openvino/model-optimizer/extensions/back/ElementwiseOpsToEltwiseOps.py

372 lines
14 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
import numpy as np
from extensions.ops.elementwise import Mul, Add, Pow
from mo.back.replacement import BackReplacementPattern
from mo.graph.graph import Node, Graph
from mo.middle.passes.fusing.helpers import get_value_in_port, get_tensor_in_port
from mo.middle.pattern_match import check_node_usages_out_of_match
from mo.ops.const import Const
from mo.ops.eltwise import Eltwise
from mo.ops.power import AttributedPower
simple_eltwise_types = ['Power', 'Add', 'Multiply', 'Maximum', 'LogicalAnd', 'LogicalOr', 'LogicalXor', 'Less',
'LessEqual',
'Greater', 'GreaterEqual', 'Equal', 'NotEqual', 'FloorMod']
type_to_operation_map = {
'Power': 'pow',
'Add': 'sum',
'Multiply': 'mul',
'Maximum': 'max',
'LogicalAnd': 'logical_and',
'LogicalOr': 'logical_or',
'LogicalXor': 'logical_xor',
'Less': 'less',
'LessEqual': 'less_equal',
'Greater': 'greater',
'GreaterEqual': 'greater_equal',
'Equal': 'equal',
'NotEqual': 'not_equal',
'FloorMod': 'floor_mod',
}
class SimpleEltwiseToEltwiseOp(BackReplacementPattern):
enabled = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
@staticmethod
def pattern():
return dict(
nodes=[
('op', dict(type=lambda t: t in type_to_operation_map.keys(), op=lambda op: op != 'AttributedPower'))],
edges=[],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
Eltwise.update_node_stat(match['op'], {'operation': type_to_operation_map[match['op'].type]})
class DivideToEltwises(BackReplacementPattern):
enabled = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
def run_before(self):
return [SimpleEltwiseToEltwiseOp]
@staticmethod
def pattern():
return dict(
nodes=[('div', {'type': 'Divide'})],
edges=[],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
node = match['div']
power_of_exponent = Const(graph, {'value': np.float64(-1)}).create_node()
reciprocal = Pow(graph, {'name': node.name + '/reciprocal_'}).create_node()
mul = Mul(graph, {'name': node.name + '/mul_'}).create_node()
# Connect nodes
node.in_port(1).get_connection().set_destination(reciprocal.in_port(0))
power_of_exponent.out_port(0).connect(reciprocal.in_port(1))
node.in_port(0).get_connection().set_destination(mul.in_port(1))
reciprocal.out_port(0).connect(mul.in_port(0))
node.out_port(0).get_connection().set_source(mul.out_port(0))
class SubtractToEltwises(BackReplacementPattern):
enabled = True
force_clean_up = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
def run_before(self):
return [SimpleEltwiseToEltwiseOp]
@staticmethod
def pattern():
return dict(
nodes=[('sub', {'type': 'Subtract'})],
edges=[],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
node = match['sub']
# Add new nodes
negate_const = Const(graph, dict(name=node.name + '/negate_const', value=np.array(-1))).create_node()
negate = Mul(graph, {'name': node.name + '/negate_'}).create_node()
add = Add(graph, {'name': node.name + '/add_'}).create_node()
# Connect nodes
node.in_port(1).get_connection().set_destination(negate.in_port(0))
negate_const.out_port(0).connect(negate.in_port(1))
node.in_port(0).get_connection().set_destination(add.in_port(1))
negate.out_port(0).connect(add.in_port(0))
node.out_port(0).get_connection().set_source(add.out_port(0))
class EltwisesWithScalarInputToPower(BackReplacementPattern):
enabled = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
force_clean_up = True
eltw_types = ['Add', 'Multiply', 'Power']
def run_before(self):
return [SimpleEltwiseToEltwiseOp]
@staticmethod
def pattern():
return dict(
nodes=[('op', {'type': lambda type: type in EltwisesWithScalarInputToPower.eltw_types})],
edges=[],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
op = match['op']
op_type = op.type
if op.has_and_set('stop_value_propagation'):
return
const_port, tensor_port = get_value_in_port(op), get_tensor_in_port(op)
if const_port is None or tensor_port is None:
return
value = const_port.data.get_value()
assert value is not None
if value.size != 1:
return
value = value.item(0)
assert op_type in EltwisesWithScalarInputToPower.eltw_types
if op_type == 'Add':
delete_node = value == 0
AttributedPower.update_node_stat(op, {'shift': value})
elif op_type == 'Multiply':
delete_node = value == 1
AttributedPower.update_node_stat(op, {'scale': value})
elif op_type == 'Power':
delete_node = value == 1
AttributedPower.update_node_stat(op, {'power': value})
op.type_infer = AttributedPower.type_infer
const_port.disconnect()
if tensor_port.idx != 0:
tensor_port.get_connection().set_destination(op.in_port(0))
op.delete_input_port(1)
# TODO: uncomment this lines in future to allow useless operations deleting
# if delete_node:
# op.out_port(0).get_connection().set_source(op.in_port(0).get_connection().get_source())
class MulAddPowerMerge(BackReplacementPattern):
enabled = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
force_clean_up = True
def run_after(self):
return [EltwisesWithScalarInputToPower]
@staticmethod
def pattern():
return dict(
nodes=[
('mul', dict(op='AttributedPower', shift=lambda x: np.all(x == 0), power=lambda x: np.all(x == 1))),
('mul_d', dict()),
('add', dict(op='AttributedPower', scale=lambda x: np.all(x == 1), power=lambda x: np.all(x == 1))),
('add_d', dict())
],
edges=[
('mul', 'mul_d'),
('mul_d', 'add'),
('add', 'add_d'),
],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
consumers = [n for n in match if n not in ['mul', 'add_d'] and not check_node_usages_out_of_match(match, n)]
if consumers:
log.warning('Power(mul,add) pattern was detected. Non pattern consumers of nodes: "{}" were found.'
