Result rename operation (openvinotoolkit#4242)

* Added result rename operation * Optimize imports * Added ResultRename to package_BOM * ResultRename moved to the end of back phase, code refactoring * Revert incorrect changes * Optimize imports * Added comments and optimized imports.
tomdol · Feb 12, 2021 · eeb7835 · eeb7835
1 parent 2bf0e88
commit eeb7835
Show file tree

Hide file tree

Showing 8 changed files with 113 additions and 25 deletions.
diff --git a/inference-engine/src/readers/ir_reader/ie_ir_parser.cpp b/inference-engine/src/readers/ir_reader/ie_ir_parser.cpp
@@ -392,7 +392,7 @@ std::shared_ptr<ngraph::Function> V10Parser::XmlDeserializer::parse_function(con
     // Read all layers and store their parameters in params map
     FOREACH_CHILD(node, root.child("layers"), "layer") {
         auto node_param = parseGenericParams(node);
-        if (opName.find(node_param.name) != opName.end())
+        if (opName.find(node_param.name) != opName.end() && node_param.type != "Result")
             THROW_IE_EXCEPTION << "Invalid IR! " << node_param.name << " name is not unique!";
         opName.insert(node_param.name);
         params[node_param.layerId] = {node, node_param};

diff --git a/model-optimizer/automation/package_BOM.txt b/model-optimizer/automation/package_BOM.txt
@@ -47,6 +47,7 @@ extensions/back/RemoveUselessConvert.py
 extensions/back/Reshape0DToSqueeze.py
 extensions/back/ReshapeMutation.py
 extensions/back/ResultNormalizer.py
+extensions/back/ResultRename.py
 extensions/back/ReverseInputChannels.py
 extensions/back/RNNSequenceTypeRename.py
 extensions/back/ScalarConstNormalize.py

diff --git a/model-optimizer/extensions/back/ResultRename.py b/model-optimizer/extensions/back/ResultRename.py
@@ -0,0 +1,38 @@
+"""
+ Copyright (C) 2018-2021 Intel Corporation
+
+ Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+ You may obtain a copy of the License at
+
+      http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
+"""
+
+from mo.back.replacement import BackReplacementPattern
+from mo.graph.graph import Graph
+
+
+class ResultRename(BackReplacementPattern):
+    # This transformation sets the Result operation name equal to the incoming tensor name.
+    # For some frameworks like kaldi and onnx this may result in appearance of nodes with identical names,
+    # which can lead to errors in other transformations.
+    # So ResultRename should be launched at the end of back phase.
+    enabled = False
+
+    def find_and_replace_pattern(self, graph: Graph):
+        for node in graph.get_op_nodes(type='Result'):
+            if node.in_ports():
+                prev_node_out_port = node.in_port(0).get_connection().get_source()
+                tensor_names = prev_node_out_port.get_tensor_names()
+                if tensor_names:
+                    result_name = tensor_names[0]
+                else:
+                    result_name = prev_node_out_port.node.soft_get('name', prev_node_out_port.node.id) + \
+                                  '/sink_port_' + str(prev_node_out_port.idx)
+                node['name'] = result_name
diff --git a/model-optimizer/extensions/back/ResultRename_test.py b/model-optimizer/extensions/back/ResultRename_test.py
@@ -0,0 +1,48 @@
+"""
+ Copyright (C) 2018-2021 Intel Corporation
+
+ Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+ You may obtain a copy of the License at
+
+      http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
+"""
+
+import unittest
+
+from extensions.back.ResultRename import ResultRename
+from mo.graph.graph import Node
+from mo.utils.ir_engine.compare_graphs import compare_graphs
+from mo.utils.unittest.graph import build_graph, regular_op, result
+
+nodes = {
+    **regular_op('Op1', {'type': 'Op1', 'kind': 'op', 'op': 'Op1'}),
+    **result('result'),
+    'Op1_data': {'kind': 'data', 'fw_tensor_debug_info': [('Op1', 0, 'Op1_tensor')]}
+}
+
+
+class ResultRenameTest(unittest.TestCase):
+    def test_case1(self):
+        graph = build_graph(nodes, [('Op1', 'Op1_data'), ('Op1_data', 'result')])
+        graph_ref = build_graph(nodes, [('Op1', 'Op1_data'), ('Op1_data', 'result')])
+        res_node = Node(graph_ref, 'result')
+        res_node['name'] = 'Op1_tensor'
+
+        ResultRename().find_and_replace_pattern(graph)
+        (flag, resp) = compare_graphs(graph, graph_ref, 'result', check_op_attrs=True)
+        self.assertTrue(flag, resp)
+
+    def test_case2(self):
+        graph = build_graph(nodes, [])
+        graph_ref = build_graph(nodes, [])
+
+        ResultRename().find_and_replace_pattern(graph)
+        (flag, resp) = compare_graphs(graph, graph_ref, 'result', check_op_attrs=True)
+        self.assertTrue(flag, resp)
diff --git a/model-optimizer/mo/back/ie_ir_ver_2/emitter.py b/model-optimizer/mo/back/ie_ir_ver_2/emitter.py
@@ -174,8 +174,8 @@ def xml_ports(node: Node, element: Element, edges: Element):
             assert node.graph.node[v]['shape'] is not None, 'Output shape is not calculated properly for node {}' \
                                                             ''.format(node.id)
             tensor_names = node.out_port(port_id).get_tensor_names(port_renumber=True)
-            if tensor_names is not None:
-                port.set('names', tensor_names)
+            if tensor_names:
+                port.set('names', ','.join(tensor_names))
             xml_shape(node.graph.node[v]['shape'], port)
 
