AutoML Forecasting¶

Experimental AutoML forecasting components.

Components:

`ForecastingEnsembleOp`(project, location, ...)	Ensembles AutoML Forecasting models.
`ForecastingStage1TunerOp`(project, location, ...)	Searches AutoML Forecasting architectures and selects the top trials.
`ForecastingStage2TunerOp`(project, location, ...)	Tunes AutoML Forecasting models and selects top trials.

preview.automl.forecasting.ForecastingEnsembleOp(project: str, location: str, root_dir: str, transform_output: dsl.Input[system.Artifact], metadata: dsl.Input[system.Artifact], tuning_result_input: dsl.Input[system.Artifact], instance_baseline: dsl.Input[system.Artifact], instance_schema_path: dsl.Input[system.Artifact], prediction_image_uri: str, gcp_resources: dsl.OutputPath(str), model_architecture: dsl.Output[system.Artifact], example_instance: dsl.Output[system.Artifact], unmanaged_container_model: dsl.Output[google.UnmanagedContainerModel], explanation_metadata: dsl.OutputPath(dict), explanation_metadata_artifact: dsl.Output[system.Artifact], explanation_parameters: dsl.OutputPath(dict), encryption_spec_key_name: str | None = '')¶

Ensembles AutoML Forecasting models.

Parameters¶

project: str¶: Project to run the job in.
location: str¶: Region to run the job in.
root_dir: str¶: The Cloud Storage path to store the output.
transform_output: dsl.Input[system.Artifact]¶: The transform output artifact.
metadata: dsl.Input[system.Artifact]¶: The tabular example gen metadata.
tuning_result_input: dsl.Input[system.Artifact]¶: AutoML Tabular tuning result.
instance_baseline: dsl.Input[system.Artifact]¶: The instance baseline used to calculate explanations.
instance_schema_path: dsl.Input[system.Artifact]¶: The path to the instance schema, describing the input data for the tf_model at serving time.
encryption_spec_key_name: str | None = ''¶: Customer-managed encryption key.
prediction_image_uri: str¶: URI of the Docker image to be used as the container for serving predictions. This URI must identify an image in Artifact Registry or Container Registry.

Returns¶

gcp_resources: dsl.OutputPath(str): GCP resources created by this component. For more details, see https://github.com/kubeflow/pipelines/blob/master/components/google-cloud/google_cloud_pipeline_components/proto/README.md.
model_architecture: dsl.Output[system.Artifact]: The architecture of the output model.
nmanaged_container_model: dsl.Output[google.UnmanagedContainerModel]: Model information needed to perform batch prediction.
explanation_metadata: dsl.OutputPath(dict): The explanation metadata used by Vertex online and batch explanations.
explanation_metadata_artifact: dsl.Output[system.Artifact]: The explanation metadata used by Vertex online and batch explanations in the format of a KFP Artifact.
explanation_parameters: dsl.OutputPath(dict): The explanation parameters used by Vertex online and batch explanations.
example_instance: dsl.Output[system.Artifact]: An example instance which may be used as an input for predictions.

preview.automl.forecasting.ForecastingStage1TunerOp(project: str, location: str, root_dir: str, num_selected_trials: int, deadline_hours: float, num_parallel_trials: int, single_run_max_secs: int, metadata: dsl.Input[system.Artifact], transform_output: dsl.Input[system.Artifact], materialized_train_split: dsl.Input[system.Artifact], materialized_eval_split: dsl.Input[system.Artifact], gcp_resources: dsl.OutputPath(str), tuning_result_output: dsl.Output[system.Artifact], study_spec_parameters_override: list | None = [], worker_pool_specs_override_json: list | None = [], reduce_search_space_mode: str | None = 'regular', encryption_spec_key_name: str | None = '')¶

Searches AutoML Forecasting architectures and selects the top trials.

