1 /* 2 * Copyright (c) 2022 Huawei Device Co., Ltd. 3 * Licensed under the Apache License, Version 2.0 (the "License"); 4 * you may not use this file except in compliance with the License. 5 * You may obtain a copy of the License at 6 * 7 * http://www.apache.org/licenses/LICENSE-2.0 8 * 9 * Unless required by applicable law or agreed to in writing, software 10 * distributed under the License is distributed on an "AS IS" BASIS, 11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12 * See the License for the specific language governing permissions and 13 * limitations under the License. 14 */ 15 #ifndef NEURAL_NETWORK_RUNTIME_COMPILATION_H 16 #define NEURAL_NETWORK_RUNTIME_COMPILATION_H 17 18 #include "inner_model.h" 19 #include "execution_plan.h" 20 21 #include "device.h" 22 #include "cpp_type.h" 23 24 namespace OHOS { 25 namespace NeuralNetworkRuntime { 26 struct ModelCacheInfo { 27 uint64_t fileNumber = 0; 28 uint64_t version = 0; 29 uint64_t deviceId = 0; 30 std::vector<unsigned short> modelCheckSum; 31 }; 32 33 class Compilation { 34 public: 35 explicit Compilation(const InnerModel* innerModel); 36 37 OH_NN_ReturnCode SetDevice(size_t deviceId); 38 OH_NN_ReturnCode SetCacheDir(const std::string& cacheModelPath, uint32_t version); 39 OH_NN_ReturnCode SetPerformance(OH_NN_PerformanceMode performance); 40 OH_NN_ReturnCode SetPriority(OH_NN_Priority priority); 41 OH_NN_ReturnCode SetEnableFp16(bool isFp16); 42 43 OH_NN_ReturnCode Build(); 44 45 bool IsBuild() const; 46 bool IsDynamicShape() const; 47 std::vector<std::shared_ptr<NNTensor>>GetInputTensors() const; 48 std::vector<std::shared_ptr<NNTensor>>GetOutputTensors() const; 49 std::shared_ptr<ExecutionPlan> GetExecutionPlan() const; 50 51 private: 52 std::shared_ptr<mindspore::lite::LiteGraph> m_liteGraph {nullptr}; 53 OH_NN_Priority m_priority {OH_NN_PRIORITY_NONE}; 54 OH_NN_PerformanceMode m_performance {OH_NN_PERFORMANCE_NONE}; 55 bool m_enableFp16 {false}; 56 std::shared_ptr<Device> m_device {nullptr}; 57 std::string m_cachePath; 58 uint32_t m_version {0}; 59 size_t m_deviceId {0}; 60 bool m_isBuild {false}; 61 std::shared_ptr<ExecutionPlan> m_executionPlan {nullptr}; 62 std::vector<std::shared_ptr<NNTensor>> m_inputTensors; 63 std::vector<std::shared_ptr<NNTensor>> m_outputTensors; 64 void* m_metaGraph {nullptr}; 65 Buffer m_quantBuffer; 66 std::string m_modelName; 67 68 private: 69 OH_NN_ReturnCode GenerateCacheFiles(const std::vector<Buffer>& modelBuffer) const; 70 OH_NN_ReturnCode GenerateCacheModel(size_t cacheNumber, std::unique_ptr<uint64_t[]>& cacheInfo, 71 std::vector<Buffer> modelBuffer) const; 72 OH_NN_ReturnCode GenerateCacheInfo(uint32_t cacheSize, std::unique_ptr<uint64_t[]>& cacheInfo) const; 73 OH_NN_ReturnCode CheckCacheInfo(ModelCacheInfo& modelCacheInfo, const std::string& cacheInfoPath) const; 74 OH_NN_ReturnCode ReadCacheModelFile(const std::string& file, Buffer& modelBuffer) const; 75 OH_NN_ReturnCode RemoveCacheFiles(uint32_t fileNumber) const; 76 unsigned short GetCrc16(const unsigned char* buffer, size_t length) const; 77 OH_NN_ReturnCode CheckCacheModel(const ModelCacheInfo& modelCacheInfo, 78 std::vector<Buffer>& modelBuffers) const; 79 OH_NN_ReturnCode NormalBuild(std::shared_ptr<PreparedModel>& preparedModel); 80 OH_NN_ReturnCode BuildCacheModel(std::shared_ptr<PreparedModel>& preparedModel); 81 OH_NN_ReturnCode GenCacheBuild(std::shared_ptr<PreparedModel>& preparedModel); 82 OH_NN_ReturnCode ReGenCacheBuild(uint32_t fileNumber, std::shared_ptr<PreparedModel>& preparedModel); 83 OH_NN_ReturnCode LoadCacheBuild(std::shared_ptr<PreparedModel>& preparedModel, const ModelCacheInfo& cacheInfo); 84 OH_NN_ReturnCode InnerBuild(); 85 OH_NN_ReturnCode GetCacheFileLength(std::ifstream& ifs, int& fsize) const; 86 OH_NN_ReturnCode IsOfflineModel(bool& isOfflineModel) const; 87 OH_NN_ReturnCode BuildOfflineModel(std::shared_ptr<PreparedModel>& preparedModel); 88 }; 89 } // namespace NeuralNetworkRuntime 90 } // namespace OHOS 91 92 #endif // NEURAL_NETWORK_RUNTIME_COMPILATION_H