saved_model_warmup_test_util.cc

/* Copyright 2020 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/servables/tensorflow/saved_model_warmup_test_util.h"
 
#include <memory>
#include <vector>
 
#include "tensorflow/cc/saved_model/constants.h"
#include "tensorflow/cc/saved_model/signature_constants.h"
#include "tensorflow/core/framework/tensor.pb.h"
#include "tensorflow/core/framework/tensor_shape.pb.h"
#include "tensorflow/core/framework/types.pb.h"
#include "tensorflow/core/platform/env.h"
#include "tensorflow_serving/apis/prediction_log.pb.h"
 
namespace tensorflow {
namespace serving {
 
void PopulateInferenceTask(const string& model_name,
                           const string& signature_name,
                           const string& method_name, InferenceTask* task) {
  ModelSpec model_spec;
  model_spec.set_name(model_name);
  model_spec.set_signature_name(signature_name);
  *task->mutable_model_spec() = model_spec;
  task->set_method_name(method_name);
}
 
void PopulateMultiInferenceRequest(MultiInferenceRequest* request) {
  request->mutable_input()->mutable_example_list()->add_examples();
  PopulateInferenceTask("test_model", kRegressMethodName, kRegressMethodName,
                        request->add_tasks());
  PopulateInferenceTask("test_model", kClassifyMethodName, kClassifyMethodName,
                        request->add_tasks());
}
 
void PopulatePredictRequest(PredictRequest* request) {
  request->mutable_model_spec()->set_signature_name(kPredictMethodName);
  TensorProto tensor_proto;
  tensor_proto.add_string_val("input_value");
  tensor_proto.set_dtype(tensorflow::DT_STRING);
  tensor_proto.mutable_tensor_shape()->add_dim()->set_size(1);
  (*request->mutable_inputs())[kPredictInputs] = tensor_proto;
}
 
void PopulateClassificationRequest(ClassificationRequest* request) {
  request->mutable_input()->mutable_example_list()->add_examples();
  request->mutable_model_spec()->set_signature_name(kClassifyMethodName);
}
 
void PopulateRegressionRequest(RegressionRequest* request) {
  request->mutable_input()->mutable_example_list()->add_examples();
  request->mutable_model_spec()->set_signature_name(kRegressMethodName);
}
 
Status PopulatePredictionLog(PredictionLog* prediction_log,
                             PredictionLog::LogTypeCase log_type,
                             int num_repeated_values) {
  if ((num_repeated_values > 1) &&
      (log_type != PredictionLog::kPredictStreamedLog)) {
    return errors::InvalidArgument(
        "Only predict_streamed_log supports num_repeated_values > 1.");
  }
  switch (log_type) {
    case PredictionLog::kRegressLog: {
      PopulateRegressionRequest(
          prediction_log->mutable_regress_log()->mutable_request());
    } break;
    case PredictionLog::kClassifyLog: {
      PopulateClassificationRequest(
          prediction_log->mutable_classify_log()->mutable_request());
    } break;
    case PredictionLog::kPredictLog: {
      PopulatePredictRequest(
          prediction_log->mutable_predict_log()->mutable_request());
    } break;
    case PredictionLog::kPredictStreamedLog: {
      for (int i = 0; i < num_repeated_values; ++i) {
        PopulatePredictRequest(
            prediction_log->mutable_predict_streamed_log()->add_request());
      }
    } break;
    case PredictionLog::kMultiInferenceLog: {
      PopulateMultiInferenceRequest(
          prediction_log->mutable_multi_inference_log()->mutable_request());
    } break;
    case PredictionLog::kSessionRunLog:
      prediction_log->mutable_session_run_log();
      TF_FALLTHROUGH_INTENDED;
    default:
      break;
  }
  return absl::OkStatus();
}
 
Status WriteWarmupData(const string& fname,
                       const std::vector<string>& warmup_records,
                       int num_warmup_records) {
  Env* env = Env::Default();
  std::unique_ptr<WritableFile> file;
  TF_RETURN_IF_ERROR(env->NewWritableFile(fname, &file));
 
  io::RecordWriterOptions options;
  io::RecordWriter writer(file.get(), options);
  for (int i = 0; i < num_warmup_records; ++i) {
    for (const string& warmup_record : warmup_records) {
      TF_RETURN_IF_ERROR(writer.WriteRecord(warmup_record));
    }
  }
  TF_RETURN_IF_ERROR(writer.Flush());
  return absl::OkStatus();
}
 
Status WriteWarmupDataAsSerializedProtos(
    const string& fname, const std::vector<string>& warmup_records,
    int num_warmup_records) {
  Env* env = Env::Default();
  std::unique_ptr<WritableFile> file;
  TF_RETURN_IF_ERROR(env->NewWritableFile(fname, &file));
  for (int i = 0; i < num_warmup_records; ++i) {
    for (const string& warmup_record : warmup_records) {
      TF_RETURN_IF_ERROR(file->Append(warmup_record));
    }
  }
  TF_RETURN_IF_ERROR(file->Close());
  return absl::OkStatus();
}
 
Status AddMixedWarmupData(
    std::vector<string>* warmup_records,
    const std::vector<PredictionLog::LogTypeCase>& log_types) {
  for (auto& log_type : log_types) {
    TF_RETURN_IF_ERROR(AddToWarmupData(warmup_records, log_type, 1));
  }
  return absl::OkStatus();
}
 
Status AddToWarmupData(std::vector<string>* warmup_records,
                       PredictionLog::LogTypeCase log_type,
                       int num_repeated_values) {
  PredictionLog prediction_log;
  TF_RETURN_IF_ERROR(
      PopulatePredictionLog(&prediction_log, log_type, num_repeated_values));
  warmup_records->push_back(prediction_log.SerializeAsString());
  return absl::OkStatus();
}
 
// Creates a test SignatureDef with the given parameters
SignatureDef CreateSignatureDef(const string& method_name,
                                const std::vector<string>& input_names,
                                const std::vector<string>& output_names) {
  SignatureDef signature_def;
  signature_def.set_method_name(method_name);
  for (const string& input_name : input_names) {
    TensorInfo input;
    input.set_name(input_name);
    (*signature_def.mutable_inputs())[input_name] = input;
  }
  for (const string& output_name : output_names) {
    TensorInfo output;
    output.set_name(output_name);
    (*signature_def.mutable_outputs())[output_name] = output;
  }
  return signature_def;
}
 
void AddSignatures(MetaGraphDef* meta_graph_def) {
  (*meta_graph_def->mutable_signature_def())[kRegressMethodName] =
      CreateSignatureDef(kRegressMethodName, {kRegressInputs},
                         {kRegressOutputs});
  (*meta_graph_def->mutable_signature_def())[kClassifyMethodName] =
      CreateSignatureDef(kClassifyMethodName, {kClassifyInputs},
                         {kClassifyOutputClasses, kClassifyOutputScores});
  (*meta_graph_def->mutable_signature_def())[kPredictMethodName] =
      CreateSignatureDef(kPredictMethodName, {kPredictInputs},
                         {kPredictOutputs});
}
 
}  // namespace serving
}  // namespace tensorflow