 

diff --git a/model-optimizer/mo/graph/port.py b/model-optimizer/mo/graph/port.py
@@ -1,5 +1,5 @@
 """
- Copyright (C) 2018-2020 Intel Corporation
+ Copyright (C) 2018-2021 Intel Corporation
 
  Licensed under the Apache License, Version 2.0 (the "License");
  you may not use this file except in compliance with the License.
@@ -310,9 +310,7 @@ def get_tensor_names_list(attrs):
             if node_idx in self.node.out_nodes():
                 out_node = self.node.out_node(node_idx)
                 fw_names += get_tensor_names_list(out_node.attrs())
-        if len(fw_names) > 0:
-            return ','.join(fw_names)
-        return None
+        return fw_names
 
     def disconnect(self):
         if self.type == 'out':

diff --git a/model-optimizer/mo/graph/port_test.py b/model-optimizer/mo/graph/port_test.py
@@ -39,50 +39,50 @@ def test_front(self):
                                                                                            ('Op1', 0, 'Op1,Op2')]})])
         graph.stage = 'front'
         input_node = Node(graph, 'input')
-        self.assertTrue(input_node.out_port(0).get_tensor_names() == 'input,Op1\\,Op2')
+        self.assertTrue(input_node.out_port(0).get_tensor_names() == ['input', 'Op1\\,Op2'])
 
         op1_node = Node(graph, 'Op1')
         op1_node.add_output_port(0)
-        self.assertTrue(op1_node.out_port(0).get_tensor_names() is None)
+        self.assertTrue(op1_node.out_port(0).get_tensor_names() == [])
 
     def test_middle(self):
         graph = build_graph(nodes, [('input', 'input_data'), ('input_data', 'Op1'),
                                     ('input_data', 'Op2')])
 
         input_node = Node(graph, 'input')
-        self.assertTrue(input_node.out_port(0).get_tensor_names() == 'input,Op1\\,Op2')
+        self.assertTrue(input_node.out_port(0).get_tensor_names() == ['input', 'Op1\\,Op2'])
 
         op1_node = Node(graph, 'Op1')
         op1_node.add_output_port(0)
-        self.assertTrue(op1_node.out_port(0).get_tensor_names() is None)
+        self.assertTrue(op1_node.out_port(0).get_tensor_names() == [])
 
         op2_node = Node(graph, 'Op2')
         op2_node.add_output_port(0)
-        self.assertTrue(op2_node.out_port(0).get_tensor_names() is None)
+        self.assertTrue(op2_node.out_port(0).get_tensor_names() == [])
 
     def test_port_renumber(self):
         graph = build_graph(nodes, [('input', 'input_data'), ('input_data', 'Op1'),
                                     ('Op1', 'Op1_data', {'out': 1}), ('Op1_data', 'Op2')])
         input_node = Node(graph, 'input')
-        self.assertTrue(input_node.out_port(0).get_tensor_names(port_renumber=True) == 'input,Op1\\,Op2')
+        self.assertTrue(input_node.out_port(0).get_tensor_names(port_renumber=True) == ['input', 'Op1\\,Op2'])
 
         op1_node = Node(graph, 'Op1')
         op1_node.add_output_port(0)
 
-        self.assertTrue(op1_node.out_port(0).get_tensor_names(port_renumber=True) == 'Op1\\,Op2')
+        self.assertTrue(op1_node.out_port(0).get_tensor_names(port_renumber=True) == ['Op1\\,Op2'])
 
     def test_reconnect_middle_case1(self):
         graph = build_graph(nodes, [('input', 'input_data'), ('input_data', 'Op1'), ('Op3', 'Op3_data')])
         input_node = Node(graph, 'input')
 
         input_node_out_port = input_node.out_port(0)
-        self.assertTrue(input_node_out_port.get_tensor_names() == 'input,Op1\\,Op2')
+        self.assertTrue(input_node_out_port.get_tensor_names() == ['input', 'Op1\\,Op2'])
 
