[ MO ] Random Uniform Replacer (#1814)

This commit is contained in:
Yegor Kruglov
2020-08-19 12:32:16 +03:00
committed by GitHub
parent 8b98f20480
commit b3e7cc55a6
3 changed files with 165 additions and 0 deletions

View File

@@ -540,6 +540,7 @@ extensions/middle/pass_separator.py
extensions/middle/permute_tensor_iterator.py
extensions/middle/preprocessing.py
extensions/middle/quantize_fuses.py
extensions/middle/RandomUniformReplacer.py
extensions/middle/ReluQuantizeFuse.py
extensions/middle/RemoveDuplicationMemory.py
extensions/middle/RemoveIdentity.py

View File

@@ -0,0 +1,66 @@
"""
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 numpy as np
from typing import Dict
from mo.front.tf.graph_utils import create_op_with_const_inputs
from mo.graph.graph import Graph, Node
from mo.middle.replacement import MiddleReplacementPattern
from mo.ops.broadcast import Broadcast
class RandomUniformReplacer(MiddleReplacementPattern):
"""
Replaces RandomUniform operation with Broadcast of ones in sub-graph:
ShapeOf ---> RandomUniform ---> Mul
"""
enabled = True
@staticmethod
def pattern():
return dict(
nodes=[
('shape', dict(op='ShapeOf')),
('shape_data', dict()),
('random_uniform', dict(op='RandomUniform')),
('random_uniform_data', dict()),
('mul', dict(op='Mul')),
('mul_const', dict(op='Const')),
('mul_const_data', dict())
],
edges=[
('shape', 'shape_data'),
('shape_data', 'random_uniform'),
('random_uniform', 'random_uniform_data'),
('random_uniform_data', 'mul'),
('mul_const', 'mul_const_data'),
('mul_const_data', 'mul')
]
)
@staticmethod
def replace_pattern(graph: Graph, match: Dict[str, Node]):
node = match['random_uniform']
node_name = node.soft_get('name', node.id)
data_type = match['mul_const'].out_port(0).get_data_type()
broadcast_node = create_op_with_const_inputs(graph, Broadcast, port_value_dict={0: np.array([1], dtype=data_type)},
op_attrs={'name': node_name + '/Broadcast', 'mode': 'numpy'})
node.in_port(0).get_connection().set_destination(broadcast_node.in_port(1))
node.out_port(0).get_connection().set_source(broadcast_node.out_port(0))

View File

@@ -0,0 +1,98 @@
"""
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 unittest
import numpy as np
from extensions.middle.RandomUniformReplacer import RandomUniformReplacer
from mo.utils.ir_engine.compare_graphs import compare_graphs
from mo.utils.unittest.graph import build_graph
nodes_attributes = {
'shape': {'kind': 'op', 'op': 'ShapeOf'},
'shape_data': {'kind': 'data'},
'random_uniform': {'kind': 'op', 'op': 'RandomUniform'},
'random_uniform_data': {'kind': 'data'},
'mul': {'kind': 'op', 'op': 'Mul'},
'mul_const': {'kind': 'op', 'op': 'Const'},
'mul_const_data': {'kind': 'data', 'value': np.array([1], dtype=np.int32)},
'broadcast': {'kind': 'op', 'op': 'Broadcast'},
'broadcast_const': {'kind': 'op', 'op': 'Const'},
'broadcast_const_data': {'kind': 'data', 'value': np.array([1], dtype=np.int32)},
}
class RandomUniformReplacerTest(unittest.TestCase):
def test_1(self):
graph = build_graph(nodes_attributes,
edges=[
('shape', 'shape_data'),
('shape_data', 'random_uniform'),
('random_uniform', 'random_uniform_data'),
('random_uniform_data', 'mul', {'in': 0}),
('mul_const', 'mul_const_data'),
('mul_const_data', 'mul', {'in': 1})
],
nodes_with_edges_only=True)
ref_graph = build_graph(nodes_attributes,
edges=[
('shape', 'shape_data'),
('shape_data', 'broadcast', {'in': 1}),
('broadcast_const', 'broadcast_const_data'),
('broadcast_const_data', 'broadcast', {'in': 0}),
('broadcast', 'random_uniform_data'),
('random_uniform_data', 'mul', {'in': 0}),
('mul_const', 'mul_const_data'),
('mul_const_data', 'mul', {'in': 1})
],
nodes_with_edges_only=True)
RandomUniformReplacer().find_and_replace_pattern(graph)
flag, resp = compare_graphs(graph, ref_graph, 'mul', check_op_attrs=True)
self.assertTrue(flag, resp)
def test_2(self):
graph = build_graph(nodes_attributes,
edges=[
('shape', 'shape_data'),
('shape_data', 'random_uniform'),
('random_uniform', 'random_uniform_data'),
('random_uniform_data', 'mul', {'in': 1}),
('mul_const', 'mul_const_data'),
('mul_const_data', 'mul', {'in': 0})
],
nodes_with_edges_only=True)
ref_graph = build_graph(nodes_attributes,
edges=[
('shape', 'shape_data'),
('shape_data', 'broadcast', {'in': 1}),
('broadcast_const', 'broadcast_const_data'),
('broadcast_const_data', 'broadcast', {'in': 0}),
('broadcast', 'random_uniform_data'),
('random_uniform_data', 'mul', {'in': 1}),
('mul_const', 'mul_const_data'),
('mul_const_data', 'mul', {'in': 0})
],
nodes_with_edges_only=True)
RandomUniformReplacer().find_and_replace_pattern(graph)
flag, resp = compare_graphs(graph, ref_graph, 'mul', check_op_attrs=True)
self.assertTrue(flag, resp)