diff --git a/paddle/fluid/inference/CMakeLists.txt b/paddle/fluid/inference/CMakeLists.txt index 88723e24184cce240d4bccd1b4c15a1ea8f14ed6..19749d3590a249a2d219bad3578d69c8f7364c96 100644 --- a/paddle/fluid/inference/CMakeLists.txt +++ b/paddle/fluid/inference/CMakeLists.txt @@ -55,12 +55,8 @@ endif() # C inference API add_subdirectory(capi) -if(WITH_TESTING) - # tests/book depends the models that generated by python/paddle/fluid/tests/book - add_subdirectory(tests/book) - if(WITH_INFERENCE_API_TEST) +if(WITH_TESTING AND WITH_INFERENCE_API_TEST) add_subdirectory(tests/api) - endif() endif() if(NOT ON_INFER) diff --git a/paddle/fluid/inference/tests/book/CMakeLists.txt b/paddle/fluid/inference/tests/book/CMakeLists.txt deleted file mode 100644 index ced9584aecf36d68a1ecf8de8a94f3a0d2db0b7c..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/CMakeLists.txt +++ /dev/null @@ -1,50 +0,0 @@ -function(inference_test TARGET_NAME) - set(options "") - set(oneValueArgs "") - set(multiValueArgs ARGS) - cmake_parse_arguments(inference_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN}) - - set(arg_list "") - if(inference_test_ARGS) - foreach(arg ${inference_test_ARGS}) - list(APPEND arg_list "_${arg}") - endforeach() - else() - list(APPEND arg_list "_") - endif() - foreach(arg ${arg_list}) - string(REGEX REPLACE "^_$" "" arg "${arg}") - cc_test(test_inference_${TARGET_NAME}${arg} - SRCS test_inference_${TARGET_NAME}.cc - DEPS paddle_fluid_api - ARGS --dirname=${PYTHON_TESTS_DIR}/book/${TARGET_NAME}${arg}.inference.model) - set_tests_properties(test_inference_${TARGET_NAME}${arg} - PROPERTIES DEPENDS test_${TARGET_NAME}) - set_tests_properties(test_inference_${TARGET_NAME}${arg} - PROPERTIES LABELS "RUN_TYPE=DIST") - endforeach() -endfunction(inference_test) - -#################### -# Inference tests here depend on fluid/tests/book. If users want to run -# individual test with ctest, they need to run tests in fluid/tests/book -# first to generate saved model. -#################### -# This unittest is buggy! -#inference_test(fit_a_line) -inference_test(image_classification ARGS vgg resnet) -inference_test(label_semantic_roles) -inference_test(recognize_digits ARGS mlp conv) -inference_test(recommender_system) -#inference_test(rnn_encoder_decoder) -#inference_test(understand_sentiment ARGS conv) -inference_test(word2vec) - -# This is an unly work around to make this test run -# TODO(TJ): clean me up -cc_test(test_inference_nlp - SRCS test_inference_nlp.cc - DEPS paddle_fluid_api - ARGS - --model_path=${PADDLE_BINARY_DIR}/python/paddle/fluid/tests/book/recognize_digits_mlp.inference.model) -set_tests_properties(test_inference_nlp PROPERTIES LABELS "RUN_TYPE=DIST") diff --git a/paddle/fluid/inference/tests/book/test_inference_fit_a_line.cc b/paddle/fluid/inference/tests/book/test_inference_fit_a_line.cc deleted file mode 100644 index 52ef648913f4a12d04b259dd1e9ed29a9cb6eefe..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_fit_a_line.cc +++ /dev/null @@ -1,91 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - http://www.apache.org/licenses/LICENSE-2.0 -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" -#include "paddle/fluid/inference/tests/test_multi_thread_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); - -TEST(inference, fit_a_line) { - if (FLAGS_dirname.empty()) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - for (int num_threads : {1, 2}) { - std::vector> cpu_feeds; - cpu_feeds.resize(num_threads); - for (int i = 0; i < num_threads; ++i) { - auto* input = new paddle::framework::LoDTensor(); - // The