::PinTypesOnehot>, machine_learning::edge_feature::EdgeFeature> py_pin_types_onehot(
+            py_edge_feature,
+            "PinTypesOnehot",
+            R"(
+One-hot encoding feature for pin types on an edge.
+)");
+
+        py_pin_types_onehot.def(py::init<>(), R"(
+Construct a PinTypesOnehot feature.
+)");
+
+        py_pin_types_onehot.def(
+            "calculate_feature",
+            [](machine_learning::edge_feature::PinTypesOnehot& self, machine_learning::Context& ctx, const Endpoint* source, const Endpoint* destination) -> std::optional<std::vector<FEATURE_TYPE>> {
+                auto res = self.calculate_feature(ctx, source, destination);
+                if (res.is_ok())
+                {
+                    return res.get();
+                }
+                else
+                {
+                    log_error("python_context", "Error in PinTypesOnehot::calculate_feature:\n{}", res.get_error().get());
+                    return std::nullopt;
+                }
+            },
+            py::arg("ctx"),
+            py::arg("source"),
+            py::arg("destination"),
+            R"(
+One-hot encoding feature vector for pin types.
+
+:param hal_py.Context ctx: The context.
+:param hal_py.Endpoint source: The source endpoint.
+:param hal_py.Endpoint destination: The destination endpoint.
+:returns: The feature vector or None.
+:rtype: list[hal_py.FEATURE_TYPE] or None
+)");
+
+        py_pin_types_onehot.def("to_string",
+                                &machine_learning::edge_feature::PinTypesOnehot::to_string,
+                                R"(
+Get a string representation.
+
+:returns: A string.
+:rtype: str
+)");
+
+        py::class_<machine_learning::edge_feature::PinDirectionOnehot, RawPtrWrapper<machine_learning::edge_feature::PinDirectionOnehot>, machine_learning::edge_feature::EdgeFeature>
+            py_pin_direction_onehot(py_edge_feature,
+                                    "PinDirectionOnehot",
+                                    R"(
+One-hot encoding feature for pin directions on an edge.
+)");
+
+        py_pin_direction_onehot.def(py::init<>(), R"(
+Construct a PinDirectionOnehot feature.
+)");
+
+        py_pin_direction_onehot.def(
+            "calculate_feature",
+            [](machine_learning::edge_feature::PinDirectionOnehot& self, machine_learning::Context& ctx, const Endpoint* source, const Endpoint* destination)
+                -> std::optional<std::vector<FEATURE_TYPE>> {
+                auto res = self.calculate_feature(ctx, source, destination);
+                if (res.is_ok())
+                {
+                    return res.get();
+                }
+                else
+                {
+                    log_error("python_context", "Error in PinDirectionOnehot::calculate_feature:\n{}", res.get_error().get());
+                    return std::nullopt;
+                }
+            },
+            py::arg("ctx"),
+            py::arg("source"),
+            py::arg("destination"),
+            R"(
+One-hot encoding feature vector for pin directions.
+
+:param hal_py.Context ctx: The context.
+:param hal_py.Endpoint source: The source endpoint.
+:param hal_py.Endpoint destination: The destination endpoint.
+:returns: The feature vector or None.
+:rtype: list[hal_py.FEATURE_TYPE] or None
+)");
+
+        py_pin_direction_onehot.def("to_string",
+                                    &machine_learning::edge_feature::PinDirectionOnehot::to_string,
+                                    R"(
+Get a string representation.
+
+:returns: A string.
+:rtype: str
+)");
+
+        py_edge_feature.def(
+            "build_feature_vec",
+            [](machine_learning::Context& ctx, const std::vector<const machine_learning::edge_feature::EdgeFeature*>& features, const Endpoint* source, const Endpoint* destination)
+                -> std::optional<std::vector<FEATURE_TYPE>> {
+                auto res = machine_learning::edge_feature::build_feature_vec(ctx, features, source, destination);
+                if (res.is_ok())
+                {
+                    return res.get();
+                }
+                else
+                {
+                    log_error("python_context", "Error in build_feature_vec:\n{}", res.get_error().get());
+                    return std::nullopt;
+                }
+            },
+            py::arg("ctx"),
+            py::arg("features"),
+            py::arg("source"),
+            py::arg("destination"),
+            R"(
+Build a combined feature vector from multiple features.
+
+:param hal_py.Context ctx: The context.
+:param list[hal_py.edge_feature.EdgeFeature] features: A list of edge features.
+:param hal_py.Endpoint source: The source endpoint.
+:param hal_py.Endpoint destination: The destination endpoint.
+:returns: The combined feature vector or None.
+:rtype: list[hal_py.FEATURE_TYPE] or None
 )");
 
 #ifndef PYBIND11_MODULE