From 3da1dded89e42ea82e4d2da53b26fbbef72a7aee Mon Sep 17 00:00:00 2001 From: unknown Date: Thu, 27 May 2021 10:44:56 +0800 Subject: [PATCH] deepfm && wide_and_deep 310 infer modified: deepfm/README.md modified: deepfm/README_CN.md modified: deepfm/postprocess.py modified: deepfm/preprocess.py modified: wide_and_deep/postprocess.py modified: wide_and_deep/preprocess.py --- model_zoo/official/recommend/deepfm/README.md | 35 +++++ .../official/recommend/deepfm/README_CN.md | 35 +++++ .../deepfm/ascend310_infer/CMakeLists.txt | 14 ++ .../recommend/deepfm/ascend310_infer/build.sh | 29 ++++ .../deepfm/ascend310_infer/inc/utils.h | 32 ++++ .../deepfm/ascend310_infer/src/main.cc | 140 ++++++++++++++++++ .../deepfm/ascend310_infer/src/utils.cc | 129 ++++++++++++++++ .../official/recommend/deepfm/postprocess.py | 52 +++++++ .../official/recommend/deepfm/preprocess.py | 57 +++++++ .../recommend/deepfm/scripts/run_infer_310.sh | 122 +++++++++++++++ .../recommend/wide_and_deep/README.md | 36 +++++ .../recommend/wide_and_deep/README_CN.md | 44 ++++++ .../ascend310_infer/CMakeLists.txt | 14 ++ .../wide_and_deep/ascend310_infer/build.sh | 29 ++++ .../wide_and_deep/ascend310_infer/inc/utils.h | 32 ++++ .../wide_and_deep/ascend310_infer/src/main.cc | 140 ++++++++++++++++++ .../ascend310_infer/src/utils.cc | 129 ++++++++++++++++ .../recommend/wide_and_deep/postprocess.py | 53 +++++++ .../recommend/wide_and_deep/preprocess.py | 63 ++++++++ .../wide_and_deep/script/run_infer_310.sh | 125 ++++++++++++++++ 20 files changed, 1310 insertions(+) create mode 100644 model_zoo/official/recommend/deepfm/ascend310_infer/CMakeLists.txt create mode 100644 model_zoo/official/recommend/deepfm/ascend310_infer/build.sh create mode 100644 model_zoo/official/recommend/deepfm/ascend310_infer/inc/utils.h create mode 100644 model_zoo/official/recommend/deepfm/ascend310_infer/src/main.cc create mode 100644 model_zoo/official/recommend/deepfm/ascend310_infer/src/utils.cc create mode 100644 model_zoo/official/recommend/deepfm/postprocess.py create mode 100644 model_zoo/official/recommend/deepfm/preprocess.py create mode 100644 model_zoo/official/recommend/deepfm/scripts/run_infer_310.sh create mode 100644 model_zoo/official/recommend/wide_and_deep/ascend310_infer/CMakeLists.txt create mode 100644 model_zoo/official/recommend/wide_and_deep/ascend310_infer/build.sh create mode 100644 model_zoo/official/recommend/wide_and_deep/ascend310_infer/inc/utils.h create mode 100644 model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/main.cc create mode 100644 model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/utils.cc create mode 100644 model_zoo/official/recommend/wide_and_deep/postprocess.py create mode 100644 model_zoo/official/recommend/wide_and_deep/preprocess.py create mode 100644 model_zoo/official/recommend/wide_and_deep/script/run_infer_310.sh diff --git a/model_zoo/official/recommend/deepfm/README.md b/model_zoo/official/recommend/deepfm/README.md index cb9f5fd60a4..2cb73859c01 100644 --- a/model_zoo/official/recommend/deepfm/README.md +++ b/model_zoo/official/recommend/deepfm/README.md @@ -13,6 +13,10 @@ - [Distributed Training](#distributed-training) - [Evaluation Process](#evaluation-process) - [Evaluation](#evaluation) + - [Inference Process](#inference-process) + - [Export MindIR](#export-mindir) + - [Infer on Ascend310](#infer-on-ascend310) + - [result](#result) - [Model Description](#model-description) - [Performance](#performance) - [Evaluation Performance](#evaluation-performance) @@ -269,6 +273,37 @@ Parameters for both training and evaluation can be set in config.py To do. +## Inference Process + +### [Export MindIR](#contents) + +```shell +python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT] +``` + +The ckpt_file parameter is required, +`EXPORT_FORMAT` should be in ["AIR", "MINDIR"] + +### Infer on Ascend310 + +Before performing inference, the mindir file must be exported by `export.py` script. We only provide an example of inference using MINDIR model. + +```shell +# Ascend310 inference +bash run_infer_310.sh [MINDIR_PATH] [DATASET_PATH] [NEED_PREPROCESS] [DEVICE_ID] +``` + +- `NEED_PREPROCESS` means weather need preprocess or not, it's value is 'y' or 'n'. +- `DEVICE_ID` is optional, default value is 0. + +### result + +Inference result is saved in current path, you can find result like this in acc.log file. + +```bash +auc : 0.8057789065281104 +``` + # [Model Description](#contents) ## [Performance](#contents) diff --git a/model_zoo/official/recommend/deepfm/README_CN.md b/model_zoo/official/recommend/deepfm/README_CN.md index 1bc9b8940db..c549570131c 100644 --- a/model_zoo/official/recommend/deepfm/README_CN.md +++ b/model_zoo/official/recommend/deepfm/README_CN.md @@ -16,6 +16,10 @@ - [分布式训练](#分布式训练) - [评估过程](#评估过程) - [评估](#评估) + - [推理过程](#推理过程) + - [导出MindIR](#导出mindir) + - [在Ascend310执行推理](#在ascend310执行推理) + - [结果](#结果) - [模型描述](#模型描述) - [性能](#性能) - [评估性能](#评估性能) @@ -251,6 +255,37 @@ FM和深度学习部分拥有相同的输入原样特征向量，让DeepFM能从 - 在GPU运行时评估数据集 待运行。 +## 推理过程 + +### [导出MindIR](#contents) + +```shell +python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT] +``` + +参数ckpt_file为必填项， +`EXPORT_FORMAT` 必须在 ["AIR", "MINDIR"]中选择。 + +### 在Ascend310执行推理 + +在执行推理前，mindir文件必须通过`export.py`脚本导出。以下展示了使用minir模型执行推理的示例。 + +```shell +# Ascend310 推理 +bash run_infer_310.sh [MINDIR_PATH] [DATASET_PATH] [NEED_PREPROCESS] [DEVICE_ID] +``` + +- `NEED_PREPROCESS` 表示数据是否需要预处理，取值范围为 'y' 或者 'n'。 +- `DEVICE_ID` 可选，默认值为0。 + +### 结果 + +推理结果保存在脚本执行的当前路径，你可以在acc.log中看到以下精度计算结果。 + +```bash +auc : 0.8057789065281104 +``` + ## 模型描述 ## 性能 diff --git a/model_zoo/official/recommend/deepfm/ascend310_infer/CMakeLists.txt b/model_zoo/official/recommend/deepfm/ascend310_infer/CMakeLists.txt new file mode 100644 index 00000000000..170e6c5275e --- /dev/null +++ b/model_zoo/official/recommend/deepfm/ascend310_infer/CMakeLists.txt @@ -0,0 +1,14 @@ +cmake_minimum_required(VERSION 3.14.1) +project(Ascend310Infer) +add_compile_definitions(_GLIBCXX_USE_CXX11_ABI=0) +set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O2 -g -std=c++17 -Werror -Wall -fPIE -Wl,--allow-shlib-undefined") +set(PROJECT_SRC_ROOT ${CMAKE_CURRENT_LIST_DIR}/) +option(MINDSPORE_PATH "mindspore install path" "") +include_directories(${MINDSPORE_PATH}) +include_directories(${MINDSPORE_PATH}/include) +include_directories(${PROJECT_SRC_ROOT}) +find_library(MS_LIB libmindspore.so ${MINDSPORE_PATH}/lib) +file(GLOB_RECURSE MD_LIB ${MINDSPORE_PATH}/_c_dataengine*) + +add_executable(main src/main.cc src/utils.cc) +target_link_libraries(main ${MS_LIB} ${MD_LIB} gflags) diff --git a/model_zoo/official/recommend/deepfm/ascend310_infer/build.sh b/model_zoo/official/recommend/deepfm/ascend310_infer/build.sh new file mode 100644 index 00000000000..285514e19f2 --- /dev/null +++ b/model_zoo/official/recommend/deepfm/ascend310_infer/build.sh @@ -0,0 +1,29 @@ +#!/bin/bash +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ +if [ -d out ]; then + rm -rf out +fi + +mkdir out +cd out || exit + +if [ -f "Makefile" ]; then + make clean +fi + +cmake .. \ + -DMINDSPORE_PATH="`pip3.7 show mindspore-ascend | grep Location | awk '{print $2"/mindspore"}' | xargs realpath`" +make diff --git a/model_zoo/official/recommend/deepfm/ascend310_infer/inc/utils.h b/model_zoo/official/recommend/deepfm/ascend310_infer/inc/utils.h new file mode 100644 index 00000000000..efebe03a8c1 --- /dev/null +++ b/model_zoo/official/recommend/deepfm/ascend310_infer/inc/utils.h @@ -0,0 +1,32 @@ +/** + * Copyright 2021 Huawei Technologies Co., Ltd + * + * 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. + */ + +#ifndef MINDSPORE_INFERENCE_UTILS_H_ +#define MINDSPORE_INFERENCE_UTILS_H_ + +#include +#include +#include +#include +#include +#include "include/api/types.h" + +std::vector GetAllFiles(std::string_view dirName); +DIR *OpenDir(std::string_view dirName); +std::string RealPath(std::string_view path); +mindspore::MSTensor ReadFileToTensor(const std::string &file); +int WriteResult(const std::string& imageFile, const std::vector &outputs); +#endif diff --git a/model_zoo/official/recommend/deepfm/ascend310_infer/src/main.cc b/model_zoo/official/recommend/deepfm/ascend310_infer/src/main.cc new file mode 100644 index 00000000000..3879ed4d9b9 --- /dev/null +++ b/model_zoo/official/recommend/deepfm/ascend310_infer/src/main.cc @@ -0,0 +1,140 @@ +/** + * Copyright 2021 Huawei Technologies Co., Ltd + * + * 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 +#include +#include +#include +#include +#include +#include +#include + +#include "include/api/model.h" +#include "include/api/context.h" +#include "include/api/types.h" +#include "include/api/serialization.h" +#include "include/dataset/execute.h" +#include "include/dataset/vision.h" +#include "inc/utils.h" + +using mindspore::Context; +using mindspore::Serialization; +using mindspore::Model; +using mindspore::Status; +using mindspore::MSTensor; +using mindspore::dataset::Execute; +using mindspore::ModelType; +using mindspore::GraphCell; +using mindspore::kSuccess; + +DEFINE_string(mindir_path, "", "mindir path"); +DEFINE_string(input0_path, ".", "input0 path"); +DEFINE_string(input1_path, ".", "input1 path"); +DEFINE_string(input2_path, ".", "input2 path"); +DEFINE_int32(device_id, 0, "device id"); + +int main(int argc, char **argv) { + gflags::ParseCommandLineFlags(&argc, &argv, true); + if (RealPath(FLAGS_mindir_path).empty()) { + std::cout << "Invalid mindir" << std::endl; + return 1; + } + + auto context = std::make_shared(); + auto ascend310 = std::make_shared(); + ascend310->SetDeviceID(FLAGS_device_id); + context->MutableDeviceInfo().push_back(ascend310); + mindspore::Graph graph; + Serialization::Load(FLAGS_mindir_path, ModelType::kMindIR, &graph); + + Model model; + Status ret = model.Build(GraphCell(graph), context); + if (ret != kSuccess) { + std::cout << "ERROR: Build failed." << std::endl; + return 1; + } + + std::vector model_inputs = model.GetInputs(); + if (model_inputs.empty()) { + std::cout << "Invalid model, inputs is empty." << std::endl; + return 1; + } + + auto input0_files = GetAllFiles(FLAGS_input0_path); + auto input1_files = GetAllFiles(FLAGS_input1_path); + auto input2_files = GetAllFiles(FLAGS_input2_path); + + if (input0_files.empty() || input1_files.empty() || input2_files.empty()) { + std::cout << "ERROR: input data empty." << std::endl; + return 1; + } + + std::map costTime_map; + size_t size = input0_files.size(); + + for (size_t i = 0; i < size; ++i) { + struct timeval start = {0}; + struct timeval end = {0}; + double startTimeMs; + double endTimeMs; + std::vector inputs; + std::vector outputs; + std::cout << "Start predict input files:" << input0_files[i] << std::endl; + + auto input0 = ReadFileToTensor(input0_files[i]); + auto input1 = ReadFileToTensor(input1_files[i]); + auto input2 = ReadFileToTensor(input2_files[i]); + inputs.emplace_back(model_inputs[0].Name(), model_inputs[0].DataType(), model_inputs[0].Shape(), + input0.Data().get(), input0.DataSize()); + inputs.emplace_back(model_inputs[1].Name(), model_inputs[1].DataType(), model_inputs[1].Shape(), + input1.Data().get(), input1.DataSize()); + inputs.emplace_back(model_inputs[2].Name(), model_inputs[2].DataType(), model_inputs[2].Shape(), + input2.Data().get(), input2.DataSize()); + + gettimeofday(&start, nullptr); + ret = model.Predict(inputs, &outputs); + gettimeofday(&end, nullptr); + if (ret != kSuccess) { + std::cout << "Predict " << input0_files[i] << " failed." << std::endl; + return 1; + } + startTimeMs = (1.0 * start.tv_sec * 1000000 + start.tv_usec) / 1000; + endTimeMs = (1.0 * end.tv_sec * 1000000 + end.tv_usec) / 1000; + costTime_map.insert(std::pair(startTimeMs, endTimeMs)); + WriteResult(input0_files[i], outputs); + } + double average = 0.0; + int inferCount = 0; + + for (auto iter = costTime_map.begin(); iter != costTime_map.end(); iter++) { + double diff = 0.0; + diff = iter->second - iter->first; + average += diff; + inferCount++; + } + average = average / inferCount; + std::stringstream timeCost; + timeCost << "NN inference cost average time: "<< average << " ms of infer_count " << inferCount << std::endl; + std::cout << "NN inference cost average time: "<< average << "ms of infer_count " << inferCount << std::endl; + std::string fileName = "./time_Result" + std::string("/test_perform_static.txt"); + std::ofstream fileStream(fileName.c_str(), std::ios::trunc); + fileStream << timeCost.str(); + fileStream.close(); + costTime_map.clear(); + return 0; +} diff --git a/model_zoo/official/recommend/deepfm/ascend310_infer/src/utils.cc b/model_zoo/official/recommend/deepfm/ascend310_infer/src/utils.cc new file mode 100644 index 00000000000..c947e4d5f45 --- /dev/null +++ b/model_zoo/official/recommend/deepfm/ascend310_infer/src/utils.cc @@ -0,0 +1,129 @@ +/** + * Copyright 2021 Huawei Technologies Co., Ltd + * + * 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 +#include "inc/utils.h" + +using mindspore::MSTensor; +using mindspore::DataType; + +std::vector GetAllFiles(std::string_view dirName) { + struct dirent *filename; + DIR *dir = OpenDir(dirName); + if (dir == nullptr) { + return {}; + } + std::vector res; + while ((filename = readdir(dir)) != nullptr) { + std::string dName = std::string(filename->d_name); + if (dName == "." || dName == ".." || filename->d_type != DT_REG) { + continue; + } + res.emplace_back(std::string(dirName) + "/" + filename->d_name); + } + std::sort(res.begin(), res.end()); + for (auto &f : res) { + std::cout << "image file: " << f << std::endl; + } + return res; +} + +int WriteResult(const std::string& imageFile, const std::vector &outputs) { + std::string homePath = "./result_Files"; + for (size_t i = 0; i < outputs.size(); ++i) { + size_t outputSize; + std::shared_ptr netOutput; + netOutput = outputs[i].Data(); + outputSize = outputs[i].DataSize(); + int pos = imageFile.rfind('/'); + std::string fileName(imageFile, pos + 1); + fileName.replace(fileName.find('.'), fileName.size() - fileName.find('.'), '_' + std::to_string(i) + ".bin"); + std::string outFileName = homePath + "/" + fileName; + FILE * outputFile = fopen(outFileName.c_str(), "wb"); + fwrite(netOutput.get(), outputSize, sizeof(char), outputFile); + fclose(outputFile); + outputFile = nullptr; + } + return 0; +} + +mindspore::MSTensor ReadFileToTensor(const std::string &file) { + if (file.empty()) { + std::cout << "Pointer file is nullptr" << std::endl; + return mindspore::MSTensor(); + } + + std::ifstream ifs(file); + if (!ifs.good()) { + std::cout << "File: " << file << " is not exist" << std::endl; + return mindspore::MSTensor(); + } + + if (!ifs.is_open()) { + std::cout << "File: " << file << "open failed" << std::endl; + return mindspore::MSTensor(); + } + + ifs.seekg(0, std::ios::end); + size_t size = ifs.tellg(); + mindspore::MSTensor buffer(file, mindspore::DataType::kNumberTypeUInt8, {static_cast(size)}, nullptr, size); + + ifs.seekg(0, std::ios::beg); + ifs.read(reinterpret_cast(buffer.MutableData()), size); + ifs.close(); + + return buffer; +} + + +DIR *OpenDir(std::string_view dirName) { + if (dirName.empty()) { + std::cout << " dirName is null ! " << std::endl; + return nullptr; + } + std::string realPath = RealPath(dirName); + struct stat s; + lstat(realPath.c_str(), &s); + if (!S_ISDIR(s.st_mode)) { + std::cout << "dirName is not a valid directory !" << std::endl; + return nullptr; + } + DIR *dir; + dir = opendir(realPath.c_str()); + if (dir == nullptr) { + std::cout << "Can not open dir " << dirName << std::endl; + return nullptr; + } + std::cout << "Successfully opened the dir " << dirName << std::endl; + return dir; +} + +std::string RealPath(std::string_view path) { + char realPathMem[PATH_MAX] = {0}; + char *realPathRet = nullptr; + realPathRet = realpath(path.data(), realPathMem); + + if (realPathRet == nullptr) { + std::cout << "File: " << path << " is not exist."; + return ""; + } + + std::string realPath(realPathMem); + std::cout << path << " realpath is: " << realPath << std::endl; + return realPath; +} diff --git a/model_zoo/official/recommend/deepfm/postprocess.py b/model_zoo/official/recommend/deepfm/postprocess.py new file mode 100644 index 00000000000..98da65046c3 --- /dev/null +++ b/model_zoo/official/recommend/deepfm/postprocess.py @@ -0,0 +1,52 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ +"""postprocess.""" +import os +import argparse +import numpy as np +from mindspore import Tensor +from src.deepfm import AUCMetric +from src.config import TrainConfig + +parser = argparse.ArgumentParser(description='postprocess') +parser.add_argument('--result_path', type=str, default="./result_Files", help='result path') +parser.add_argument('--label_path', type=str, default=None, help='label path') +args_opt, _ = parser.parse_known_args() + +def get_acc(): + ''' get accuracy ''' + auc_metric = AUCMetric() + train_config = TrainConfig() + files = os.listdir(args_opt.label_path) + batch_size = train_config.batch_size + + for f in files: + rst_file = os.path.join(args_opt.result_path, f.split('.')