         op3_node = Node(graph, 'Op3')
         input_node_out_port.get_connection().set_source(op3_node.out_port(0))
 
         self.assertTrue(input_node_out_port.get_tensor_names() is None)
-        self.assertTrue(op3_node.out_port(0).get_tensor_names() == 'Op3,input,Op1\\,Op2')
+        self.assertTrue(op3_node.out_port(0).get_tensor_names() == ['Op3', 'input', 'Op1\\,Op2'])
 
     def test_reconnect_front_case1(self):
         graph = build_graph(nodes, [('input', 'Op1', {'in': 0, 'out': 0, 'fw_tensor_debug_info': [('input', 0, 'input'),
@@ -93,37 +93,37 @@ def test_reconnect_front_case1(self):
         input_node = Node(graph, 'input')
 
         input_node_out_port = input_node.out_port(0)
-        self.assertTrue(input_node_out_port.get_tensor_names() == 'input,Op1\\,Op2')
+        self.assertTrue(input_node_out_port.get_tensor_names() == ['input', 'Op1\\,Op2'])
 
         op3_node = Node(graph, 'Op3')
         input_node_out_port.get_connection().set_source(op3_node.out_port(0))
 
-        self.assertTrue(input_node_out_port.get_tensor_names() is None)
-        self.assertTrue(op3_node.out_port(0).get_tensor_names() == 'Op3,input,Op1\\,Op2')
+        self.assertTrue(input_node_out_port.get_tensor_names() == [])
+        self.assertTrue(op3_node.out_port(0).get_tensor_names() == ['Op3', 'input', 'Op1\\,Op2'])
 
     def test_reconnect_middle_case1(self):
         graph = build_graph(nodes, [('input', 'input_data'), ('input_data', 'Op1'), ('Op3', 'Op3_data')])
         input_node = Node(graph, 'input')
 
         input_node_out_port = input_node.out_port(0)
-        self.assertTrue(input_node_out_port.get_tensor_names() == 'input,Op1\\,Op2')
+        self.assertTrue(input_node_out_port.get_tensor_names() == ['input', 'Op1\\,Op2'])
 
         op3_node = Node(graph, 'Op3')
         input_node_out_port.get_connection().set_source(op3_node.out_port(0))
 
-        self.assertTrue(input_node_out_port.get_tensor_names() is None)
-        self.assertTrue(op3_node.out_port(0).get_tensor_names() == 'Op3,input,Op1\\,Op2')
+        self.assertTrue(input_node_out_port.get_tensor_names() == [])
+        self.assertTrue(op3_node.out_port(0).get_tensor_names() == ['Op3', 'input', 'Op1\\,Op2'])
 
     def test_reconnect_middle_case2(self):
         graph = build_graph(nodes, [('input', 'input_data'), ('input_data', 'Op1', {'out': 0}),
                                     ('input_data', 'Op1', {'out': 1}), ('Op3', 'Op3_data')])
         input_node = Node(graph, 'input')
 
         input_node_out_port = input_node.out_port(0)
-        self.assertTrue(input_node_out_port.get_tensor_names() == 'input,Op1\\,Op2')
+        self.assertTrue(input_node_out_port.get_tensor_names() == ['input', 'Op1\\,Op2'])
 
         op3_node = Node(graph, 'Op3')
         input_node_out_port.get_connection().set_source(op3_node.out_port(0))
 
-        self.assertTrue(input_node_out_port.get_tensor_names() is None)
-        self.assertTrue(op3_node.out_port(0).get_tensor_names() == 'Op3,input,Op1\\,Op2')
+        self.assertTrue(input_node_out_port.get_tensor_names() == [])
+        self.assertTrue(op3_node.out_port(0).get_tensor_names() == ['Op3', 'input', 'Op1\\,Op2'])
diff --git a/model-optimizer/mo/pipeline/common.py b/model-optimizer/mo/pipeline/common.py
@@ -22,6 +22,7 @@
 import networkx as nx
 
 from extensions.back.RemoveUselessConvert import RemoveUselessConvert
+from extensions.back.ResultRename import ResultRename
 from extensions.back.op_versioning import OpVersioning
 from extensions.ops.Cast import Cast
 from mo.back.ie_ir_ver_2.emitter import port_renumber, serialize_constants, generate_ie_ir, serialize_mean_image
@@ -208,6 +209,8 @@ def prepare_emit_ir(graph: Graph, data_type: str, output_dir: str, output_model_
     type_infer(graph)
     RemoveUselessConvert().find_and_replace_pattern(graph)
 
+    ResultRename().find_and_replace_pattern(graph)
+
     for sub_graph in [graph] + collect_sub_graphs(graph):
         op_order, data_order = determined_sort(get_sorted_outputs(sub_graph))
         mapping = {v: u for u, v in enumerate(op_order)}