second dim of the input tensor should be 13 - // The input data should be >= 0 - int64_t batch_size = 10; - SetupTensor(input, {batch_size, 13}, static_cast(0), - static_cast(10)); - cpu_feeds[i].push_back(input); - } - - std::vector> cpu_fetchs1; - cpu_fetchs1.resize(num_threads); - for (int i = 0; i < num_threads; ++i) { - auto* output = new paddle::framework::FetchType(); - cpu_fetchs1[i].push_back(output); - } - - // Run inference on CPU - LOG(INFO) << "--- CPU Runs (num_threads: " << num_threads << "): ---"; - if (num_threads == 1) { - TestInference(dirname, cpu_feeds[0], - cpu_fetchs1[0]); - } else { - TestMultiThreadInference( - dirname, cpu_feeds, cpu_fetchs1, num_threads); - } - -#ifdef PADDLE_WITH_CUDA - std::vector> cpu_fetchs2; - cpu_fetchs2.resize(num_threads); - for (int i = 0; i < num_threads; ++i) { - auto* output = new paddle::framework::FetchType(); - cpu_fetchs2[i].push_back(output); - } - - // Run inference on CUDA GPU - LOG(INFO) << "--- GPU Runs (num_threads: " << num_threads << "): ---"; - if (num_threads == 1) { - TestInference(dirname, cpu_feeds[0], - cpu_fetchs2[0]); - } else { - TestMultiThreadInference( - dirname, cpu_feeds, cpu_fetchs2, num_threads); - } - - for (int i = 0; i < num_threads; ++i) { - CheckError( - BOOST_GET(paddle::framework::LoDTensor, *cpu_fetchs1[i][0]), - BOOST_GET(paddle::framework::LoDTensor, *cpu_fetchs2[i][0])); - delete cpu_fetchs2[i][0]; - } -#endif - - for (int i = 0; i < num_threads; ++i) { - delete cpu_feeds[i][0]; - delete cpu_fetchs1[i][0]; - } - } // num_threads-loop -} diff --git a/paddle/fluid/inference/tests/book/test_inference_image_classification.cc b/paddle/fluid/inference/tests/book/test_inference_image_classification.cc deleted file mode 100644 index c0f101062431f9e9e36f39047e469b70e6e02694..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_image_classification.cc +++ /dev/null @@ -1,100 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); -DEFINE_string(fp16_dirname, "", "Directory of the float16 inference model."); -DEFINE_int32(batch_size, 1, "Batch size of input data"); -DEFINE_int32(repeat, 1, "Running the inference program repeat times"); -DEFINE_bool(skip_cpu, false, "Skip the cpu test"); - -TEST(inference, image_classification) { - if (FLAGS_dirname.empty() || FLAGS_batch_size < 1 || FLAGS_repeat < 1) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model " - "--batch_size=1 --repeat=1"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - const bool is_combined = false; - std::vector> feed_target_shapes = - GetFeedTargetShapes(dirname, is_combined); - - paddle::framework::LoDTensor input; - // Use normilized image pixels as input data, - // which should be in the range [0.0, 1.0]. - feed_target_shapes[0][0] = FLAGS_batch_size; - paddle::framework::DDim input_dims = - paddle::framework::make_ddim(feed_target_shapes[0]); - LOG(INFO) << input_dims; - SetupTensor(&input, input_dims, static_cast(0), - static_cast(1)); - std::vector cpu_feeds; - cpu_feeds.push_back(&input); - - paddle::framework::FetchType output1; - if (!FLAGS_skip_cpu) { - std::vector cpu_fetchs1; - cpu_fetchs1.push_back(&output1); - - // Run inference on CPU - LOG(INFO) << "--- CPU Runs: ---"; - LOG(INFO) << "Batch size is " << FLAGS_batch_size; - TestInference( - dirname, cpu_feeds, cpu_fetchs1, FLAGS_repeat, is_combined); - LOG(INFO) << BOOST_GET(paddle::framework::LoDTensor, output1).dims(); - } - -#ifdef PADDLE_WITH_CUDA - paddle::framework::FetchType output2; - std::vector cpu_fetchs2; - cpu_fetchs2.push_back(&output2); - - // Run