openvino/model-optimizer/extensions/front/TopKNormalize.py

30 lines
1.2 KiB
Python

# Copyright (C) 2018-2021 Intel Corporation
# SPDX-License-Identifier: Apache-2.0
import logging as log
from mo.front.common.partial_infer.utils import int64_array
from mo.front.common.replacement import FrontReplacementPattern
from mo.graph.graph import Graph
from mo.ops.const import Const
class TopKNormalize(FrontReplacementPattern):
"""
This pass do TopK layer normalization:
1. Adds the second input to the TopK layer if it has just one. In this case the attribute 'k' should be defined.
2. If one of TopK ports isn't connected - adds output on this port to keep this port in IR.
"""
enabled = True
def find_and_replace_pattern(self, graph: Graph):
for topk_node in graph.get_op_nodes(op='TopK'):
if topk_node.in_port(1).disconnected():
assert topk_node.has_valid('k'), 'The TopK node "{}" misses "k" attribute'.format(topk_node.name)
k_node = Const(graph, {'name': topk_node.id + '/Dims', 'value': int64_array(topk_node.k)}).create_node()
topk_node.in_port(1).connect(k_node.out_port(0))
del topk_node['k']
else:
log.debug('The TopK node input "{}" is already normalized'.format(topk_node.name))