Implementation:Tensorflow Serving Predict Impl Test

Knowledge Sources	Tensorflow_Serving
Domains	Testing, Prediction
Last Updated	2026-02-13 00:00 GMT

Overview

Test suite validating the TensorflowPredictor predict implementation which processes prediction requests through ServerCore.

Description

This test file validates the TensorflowPredictor class which provides the top-level predict API used by the model server. The PredictImplTest fixture sets up two ServerCore instances: one with the half_plus_two SavedModel and another with a counter model. Tests exercise the full prediction pipeline from request validation through model execution to response construction.

Key areas tested include:

• Model spec validation (missing, empty, wrong model name)
• Input tensor validation (empty, mismatched, wrong type)
• Output filter validation (non-existent filters, valid filters)
• Successful prediction with default and named signatures
• Customized signatures for regression and classification
• Model spec override in responses
• Thread pool factory integration

Usage

Run these tests to validate changes to the predict implementation, including request validation, model loading, and response construction.

Code Reference

Source Location

• Repository: Tensorflow_Serving
• File: tensorflow_serving/servables/tensorflow/predict_impl_test.cc
• Lines: 1-481

Test Fixture

class PredictImplTest : public ::testing::Test {
 public:
  static void SetUpTestSuite() {
    TF_ASSERT_OK(CreateServerCore(
        test_util::TensorflowTestSrcDirPath(
            "cc/saved_model/testdata/half_plus_two"),
        &saved_model_server_core_));
    TF_ASSERT_OK(CreateServerCore(
        test_util::TestSrcDirPath(
            "/servables/tensorflow/testdata/saved_model_counter"),
        &saved_model_server_core_counter_model_));
  }
 protected:
  ServerCore* GetServerCore() { return saved_model_server_core_.get(); }
  ServerCore* GetServerCoreWithCounterModel();
  RunOptions GetRunOptions() { return RunOptions(); }
 private:
  static std::unique_ptr<ServerCore> saved_model_server_core_;
  static std::unique_ptr<ServerCore> saved_model_server_core_counter_model_;
};

Build Target

bazel test //tensorflow_serving/servables/tensorflow:predict_impl_test

Test Coverage

Key Test Cases

Test Name	Category	Description
MissingOrEmptyModelSpec	Validation	Tests error on missing or empty model spec
EmptyInputList	Validation	Tests error on empty input tensors
InputTensorsDontMatchModelSpecInputs	Validation	Tests error on input tensor key mismatch
OutputFiltersDontMatchModelSpecOutputs	Validation	Tests error on invalid output filters
InputTensorsHaveWrongType	Validation	Tests error on wrong input tensor dtype
PredictionSuccess	Integration	Tests successful prediction with half_plus_two model
PredictionWithNamedRegressionSignature	Integration	Tests prediction using a named regression signature
PredictionWithNamedClassificationSignature	Integration	Tests prediction using a named classification signature
PredictionWithCustomizedSignatures	Integration	Tests prediction with custom signature definitions
ModelSpecOverride	Integration	Tests model spec version override in response
ThreadPoolFactory	Integration	Tests integration with custom thread pool factory

Usage Examples

Test Pattern

TEST_F(PredictImplTest, PredictionSuccess) {
  PredictRequest request;
  PredictResponse response;

  ModelSpec* model_spec = request.mutable_model_spec();
  model_spec->set_name(kTestModelName);
  model_spec->mutable_version()->set_value(kTestModelVersion);

  TensorProto tensor_proto;
  tensor_proto.add_float_val(2.0);
  tensor_proto.set_dtype(tensorflow::DT_FLOAT);
  (*request.mutable_inputs())[kInputTensorKey] = tensor_proto;

  TensorflowPredictor predictor;
  TF_EXPECT_OK(
      predictor.Predict(GetRunOptions(), GetServerCore(), request, &response));
  // Expected output: 0.5 * 2 + 2 = 3
}

Related Pages

• Implementation:Tensorflow_Serving_Predict_Util_Test