accept 2d sum params for fully connected nodes

src/pyjuice/nodes/sum_nodes.py

@@ -95,6 +95,13 @@ def set_params(self, params: torch.Tensor, normalize: bool = True, pseudocount:
 
             self._params = params.clone().view(-1, 1, 1)
 
+        elif params.dim() == 2:
+            ch_num_ngroups = sum([cs.num_node_groups for cs in self.chs])
+            assert params.size(0) == self.num_nodes
+            assert params.size(1) == self.ch_group_size * ch_num_ngroups
+
+            self._params = params.reshape(self.num_node_groups, self.group_size, ch_num_ngroups, self.ch_group_size).permute(0, 2, 1, 3).flatten(0, 1).contiguous()
+
         elif params.dim() == 3:
             assert self.edge_ids.size(1) == params.size(0) and self.group_size == params.size(1) and self.ch_group_size == params.size(2)