Improve some code comments about message-passing layers.

ptgnn/neuralmodels/gnn/messagepassing/abstractmessagepassing.py

@@ -20,13 +20,19 @@ class AbstractMessagePassingLayer(nn.Module):
     ) -> torch.Tensor:
         """
         :param node_states: A [num_nodes, D] matrix containing the states of all nodes.
-        :param adjacency_lists: A list of [num_edges, 2] adjacency lists, one per edge type.
-        :param node_to_graph_idx:
-        :param reference_node_ids:
-        :param reference_node_graph_idx:
+        :param adjacency_lists: A list with as many elements as edge types. Each element in the list
+            is a pair (tuple) of [num_edges_for_edge_type]-sized tensors indicating the left/right
+            hand-side of the elements in the adjacency list.
+        :param node_to_graph_idx: a vector of shape [num_nodes] indicating the graph that each node
+            belongs to.
+        :param reference_node_ids: a dictionary that maps each reference (key) to
+            the indices of the reference nodes in `node_states`.
+        :param reference_node_graph_idx: a dictionary that maps each reference (key) to
+            the graph it belongs to. For each reference `ref_name`,
+                len(reference_node_ids[ref_name])==len(reference_node_graph_idx[ref_name])
         :param edge_features: A list of [num_edges, H] with edge features.
-            Has the size of `adjacency_lists`.
-        :return: the next node states in a [num_nodes, D'] matrix.
+            Has the same length as `adjacency_lists`, ie. len(adjacency_lists) == len(edge_features)
+        :return: the output node states in a [num_nodes, D'] matrix.
         """
 
     def _aggregate_messages(

ptgnn/neuralmodels/gnn/messagepassing/mlpmessagepassing.py

@@ -74,7 +74,9 @@ class MlpMessagePassingLayer(AbstractMessagePassingLayer):
         reference_node_graph_idx: Dict[str, torch.Tensor],
         edge_features: List[torch.Tensor],
     ) -> torch.Tensor:
-        assert len(adjacency_lists) == len(self.__edge_message_transformation_layers)
+        assert len(adjacency_lists) == len(
+            self.__edge_message_transformation_layers
+        ), "The number of adjacency lists must be equal to the number of edge types."
 
         all_message_targets, all_messages = [], []
         for edge_type_idx, (adj_list, features, edge_transformation_layer) in enumerate(