prediction_service_impl.cc

/* Copyright 2018 Google Inc. 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 "tensorflow_serving/model_servers/prediction_service_impl.h"
 
#include "grpc/grpc.h"
#include "tensorflow_serving/model_servers/grpc_status_util.h"
#include "tensorflow_serving/servables/tensorflow/classification_service.h"
#include "tensorflow_serving/servables/tensorflow/get_model_metadata_impl.h"
#include "tensorflow_serving/servables/tensorflow/multi_inference_helper.h"
#include "tensorflow_serving/servables/tensorflow/regression_service.h"
#include "tensorflow_serving/servables/tensorflow/util.h"
 
namespace tensorflow {
namespace serving {
 
namespace {
 
ScopedThreadPools GetThreadPools(ThreadPoolFactory *thread_pool_factory) {
  return thread_pool_factory == nullptr ? ScopedThreadPools()
                                        : thread_pool_factory->GetThreadPools();
}
 
}  // namespace
 
::grpc::Status PredictionServiceImpl::Predict(::grpc::ServerContext *context,
                                              const PredictRequest *request,
                                              PredictResponse *response) {
  const uint64_t start = Env::Default()->NowMicros();
  tensorflow::RunOptions run_options = tensorflow::RunOptions();
  if (enforce_session_run_timeout_) {
    run_options.set_timeout_in_ms(
        DeadlineToTimeoutMillis(context->raw_deadline()));
  }
 
  const ::tensorflow::Status tf_status =
      predictor_->Predict(run_options, core_, *request, response);
  const ::grpc::Status status = ToGRPCStatus(tf_status);
 
  if (status.ok()) {
    RecordRequestLatency(request->model_spec().name(), /*api=*/"Predict",
                         /*entrypoint=*/"GRPC",
                         Env::Default()->NowMicros() - start);
  } else {
    VLOG(1) << "Predict failed: " << status.error_message();
  }
  RecordModelRequestCount(request->model_spec().name(), tf_status);
 
  return status;
}
 
::grpc::Status PredictionServiceImpl::GetModelMetadata(
    ::grpc::ServerContext *context, const GetModelMetadataRequest *request,
    GetModelMetadataResponse *response) {
  const ::grpc::Status status = ToGRPCStatus(
      GetModelMetadataImpl::GetModelMetadata(core_, *request, response));
  if (!status.ok()) {
    VLOG(1) << "GetModelMetadata failed: " << status.error_message();
  }
  return status;
}
 
::grpc::Status PredictionServiceImpl::Classify(
    ::grpc::ServerContext *context, const ClassificationRequest *request,
    ClassificationResponse *response) {
  const uint64_t start = Env::Default()->NowMicros();
  tensorflow::RunOptions run_options = tensorflow::RunOptions();
  // By default, this is infinite which is the same default as RunOptions.
  if (enforce_session_run_timeout_) {
    run_options.set_timeout_in_ms(
        DeadlineToTimeoutMillis(context->raw_deadline()));
  }
 
  const ::tensorflow::Status tf_status =
      TensorflowClassificationServiceImpl::Classify(
          run_options, core_, GetThreadPools(thread_pool_factory_).get(),
          *request, response);
  const ::grpc::Status status = ToGRPCStatus(tf_status);
 
  if (status.ok()) {
    RecordRequestLatency(request->model_spec().name(), /*api=*/"Classify",
                         /*entrypoint=*/"GRPC",
                         Env::Default()->NowMicros() - start);
  } else {
    VLOG(1) << "Classify request failed: " << status.error_message();
  }
  RecordModelRequestCount(request->model_spec().name(), tf_status);
 
  return status;
}
 
::grpc::Status PredictionServiceImpl::Regress(::grpc::ServerContext *context,
                                              const RegressionRequest *request,
                                              RegressionResponse *response) {
  const uint64_t start = Env::Default()->NowMicros();
  tensorflow::RunOptions run_options = tensorflow::RunOptions();
  // By default, this is infinite which is the same default as RunOptions.
  if (enforce_session_run_timeout_) {
    run_options.set_timeout_in_ms(
        DeadlineToTimeoutMillis(context->raw_deadline()));
  }
 
  const ::tensorflow::Status tf_status =
      TensorflowRegressionServiceImpl::Regress(
          run_options, core_, GetThreadPools(thread_pool_factory_).get(),
          *request, response);
  const ::grpc::Status status = ToGRPCStatus(tf_status);
 
  if (status.ok()) {
    RecordRequestLatency(request->model_spec().name(), /*api=*/"Regress",
                         /*entrypoint=*/"GRPC",
                         Env::Default()->NowMicros() - start);
  } else {
    VLOG(1) << "Regress request failed: " << status.error_message();
  }
  RecordModelRequestCount(request->model_spec().name(), tf_status);
 
  return status;
}
 
::grpc::Status PredictionServiceImpl::MultiInference(
    ::grpc::ServerContext *context, const MultiInferenceRequest *request,
    MultiInferenceResponse *response) {
  tensorflow::RunOptions run_options = tensorflow::RunOptions();
  // By default, this is infinite which is the same default as RunOptions.
  if (enforce_session_run_timeout_) {
    run_options.set_timeout_in_ms(
        DeadlineToTimeoutMillis(context->raw_deadline()));
  }
  const ::grpc::Status status = ToGRPCStatus(RunMultiInferenceWithServerCore(
      run_options, core_, GetThreadPools(thread_pool_factory_).get(), *request,
      response));
  if (!status.ok()) {
    VLOG(1) << "MultiInference request failed: " << status.error_message();
  }
  return status;
}
 
}  // namespace serving
}  // namespace tensorflow