inference on CUDA GPU - LOG(INFO) << "--- GPU Runs: ---"; - LOG(INFO) << "Batch size is " << FLAGS_batch_size; - TestInference( - dirname, cpu_feeds, cpu_fetchs2, FLAGS_repeat, is_combined); - LOG(INFO) << BOOST_GET(paddle::framework::LoDTensor, output2).dims(); - - if (!FLAGS_skip_cpu) { - CheckError(BOOST_GET(paddle::framework::LoDTensor, output1), - BOOST_GET(paddle::framework::LoDTensor, output2)); - } - - // float16 inference requires cuda GPUs with >= 5.3 compute capability - if (!FLAGS_fp16_dirname.empty() && - paddle::platform::GetCUDAComputeCapability(0) >= 53) { - paddle::framework::FetchType output3; - std::vector cpu_fetchs3; - cpu_fetchs3.push_back(&output3); - - LOG(INFO) << "--- GPU Runs in float16 mode: ---"; - LOG(INFO) << "Batch size is " << FLAGS_batch_size; - - TestInference( - FLAGS_fp16_dirname, cpu_feeds, cpu_fetchs3, FLAGS_repeat); - - CheckError(BOOST_GET(paddle::framework::LoDTensor, output2), - BOOST_GET(paddle::framework::LoDTensor, output3)); - } -#endif -} diff --git a/paddle/fluid/inference/tests/book/test_inference_label_semantic_roles.cc b/paddle/fluid/inference/tests/book/test_inference_label_semantic_roles.cc deleted file mode 100644 index c5bb24860ea220d1ddedbe4c8cd7405a1f9bc758..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_label_semantic_roles.cc +++ /dev/null @@ -1,89 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); - -TEST(inference, label_semantic_roles) { - if (FLAGS_dirname.empty()) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - paddle::framework::LoDTensor word, predicate, ctx_n2, ctx_n1, ctx_0, ctx_p1, - ctx_p2, mark; - paddle::framework::LoD lod{{0, 4, 10}}; - int64_t word_dict_len = 44068; - int64_t predicate_dict_len = 3162; - int64_t mark_dict_len = 2; - - SetupLoDTensor(&word, lod, static_cast(0), - static_cast(word_dict_len - 1)); - SetupLoDTensor(&predicate, lod, static_cast(0), - static_cast(predicate_dict_len - 1)); - SetupLoDTensor(&ctx_n2, lod, static_cast(0), - static_cast(word_dict_len - 1)); - SetupLoDTensor(&ctx_n1, lod, static_cast(0), - static_cast(word_dict_len - 1)); - SetupLoDTensor(&ctx_0, lod, static_cast(0), - static_cast(word_dict_len - 1)); - SetupLoDTensor(&ctx_p1, lod, static_cast(0), - static_cast(word_dict_len - 1)); - SetupLoDTensor(&ctx_p2, lod, static_cast(0), - static_cast(word_dict_len - 1)); - SetupLoDTensor(&mark, lod, static_cast(0), - static_cast(mark_dict_len - 1)); - - std::vector cpu_feeds; - cpu_feeds.push_back(&word); - cpu_feeds.push_back(&predicate); - cpu_feeds.push_back(&ctx_n2); - cpu_feeds.push_back(&ctx_n1); - cpu_feeds.push_back(&ctx_0); - cpu_feeds.push_back(&ctx_p1); - cpu_feeds.push_back(&ctx_p2); - cpu_feeds.push_back(&mark); - - paddle::framework::FetchType output1; - std::vector cpu_fetchs1; - cpu_fetchs1.push_back(&output1); - - // Run inference on CPU - TestInference(dirname, cpu_feeds, cpu_fetchs1); - auto output1_tensor = BOOST_GET(paddle::framework::LoDTensor, output1); - LOG(INFO) << output1_tensor.lod(); - LOG(INFO) << output1_tensor.dims(); - -#ifdef PADDLE_WITH_CUDA - paddle::framework::FetchType output2; - std::vector cpu_fetchs2; - cpu_fetchs2.push_back(&output2); - - // Run inference on CUDA GPU - TestInference(dirname, cpu_feeds, cpu_fetchs2); - auto output2_tensor = BOOST_GET(paddle::framework::LoDTensor, output2); - LOG(INFO) << output2_tensor.lod(); - LOG(INFO) << output2_tensor.dims(); - - CheckError(output1_tensor, output2_tensor); -#endif -} diff --git a/paddle/fluid/inference/tests/book/test_inference_nlp.cc b/paddle/fluid/inference/tests/book/test_inference_nlp.cc deleted file mode 100644 index 76d5063dba5a830d13cb024179f06cc06946ca85..