1 /** 2 * Copyright 2019-2021 Huawei Technologies Co., Ltd 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 #include <chrono> 17 #include <cstring> 18 #include <iostream> 19 #include <memory> 20 #include <string> 21 #include "minddata/dataset/core/client.h" 22 #include "minddata/dataset/core/tensor.h" 23 #include "minddata/dataset/core/config_manager.h" 24 #include "minddata/dataset/engine/datasetops/zip_op.h" 25 #include "minddata/dataset/engine/jagged_connector.h" 26 #include "common/common.h" 27 #include "utils/ms_utils.h" 28 29 #include "gtest/gtest.h" 30 #include "utils/log_adapter.h" 31 32 namespace common = mindspore::common; 33 34 using namespace mindspore::dataset; 35 using mindspore::LogStream; 36 using mindspore::ExceptionType::NoExceptionType; 37 using mindspore::MsLogLevel::INFO; 38 39 class MindDataTestZipOp : public UT::DatasetOpTesting {}; 40 41 TEST_F(MindDataTestZipOp, MindDataTestZipOpDefault) { 42 /* Tree: 43 * 44 * 45 * OpId(2) ZipOp 46 * / \ 47 * OpId(0) TFReaderOp OpId(1) TFReaderOp 48 * Start with an empty execution tree 49 */ 50 Status rc; 51 MS_LOG(INFO) << "UT test TestZipBasic."; 52 auto my_tree = std::make_shared<ExecutionTree>(); 53 // Creating TFReaderOp 54 55 std::string dataset_path = datasets_root_path_ + "/test_tf_file_3_images/train-0000-of-0001.data"; 56 std::string dataset_path2 = datasets_root_path_ + "/testBatchDataset/test.data"; 57 std::shared_ptr<ConfigManager> config_manager = GlobalContext::config_manager(); 58 std::vector<std::string> columns_to_load = {}; 59 std::unique_ptr<DataSchema> schema = std::make_unique<DataSchema>(); 60 std::vector<std::string> files1 = {dataset_path}; 61 auto op_connector_size = config_manager->op_connector_size(); 62 std::shared_ptr<TFReaderOp> my_tfreader_op = std::make_shared<TFReaderOp>( 63 1, 16, 0, files1, std::move(schema), op_connector_size, columns_to_load, false, 1, 0, false); 64 rc = my_tfreader_op->Init(); 65 EXPECT_TRUE(rc.IsOk()); 66 rc = my_tree->AssociateNode(my_tfreader_op); 67 EXPECT_TRUE(rc.IsOk()); 68 std::vector<std::string> files2 = {dataset_path2}; 69 std::unique_ptr<DataSchema> schema2 = std::make_unique<DataSchema>(); 70 std::shared_ptr<TFReaderOp> my_tfreader_op2 = std::make_shared<TFReaderOp>( 71 1, 1, 0, files2, std::make_unique<DataSchema>(), op_connector_size, columns_to_load, false, 1, 0, false); 72 rc = my_tfreader_op2->Init(); 73 EXPECT_TRUE(rc.IsOk()); 74 rc = my_tree->AssociateNode(my_tfreader_op2); 75 EXPECT_TRUE(rc.IsOk()); 76 77 // Creating DatasetOp 78 std::shared_ptr<ZipOp> zip_op = std::make_shared<ZipOp>(); 79 80 rc = my_tree->AssociateNode(zip_op); 81 EXPECT_TRUE(rc.IsOk()); 82 rc = zip_op->AddChild(std::move(my_tfreader_op)); 83 EXPECT_TRUE(rc.IsOk()); 84 rc = zip_op->AddChild(std::move(my_tfreader_op2)); 85 EXPECT_TRUE(rc.IsOk()); 86 rc = my_tree->AssignRoot(zip_op); 87 EXPECT_TRUE(rc.IsOk()); 88 rc = my_tree->Prepare(); 89 EXPECT_TRUE(rc.IsOk()); 90 91 // Launch the tree execution to kick off threads and start running the pipeline 92 MS_LOG(INFO) << "Launching my tree."; 93 rc = my_tree->Launch(); 94 EXPECT_TRUE(rc.IsOk()); 95 96 // Simulate a parse of data from our pipeline. 97 std::shared_ptr<DatasetOp> rootNode = my_tree->root(); 98 99 DatasetIterator di(my_tree); 100 TensorRow tensor_list; 101 rc = di.FetchNextTensorRow(&tensor_list); 102 EXPECT_TRUE(rc.IsOk()); 103 104 int row_count = 0; 105 while (!tensor_list.empty()) { 106 MS_LOG(INFO) << "Row display for row #: " << row_count << "."; 107 108 // Display the tensor by calling the printer on it 109 for (int i = 0; i < tensor_list.size(); i++) { 110 std::ostringstream ss; 111 ss << "(" << tensor_list[i] << "): " << *tensor_list[i] << std::endl; 112 MS_LOG(INFO) << "Tensor print: " << common::SafeCStr(ss.str()) << "."; 113 } 114 rc = di.FetchNextTensorRow(&tensor_list); 115 