support the scenario that there are duplicate tensors in a same tuple (#3340)

nni/common/graph_utils.py

@@ -285,8 +285,8 @@ class TorchModuleGraph(TorchGraph):
         self.global_count += 1
         op_type = node.kind()
         node_group = [node]
-        inputs = set()
-        outputs = set()
+        inputs = []
+        outputs = []
         node_queue = queue.Queue()
         node_queue.put(node)
         while not node_queue.empty():
@@ -303,17 +303,17 @@ class TorchModuleGraph(TorchGraph):
                                     node_group.append(predecessor_node)
                                     node_queue.put(predecessor_node)
                             else:
-                                inputs.add(input_name)
+                                inputs.append(input_name)
                         else:
-                            inputs.add(input_name)
+                            inputs.append(input_name)
                 else:
-                    inputs.add(input_name)
+                    inputs.append(input_name)
         for output in node.outputs():
             if output.node().kind() == CONSTANT_KIND:
                 continue
-            outputs.add(output.debugName())
+            outputs.append(output.debugName())
         nodepy = NodePyGroup(node_name, unique_name, module_type, op_type,
-                             node_group, inputs=list(inputs), outputs=list(outputs), key_node=node)
+                             node_group, inputs=inputs, outputs=outputs, key_node=node)
         return nodepy
 
     def _expand_module_node(self, node, node_name, unique_name, op_type, nodes,
@@ -353,8 +353,8 @@ class TorchModuleGraph(TorchGraph):
         if not op_type:
             op_type = node.kind()
         node_group = [node]
-        inputs = set()
-        outputs = set()
+        inputs = []
+        outputs = []
         node_queue = queue.Queue()
         node_queue.put(node)
         visited = {node}
@@ -372,9 +372,9 @@ class TorchModuleGraph(TorchGraph):
                                 node_queue.put(predecessor_node)
                                 visited.add(predecessor_node)
                         else:
-                            inputs.add(input_name)
+                            inputs.append(input_name)
                 else:
-                    inputs.add(input_name)
+                    inputs.append(input_name)
             for _output in curr_node.outputs():
                 if _output.node().kind() == CONSTANT_KIND:
                     continue
@@ -387,9 +387,9 @@ class TorchModuleGraph(TorchGraph):
                                 node_queue.put(successor_node)
                                 visited.add(successor_node)
                         else:
-                            outputs.add(output_name)
+                            outputs.append(output_name)
                 else:
-                    outputs.add(output_name)
+                    outputs.append(output_name)
 
         nodepy = NodePyGroup(node_name, unique_name, module_type, op_type,
                              node_group, inputs=list(inputs), outputs=list(outputs))
@@ -562,10 +562,13 @@ class TorchModuleGraph(TorchGraph):
         for node in nodes_op:
             name_to_node[node.unique_name] = node
             for _input in node.inputs:
-                input_to_node[_input].append(node)
+                # inputs may have duplicate tensors
+                if node not in input_to_node[_input]:
+                    input_to_node[_input].append(node)
             for output in node.outputs:
-                assert not output in output_to_node, \
-                    "One output cannot be generated by multiple nodes %s" % output
+                if output in output_to_node:
+                    assert output_to_node[output] == node, \
+                        "One output cannot be generated by multiple nodes %s" % output
                 output_to_node[output] = node
         return name_to_node, input_to_node, output_to_node
 
@@ -619,8 +622,6 @@ class TorchModuleGraph(TorchGraph):
 
                     assert len(node.inputs) == len(list(last_cpp.inputs())), errmsg
                     for _debug_input, _debug_output in zip(node.inputs, node.outputs):
-                        # _debug_input = _input.debugName()
-                        # _debug_output = _output.debugName()
                         if _debug_input in self.input_to_node and _debug_output in self.input_to_node:
                             # input_to_node[_debug_input] is a list of NodePyGroup, because
                             # one tensor can be used as input for multiple nodes at the same time.
@@ -629,7 +630,8 @@ class TorchModuleGraph(TorchGraph):
                             # will be merged into the same NodePyGroup, so we remove the `node` from
                             # input_to_node[_debug_input] and directly connect this tensor to the
                             # input_to_node[_debug_output]
-                            self.input_to_node[_debug_input].remove(node)
+                            if node in self.input_to_node[_debug_input]:
+                                self.input_to_node[_debug_input].remove(node)
                             # add the following nodes of _output into the input_to_node[_debug_input]
                             self.input_to_node[_debug_input].extend(self.input_to_node[_debug_output])
                         # just remove the _debug_output from the grapgh index. So that we can also skip