Files
openvino/model-optimizer/extensions/middle/TensorIteratorBackEdge.py

128 lines
5.3 KiB
Python

"""
Copyright (C) 2018-2020 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 logging as log
from extensions.middle.TensorIteratorCondition import DynamicDecoderConditionMatcher
from extensions.ops.TensorIterator_ops import TensorIteratorBackEdge, TensorIteratorOutput
from mo.graph.graph import Graph
from mo.middle.replacement import MiddleReplacementPattern
class BackEdgesMatching(MiddleReplacementPattern):
"""
This pattern are needed for matching back edges in while loops in TF graphs.
Back edge is a chain of nodes in while loop that iterate one variable in graph over loop steps. It consist of
nodes:
Exit (optional)
^
|
Enter () -> Merge -> Switch -> Identity -> SOME OPERATIONS -> NextIteration ->
^ |
| |
------------------------------------------------------------------
The structure of pattern without Data nodes between ops (every node is named as op attribute of this node):
Data--
|
NextIteration -> Merge--
|
->Switch (out=1) -> Identity
|
TensorIteratorCondition--
"""
enabled = True
graph_condition = [lambda graph: graph.graph['is_cyclic']]
def run_after(self):
return [DynamicDecoderConditionMatcher]
def run_before(self):
from extensions.middle.TensorIteratorMerge import TensorIteratorMerge
return [TensorIteratorMerge]
@staticmethod
def pattern():
return dict(
nodes=[
('Enter_1_data', dict(kind='data')),
('Merge_1', dict(kind='op', op='Merge')),
('Merge_1_data', dict(kind='data')),
('Switch_1', dict(kind='op', op='Switch')),
('Switch_1_data', dict(kind='data')),
('Identity_1', dict(kind='op', op='Identity')),
('Identity_1_data', dict(kind='data')),
('NextIteration', dict(kind='op', op='NextIteration')),
('NextIteration_data', dict(kind='data')),
('condition', dict(kind='op', op='TensorIteratorCondition')),
('condition_cond_data', dict(kind='data')),
],
edges=[
('Enter_1_data', 'Merge_1'),
('Merge_1', 'Merge_1_data'),
('Merge_1_data', 'Switch_1'),
('Switch_1', 'Switch_1_data', {'out': 1}),
('Switch_1_data', 'Identity_1'),
('Identity_1', 'Identity_1_data'),
('NextIteration', 'NextIteration_data'),
('NextIteration_data', 'Merge_1'),
('condition', 'condition_cond_data'),
('condition_cond_data', 'Switch_1'),
]
)
def replace_pattern(self, graph: Graph, match: dict):
log.debug('================== BackEdgeFind ===============')
nodes_for_remove = []
from_body_data = match['NextIteration'].in_node()
# If Exit path is exist -> create TensorIteratorOutput for this
if 0 in match['Switch_1'].out_nodes():
Exit = match['Switch_1'].out_node(0).out_node(0) # Switch -> Switch_data -> Exit
assert Exit.has_valid('op') and Exit.op == 'Exit'
output_data = Exit.out_node(0)
nodes_for_remove.append(match['Switch_1'].out_node(0).id)
nodes_for_remove.append(Exit.id)
# Creating TensorIteratorOutput without partition
output = TensorIteratorOutput(graph, dict(external_port_id=None,
internal_layer_id=None, \
name=Exit.name + '/TensorIteratorOutput_'
))
output.create_node_with_data(inputs=[from_body_data, match['condition_cond_data']],
data_nodes=[output_data])
assert match['NextIteration_data'].id != match['Enter_1_data'].id
backedge = TensorIteratorBackEdge(graph, dict(name=match['Identity_1'].name + '/TensorIteratorBackEdge_'))
backedge.create_node_with_data(inputs=[match['Enter_1_data'], from_body_data, match['condition_cond_data']],
data_nodes=[match['Identity_1_data']])
# Delete useless nodes
safe_nodes = ['Identity_1_data', 'condition', 'condition_cond_data', 'Enter_1_data']
for node in match.keys():
if node not in safe_nodes:
nodes_for_remove.append(match[node].id)
graph.remove_nodes_from(nodes_for_remove)