Join the SIG TFX-Addons community and help make TFX even better!

tfx.extensions.google_cloud_ai_platform.runner.deploy_model_for_aip_prediction

Deploys a model for serving with AI Platform.

tfx.extensions.google_cloud_ai_platform.runner.deploy_model_for_aip_prediction(
    api: discovery.Resource,
    serving_path: Text,
    model_version: Text,
    ai_platform_serving_args: Dict[Text, Any],
    job_labels: Dict[Text, Text],
    skip_model_creation: bool = False,
    set_default_version: bool = True
) -> None

api Google API client resource.
serving_path The path to the model. Must be a GCS URI.
model_version Version of the model being deployed. Must be different from what is currently being served.
ai_platform_serving_args Dictionary containing arguments for pushing to AI Platform. The full set of parameters supported can be found at https://cloud.google.com/ml-engine/reference/rest/v1/projects.models.versions#Version Most keys are forwarded as-is, but following keys are handled specially:

  • name: this must be empty (and will be filled by pusher).
  • deployment_uri: this must be empty (and will be filled by pusher).
  • python_version: when left empty, this will be filled by python version of the environment being used.
  • runtime_version: when left empty, this will be filled by TensorFlow version from the environment.
  • labels: a list of job labels will be merged with user's input.
job_labels The dict of labels that will be attached to this job. They are merged with optional labels from ai_platform_serving_args.
skip_model_creation If true, the method assuem model already exist in AI platform, therefore skipping model creation.
set_default_version Whether set the newly deployed model version as the default version.

RuntimeError if an error is encountered when trying to push.