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 strided_slice_np_style.""" 16from __future__ import absolute_import 17from __future__ import division 18from __future__ import print_function 19 20import tensorflow.compat.v1 as tf 21from tensorflow.lite.testing.zip_test_utils import create_tensor_data 22from tensorflow.lite.testing.zip_test_utils import make_zip_of_tests 23from tensorflow.lite.testing.zip_test_utils import register_make_test_function 24 25 26# TODO(b/137615945): Expand the test coverage of this one and remove the old 27# ones. 28@register_make_test_function() 29def make_strided_slice_np_style_tests(options): 30 """Make a set of tests to test strided_slice in np style.""" 31 32 test_parameters = [ 33 { 34 "dtype": [tf.float32], 35 "shape": [[12, 7], [33, 1]], 36 "spec": [[slice(3, 7, 2), slice(None)], 37 [tf.newaxis, 38 slice(3, 7, 1), tf.newaxis, 39 slice(None)], [slice(1, 5, 1), slice(None)]], 40 }, 41 # 1-D case 42 { 43 "dtype": [tf.float32], 44 "shape": [[44]], 45 "spec": [[slice(3, 7, 2)], [tf.newaxis, slice(None)]], 46 }, 47 # Shrink mask. 48 { 49 "dtype": [tf.float32], 50 "shape": [[21, 15, 7]], 51 "spec": [[slice(3, 7, 2), slice(None), 2]], 52 }, 53 # Ellipsis 3d. 54 { 55 "dtype": [tf.float32], 56 "shape": [[21, 15, 7]], 57 "spec": [[slice(3, 7, 2), Ellipsis], 58 [slice(1, 11, 3), Ellipsis, 59 slice(3, 7, 2)]], 60 }, 61 # Ellipsis 4d. 62 { 63 "dtype": [tf.float32], 64 "shape": [[21, 15, 7, 9]], 65 "spec": [[slice(3, 7, 2), Ellipsis]], 66 }, 67 # Ellipsis 5d. 68 { 69 "dtype": [tf.float32], 70 "shape": [[11, 21, 15, 7, 9]], 71 "spec": [[ 72 slice(3, 7, 2), 73 slice(None), 74 slice(None), 75 slice(None), 76 slice(None) 77 ]], 78 }, 79 # Ellipsis + Shrink Mask 80 { 81 "dtype": [tf.float32], 82 "shape": [[22, 15, 7]], 83 "spec": [ 84 [ 85 2, # shrink before ellipsis 86 Ellipsis 87 ], 88 ], 89 }, 90 # Ellipsis + New Axis Mask 91 { 92 "dtype": [tf.float32], 93 "shape": [[23, 15, 7]], 94 "spec": [ 95 [ 96 tf.newaxis, # new_axis before ellipsis 97 slice(3, 7, 2), 98 slice(None), 99 Ellipsis 100 ], 101 [ 102 tf.newaxis, # new_axis after (and before) ellipsis 103 slice(3, 7, 2), 104 slice(None), 105 Ellipsis, 106 tf.newaxis 107 ] 108 ], 109 }, 110 ] 111 112 if options.use_experimental_converter: 113 # The case when Ellipsis is expanded to multiple dimension is only supported 114 # by MLIR converter (b/183902491). 115 test_parameters = test_parameters + [ 116 # Ellipsis 3d. 117 { 118 "dtype": [tf.float32], 119 "shape": [[21, 15, 7]], 120 "spec": [[Ellipsis, slice(3, 7, 2)]], 121 }, 122 # Ellipsis 4d. 123 { 124 "dtype": [tf.float32], 125 "shape": [[21, 15, 7, 9]], 126 "spec": [[Ellipsis, slice(3, 7, 2)], 127 [slice(1, 11, 3), Ellipsis, 128 slice(3, 7, 2)]], 129 }, 130 # Ellipsis 5d. 131 { 132 "dtype": [tf.float32], 133 "shape": [[11, 21, 15, 7, 9]], 134 "spec": [[Ellipsis, slice(3, 7, 2)]], 135 }, 136 # Ellipsis + Shrink Mask 137 { 138 "dtype": [tf.float32], 139 "shape": [[22, 15, 7]], 140 "spec": [[ 141 Ellipsis, # shrink after ellipsis 142 2 143 ]], 144 }, 145 ] 146 147 def build_graph(parameters): 148 """Build a simple graph with np style strided_slice.""" 149 input_value = tf.compat.v1.placeholder( 150 dtype=parameters["dtype"], shape=parameters["shape"]) 151 out = input_value.__getitem__(parameters["spec"]) 152 return [input_value], [out] 153 154 def build_inputs(parameters, sess, inputs, outputs): 155 input_value = create_tensor_data(parameters["dtype"], parameters["shape"]) 156 return [input_value], sess.run( 157 outputs, feed_dict=dict(zip(inputs, [input_value]))) 158 159 make_zip_of_tests(options, test_parameters, build_graph, build_inputs) 160