generated: '2026-06-20'
method: searched
source: https://www.tensorflow.org/tfx/serving/api_rest, https://www.tensorflow.org/tfx/serving/docker
note: >-
  TensorFlow Serving is self-hosted, so its "sandbox" is a published demo model plus the official
  Docker image that lets anyone stand up a working ModelServer locally. Values below are published
  verbatim in the TensorFlow Serving docs; no test values are invented.
test_model:
  name: saved_model_half_plus_two
  description: >-
    The canonical demo SavedModel used throughout the Serving REST API docs. It returns
    y = 0.5 * x + 2 for every input x, so responses are deterministic and easy to verify.
  source: https://github.com/tensorflow/serving/tree/master/tensorflow_serving/servables/tensorflow/testdata
docker:
  image: tensorflow/serving
  registry: https://hub.docker.com/r/tensorflow/serving
  quickstart: >-
    docker run -p 8501:8501 --mount type=bind,source=/path/to/saved_model_half_plus_two,target=/models/half_plus_two
    -e MODEL_NAME=half_plus_two -t tensorflow/serving
  rest_port: 8501
  grpc_port: 8500
example_calls:
  predict:
    method: POST
    url: http://localhost:8501/v1/models/half_plus_two:predict
    body: '{"instances": [1.0, 2.0, 5.0]}'
    expected: '{"predictions": [2.5, 3.0, 4.5]}'
  status:
    method: GET
    url: http://localhost:8501/v1/models/half_plus_two