This page shows you how to train a forecasting model from atabular dataset with Tabular Workflow for Forecasting.

To learn about the service accounts this workflow uses, seeService accounts for Tabular Workflows.

If you receive a quota error while running Tabular Workflow forForecasting, you might need to request a higher quota. To learn more, seeManage quotas for Tabular Workflows.

Tabular Workflow for Forecasting doesn't support model export.

Workflow APIs

This workflow uses the following APIs:

• Vertex AI
• Dataflow
• Compute Engine
• Cloud Storage

Get the URI of the previous hyperparameter tuning result

If you previously completed a Tabular Workflow for Forecasting run, youcan use the hyperparameter tuning result from the previous run to save trainingtime and resources. Find the previous hyperparameter tuning result byusing the Google Cloud console or by loading it programmatically with the API.

defget_task_detail(task_details:List[Dict[str,Any]],task_name:str)->List[Dict[str,Any]]:fortask_detailintask_details:iftask_detail.task_name==task_name:returntask_detailpipeline_task_details=job.gca_resource.job_detail.task_detailsstage_1_tuner_task=get_task_detail(pipeline_task_details,"automl-forecasting-stage-1-tuner")stage_1_tuning_result_artifact_uri=(stage_1_tuner_task.outputs["tuning_result_output"].artifacts[0].uri)

Train a model

The following sample code demonstrates how you run a model training pipeline:

job = aiplatform.PipelineJob(    ...    template_path=template_path,    parameter_values=parameter_values,    ...)job.run(service_account=SERVICE_ACCOUNT)

The optionalservice_account parameter injob.run() lets you set theVertex AI Pipelines service account to an account of your choice.

Vertex AI supports the following methods for training your model:

• Time series Dense Encoder (TiDE). To use this model training method,define your pipeline and parameter values by using the following function:

  template_path, parameter_values = automl_forecasting_utils.get_time_series_dense_encoder_forecasting_pipeline_and_parameters(...)

• Temporal Fusion Transformer (TFT). To use this model training method,define your pipeline and parameter values by using the following function:

  template_path, parameter_values = automl_forecasting_utils.get_temporal_fusion_transformer_forecasting_pipeline_and_parameters(...)

• AutoML (L2L). To use this model training method,define your pipeline and parameter values by using the following function:

  template_path, parameter_values = automl_forecasting_utils.get_learn_to_learn_forecasting_pipeline_and_parameters(...)

• Seq2Seq+. To use this model training method,define your pipeline and parameter values by using the following function:

  template_path, parameter_values = automl_forecasting_utils.get_sequence_to_sequence_forecasting_pipeline_and_parameters(...)

To learn more, seeModel training methods.

The training data can be either a CSV file in Cloud Storage or a table inBigQuery.

The following is a subset of model training parameters:

Transformations

You can provide a dictionary mapping of auto- or type-resolutions to featurecolumns. The supported types are: auto, numeric, categorical, text, and timestamp.

The following code provides a helper function for populating thetransformationsparameter. It also demonstrates how you can use this function to apply automatictransformations to a set of columns defined by afeatures variable.

defgenerate_transformation(auto_column_names:Optional[List[str]]=None,numeric_column_names:Optional[List[str]]=None,categorical_column_names:Optional[List[str]]=None,text_column_names:Optional[List[str]]=None,timestamp_column_names:Optional[List[str]]=None,)->List[Dict[str, Any]]:ifauto_column_namesisNone:auto_column_names=[]ifnumeric_column_namesisNone:numeric_column_names=[]ifcategorical_column_namesisNone:categorical_column_names=[]iftext_column_namesisNone:text_column_names=[]iftimestamp_column_namesisNone:timestamp_column_names=[]return{"auto":auto_column_names,"numeric":numeric_column_names,"categorical":categorical_column_names,"text":text_column_names,"timestamp":timestamp_column_names,}transformations=generate_transformation(auto_column_names=features)

To learn more about transformations, seeData types and transformations.

Workflow customization options

You can customize the Tabular Workflow for Forecasting by defining argumentvalues that are passed in during pipeline definition. You can customize yourworkflow in the following ways:

• Configure hardware
• Skip architecture search

Configure hardware

The following model training parameter lets you configure the machine types andthe number of machines for training. This option is a good choice if you have alarge dataset and want to optimize the machine hardware accordingly.

The following code demonstrates how to setn1-standard-8 machine type for theTensorFlow chief node andn1-standard-4 machine type for theTensorFlow evaluator node:

worker_pool_specs_override=[{"machine_spec":{"machine_type":"n1-standard-8"}},#overrideforTFchiefnode{},#overrideforTFworkernode,sinceit'snotused,leaveitempty{},#overrideforTFpsnode,sinceit'snotused,leaveitempty{"machine_spec":{"machine_type":"n1-standard-4"#overrideforTFevaluatornode}}]

Skip architecture search

The following model training parameter lets you run the pipeline without thearchitecture search and provide a set ofhyperparameters from a previous pipeline run instead.

What's next

• Learn aboutbatch inferences for forecasting models.
• Learn aboutpricing for model training.