256 lines
10 KiB
Python
256 lines
10 KiB
Python
# Copyright (C) 2018-2021 Intel Corporation
|
|
# SPDX-License-Identifier: Apache-2.0
|
|
|
|
import unittest
|
|
|
|
import numpy as np
|
|
|
|
from extensions.back.ConvolutionNormalizer import PullReshapeThroughFQ, V7ConvolutionWithGroupsResolver, \
|
|
V10ConvolutionWithGroupsResolver
|
|
from extensions.ops.fakequantize import FakeQuantize
|
|
from mo.front.common.partial_infer.utils import int64_array
|
|
from mo.ops.reshape import Reshape
|
|
from mo.utils.ir_engine.compare_graphs import compare_graphs
|
|
from unit_tests.utils.graph import build_graph, result, regular_op_with_shaped_data, regular_op_with_empty_data, \
|
|
valued_const_with_data, connect
|
|
|
|
|
|
def graph_template(weights_initial_shape, new_reshape_shape, limits_initial_shape, limits_new_shape=None):
|
|
limits_new_shape = limits_initial_shape if limits_new_shape is None else limits_new_shape
|
|
|
|
core_connections = [
|
|
*connect('input:0', '0:convolution'),
|
|
*connect('convolution:0', '0:output'),
|
|
]
|
|
|
|
core_nodes = lambda weights_shape, limit_shape, reshape_shape: {
|
|
**regular_op_with_shaped_data('input', None, {'type': 'Parameter', 'op': 'Parameter'}),
|
|
|
|
**valued_const_with_data('weights', np.ones(weights_shape)),
|
|
|
|
**valued_const_with_data('dim', int64_array(reshape_shape)),
|
|
**regular_op_with_shaped_data('reshape', reshape_shape, {'type': 'Reshape', 'infer': Reshape.infer, 'op': 'Reshape'}),
|
|
|
|
**valued_const_with_data('il', np.ones(limit_shape)),
|
|
**valued_const_with_data('ih', np.ones(limit_shape)),
|
|
**valued_const_with_data('ol', np.ones(limit_shape)),
|
|
**valued_const_with_data('oh', np.ones(limit_shape)),
|
|
|
|
**regular_op_with_shaped_data('FQ', weights_shape, {'type': 'FakeQuantize', 'infer': FakeQuantize.infer,
|
|
'stop_value_propagation': True, 'levels': 2, 'op': 'FakeQuantize'}),
|
|
|
|
**regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'op': 'Convolution'}),
|
|
|
|
**result(),
|
|
}
|
|
|
|
nodes_before = core_nodes(weights_initial_shape, limits_initial_shape, new_reshape_shape)
|
|
edges_before = [
|
|
|
|
*connect('weights:0', '0:FQ'),
|
|
*connect('il:0', '1:FQ'),
|
|
*connect('ih:0', '2:FQ'),
|
|
*connect('ol:0', '3:FQ'),
|
|
*connect('oh:0', '4:FQ'),
|
|
|
|
*connect('FQ:0', '0:reshape'),
|
|
*connect('dim:0', '1:reshape'),
|
|
*connect('reshape:0', '1:convolution'),
|
|
|
|
*core_connections,
|
|
]
|
|
graph = build_graph(nodes_attrs=nodes_before, edges=edges_before, nodes_with_edges_only=True)
|
|
|
|
nodes_after = core_nodes(new_reshape_shape, limits_new_shape, [])
|
|
edges_after = [
|
|
*connect('weights:0', '0:FQ'),
|
|
*connect('il:0', '1:FQ'),
|
|
*connect('ih:0', '2:FQ'),
|
|
*connect('ol:0', '3:FQ'),
|
|
*connect('oh:0', '4:FQ'),
|
|
*connect('FQ:0', '1:convolution'),
|
|
|
|
*core_connections,
|
|
]
|
|
graph_ref = build_graph(nodes_attrs=nodes_after, edges=edges_after, nodes_with_edges_only=True)
|
|
return graph, graph_ref
|
|
|
|
|
|
class TestPullReshapeThroughFQ(unittest.TestCase):
|
|
|
|
def test_v7_weights_reshape(self):
|
|
graph, graph_ref = graph_template([3, 8, 7, 7], [24, 1, 7, 7], [1, 1, 1, 1])
|
|
|
|
PullReshapeThroughFQ().find_and_replace_pattern(graph)
|
|
graph.clean_up()
|
|
graph_ref.clean_up()
|
|
|
|
(flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
|
|
self.assertTrue(flag, resp)
|
|
|
|
def test_reshape_reducing_tensor_rank(self):
|
|
graph, graph_ref = graph_template([3, 8, 7, 7], [24, 7, 7], [1, 1, 1])
|
|
|
|
PullReshapeThroughFQ().find_and_replace_pattern(graph)
|
|
graph.clean_up()
|
|
graph_ref.clean_up()
|
|
|
|
(flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
|
|
self.assertTrue(flag, resp)
|
|
|
|
|
|
class TestV7ConvolutionWithGroupsResolver(unittest.TestCase):
|
|
def test_v7_group_convolution_resolver(self):
|
|
nodes = {
|
|
**regular_op_with_shaped_data('input', [1, 3, 224, 224], {'type': 'Parameter'}),
|
|
|
|
**valued_const_with_data('weights', np.ones([3, 8, 7, 7])),
|
|
|
|
**valued_const_with_data('dim', int64_array([24, -1, 7, 7])),
|
|
**regular_op_with_empty_data('reshape', {'type': 'Reshape'}),
|
|
|
|
**regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 3, 'output': 24}),
|
|
|
|
**result(),
|