[0] + '_0.bin') + label_file = os.path.join(args_opt.label_path, f) + + logit = Tensor(np.fromfile(rst_file, np.float32).reshape(batch_size, 1)) + label = Tensor(np.fromfile(label_file, np.float32).reshape(batch_size, 1)) + + res = [] + res.append(logit) + res.append(logit) + res.append(label) + + auc_metric.update(*res) + auc = auc_metric.eval() + print("auc : {}".format(auc)) + +if __name__ == '__main__': + get_acc() diff --git a/model_zoo/official/recommend/deepfm/preprocess.py b/model_zoo/official/recommend/deepfm/preprocess.py new file mode 100644 index 00000000000..75a3bea0822 --- /dev/null +++ b/model_zoo/official/recommend/deepfm/preprocess.py @@ -0,0 +1,57 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ +"""preprocess.""" +import os +import argparse + +from src.config import DataConfig, TrainConfig +from src.dataset import create_dataset, DataType + +parser = argparse.ArgumentParser(description='preprocess.') +parser.add_argument('--dataset_path', type=str, default=None, help='Dataset path') +parser.add_argument('--result_path', type=str, default='./preprocess_Result', help='Result path') +args_opt, _ = parser.parse_known_args() + +def generate_bin(): + '''generate bin files''' + data_config = DataConfig() + train_config = TrainConfig() + + ds = create_dataset(args_opt.dataset_path, train_mode=False, + epochs=1, batch_size=train_config.batch_size, + data_type=DataType(data_config.data_format)) + batch_ids_path = os.path.join(args_opt.result_path, "00_batch_ids") + batch_wts_path = os.path.join(args_opt.result_path, "01_batch_wts") + labels_path = os.path.join(args_opt.result_path, "02_labels") + + os.makedirs(batch_ids_path) + os.makedirs(batch_wts_path) + os.makedirs(labels_path) + + for i, data in enumerate(ds.create_dict_iterator(output_numpy=True)): + file_name = "criteo_bs" + str(train_config.batch_size) + "_" + str(i) + ".bin" + batch_ids = data['feat_ids'] + batch_ids.tofile(os.path.join(batch_ids_path, file_name)) + + batch_wts = data['feat_vals'] + batch_wts.tofile(os.path.join(batch_wts_path, file_name)) + + labels = data['label'] + labels.tofile(os.path.join(labels_path, file_name)) + + print("=" * 20, "export bin files finished", "=" * 20) + +if __name__ == '__main__': + generate_bin() diff --git a/model_zoo/official/recommend/deepfm/scripts/run_infer_310.sh b/model_zoo/official/recommend/deepfm/scripts/run_infer_310.sh new file mode 100644 index 00000000000..e4c291e7119 --- /dev/null +++ b/model_zoo/official/recommend/deepfm/scripts/run_infer_310.sh @@ -0,0 +1,122 @@ +#!/bin/bash +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ + +if [[ $# -lt 3 || $# -gt 4 ]]; then + echo "Usage: bash run_infer_310.sh [MINDIR_PATH] [DATASET_PATH] [NEED_PREPROCESS] [DEVICE_ID] + NEED_PREPROCESS means weather need preprocess or not, it's value is 'y' or 'n'. + DEVICE_ID is optional, it can be set by environment variable device_id, otherwise the value is zero" +exit 1 +fi + +get_real_path(){ + if [ "${1:0:1}" == "/" ]; then + echo "$1" + else + echo "$(realpath -m $PWD/$1)" + fi +} +model=$(get_real_path $1) +dataset_path=$(get_real_path $2) + +if [ "$3" == "y" ] || [ "$3" == "n" ];then + need_preprocess=$3 +else + echo "weather need preprocess or not, it's value must be in [y, n]" + exit 1 +fi + +device_id=0 +if [ $# == 4 ]; then + device_id=$4 +fi + +echo "mindir name: "$model +echo "dataset path: "$dataset_path +echo "need preprocess: "$need_preprocess +echo "device id: "$device_id + +export ASCEND_HOME=/usr/local/Ascend/ +if [ -d ${ASCEND_HOME}/ascend-toolkit ]; then + export PATH=$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/ccec_compiler/bin:$ASCEND_HOME/ascend-toolkit/latest/atc/bin:$PATH + export LD_LIBRARY_PATH=/usr/local/lib:$ASCEND_HOME/ascend-toolkit/latest/atc/lib64:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH + export TBE_IMPL_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp/op_impl/built-in/ai_core/tbe + export PYTHONPATH=${TBE_IMPL_PATH}:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/python/site-packages:$PYTHONPATH + export ASCEND_OPP_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp +else + export PATH=$ASCEND_HOME/atc/ccec_compiler/bin:$ASCEND_HOME/atc/bin:$PATH + export LD_LIBRARY_PATH=/usr/local/lib:$ASCEND_HOME/atc/lib64:$ASCEND_HOME/acllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH + export PYTHONPATH=$ASCEND_HOME/atc/python/site-packages:$PYTHONPATH + export ASCEND_OPP_PATH=$ASCEND_HOME/opp +fi + +function preprocess_data() +{ + if [ -d preprocess_Result ]; then + rm -rf ./preprocess_Result + fi + mkdir preprocess_Result + python3.7 ../preprocess.py --dataset_path=$dataset_path --result_path=./preprocess_Result/ +} + +function compile_app() +{ + cd ../ascend310_infer || exit + bash build.sh &> build.log +} + +function infer() +{ + cd - || exit + if [ -d result_Files ]; then + rm -rf ./result_Files + fi + if [ -d time_Result ]; then + rm -rf ./time_Result + fi + mkdir result_Files + mkdir time_Result + + ../ascend310_infer/out/main --mindir_path=$model --input0_path=./preprocess_Result/00_batch_ids --input1_path=./preprocess_Result/01_batch_wts --input2_path=./preprocess_Result/02_labels --device_id=$device_id &> infer.log + +} + +function cal_acc() +{ + python3.7 ../postprocess.py --result_path=./result_Files --label_path=./preprocess_Result/02_labels &> acc.log +} + +if [ $need_preprocess == "y" ]; then + preprocess_data + if [ $? -ne 0 ]; then + echo "preprocess dataset failed" + exit 1 + fi +fi +compile_app +if [ $? -ne 0 ]; then + echo "compile app code failed" + exit 1 +fi +infer +if [ $? -ne 0 ]; then + echo " execute inference failed" + exit 1 +fi +cal_acc +if [ $? -ne 0 ]; then + echo "calculate accuracy failed" + exit 1 +fi \ No newline at end of file diff --git a/model_zoo/official/recommend/wide_and_deep/README.md b/model_zoo/official/recommend/wide_and_deep/README.md index 1faa920dff4..8ac4cab6231 100644 --- a/model_zoo/official/recommend/wide_and_deep/README.md +++ b/model_zoo/official/recommend/wide_and_deep/README.md @@ -19,6 +19,10 @@ - [Distribute Training](#distribute-training) - [Parameter Server](#parameter-server) - [Evaluation Process](#evaluation-process) + - [Inference Process](#inference-process) + - [Export MindIR](#export-mindir) + - [Infer on Ascend310](#infer-on-ascend310) + - [result](#result) - [Model Description](#model-description) - [Performance](#performance) - [Training Performance](#training-performance) @@ -317,6 +321,38 @@ To evaluate the model, command as follows: python eval.py ``` +## Inference Process + +### [Export MindIR](#contents) + +```shell +python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT] +``` + +The ckpt_file parameter is required, +`EXPORT_FORMAT` should be in ["AIR", "MINDIR"] + +### Infer on Ascend310 + +Before performing inference, the mindir file must be exported by `export.py` script. We only provide an example of inference using MINDIR model. + +```shell +# Ascend310 inference +bash run_infer_310.sh [MINDIR_PATH] [DATASET_PATH] [DATA_TYPE] [NEED_PREPROCESS] [DEVICE_ID] +``` + +- `DATA_TYPE` means dataset type, it's value is ['tfrecord', 'mindrecord', 'hd5']. +- `NEED_PREPROCESS` means weather need preprocess or not, it's value is 'y' or 'n'. +- `DEVICE_ID` is optional, default value is 0. + +### result + +Inference result is saved in current path, you can find result like this in acc.log file. + +```bash +================================================================================ auc : 0.8080494136248402 +``` + # [Model Description](#contents) ## [Performance](#contents) diff --git a/model_zoo/official/recommend/wide_and_deep/README_CN.md b/model_zoo/official/recommend/wide_and_deep/README_CN.md index 5b8179e738a..3da389f7dfd 100644 --- a/model_zoo/official/recommend/wide_and_deep/README_CN.md +++ b/model_zoo/official/recommend/wide_and_deep/README_CN.md @@ -19,6 +19,10 @@ - [分布式训练](#分布式训练) - [参数服务器](#参数服务器) - [评估过程](#评估过程) + - [推理过程](#推理过程) + - [导出MindIR](#导出mindir) + - [在Ascend310执行推理](#在ascend310执行推理) + - [结果](#结果) - [模型描述](#模型描述) - [性能](#性能) - [训练性能](#训练性能) @@ -319,6 +323,46 @@ bash run_parameter_server_train.sh RANK_SIZE EPOCHS DATASET RANK_TABLE_FILE SERV python eval.py ``` +## [Evaluation Process](#contents) + +To evaluate the model, command as follows: + +```python +python eval.py +``` + +## 推理过程 + +### [导出MindIR](#contents) + +```shell +python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT] +``` + +参数ckpt_file为必填项， +`EXPORT_FORMAT` 必须在 ["AIR", "MINDIR"]中选择。 + +### 在Ascend310执行推理 + +在执行推理前，mindir文件必须通过`export.py`脚本导出。以下展示了使用minir模型执行推理的示例。 + +```shell +# Ascend310 推理 +bash run_infer_310.sh [MINDIR_PATH] [DATASET_PATH] [DATA_TYPE] [NEED_PREPROCESS] [DEVICE_ID] +``` + +- `DATA_TYPE` 表示数据类型, 取值范围为 ['tfrecord', 'mindrecord', 'hd5']。 +- `NEED_PREPROCESS` 表示数据是否需要预处理，取值范围为 'y' 或者 'n'。 +- `DEVICE_ID` 可选，默认值为0。 + +### result + +推理结果保存在脚本执行的当前路径，你可以在acc.log中看到以下精度计算结果。 + +```bash +================================================================================ auc : 0.8080494136248402 +``` + # 模型描述 ## 性能 diff --git a/model_zoo/official/recommend/wide_and_deep/ascend310_infer/CMakeLists.txt b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/CMakeLists.txt new file mode 100644 index 00000000000..170e6c5275e --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/CMakeLists.txt @@ -0,0 +1,14 @@ +cmake_minimum_required(VERSION 3.14.1) +project(Ascend310Infer) +add_compile_definitions(_GLIBCXX_USE_CXX11_ABI=0) +set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O2 -g -std=c++17 -Werror -Wall -fPIE -Wl,--allow-shlib-undefined") +set(PROJECT_SRC_ROOT ${CMAKE_CURRENT_LIST_DIR}/) +option(MINDSPORE_PATH "mindspore install path" "") +include_directories(${MINDSPORE_PATH}) +include_directories(${MINDSPORE_PATH}/include) +include_directories(${PROJECT_SRC_ROOT}) +find_library(MS_LIB libmindspore.so ${MINDSPORE_PATH}/lib) +file(GLOB_RECURSE MD_LIB ${MINDSPORE_PATH}/_c_dataengine*) + +add_executable(main src/main.cc src/utils.cc) +target_link_libraries(main ${MS_LIB} ${MD_LIB} gflags) diff --git a/model_zoo/official/recommend/wide_and_deep/ascend310_infer/build.sh b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/build.sh new file mode 100644 index 00000000000..285514e19f2 --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/build.sh @@ -0,0 +1,29 @@ +#!