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_nlp.cc +++ /dev/null @@ -1,244 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include -#include -#include // NOLINT -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" -#include "paddle/fluid/platform/cpu_helper.h" - -#include "paddle/fluid/framework/feed_fetch_method.h" - -DEFINE_string(model_path, "", "Directory of the inference model."); -DEFINE_string(data_file, "", "File of input index data."); -DEFINE_int32(repeat, 100, "Running the inference program repeat times"); -DEFINE_bool(prepare_vars, true, "Prepare variables before executor"); -DEFINE_int32(num_threads, 1, "Number of threads should be used"); -DECLARE_bool(use_mkldnn); -DECLARE_int32(paddle_num_threads); - -inline double GetCurrentMs() { - struct timeval time; - gettimeofday(&time, NULL); - return 1e+3 * time.tv_sec + 1e-3 * time.tv_usec; -} - -// This function just give dummy data for recognize_digits model. -size_t DummyData(std::vector* out) { - paddle::framework::LoDTensor input; - SetupTensor(&input, {1, 1, 28, 28}, -1.f, 1.f); - out->emplace_back(input); - return 1; -} - -// Load the input word index data from file and save into LodTensor. -// Return the size of words. -size_t LoadData(std::vector* out, - const std::string& filename) { - if (filename.empty()) { - return DummyData(out); - } - - size_t sz = 0; - std::fstream fin(filename); - std::string line; - out->clear(); - while (getline(fin, line)) { - std::istringstream iss(line); - std::vector ids; - std::string field; - while (getline(iss, field, ' ')) { - ids.push_back(stoi(field)); - } - if (ids.size() >= 1024) { - // Synced with NLP guys, they will ignore input larger then 1024 - continue; - } - - paddle::framework::LoDTensor words; - paddle::framework::LoD lod{{0, ids.size()}}; - words.set_lod(lod); - int64_t* pdata = words.mutable_data( - {static_cast(ids.size()), 1}, paddle::platform::CPUPlace()); - memcpy(pdata, ids.data(), words.numel() * sizeof(int64_t)); - out->emplace_back(words); - sz += ids.size(); - } - return sz; -} - -// Split input data samples into small pieces jobs as balanced as possible, -// according to the number of threads. -void SplitData( - const std::vector& datasets, - std::vector>* jobs, - const int num_threads) { - size_t s = 0; - jobs->resize(num_threads); - while (s < datasets.size()) { - for (auto it = jobs->begin(); it != jobs->end(); it++) { - it->emplace_back(&datasets[s]); - s++; - if (s >= datasets.size()) { - break; - } - } - } -} - -void ThreadRunInfer( - const int tid, paddle::framework::Scope* scope, - const std::vector>& jobs) { - // maybe framework:ProgramDesc is not thread-safe - paddle::platform::CPUPlace place; - paddle::framework::Executor executor(place); - auto& sub_scope = scope->NewScope(); - auto inference_program = - paddle::inference::Load(&executor, scope, FLAGS_model_path); - - auto ctx = executor.Prepare(*inference_program, /*block_id*/ 0); - executor.CreateVariables(*inference_program, &sub_scope, /*block_id*/ 0); - - const std::vector& feed_target_names = - inference_program->GetFeedTargetNames(); - const std::vector& fetch_target_names = - inference_program->GetFetchTargetNames(); - - PADDLE_ENFORCE_EQ(fetch_target_names.size(), 1UL); - std::map fetch_targets; - paddle::framework::FetchType outtensor; - fetch_targets[fetch_target_names[0]] = &outtensor; - - std::map feed_targets; - PADDLE_ENFORCE_EQ(feed_target_names.size(), 1UL); - - // map the data of feed_targets to feed_holder - for (auto* op : inference_program->Block(0).AllOps()) { - if (op->Type() == "feed") { - std::string feed_target_name = op->Output("Out")[0]; - int idx = BOOST_GET_CONST(int, op->GetAttr("col")); - paddle::framework::SetFeedVariable(scope, *feed_targets[feed_target_name], - "feed", idx); - } - } - - auto& inputs = jobs[tid]; - auto start_ms = GetCurrentMs(); - for (size_t i = 0; i < inputs.size(); ++i) { - feed_targets[feed_target_names[0]] = inputs[i]; - executor.RunPreparedContext(ctx.get(), &sub_scope, - false /*create_local_scope*/); - } - auto stop_ms = GetCurrentMs(); - - // obtain the data of fetch_targets from fetch_holder - for (auto* op : inference_program->Block(0).AllOps()) { - if (op->Type() == "fetch") { - std::string fetch_target_name = op->Input("X")[0]; - int idx = BOOST_GET_CONST(int, op->GetAttr("col")); - *fetch_targets[fetch_target_name] = - BOOST_GET(paddle::framework::LoDTensor, - paddle::framework::GetFetchVariable(*scope, "fetch", idx)); - } - } - - scope->DeleteScope(&sub_scope); - LOG(INFO) << "Tid: " << tid << ", process " << inputs.size() - << " samples, avg time per sample: " - << (stop_ms - start_ms) / inputs.size() << " ms"; -} - -TEST(inference, nlp) { - if (FLAGS_model_path.empty()) { - LOG(FATAL) << "Usage: ./example --model_path=path/to/your/model"; - } - if (FLAGS_data_file.empty()) { - LOG(WARNING) << "No data file provided, will use dummy data!" - << "Note: if you use nlp model, please provide data file."; - } - LOG(INFO) << "Model Path: " << FLAGS_model_path; - LOG(INFO) << "Data File: " << FLAGS_data_file; - - std::vector datasets; - size_t num_total_words = LoadData(&datasets, FLAGS_data_file); - LOG(INFO) << "Number of samples (seq_len<1024): " << datasets.size(); - LOG(INFO) << "Total number of words: " << num_total_words; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - std::unique_ptr scope( - new paddle::framework::Scope()); - - paddle::platform::SetNumThreads(FLAGS_paddle_num_threads); - - double start_ms = 0, stop_ms = 0; - if (FLAGS_num_threads > 1) { - std::vector> jobs; - SplitData(datasets, &jobs, FLAGS_num_threads); - std::vector> threads; - start_ms = GetCurrentMs(); - for (int i = 0; i < FLAGS_num_threads; ++i) { - threads.emplace_back( - new std::thread(ThreadRunInfer, i, scope.get(), std::ref(jobs))); - } - for (int i = 0; i < FLAGS_num_threads; ++i) { - threads[i]->join(); - } - stop_ms = GetCurrentMs(); - } else { - // 1. Define place, executor, scope - paddle::platform::CPUPlace place; - paddle::framework::Executor executor(place); - - // 2. Initialize the inference_program and load parameters - std::unique_ptr inference_program; - inference_program = InitProgram(&executor, scope.get(), FLAGS_model_path, - /*model combined*/ false); - // always prepare context - std::unique_ptr ctx; - ctx = executor.Prepare(*inference_program, 0); - if (FLAGS_prepare_vars) { - executor.CreateVariables(*inference_program, scope.get(), 0); - } - // preapre fetch - const std::vector& fetch_target_names = - inference_program->GetFetchTargetNames(); - PADDLE_ENFORCE_EQ(fetch_target_names.size(), 1UL); - std::map fetch_targets; - paddle::framework::FetchType outtensor; - fetch_targets[fetch_target_names[0]] = &outtensor; - - // prepare feed - const std::vector& feed_target_names = - inference_program->GetFeedTargetNames(); - PADDLE_ENFORCE_EQ(feed_target_names.size(), 1UL); - std::map feed_targets; - - // feed data and run - start_ms = GetCurrentMs(); - for (size_t i = 0; i < datasets.size(); ++i) { - feed_targets[feed_target_names[0]] = &(datasets[i]); - executor.RunPreparedContext(ctx.get(), scope.get(), &feed_targets, - &fetch_targets, !FLAGS_prepare_vars); - } - stop_ms = GetCurrentMs(); - LOG(INFO) << "Tid: 0, process " << datasets.size() - << " samples, avg time per sample: " - << (stop_ms - start_ms) / datasets.size() << " ms"; - } - LOG(INFO) << "Total inference time with " << FLAGS_num_threads - << " threads : " << (stop_ms - start_ms) / 1000.0 - << " sec, QPS: " << datasets.size() / ((stop_ms - start_ms) / 1000); -} diff --git a/paddle/fluid/inference/tests/book/test_inference_recognize_digits.cc b/paddle/fluid/inference/tests/book/test_inference_recognize_digits.cc deleted file mode 100644 index 423cfa2f500951753249a04cce3018d1c0ec725e..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_recognize_digits.cc +++ /dev/null @@ -1,70 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); -DEFINE_int32(batch_size, 1, "Batch size of input data"); -DEFINE_int32(repeat, 1, "Running the inference program repeat times"); - -TEST(inference, recognize_digits) { - if (FLAGS_dirname.empty() || FLAGS_batch_size < 1 || FLAGS_repeat < 1) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model " - "--batch_size=1 --repeat=1"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - paddle::framework::LoDTensor input; - // Use normilized image pixels as input data, - // which should be in the range [-1.0, 1.0]. - SetupTensor(&input, {FLAGS_batch_size, 1, 28, 28}, - static_cast(-1), static_cast(1)); - std::vector cpu_feeds; - cpu_feeds.push_back(&input); - - for (auto is_combined : {false, true}) { - paddle::framework::FetchType output1; - std::vector cpu_fetchs1; - cpu_fetchs1.push_back(&output1); - - // Run inference on CPU - LOG(INFO) << "--- CPU Runs: is_combined=" << is_combined << " ---"; - TestInference(dirname, cpu_feeds, cpu_fetchs1, - FLAGS_repeat, is_combined); - auto output1_tensor = BOOST_GET(paddle::framework::LoDTensor, output1); - LOG(INFO) << output1_tensor.dims(); - -#ifdef PADDLE_WITH_CUDA - paddle::framework::FetchType output2; - std::vector cpu_fetchs2; - cpu_fetchs2.push_back(&output2); - - // Run inference on CUDA GPU - LOG(INFO) << "--- GPU Runs: is_combined=" << is_combined << " ---"; - TestInference(dirname, cpu_feeds, cpu_fetchs2, - FLAGS_repeat, is_combined); - auto output2_tensor = BOOST_GET(paddle::framework::LoDTensor, output2); - LOG(INFO) << output2_tensor.dims(); - - CheckError(output1_tensor, output2_tensor); -#endif - } -} diff --git a/paddle/fluid/inference/tests/book/test_inference_recommender_system.cc b/paddle/fluid/inference/tests/book/test_inference_recommender_system.cc deleted file mode 100644 index 0315f797b70a3078932dd524fae618cfd2688a25..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_recommender_system.cc +++ /dev/null @@ -1,89 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); - -TEST(inference, recommender_system) { - if (FLAGS_dirname.empty()) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - int64_t batch_size = 1; - - paddle::framework::LoDTensor user_id, gender_id, age_id, job_id, movie_id, - category_id, movie_title; - - // Use the first data from paddle.dataset.movielens.test() as input - std::vector user_id_data = {1}; - SetupTensor(&user_id, {batch_size, 1}, user_id_data); - - std::vector gender_id_data = {1}; - SetupTensor(&gender_id, {batch_size, 1}, gender_id_data); - - std::vector age_id_data = {0}; - SetupTensor(&age_id, {batch_size, 1}, age_id_data); - - std::vector job_id_data = {10}; - SetupTensor(&job_id, {batch_size, 1}, job_id_data); - - std::vector movie_id_data = {783}; - SetupTensor(&movie_id, {batch_size, 