openvino/model-optimizer/extensions/middle/LayoutChangeForConstantShap...

122 lines
5.9 KiB
Python

# Copyright (C) 2018-2021 Intel Corporation
# SPDX-License-Identifier: Apache-2.0
from collections import deque
from typing import List, Set
from extensions.middle.InsertLayoutPropagationTransposes import is_output_data_in_correct_layout, \
InsertLayoutPropagationTranspose, mark_input_as_in_correct_layout, mark_output_as_in_correct_layout
from extensions.ops.gather import Gather
from extensions.ops.transpose import Transpose
from mo.front.common.partial_infer.utils import int64_array
from mo.front.tf.graph_utils import create_op_with_const_inputs
from mo.graph.graph import Graph
from mo.graph.port import Port
from mo.middle.replacement import MiddleReplacementPattern
class LayoutChangeForConstantShapePaths(MiddleReplacementPattern):
enabled = True
graph_condition = [lambda graph: graph.graph['layout'] == 'NHWC',
lambda graph: not graph.graph['cmd_params'].static_shape]
force_clean_up = True
def run_after(self):
return [InsertLayoutPropagationTranspose]
def run_before(self):
return []
@staticmethod
def get_next_in_ports(in_port: Port) -> Set[Port]:
next_in_ports = set()
for out_port in in_port.node.out_ports().values():
next_in_ports.update(out_port.get_destinations())
return next_in_ports
def find_shape_subgraph_endpoints(self, out_ports: List[Port], visited: set = None,
action: callable = None) -> Set[Port]:
"""
Searches for input ports of data dependent operations starting from output ports passed to the function.
Condition for data dependent operations is absence of node output value.
:param out_ports: list of output ports to start search from
:param visited: set of input ports that were visited to avoid visiting them more than once
:param action: function to call on the each input port of shape sub-graph
:return: set of input ports of data dependent operations
"""
if visited is None:
visited = set()
deque_of_in_ports = deque()
for out_port in out_ports:
deque_of_in_ports.extend(out_port.get_destinations())
end_points_in_ports = set()
while len(deque_of_in_ports):
in_port = deque_of_in_ports.popleft()
if in_port in visited:
continue
next_in_ports = self.get_next_in_ports(in_port)
if any([port.data.get_value() is None for port in next_in_ports]):
end_points_in_ports.add(in_port)
else:
deque_of_in_ports.extend(next_in_ports)
if action is not None:
action(in_port)
visited.add(in_port)
return end_points_in_ports
def find_and_replace_pattern(self, graph: Graph):
shape_ops = graph.get_op_nodes(op='ShapeOf')
# 1. Inserting Gather to N*C format on constant shape paths
for shape in shape_ops:
source_port = shape.in_port(0).get_source()
if is_output_data_in_correct_layout(source_port.node, source_port.idx):
continue # data is already in N*C format
name = shape.soft_get('name', shape.id)
rank = source_port.data.get_shape().size
if rank in [4, 5]:
index = int64_array([0, *list(range(2, rank)), 1])
else:
continue # data is layout independent
gather = create_op_with_const_inputs(graph, op=Gather, port_value_dict={1: index, 2: int64_array(0)},
op_attrs={'name': name + '/GatherNCHWtoNHWC'})
shape.out_port(0).get_connection().insert_node(gather)
# 2. Inserting Gather/Transpose to NC* format
shape_sub_graph_end_points = self.find_shape_subgraph_endpoints([shape.out_port(0) for shape in shape_ops])
for in_port in shape_sub_graph_end_points:
name = in_port.node.soft_get('name', in_port.node.id)
shape = in_port.data.get_shape()
should_switch_layout = not any([is_output_data_in_correct_layout(port.node, port.idx)
for port in in_port.node.out_ports().values() if not port.disconnected()])
should_insert_gather = should_switch_layout and len(shape) == 1 and shape.item(0) in [4, 5]
should_insert_transpose = should_switch_layout and len(shape) in [4, 5]
if should_insert_gather:
# we should turn input permutation off to perform it with the following gather insertion
in_port.__setattr__('input_permutation', None)
index = int64_array([0, shape.item(0) - 1, *list(range(1, shape.item(0) - 1))])
gather = create_op_with_const_inputs(graph, op=Gather,
port_value_dict={1: index, 2: int64_array(0)},
op_attrs={'name': name + '/GatherNHWCtoNCHW'})
in_port.get_connection().insert_node(gather)
elif should_insert_transpose:
# we should turn input permutation off to perform it with the following transpose insertion
in_port.__setattr__('input_permutation', None)
order = int64_array([0, len(shape) - 1, *list(range(1, len(shape) - 1))])
transpose = create_op_with_const_inputs(graph, op=Transpose, port_value_dict={1: order},
op_attrs={'name': name + '/TransposeNHWCtoNCHW',
'override_output_shape': True})
mark_input_as_in_correct_layout(transpose, 0)
mark_output_as_in_correct_layout(transpose, 0)
in_port.get_connection().insert_node(transpose)
else:
continue # data is layout independent