Program Listing for File inference_client_stage.hpp

            

            
/*
* SPDX-FileCopyrightText: Copyright (c) 2024, 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.
*/

#pragma once

#include "morpheus/export.h"
#include "morpheus/messages/multi_inference.hpp"
#include "morpheus/messages/multi_response.hpp"
#include "morpheus/types.hpp"

#include <mrc/coroutines/async_generator.hpp>
#include <mrc/coroutines/scheduler.hpp>
#include <mrc/coroutines/task.hpp>
#include <mrc/segment/builder.hpp>
#include <mrc/segment/object.hpp>
#include <pybind11/pybind11.h>
#include <pymrc/asyncio_runnable.hpp>

#include <cstdint>
#include <map>
#include <memory>
#include <mutex>
#include <string>
#include <vector>

namespace morpheus {

struct MORPHEUS_EXPORT TensorModelMapping
{
    std::string model_field_name;

    std::string tensor_field_name;
};

class MORPHEUS_EXPORT IInferenceClientSession
{
  public:
    virtual ~IInferenceClientSession() = default;
    virtual std::vector<TensorModelMapping> get_input_mappings(std::vector<TensorModelMapping> input_map_overrides) = 0;

    virtual std::vector<TensorModelMapping> get_output_mappings(
        std::vector<TensorModelMapping> output_map_overrides) = 0;

    virtual mrc::coroutines::Task<TensorMap> infer(TensorMap&& inputs) = 0;
};

class MORPHEUS_EXPORT IInferenceClient
{
  public:
    virtual ~IInferenceClient() = default;
    virtual std::unique_ptr<IInferenceClientSession> create_session() = 0;
};

class MORPHEUS_EXPORT InferenceClientStage
  : public mrc::pymrc::AsyncioRunnable<std::shared_ptr<MultiInferenceMessage>, std::shared_ptr<MultiResponseMessage>>
{
  public:
    using sink_type_t   = std::shared_ptr<MultiInferenceMessage>;
    using source_type_t = std::shared_ptr<MultiResponseMessage>;

    InferenceClientStage(std::unique_ptr<IInferenceClient>&& client,
                         std::string model_name,
                         bool needs_logits,
                         std::vector<TensorModelMapping> input_mapping,
                         std::vector<TensorModelMapping> output_mapping);

    mrc::coroutines::AsyncGenerator<std::shared_ptr<MultiResponseMessage>> on_data(
        std::shared_ptr<MultiInferenceMessage>&& data, std::shared_ptr<mrc::coroutines::Scheduler> on) override;

  private:
    std::string m_model_name;
    std::shared_ptr<IInferenceClient> m_client;
    std::shared_ptr<IInferenceClientSession> m_session;
    bool m_needs_logits{true};
    std::vector<TensorModelMapping> m_input_mapping;
    std::vector<TensorModelMapping> m_output_mapping;
    std::mutex m_session_mutex;

    int32_t m_retry_max = 10;
};

/****** InferenceClientStageInferenceProxy******************/
struct MORPHEUS_EXPORT InferenceClientStageInterfaceProxy
{
    static std::shared_ptr<mrc::segment::Object<InferenceClientStage>> init(
        mrc::segment::Builder& builder,
        const std::string& name,
        std::string model_name,
        std::string server_url,
        bool needs_logits,
        std::map<std::string, std::string> input_mapping,
        std::map<std::string, std::string> output_mapping);
};  // end of group

}  // namespace morpheus