• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7#     http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14# ==============================================================================
15"""Test configs for conv."""
16from __future__ import absolute_import
17from __future__ import division
18from __future__ import print_function
19
20import numpy as np
21import tensorflow.compat.v1 as tf
22from tensorflow.lite.testing.zip_test_utils import create_tensor_data
23from tensorflow.lite.testing.zip_test_utils import make_zip_of_tests
24from tensorflow.lite.testing.zip_test_utils import register_make_test_function
25
26
27@register_make_test_function()
28def make_conv_to_depthwiseconv_with_shared_weights_tests(options):
29  """Make a test where 2 Conv ops shared the same constant weight tensor."""
30
31  test_parameters = [{
32      "input_shape": [[1, 10, 10, 1]],
33      "filter_shape": [[3, 3]],
34      "strides": [[1, 1, 1, 1]],
35      "dilations": [[1, 1, 1, 1]],
36      "padding": ["SAME"],
37      "data_format": ["NHWC"],
38      "channel_multiplier": [3],
39  }]
40
41  def get_tensor_shapes(parameters):
42    input_shape = parameters["input_shape"]
43    filter_size = parameters["filter_shape"]
44    filter_shape = filter_size + [
45        input_shape[3], parameters["channel_multiplier"]
46    ]
47    return [input_shape, filter_shape]
48
49  def build_graph(parameters):
50    """Build a conv graph given `parameters`."""
51    input_shape, filter_shape = get_tensor_shapes(parameters)
52    input_tensor = tf.compat.v1.placeholder(
53        dtype=tf.float32, name="input", shape=input_shape)
54
55    # Construct a constant weights tensor which will be used by both Conv2D.
56    filter_tensor = tf.constant(
57        create_tensor_data(np.float32, filter_shape), dtype=tf.float32)
58    input_tensors = [input_tensor]
59
60    # Construct 2 Conv2D operations which use exactly the same input and
61    # weights.
62    result1 = tf.nn.conv2d(
63        input_tensor,
64        filter_tensor,
65        strides=parameters["strides"],
66        dilations=parameters["dilations"],
67        padding=parameters["padding"],
68        data_format=parameters["data_format"])
69    result2 = tf.nn.conv2d(
70        input_tensor,
71        filter_tensor,
72        strides=parameters["strides"],
73        dilations=parameters["dilations"],
74        padding=parameters["padding"],
75        data_format=parameters["data_format"])
76    # Add the 2 results up.
77    out = result1 + result2
78    return input_tensors, [out]
79
80  def build_inputs(parameters, sess, inputs, outputs):
81    # Build list of input values either containing 1 tensor (input) or 2 tensors
82    # (input, filter) based on whether filter is constant or variable input.
83    input_shape, unused_filter_shape = get_tensor_shapes(parameters)
84    values = [create_tensor_data(np.float32, input_shape)]
85    return values, sess.run(outputs, feed_dict=dict(zip(inputs, values)))
86
87  make_zip_of_tests(options, test_parameters, build_graph, build_inputs)
88