gigl.common.services.VertexAIService

class gigl.common.services.vertex_ai.VertexAIService(project: str, location: str, service_account: str, staging_bucket: str)

Bases: object

A class representing a Vertex AI service.

Args:

project (str): The project ID. location (str): The location of the service. service_account (str): The service account to use for authentication. staging_bucket (str): The staging bucket for the service.

Methods

__init__
get_pipeline_job_from_job_name	Fetches the pipeline job with the given job name.
get_pipeline_run_url	Returns the URL for the pipeline run.
launch_job	Launch a Vertex AI CustomJob.
run_pipeline	Runs a pipeline using the Vertex AI Pipelines service.
wait_for_run_completion	Waits for a run to complete.

__init__(project: str, location: str, service_account: str, staging_bucket: str)

__weakref__

list of weak references to the object (if defined)

get_pipeline_job_from_job_name(job_name: str) → PipelineJob

Fetches the pipeline job with the given job name.

static get_pipeline_run_url(project: str, location: str, job_name: str) → str

Returns the URL for the pipeline run.

launch_job(job_config: VertexAiJobConfig) → None

Launch a Vertex AI CustomJob. See the docs for more info. https://cloud.google.com/python/docs/reference/aiplatform/latest/google.cloud.aiplatform.CustomJob

Args:

job_config (VertexAiJobConfig): The configuration for the job.

property project: str

The GCP project that is being used for this service.

run_pipeline(display_name: str, template_path: Uri, run_keyword_args: Dict[str, str], job_id: str | None = None, experiment: str | None = None) → PipelineJob

Runs a pipeline using the Vertex AI Pipelines service. For more info, see the Vertex AI docs https://cloud.google.com/python/docs/reference/aiplatform/latest/google.cloud.aiplatform.PipelineJob#google_cloud_aiplatform_PipelineJob_submit

Args:

display_name (str): The display of the pipeline. template_path (Uri): The path to the compiled pipeline YAML. run_keyword_args (Dict[str, str]): Runtime arguements passed to your pipeline. job_id (Optional[str]): The ID of the job. If not provided will be the pipeline_name + datetime.

Note: The pipeline_name and display_name are not the same. Note: pipeline_name comes is defined in the template_path and ultimately comes from Python pipeline definition. If provided, must be unique.

experiment (Optional[str]): The name of the experiment to associate the run with.

Returns:

The PipelineJob created.

static wait_for_run_completion(resource_name: str, timeout: float = 129600, polling_period_s: int = 60) → None

Waits for a run to complete.

Args:

resource_name (str): The resource name of the run. timeout (float): The maximum time to wait for the run to complete, in seconds. Defaults to 7200. polling_period_s (int): The time to wait between polling the run status, in seconds. Defaults to 60.

Returns:

None