EXPECT_TRUE(rc.IsOk()); 116 row_count++; 117 } 118 ASSERT_EQ(row_count, 3); // Should be 3 rows fetched 119 } 120 121 TEST_F(MindDataTestZipOp, MindDataTestZipOpRepeat) { 122 /* Tree: 123 * OpId(3) Repeat(3) 124 * 125 * OpId(2) ZipOp 126 * / \ 127 * OpId(0) TFReaderOp OpId(1) TFReaderOp 128 * 129 * Start with an empty execution tree 130 */ 131 Status rc; 132 MS_LOG(INFO) << "UT test TestZipRepeat."; 133 auto my_tree = std::make_shared<ExecutionTree>(); 134 135 uint32_t num_repeats = 3; 136 std::string dataset_path = datasets_root_path_ + "/test_tf_file_3_images/train-0000-of-0001.data"; 137 std::string dataset_path2 = datasets_root_path_ + "/testBatchDataset/test.data"; 138 std::shared_ptr<ConfigManager> config_manager = GlobalContext::config_manager(); 139 auto op_connector_size = config_manager->op_connector_size(); 140 std::vector<std::string> columns_to_load = {}; 141 std::vector<std::string> files1 = {dataset_path}; 142 std::unique_ptr<DataSchema> schema1 = std::make_unique<DataSchema>(); 143 std::shared_ptr<TFReaderOp> my_tfreader_op = std::make_shared<TFReaderOp>( 144 1, 16, 0, files1, std::move(schema1), op_connector_size, columns_to_load, false, 1, 0, false); 145 rc = my_tfreader_op->Init(); 146 EXPECT_TRUE(rc.IsOk()); 147 148 rc = my_tree->AssociateNode(my_tfreader_op); 149 150 rc = my_tree->AssociateNode(my_tfreader_op); 151 EXPECT_TRUE(rc.IsOk()); 152 std::vector<std::string> files2 = {dataset_path2}; 153 std::unique_ptr<DataSchema> schema2 = std::make_unique<DataSchema>(); 154 std::shared_ptr<TFReaderOp> my_tfreader_op2 = std::make_shared<TFReaderOp>( 155 1, 1, 0, files2, std::move(schema2), op_connector_size, columns_to_load, false, 1, 0, false); 156 rc = my_tfreader_op2->Init(); 157 EXPECT_TRUE(rc.IsOk()); 158 159 rc = my_tree->AssociateNode(my_tfreader_op2); 160 EXPECT_TRUE(rc.IsOk()); 161 // Creating DatasetOp 162 std::shared_ptr<ZipOp> zip_op = std::make_shared<ZipOp>(); 163 rc = my_tree->AssociateNode(zip_op); 164 EXPECT_TRUE(rc.IsOk()); 165 my_tfreader_op->SetTotalRepeats(num_repeats); 166 my_tfreader_op->SetNumRepeatsPerEpoch(num_repeats); 167 rc = zip_op->AddChild(std::move(my_tfreader_op)); 168 EXPECT_TRUE(rc.IsOk()); 169 my_tfreader_op2->SetTotalRepeats(num_repeats); 170 my_tfreader_op2->SetNumRepeatsPerEpoch(num_repeats); 171 rc = zip_op->AddChild(std::move(my_tfreader_op2)); 172 EXPECT_TRUE(rc.IsOk()); 173 174 std::shared_ptr<RepeatOp> my_repeat_op = std::make_shared<RepeatOp>(num_repeats); 175 rc = my_tree->AssociateNode(my_repeat_op); 176 EXPECT_TRUE(rc.IsOk()); 177 zip_op->SetTotalRepeats(num_repeats); 178 zip_op->SetNumRepeatsPerEpoch(num_repeats); 179 rc = my_repeat_op->AddChild(zip_op); 180 EXPECT_TRUE(rc.IsOk()); 181 rc = my_tree->AssignRoot(my_repeat_op); 182 EXPECT_TRUE(rc.IsOk()); 183 rc = my_tree->Prepare(); 184 EXPECT_TRUE(rc.IsOk()); 185 186 // Launch the tree execution to kick off threads and start running the pipeline 187 MS_LOG(INFO) << "Launching my tree."; 188 rc = my_tree->Launch(); 189 EXPECT_TRUE(rc.IsOk()); 190 191 // Simulate a parse of data from our pipeline. 192 std::shared_ptr<DatasetOp> rootNode = my_tree->root(); 193 194 DatasetIterator di(my_tree); 195 TensorRow tensor_list; 196 rc = di.FetchNextTensorRow(&tensor_list); 197 EXPECT_TRUE(rc.IsOk()); 198 199 int row_count = 0; 200 while (!tensor_list.empty()) { 201 MS_LOG(INFO) << "Row display for row #: " << row_count << "."; 202 203 // Display the tensor by calling the printer on it 204 for (int i = 0; i < tensor_list.size(); i++) { 205 std::ostringstream ss; 206 ss << "(" << tensor_list[i] << "): " << *tensor_list[i] << std::endl; 207 MS_LOG(INFO) << "Tensor print: " << common::SafeCStr(ss.str()) << "."; 208 } 209 rc = di.FetchNextTensorRow(&tensor_list); 210 EXPECT_TRUE(rc.IsOk()); 211 row_count++; 212 } 213 ASSERT_EQ(row_count, 9); // Should be 9 rows fetched 214 } 215