|
}
|
|
graph = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
V7ConvolutionWithGroupsResolver().find_and_replace_pattern(graph)
|
|
graph_ref = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '0:reshape'),
|
|
*connect('dim', '1:reshape'),
|
|
*connect('reshape', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
(flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
|
|
self.assertTrue(flag, resp)
|
|
|
|
def test_v7_group_convolution_resolver_weight_are_in_the_right_layout(self):
|
|
nodes = {
|
|
**regular_op_with_shaped_data('input', [1, 3, 224, 224], {'type': 'Parameter'}),
|
|
**valued_const_with_data('weights', np.ones([24, 1, 7, 7])),
|
|
**regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 3, 'output': 24}),
|
|
**result(),
|
|
}
|
|
edges = [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
]
|
|
graph = build_graph(nodes, edges)
|
|
V7ConvolutionWithGroupsResolver().find_and_replace_pattern(graph)
|
|
graph_ref = build_graph(nodes, edges)
|
|
(flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
|
|
self.assertTrue(flag, resp)
|
|
|
|
def test_v7_group_convolution_resolver_depthwise_conv2d(self):
|
|
nodes = {
|
|
**regular_op_with_shaped_data('input', [1, 1, 224, 224], {'type': 'Parameter'}),
|
|
|
|
**valued_const_with_data('weights', np.ones([1, 8, 7, 7])),
|
|
|
|
**valued_const_with_data('dim', int64_array([8, -1, 7, 7])),
|
|
**regular_op_with_empty_data('reshape', {'type': 'Reshape'}),
|
|
|
|
**regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 1, 'output': 8,
|
|
'op': 'DepthwiseConv2dNative'}),
|
|
|
|
**result(),
|
|
}
|
|
graph = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
V7ConvolutionWithGroupsResolver().find_and_replace_pattern(graph)
|
|
graph_ref = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '0:reshape'),
|
|
*connect('dim', '1:reshape'),
|
|
*connect('reshape', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
(flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
|
|
self.assertTrue(flag, resp)
|
|
|
|
|
|
class TestV10ConvolutionWithGroupsResolver(unittest.TestCase):
|
|
def test_v10_group_convolution_resolver(self):
|
|
nodes = {
|
|
**regular_op_with_shaped_data('input', [1, 3, 224, 224], {'type': 'Parameter'}),
|
|
|
|
**valued_const_with_data('weights', np.ones([3, 8, 7, 7])),
|
|
|
|
**valued_const_with_data('dim', int64_array([3, 8, 1, 7, 7])),
|
|
**regular_op_with_empty_data('reshape', {'type': 'Reshape'}),
|
|
|
|
**regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 3, 'output': 24}),
|
|
|
|
**result(),
|
|
}
|
|
graph = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
V10ConvolutionWithGroupsResolver().find_and_replace_pattern(graph)
|
|
|
|
nodes['convolution']['type'] = 'GroupConvolution'
|
|
del nodes['convolution']['group']
|
|
|
|
graph_ref = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '0:reshape'),
|
|
*connect('dim', '1:reshape'),
|
|
*connect('reshape', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
(flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
|
|
self.assertTrue(flag, resp)
|
|
|
|
def test_v10_group_convolution_resolver_depthwise_conv2d(self):
|
|
nodes = {
|
|
**regular_op_with_shaped_data('input', [1, 1, 224, 224], {'type': 'Parameter'}),
|
|
|
|
**valued_const_with_data('weights', np.ones([1, 8, 7, 7])),
|
|
|
|
**valued_const_with_data('dim', int64_array([1, 8, 1, 7, 7])),
|
|
**regular_op_with_empty_data('reshape', {'type': 'Reshape'}),
|
|
|
|
**regular_op_with_shaped_data('convolution', None, {'type': 'Convolution', 'group': 1, 'output': 8,
|
|
'op': 'DepthwiseConv2dNative'}),
|
|
|
|
**result(),
|
|
}
|
|
graph = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
V10ConvolutionWithGroupsResolver().find_and_replace_pattern(graph)
|
|
|
|
nodes['convolution']['type'] = 'GroupConvolution'
|
|
del nodes['convolution']['group']
|
|
|
|
graph_ref = build_graph(nodes, [
|
|
*connect('input', '0:convolution'),
|
|
*connect('weights', '0:reshape'),
|
|
*connect('dim', '1:reshape'),
|
|
*connect('reshape', '1:convolution'),
|
|
*connect('convolution', 'output'),
|
|
], nodes_with_edges_only=True)
|
|
|
|
(flag, resp) = compare_graphs(graph, graph_ref, last_node='output', check_op_attrs=True)
|
|
self.assertTrue(flag, resp)
|