Program Listing for File infer_manager.hpp

            

            
/*
* SPDX-FileCopyrightText: Copyright (c) 2022-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* 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 _HOLOSCAN_INFER_MANAGER_H
#define _HOLOSCAN_INFER_MANAGER_H

#include <functional>
#include <future>
#include <iostream>
#include <map>
#include <memory>
#include <set>
#include <string>

#include <holoinfer.hpp>
#include <holoinfer_buffer.hpp>
#include <holoinfer_constants.hpp>
#include <holoinfer_utils.hpp>
#include <infer/infer.hpp>

#if __has_include(<onnxruntime_c_api.h>)
#define use_onnxruntime 1
#include <infer/onnx/core.hpp>
#endif

#ifdef use_torch
#include <infer/torch/core.hpp>
#endif

#include <infer/trt/core.hpp>
#include <params/infer_param.hpp>

namespace holoscan {
namespace inference {
class ManagerInfer {
 public:
  ManagerInfer();

  ~ManagerInfer();

  InferStatus set_inference_params(std::shared_ptr<InferenceSpecs>& inference_specs);

  InferStatus execute_inference(DataMap& preprocess_data_map, DataMap& output_data_map);

  InferStatus run_core_inference(const std::string& model_name, DataMap& permodel_preprocess_data,
                                 DataMap& permodel_output_data);
  void cleanup();

  DimType get_input_dimensions() const;

  DimType get_output_dimensions() const;

 private:
  bool parallel_processing_ = false;

  bool cuda_buffer_in_ = false;

  bool cuda_buffer_out_ = false;

  bool mgpu_p2p_transfer = true;

  std::map<std::string, std::map<std::string, cudaStream_t>> input_streams_gpudt;

  std::map<std::string, std::map<std::string, cudaStream_t>> output_streams_gpudt;

  std::map<std::string, std::map<std::string, cudaStream_t>> input_streams_device;

  std::map<std::string, std::map<std::string, cudaStream_t>> output_streams_device;

  std::map<std::string, std::unique_ptr<Params>> infer_param_;

  std::map<std::string, std::unique_ptr<InferBase>> holo_infer_context_;

  DimType models_input_dims_;

  std::map<std::string, DataMap> mgpu_output_buffer_;

  std::map<std::string, DataMap> mgpu_input_buffer_;

  int device_gpu_dt = 0;

  DimType models_output_dims_;

  inline static std::map<std::string, holoinfer_backend> supported_backend_{
      {"onnxrt", holoinfer_backend::h_onnx},
      {"trt", holoinfer_backend::h_trt},
      {"torch", holoinfer_backend::h_torch}};
};

std::unique_ptr<ManagerInfer> manager;

}  // namespace inference
}  // namespace holoscan

#endif