1}, movie_id_data); - - std::vector category_id_data = {10, 8, 9}; - SetupLoDTensor(&category_id, {3, 1}, {{0, 3}}, category_id_data); - - std::vector movie_title_data = {1069, 4140, 2923, 710, 988}; - SetupLoDTensor(&movie_title, {5, 1}, {{0, 5}}, movie_title_data); - - std::vector cpu_feeds; - cpu_feeds.push_back(&user_id); - cpu_feeds.push_back(&gender_id); - cpu_feeds.push_back(&age_id); - cpu_feeds.push_back(&job_id); - cpu_feeds.push_back(&movie_id); - cpu_feeds.push_back(&category_id); - cpu_feeds.push_back(&movie_title); - - paddle::framework::FetchType output1; - std::vector cpu_fetchs1; - cpu_fetchs1.push_back(&output1); - - // Run inference on CPU - TestInference(dirname, cpu_feeds, cpu_fetchs1); - auto output1_tensor = BOOST_GET(paddle::framework::LoDTensor, output1); - LOG(INFO) << output1_tensor.dims(); - -#ifdef PADDLE_WITH_CUDA - paddle::framework::FetchType output2; - std::vector cpu_fetchs2; - cpu_fetchs2.push_back(&output2); - - // Run inference on CUDA GPU - TestInference(dirname, cpu_feeds, cpu_fetchs2); - auto output2_tensor = BOOST_GET(paddle::framework::LoDTensor, output2); - LOG(INFO) << output2_tensor.dims(); - - CheckError(output1_tensor, output2_tensor); -#endif -} diff --git a/paddle/fluid/inference/tests/book/test_inference_rnn_encoder_decoder.cc b/paddle/fluid/inference/tests/book/test_inference_rnn_encoder_decoder.cc deleted file mode 100644 index 9a6ed68641120dad22fc07aa9bae61f18d9d1a22..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_rnn_encoder_decoder.cc +++ /dev/null @@ -1,67 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); - -TEST(inference, rnn_encoder_decoder) { - if (FLAGS_dirname.empty()) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - paddle::framework::LoDTensor word_data, trg_word; - paddle::framework::LoD lod{{0, 4, 10}}; - - SetupLoDTensor(&word_data, lod, static_cast(0), - static_cast(1)); - SetupLoDTensor(&trg_word, lod, static_cast(0), - static_cast(1)); - - std::vector cpu_feeds; - cpu_feeds.push_back(&word_data); - cpu_feeds.push_back(&trg_word); - - paddle::framework::FetchType output1; - std::vector cpu_fetchs1; - cpu_fetchs1.push_back(&output1); - - // Run inference on CPU - TestInference(dirname, cpu_feeds, cpu_fetchs1); - auto output1_tensor = BOOST_GET(paddle::framework::LoDTensor, output1); - LOG(INFO) << output1_tensor.lod(); - LOG(INFO) << output1_tensor.dims(); - -#ifdef PADDLE_WITH_CUDA - paddle::framework::FetchType output2; - std::vector cpu_fetchs2; - cpu_fetchs2.push_back(&output2); - - // Run inference on CUDA GPU - TestInference(dirname, cpu_feeds, cpu_fetchs2); - auto output2_tensor = BOOST_GET(paddle::framework::LoDTensor, output2); - LOG(INFO) << output2_tensor.lod(); - LOG(INFO) << output2_tensor.dims(); - - CheckError(output1_tensor, output2_tensor); -#endif -} diff --git a/paddle/fluid/inference/tests/book/test_inference_understand_sentiment.cc b/paddle/fluid/inference/tests/book/test_inference_understand_sentiment.cc deleted file mode 100644 index cd86838ad1ea2220bad76450183308aa4f4bb50f..