/bin/bash +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ +if [ -d out ]; then + rm -rf out +fi + +mkdir out +cd out || exit + +if [ -f "Makefile" ]; then + make clean +fi + +cmake .. \ + -DMINDSPORE_PATH="`pip3.7 show mindspore-ascend | grep Location | awk '{print $2"/mindspore"}' | xargs realpath`" +make diff --git a/model_zoo/official/recommend/wide_and_deep/ascend310_infer/inc/utils.h b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/inc/utils.h new file mode 100644 index 00000000000..efebe03a8c1 --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/inc/utils.h @@ -0,0 +1,32 @@ +/** + * Copyright 2021 Huawei Technologies Co., Ltd + * + * 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. + */ + +#ifndef MINDSPORE_INFERENCE_UTILS_H_ +#define MINDSPORE_INFERENCE_UTILS_H_ + +#include +#include +#include +#include +#include +#include "include/api/types.h" + +std::vector GetAllFiles(std::string_view dirName); +DIR *OpenDir(std::string_view dirName); +std::string RealPath(std::string_view path); +mindspore::MSTensor ReadFileToTensor(const std::string &file); +int WriteResult(const std::string& imageFile, const std::vector &outputs); +#endif diff --git a/model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/main.cc b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/main.cc new file mode 100644 index 00000000000..3879ed4d9b9 --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/main.cc @@ -0,0 +1,140 @@ +/** + * Copyright 2021 Huawei Technologies Co., Ltd + * + * 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 +#include +#include +#include +#include +#include +#include +#include + +#include "include/api/model.h" +#include "include/api/context.h" +#include "include/api/types.h" +#include "include/api/serialization.h" +#include "include/dataset/execute.h" +#include "include/dataset/vision.h" +#include "inc/utils.h" + +using mindspore::Context; +using mindspore::Serialization; +using mindspore::Model; +using mindspore::Status; +using mindspore::MSTensor; +using mindspore::dataset::Execute; +using mindspore::ModelType; +using mindspore::GraphCell; +using mindspore::kSuccess; + +DEFINE_string(mindir_path, "", "mindir path"); +DEFINE_string(input0_path, ".", "input0 path"); +DEFINE_string(input1_path, ".", "input1 path"); +DEFINE_string(input2_path, ".", "input2 path"); +DEFINE_int32(device_id, 0, "device id"); + +int main(int argc, char **argv) { + gflags::ParseCommandLineFlags(&argc, &argv, true); + if (RealPath(FLAGS_mindir_path).empty()) { + std::cout << "Invalid mindir" << std::endl; + return 1; + } + + auto context = std::make_shared(); + auto ascend310 = std::make_shared(); + ascend310->SetDeviceID(FLAGS_device_id); + context->MutableDeviceInfo().push_back(ascend310); + mindspore::Graph graph; + Serialization::Load(FLAGS_mindir_path, ModelType::kMindIR, &graph); + + Model model; + Status ret = model.Build(GraphCell(graph), context); + if (ret != kSuccess) { + std::cout << "ERROR: Build failed." << std::endl; + return 1; + } + + std::vector model_inputs = model.GetInputs(); + if (model_inputs.empty()) { + std::cout << "Invalid model, inputs is empty." << std::endl; + return 1; + } + + auto input0_files = GetAllFiles(FLAGS_input0_path); + auto input1_files = GetAllFiles(FLAGS_input1_path); + auto input2_files = GetAllFiles(FLAGS_input2_path); + + if (input0_files.empty() || input1_files.empty() || input2_files.empty()) { + std::cout << "ERROR: input data empty." << std::endl; + return 1; + } + + std::map costTime_map; + size_t size = input0_files.size(); + + for (size_t i = 0; i < size; ++i) { + struct timeval start = {0}; + struct timeval end = {0}; + double startTimeMs; + double endTimeMs; + std::vector inputs; + std::vector outputs; + std::cout << "Start predict input files:" << input0_files[i] << std::endl; + + auto input0 = ReadFileToTensor(input0_files[i]); + auto input1 = ReadFileToTensor(input1_files[i]); + auto input2 = ReadFileToTensor(input2_files[i]); + inputs.emplace_back(model_inputs[0].Name(), model_inputs[0].DataType(), model_inputs[0].Shape(), + input0.Data().get(), input0.DataSize()); + inputs.emplace_back(model_inputs[1].Name(), model_inputs[1].DataType(), model_inputs[1].Shape(), + input1.Data().get(), input1.DataSize()); + inputs.emplace_back(model_inputs[2].Name(), model_inputs[2].DataType(), model_inputs[2].Shape(), + input2.Data().get(), input2.DataSize()); + + gettimeofday(&start, nullptr); + ret = model.Predict(inputs, &outputs); + gettimeofday(&end, nullptr); + if (ret != kSuccess) { + std::cout << "Predict " << input0_files[i] << " failed." << std::endl; + return 1; + } + startTimeMs = (1.0 * start.tv_sec * 1000000 + start.tv_usec) / 1000; + endTimeMs = (1.0 * end.tv_sec * 1000000 + end.tv_usec) / 1000; + costTime_map.insert(std::pair(startTimeMs, endTimeMs)); + WriteResult(input0_files[i], outputs); + } + double average = 0.0; + int inferCount = 0; + + for (auto iter = costTime_map.begin(); iter != costTime_map.end(); iter++) { + double diff = 0.0; + diff = iter->second - iter->first; + average += diff; + inferCount++; + } + average = average / inferCount; + std::stringstream timeCost; + timeCost << "NN inference cost average time: "<< average << " ms of infer_count " << inferCount << std::endl; + std::cout << "NN inference cost average time: "<< average << "ms of infer_count " << inferCount << std::endl; + std::string fileName = "./time_Result" + std::string("/test_perform_static.txt"); + std::ofstream fileStream(fileName.c_str(), std::ios::trunc); + fileStream << timeCost.str(); + fileStream.close(); + costTime_map.clear(); + return 