Parameters¶

project: str¶: Project to run hyperparameter tuning.
location: str¶: Location for running the hyperparameter tuning.
root_dir: str¶: The Cloud Storage location to store the output.
study_spec_parameters_override: list | None = []¶: JSON study spec. E.g., [{“parameter_id”: “activation”,”categorical_value_spec”: {“values”: [“tanh”]}}]
worker_pool_specs_override_json: list | None = []¶: JSON worker pool specs. E.g., [{“machine_spec”: {“machine_type”: “n1-standard-16”}},{},{},{“machine_spec”: {“machine_type”: “n1-standard-16”}}]
reduce_search_space_mode: str | None = 'regular'¶: The reduce search space mode. Possible values: “regular” (default), “minimal”, “full”.
num_selected_trials: int¶: Number of selected trials. The number of weak learners in the final model is 5 * num_selected_trials.
deadline_hours: float¶: Number of hours the hyperparameter tuning should run.
num_parallel_trials: int¶: Number of parallel training trials.
single_run_max_secs: int¶: Max number of seconds each training trial runs.
metadata: dsl.Input[system.Artifact]¶: The tabular example gen metadata.
transform_output: dsl.Input[system.Artifact]¶: The transform output artifact.
materialized_train_split: dsl.Input[system.Artifact]¶: The materialized train split.
materialized_eval_split: dsl.Input[system.Artifact]¶: The materialized eval split.
encryption_spec_key_name: str | None = ''¶: Customer-managed encryption key.

Returns¶

gcp_resources: dsl.OutputPath(str): GCP resources created by this component. For more details, see https://github.com/kubeflow/pipelines/blob/master/components/google-cloud/google_cloud_pipeline_components/proto/README.md.
ning_result_output: dsl.Output[system.Artifact]: The trained model and architectures.

preview.automl.forecasting.ForecastingStage2TunerOp(project: str, location: str, root_dir: str, num_selected_trials: int, deadline_hours: float, num_parallel_trials: int, single_run_max_secs: int, metadata: dsl.Input[system.Artifact], transform_output: dsl.Input[system.Artifact], materialized_train_split: dsl.Input[system.Artifact], materialized_eval_split: dsl.Input[system.Artifact], tuning_result_input_path: dsl.Input[system.Artifact], gcp_resources: dsl.OutputPath(str), tuning_result_output: dsl.Output[system.Artifact], worker_pool_specs_override_json: list | None = [], encryption_spec_key_name: str | None = '')¶

Tunes AutoML Forecasting models and selects top trials.

Parameters¶

project: str¶: Project to run stage 2 tuner.
location: str¶: Cloud region for running the component: us-central1).
root_dir: str¶: The Cloud Storage location to store the output.
worker_pool_specs_override_json: list | None = []¶: JSON worker pool specs. E.g., [{“machine_spec”: {“machine_type”: “n1-standard-16”}},{},{},{“machine_spec”: {“machine_type”: “n1-standard-16”}}]
num_selected_trials: int¶: Number of selected trials. The number of weak learners in the final model.
deadline_hours: float¶: Number of hours the cross-validation trainer should run.
num_parallel_trials: int¶: Number of parallel training trials.
single_run_max_secs: int¶: Max number of seconds each training trial runs.
metadata: dsl.Input[system.Artifact]¶: The forecasting example gen metadata.
transform_output: dsl.Input[system.Artifact]¶: The transform output artifact.
materialized_train_split: dsl.Input[system.Artifact]¶: The materialized train split.
materialized_eval_split: dsl.Input[system.Artifact]¶: The materialized eval split.
encryption_spec_key_name: str | None = ''¶: Customer-managed encryption key.
tuning_result_input_path: dsl.Input[system.Artifact]¶: Path to the json of hyperparameter tuning results to use when evaluating models.

Returns¶

gcp_resources: dsl.OutputPath(str): GCP resources created by this component. For more details, see https://github.com/kubeflow/pipelines/blob/master/components/google-cloud/google_cloud_pipeline_components/proto/README.md.
ning_result_output: dsl.Output[system.Artifact]: The trained (private) model artifact paths and their hyperparameters.