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_understand_sentiment.cc +++ /dev/null @@ -1,65 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); - -TEST(inference, understand_sentiment) { - if (FLAGS_dirname.empty()) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - paddle::framework::LoDTensor words; - paddle::framework::LoD lod{{0, 4, 10}}; - int64_t word_dict_len = 5147; - - SetupLoDTensor(&words, lod, static_cast(0), - static_cast(word_dict_len - 1)); - - std::vector cpu_feeds; - cpu_feeds.push_back(&words); - - paddle::framework::FetchType output1; - std::vector cpu_fetchs1; - cpu_fetchs1.push_back(&output1); - - // Run inference on CPU - TestInference(dirname, cpu_feeds, cpu_fetchs1); - auto output1_tensor = BOOST_GET(paddle::framework::LoDTensor, output1); - LOG(INFO) << output1_tensor.lod(); - LOG(INFO) << output1_tensor.dims(); - -#ifdef PADDLE_WITH_CUDA - paddle::framework::FetchType output2; - std::vector cpu_fetchs2; - cpu_fetchs2.push_back(&output2); - - // Run inference on CUDA GPU - TestInference(dirname, cpu_feeds, cpu_fetchs2); - auto output2_tensor = BOOST_GET(paddle::framework::LoDTensor, output2); - LOG(INFO) << output2_tensor.lod(); - LOG(INFO) << output2_tensor.dims(); - - CheckError(output1_tensor, output2_tensor); -#endif -} diff --git a/paddle/fluid/inference/tests/book/test_inference_word2vec.cc b/paddle/fluid/inference/tests/book/test_inference_word2vec.cc deleted file mode 100644 index 3faca4126bc4ec116621e0ed3c3255425f3e45b8..0000000000000000000000000000000000000000 --- a/paddle/fluid/inference/tests/book/test_inference_word2vec.cc +++ /dev/null @@ -1,70 +0,0 @@ -/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. - -Licensed under the Apache License, Version 2.0 (the "License"); -you may not use this file except in compliance with the License. -You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - -Unless required by applicable law or agreed to in writing, software -distributed under the License is distributed on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -See the License for the specific language governing permissions and -limitations under the License. */ - -#include "gflags/gflags.h" -#include "gtest/gtest.h" -#include "paddle/fluid/inference/tests/test_helper.h" - -DEFINE_string(dirname, "", "Directory of the inference model."); - -TEST(inference, word2vec) { - if (FLAGS_dirname.empty()) { - LOG(FATAL) << "Usage: ./example --dirname=path/to/your/model"; - } - - LOG(INFO) << "FLAGS_dirname: " << FLAGS_dirname << std::endl; - std::string dirname = FLAGS_dirname; - - // 0. Call `paddle::framework::InitDevices()` initialize all the devices - // In unittests, this is done in paddle/testing/paddle_gtest_main.cc - - paddle::framework::LoDTensor first_word, second_word, third_word, fourth_word; - paddle::framework::LoD lod{{0, 1}}; - int64_t dict_size = 2073; // The size of dictionary - - SetupLoDTensor(&first_word, lod, static_cast(0), dict_size - 1); - SetupLoDTensor(&second_word, lod, static_cast(0), dict_size - 1); - SetupLoDTensor(&third_word, lod, static_cast(0), dict_size - 1); - SetupLoDTensor(&fourth_word, lod, static_cast(0), dict_size - 1); - - std::vector cpu_feeds; - cpu_feeds.push_back(&first_word); - cpu_feeds.push_back(&second_word); - cpu_feeds.push_back(&third_word); - cpu_feeds.push_back(&fourth_word); - - paddle::framework::FetchType output1; - std::vector cpu_fetchs1; - cpu_fetchs1.push_back(&output1); - - // Run inference on CPU - TestInference(dirname, cpu_feeds, cpu_fetchs1); - auto output1_tensor = BOOST_GET(paddle::framework::LoDTensor, output1); - LOG(INFO) << output1_tensor.lod(); - LOG(INFO) << output1_tensor.dims(); - -#ifdef PADDLE_WITH_CUDA - paddle::framework::FetchType output2; - std::vector cpu_fetchs2; - cpu_fetchs2.push_back(&output2); - - // Run inference on CUDA GPU - TestInference(dirname, cpu_feeds, cpu_fetchs2); - auto output2_tensor = BOOST_GET(paddle::framework::LoDTensor, output2); - LOG(INFO) << output2_tensor.lod(); - LOG(INFO) << output2_tensor.dims(); - - CheckError(output1_tensor, output2_tensor); -#endif -}