0; +} diff --git a/model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/utils.cc b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/utils.cc new file mode 100644 index 00000000000..c947e4d5f45 --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/ascend310_infer/src/utils.cc @@ -0,0 +1,129 @@ +/** + * Copyright 2021 Huawei Technologies Co., Ltd + * + * 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 +#include "inc/utils.h" + +using mindspore::MSTensor; +using mindspore::DataType; + +std::vector GetAllFiles(std::string_view dirName) { + struct dirent *filename; + DIR *dir = OpenDir(dirName); + if (dir == nullptr) { + return {}; + } + std::vector res; + while ((filename = readdir(dir)) != nullptr) { + std::string dName = std::string(filename->d_name); + if (dName == "." || dName == ".." || filename->d_type != DT_REG) { + continue; + } + res.emplace_back(std::string(dirName) + "/" + filename->d_name); + } + std::sort(res.begin(), res.end()); + for (auto &f : res) { + std::cout << "image file: " << f << std::endl; + } + return res; +} + +int WriteResult(const std::string& imageFile, const std::vector &outputs) { + std::string homePath = "./result_Files"; + for (size_t i = 0; i < outputs.size(); ++i) { + size_t outputSize; + std::shared_ptr netOutput; + netOutput = outputs[i].Data(); + outputSize = outputs[i].DataSize(); + int pos = imageFile.rfind('/'); + std::string fileName(imageFile, pos + 1); + fileName.replace(fileName.find('.'), fileName.size() - fileName.find('.'), '_' + std::to_string(i) + ".bin"); + std::string outFileName = homePath + "/" + fileName; + FILE * outputFile = fopen(outFileName.c_str(), "wb"); + fwrite(netOutput.get(), outputSize, sizeof(char), outputFile); + fclose(outputFile); + outputFile = nullptr; + } + return 0; +} + +mindspore::MSTensor ReadFileToTensor(const std::string &file) { + if (file.empty()) { + std::cout << "Pointer file is nullptr" << std::endl; + return mindspore::MSTensor(); + } + + std::ifstream ifs(file); + if (!ifs.good()) { + std::cout << "File: " << file << " is not exist" << std::endl; + return mindspore::MSTensor(); + } + + if (!ifs.is_open()) { + std::cout << "File: " << file << "open failed" << std::endl; + return mindspore::MSTensor(); + } + + ifs.seekg(0, std::ios::end); + size_t size = ifs.tellg(); + mindspore::MSTensor buffer(file, mindspore::DataType::kNumberTypeUInt8, {static_cast(size)}, nullptr, size); + + ifs.seekg(0, std::ios::beg); + ifs.read(reinterpret_cast(buffer.MutableData()), size); + ifs.close(); + + return buffer; +} + + +DIR *OpenDir(std::string_view dirName) { + if (dirName.empty()) { + std::cout << " dirName is null ! " << std::endl; + return nullptr; + } + std::string realPath = RealPath(dirName); + struct stat s; + lstat(realPath.c_str(), &s); + if (!S_ISDIR(s.st_mode)) { + std::cout << "dirName is not a valid directory !" << std::endl; + return nullptr; + } + DIR *dir; + dir = opendir(realPath.c_str()); + if (dir == nullptr) { + std::cout << "Can not open dir " << dirName << std::endl; + return nullptr; + } + std::cout << "Successfully opened the dir " << dirName << std::endl; + return dir; +} + +std::string RealPath(std::string_view path) { + char realPathMem[PATH_MAX] = {0}; + char *realPathRet = nullptr; + realPathRet = realpath(path.data(), realPathMem); + + if (realPathRet == nullptr) { + std::cout << "File: " << path << " is not exist."; + return ""; + } + + std::string realPath(realPathMem); + std::cout << path << " realpath is: " << realPath << std::endl; + return realPath; +} diff --git a/model_zoo/official/recommend/wide_and_deep/postprocess.py b/model_zoo/official/recommend/wide_and_deep/postprocess.py new file mode 100644 index 00000000000..c2f53b58359 --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/postprocess.py @@ -0,0 +1,53 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ +"""postprocess.""" +import os +import argparse +import numpy as np +from mindspore import Tensor +from src.metrics import AUCMetric + +parser = argparse.ArgumentParser(description='postprocess') +parser.add_argument('--result_path', type=str, default="./result_Files", help='result path') +parser.add_argument('--label_path', type=str, default=None, help='label path') +parser.add_argument("--batch_size", type=int, default=16000, help="infer batch size.") +args_opt, _ = parser.parse_known_args() + +def get_acc(): + ''' get accuracy ''' + auc_metric = AUCMetric() + auc_metric.clear() + batch_size = args_opt.batch_size + + files = os.listdir(args_opt.label_path) + + for f in files: + rst_file = os.path.join(args_opt.result_path, f.split('.')[0] + '_0.bin') + label_file = os.path.join(args_opt.label_path, f) + + predict = Tensor(np.fromfile(rst_file, np.float32).reshape(batch_size, 1)) + label = Tensor(np.fromfile(label_file, np.float32).reshape(batch_size, 1)) + + res = [] + res.append(predict) + res.append(predict) + res.append(label) + + auc_metric.update(*res) + + auc_metric.eval() + +if __name__ == '__main__': + get_acc() diff --git a/model_zoo/official/recommend/wide_and_deep/preprocess.py b/model_zoo/official/recommend/wide_and_deep/preprocess.py new file mode 100644 index 00000000000..102f5781198 --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/preprocess.py @@ -0,0 +1,63 @@ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ +"""preprocess.""" +import os +import argparse + +from src.datasets import create_dataset, DataType + +parser = argparse.ArgumentParser(description='preprocess.') +parser.add_argument('--dataset_path', type=str, default=None, help='Dataset path') +parser.add_argument('--result_path', type=str, default='./preprocess_Result', help='Result path') +parser.add_argument("--batch_size", type=int, default=16000, help="infer batch size.") +parser.add_argument("--dataset_type", type=str, default="mindrecord", choices=["tfrecord", "mindrecord", "hd5"]) + +args_opt, _ = parser.parse_known_args() + +def generate_bin(): + '''generate bin files''' + data_path = args_opt.dataset_path + batch_size = args_opt.batch_size + if args_opt.dataset_type == "tfrecord": + dataset_type = DataType.TFRECORD + elif args_opt.dataset_type == "mindrecord": + dataset_type = DataType.MINDRECORD + else: + dataset_type = DataType.H5 + ds = create_dataset(data_path, train_mode=False, epochs=1, + batch_size=batch_size, data_type=dataset_type) + feat_ids_path = os.path.join(args_opt.result_path, "00_feat_ids") + feat_vals_path = os.path.join(args_opt.result_path, "01_feat_vals") + label_path = os.path.join(args_opt.result_path, "02_labels") + + os.makedirs(feat_ids_path) + os.makedirs(feat_vals_path) + os.makedirs(label_path) + + for i, data in enumerate(ds.create_dict_iterator(output_numpy=True)): + file_name = "criteo_bs" + str(batch_size) + "_" + str(i) + ".bin" + batch_ids = data['feat_ids'] + batch_ids.tofile(os.path.join(feat_ids_path, file_name)) + + batch_wts = data['feat_vals'] + batch_wts.tofile(os.path.join(feat_vals_path, file_name)) + + labels = data['label'] + labels.tofile(os.path.join(label_path, file_name)) + + print("=" * 20, "export bin files finished", "=" * 20) + +if __name__ == '__main__': + generate_bin() diff --git a/model_zoo/official/recommend/wide_and_deep/script/run_infer_310.sh b/model_zoo/official/recommend/wide_and_deep/script/run_infer_310.sh new file mode 100644 index 00000000000..a0c45f02f6e --- /dev/null +++ b/model_zoo/official/recommend/wide_and_deep/script/run_infer_310.sh @@ -0,0 +1,125 @@ +#!/bin/bash +# Copyright 2021 Huawei Technologies Co., Ltd +# +# 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. +# ============================================================================ + +if [[ $# -lt 4 || $# -gt 5 ]]; then + echo "Usage: bash run_infer_310.sh [MINDIR_PATH] [DATASET_PATH] [DATA_TYPE] [NEED_PREPROCESS] [DEVICE_ID] + DATA_TYPE means dataset type, it's value is ['tfrecord', 'mindrecord', 'hd5']. + NEED_PREPROCESS means weather need preprocess or not, it's value is 'y' or 'n'. + DEVICE_ID is optional, it can be set by environment variable device_id, otherwise the value is zero" +exit 1 +fi + +get_real_path(){ + if [ "${1:0:1}" == "/" ]; then + echo "$1" + else + echo "$(realpath -m $PWD/$1)" + fi +} +model=$(get_real_path $1) +dataset_path=$(get_real_path $2) +dataset_type=$3 + +if [ "$4" == "y" ] || [ "$4" == "n" ];then + need_preprocess=$4 +else + echo "weather need preprocess or not, it's value must be in [y, n]" + exit 1 +fi + +device_id=0 +if [ $# == 5 ]; then + device_id=$5 +fi + +echo "mindir name: "$model +echo "dataset path: "$dataset_path +echo "data type: "$dataset_type +echo "need preprocess: "$need_preprocess +echo "device id: "$device_id + +export ASCEND_HOME=/usr/local/Ascend/ +if [ -d ${ASCEND_HOME}/ascend-toolkit ]; then + export PATH=$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/ccec_compiler/bin:$ASCEND_HOME/ascend-toolkit/latest/atc/bin:$PATH + export LD_LIBRARY_PATH=/usr/local/lib:$ASCEND_HOME/ascend-toolkit/latest/atc/lib64:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH + export TBE_IMPL_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp/op_impl/built-in/ai_core/tbe + export PYTHONPATH=${TBE_IMPL_PATH}:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/python/site-packages:$PYTHONPATH + export ASCEND_OPP_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp +else + export PATH=$ASCEND_HOME/atc/ccec_compiler/bin:$ASCEND_HOME/atc/bin:$PATH + export LD_LIBRARY_PATH=/usr/local/lib:$ASCEND_HOME/atc/lib64:$ASCEND_HOME/acllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH + export PYTHONPATH=$ASCEND_HOME/atc/python/site-packages:$PYTHONPATH + export ASCEND_OPP_PATH=$ASCEND_HOME/opp +fi + +function preprocess_data() +{ + if [ -d preprocess_Result ]; then + rm -rf ./preprocess_Result + fi + mkdir preprocess_Result + python3.7 ../preprocess.py --dataset_path=$dataset_path --dataset_type=$dataset_type --result_path=./preprocess_Result/ +} + +function compile_app() +{ + cd ../ascend310_infer || exit + bash build.sh &> build.log +} + +function infer() +{ + cd - || exit + if [ -d result_Files ]; then + rm -rf ./result_Files + fi + if [ -d time_Result ]; then + rm -rf ./time_Result + fi + mkdir result_Files + mkdir time_Result + + ../ascend310_infer/out/main --mindir_path=$model --input0_path=./preprocess_Result/00_feat_ids --input1_path=./preprocess_Result/01_feat_vals --input2_path=./preprocess_Result/02_labels --device_id=$device_id &> infer.log + +} + +function cal_acc() +{ + python3.7 ../postprocess.py --result_path=./result_Files --label_path=./preprocess_Result/02_labels &> acc.log +} + +if [ $need_preprocess == "y" ]; then + preprocess_data + if [ $? -ne 0 ]; then + echo "preprocess dataset failed" + exit 1 + fi +fi +compile_app +if [ $? -ne 0 ]; then + echo "compile app code failed" + exit 1 +fi +infer +if [ $? -ne 0 ]; then + echo " execute inference failed" + exit 1 +fi +cal_acc +if [ $? -ne 0 ]; then + echo "calculate accuracy failed" + exit 1 +fi \ No newline at end of file