' Won\'t replace'.format(', '.join([match[n].id for n in consumers])))
return
mul = match['mul']
add = match['add']
new_power = AttributedPower(graph, {'name': mul.name + '/fused_power', 'scale': mul.scale,
'shift': add.shift}).create_node()
source = mul.in_port(0).get_connection().get_source()
mul.in_port(0).disconnect()
new_power.in_port(0).connect(source)
add.out_port(0).get_connection().set_source(new_power.out_port(0))
log.debug('Power nodes {} and {} were fused to single Power node {}'.format(mul.name, add.name, new_power.name))
class MulPowPowerMerge(BackReplacementPattern):
enabled = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
force_clean_up = True
def run_after(self):
return [MulAddPowerMerge]
def run_before(self):
return [SimpleEltwiseToEltwiseOp]
@staticmethod
def pattern():
return dict(
nodes=[
('mul', dict(op='AttributedPower', shift=lambda x: np.all(x == 0), power=lambda x: np.all(x == 1))),
('mul_d', dict()),
('pow', dict(op='AttributedPower', scale=lambda x: np.all(x == 1), shift=lambda x: np.all(x == 0))),
('pow_d', dict())
],
edges=[
('mul', 'mul_d'),
('mul_d', 'pow'),
('pow', 'pow_d'),
],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
consumers = [n for n in match if n not in ['mul', 'pow_d'] and not check_node_usages_out_of_match(match, n)]
if consumers:
log.warning('Power(mul,pow) pattern was detected. Non pattern consumers of nodes: "{}" were found.'
' Won\'t replace'.format(', '.join([match[n].id for n in consumers])))
return
mul = match['mul']
pow = match['pow']
new_power = AttributedPower(graph, {'name': mul.name + '/fused_power', 'scale': mul.scale,
'power': pow.power}).create_node()
source = mul.in_port(0).get_connection().get_source()
mul.in_port(0).disconnect()
new_power.in_port(0).connect(source)
pow.out_port(0).get_connection().set_source(new_power.out_port(0))
log.debug('Power nodes {} and {} were fused to single Power node {}'.format(mul.name, pow.name, new_power.name))
class AddPowPowerMerge(BackReplacementPattern):
enabled = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
force_clean_up = True
def run_after(self):
return [MulAddPowerMerge]
def run_before(self):
return [SimpleEltwiseToEltwiseOp]
@staticmethod
def pattern():
return dict(
nodes=[
('add', dict(op='AttributedPower', scale=lambda x: np.all(x == 1), power=lambda x: np.all(x == 1))),
('add_d', dict()),
('pow', dict(op='AttributedPower', scale=lambda x: np.all(x == 1), shift=lambda x: np.all(x == 0))),
('pow_d', dict())
],
edges=[
('add', 'add_d'),
('add_d', 'pow'),
('pow', 'pow_d'),
],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
consumers = [n for n in match if n not in ['add', 'pow_d'] and not check_node_usages_out_of_match(match, n)]
if consumers:
log.warning('Power(add,pow) pattern was detected. Non pattern consumers of nodes: "{}" were found.'
' Won\'t replace'.format(', '.join([match[n].id for n in consumers])))
return
add = match['add']
pow = match['pow']
new_power = AttributedPower(graph, {'name': add.name + '/fused_power', 'shift': add.shift,
'power': pow.power}).create_node()
source = add.in_port(0).get_connection().get_source()
add.in_port(0).disconnect()
new_power.in_port(0).connect(source)
pow.out_port(0).get_connection().set_source(new_power.out_port(0))
log.debug('Power nodes {} and {} were fused to single Power node {}'.format(add.name, pow.name, new_power.name))
class MulAddPowPowerMerge(BackReplacementPattern):
enabled = True
graph_condition = [lambda graph: not graph.graph['cmd_params'].generate_experimental_IR_V10]
force_clean_up = True
def run_after(self):
return [MulAddPowerMerge, MulPowPowerMerge, AddPowPowerMerge]
def run_before(self):
return [SimpleEltwiseToEltwiseOp]
@staticmethod
def pattern():
return dict(
nodes=[
('mul_add', dict(op='AttributedPower', power=lambda x: np.all(x == 1))),
('mul_add_d', dict()),
('pow', dict(op='AttributedPower', scale=lambda x: np.all(x == 1), shift=lambda x: np.all(x == 0))),
('pow_d', dict())
],
edges=[
('mul_add', 'mul_add_d'),
('mul_add_d', 'pow'),
('pow', 'pow_d'),
],
)
@staticmethod
def replace_pattern(graph: Graph, match: [str, Node]):
consumers = [n for n in match if n not in ['mul_add', 'pow_d'] and not check_node_usages_out_of_match(match, n)]
if consumers:
log.warning('Power(mul_add,pow) pattern was detected. Non pattern consumers of nodes: "{}" were found.'
' Won\'t replace'.format(', '.join([match[n].id for n in consumers])))
return
mul_add = match['mul_add']
pow = match['pow']
new_power = AttributedPower(graph, {'name': mul_add.name + '/fused_power', 'shift': mul_add.shift,
'scale': mul_add.scale,
'power': pow.power}).create_node()
source = mul_add.in_port(0).get_connection().get_source()
mul_add.in_port(0).disconnect()
new_power.in_port(0).connect(source)
pow.out_port(0).get_connection().set_source(new_power.out_port(0))
log.debug('Power nodes {} and {} were fused to single Power node {}'.format(mul_add.name, pow.name,
new_power.name))