model-optimizer/mo/front/common/partial_infer/eltwise_test.py

   1 """
   2  Copyright (c) 2018-2019 Intel Corporation
   3
   4  Licensed under the Apache License, Version 2.0 (the "License");
   5  you may not use this file except in compliance with the License.
   6  You may obtain a copy of the License at
   7
   8       http://www.apache.org/licenses/LICENSE-2.0
   9
  10  Unless required by applicable law or agreed to in writing, software
  11  distributed under the License is distributed on an "AS IS" BASIS,
  12  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  See the License for the specific language governing permissions and
  14  limitations under the License.
  15 """
  16
  17 import unittest
  18
  19 import numpy as np
  20
  21 from mo.front.common.partial_infer.eltwise import eltwise_infer
  22 from mo.graph.graph import Node
  23 from mo.utils.unittest.graph import build_graph
  24
  25 nodes_attributes = {'node_1': {'value': 2, 'kind': 'data'},
  26                     'node_2': {'value': 3, 'kind': 'data'},
  27                     'eltw_1': {'type': 'Eltwise', 'kind': 'op'},
  28                     'node_3': {'value': None, 'kind': 'data'},
  29                     'op_output': { 'kind': 'op', 'op': 'OpOutput'},
  30                     }
  31
  32
  33 class TestEltwiseInfer(unittest.TestCase):
  34     def test_eltwise_infer_max(self):
  35         graph = build_graph(nodes_attributes,
  36                             [('node_1', 'eltw_1'),
  37                              ('node_2', 'eltw_1'),
  38                              ('eltw_1', 'node_3'),
  39                              ('node_3', 'op_output')
  40                              ],
  41                             {'node_3': {'shape': None},
  42                              'node_1': {'shape': np.array([1, 3, 256, 256])},
  43                              'node_2': {'shape': np.array([1, 3, 256, 256])},
  44                              'eltw_1': {}
  45                              })
  46
  47         graph.graph['layout'] = 'NCHW'
  48
  49         eltwise_node = Node(graph, 'eltw_1')
  50
  51         eltwise_infer(eltwise_node, lambda a, b: np.maximum(a, b))
  52         exp_shape = np.array([1, 3, 256, 256])
  53         exp_value = 3
  54         res_shape = graph.node['node_3']['shape']
  55         res_value = eltwise_node.out_node().value
  56         for i in range(0, len(exp_shape)):
  57             self.assertEqual(exp_shape[i], res_shape[i])
  58
  59         self.assertEqual(exp_value, res_value)
  60
  61     def test_eltwise_infer_sum(self):
  62         graph = build_graph(nodes_attributes,
  63                             [('node_1', 'eltw_1'),
  64                              ('node_2', 'eltw_1'),
  65                              ('eltw_1', 'node_3'),
  66                              ('node_3', 'op_output')
  67                              ],
  68                             {'node_3': {'shape': None},
  69                              'node_1': {'shape': np.array([1, 3, 256, 256])},
  70                              'node_2': {'shape': np.array([1, 3, 256, 256])}
  71                              })
  72         graph.graph['layout'] = 'NCHW'
  73         eltwise_node = Node(graph, 'eltw_1')
  74
  75         eltwise_infer(eltwise_node, lambda a, b: a + b)
  76         exp_shape = np.array([1, 3, 256, 256])
  77         exp_value = 5
  78         res_shape = graph.node['node_3']['shape']
  79         res_value = eltwise_node.out_node().value
  80         for i in range(0, len(exp_shape)):
  81             self.assertEqual(exp_shape[i], res_shape[i])
  82
  83         self.assertEqual(exp_value, res_value)
  84
  85     def test_eltwise_infer_mul(self):
  86         graph = build_graph(nodes_attributes,
  87                             [('node_1', 'eltw_1'),
  88                              ('node_2', 'eltw_1'),
  89                              ('eltw_1', 'node_3'),
  90                              ('node_3', 'op_output')
  91                              ],
  92                             {'node_3': {'shape': None},
  93                              'node_1': {'shape': np.array([1, 3, 256, 256])},
  94                              'node_2': {'shape': np.array([1, 3, 256, 256])}
  95                              })
  96         graph.graph['layout'] = 'NCHW'
  97         eltwise_node = Node(graph, 'eltw_1')
  98
  99         eltwise_infer(eltwise_node, lambda a, b: a * b)
 100         exp_shape = np.array([1, 3, 256, 256])
 101         exp_value = 6
 102         res_shape = graph.node['node_3']['shape']
 103         res_value = eltwise_node.out_node().value
 104         for i in range(0, len(exp_shape)):
 105             self.assertEqual(exp_shape[i], res_shape[i])
 106
 107         self.assertEqual(exp_value, res_value)
 108
 109     def test_eltwise_infer_none_val(self):
 110         graph = build_graph(nodes_attributes,
 111                             [('node_1', 'eltw_1'),
 112                              ('node_2', 'eltw_1'),
 113                              ('eltw_1', 'node_3'),
 114                              ('node_3', 'op_output')
 115                              ],
 116                             {'node_3': {'shape': None},
 117                              'node_1': {'shape': np.array([1, 3, 256, 256]), 'value': None},
 118                              'node_2': {'shape': np.array([1, 3, 256, 256])}
 119                              })
 120         graph.graph['layout'] = 'NCHW'
 121         eltwise_node = Node(graph, 'eltw_1')
 122
 123         eltwise_infer(eltwise_node, lambda a, b: a * b)
 124         exp_shape = np.array([1, 3, 256, 256])
 125         res_shape = graph.node['node_3']['shape']
 126         res_value = eltwise_node.out_node().value
 127         for i in range(0, len(exp_shape)):
 128             self.assertEqual(exp_shape[i], res_shape[i])
 129
 130         self.assertIsNone(res_value)
 131
 132     def test_eltwise_infer_none_min_max(self):
 133         graph = build_graph(nodes_attributes,
 134                             [('node_1', 'eltw_1'),
 135                              ('node_2', 'eltw_1'),
 136                              ('eltw_1', 'node_3'),
 137                              ('node_3', 'op_output')
 138                              ],
 139                             {'node_3': {'shape': None},
 140                              'node_1': {'shape': np.array([1, 3, 257, 256])},
 141                              'node_2': {'shape': np.array([1, 3, 256, 257])}
 142                              })
 143         graph.graph['layout'] = 'NCHW'
 144         eltwise_node = Node(graph, 'eltw_1')
 145
 146         eltwise_infer(eltwise_node)
 147         exp_shape = np.array([1, 3, -1, -1])
 148         res_shape = graph.node['node_3']['shape']
 149         for i in range(0, len(exp_shape)):
 150             self.assertEqual(exp_shape[i], res_shape[i])