Endpoint¶
Note
New in version 0.15.0.
-
class
verta.endpoint._endpoint.
Endpoint
¶ Object representing an endpoint for deployment.
There should not be a need to instantiate this class directly; please use
Client.get_or_create_endpoint()
.Variables: - id (int) – ID of this Endpoint.
- path (str) – Path of this Endpoint.
-
update
(model_reference, strategy=None, wait=False, resources=None, autoscaling=None, env_vars=None)¶ Updates the Endpoint with a model logged in an Experiment Run or a Model Version.
Parameters: - model_reference (
ExperimentRun
orRegisteredModelVersion
) – An Experiment Run or a Model Version with a model logged. - strategy (update strategy, default DirectUpdateStrategy()) – Strategy (direct or canary) for updating the Endpoint.
- wait (bool, default False) – Whether to wait for the Endpoint to finish updating before returning.
- resources (
Resources
, optional) – Resources allowed for the updated Endpoint. - autoscaling (
Autoscaling
, optional) – Autoscaling condition for the updated Endpoint. - env_vars (dict of str to str, optional) – Environment variables.
Returns: status (dict of str to {None, bool, float, int, str, list, dict})
- model_reference (
-
update_from_config
(filepath, wait=False)¶ Updates the Endpoint via a YAML or JSON config file.
Parameters: - filepath (str) – Path to the YAML or JSON config file.
- wait (bool, default False) – Whether to wait for the Endpoint to finish updating before returning.
Returns: status (dict of str to {None, bool, float, int, str, list, dict})
Raises: ValueError
– If the file is not JSON or YAML.
-
get_status
()¶ Gets status on the endpoint.
Returns: status (dict of str to {None, bool, float, int, str, list, dict})
-
get_access_token
()¶ Gets the access token of the Endpoint.
Returns: str or None
-
create_access_token
(token)¶ Creates an access token for the Endpoint.
Parameters: token (str) – Token to create.
-
get_deployed_model
()¶ Returns an object for making predictions against the deployed model.
Returns: DeployedModel
Raises: RuntimeError
– If the model is not currently deployed.
-
get_update_status
()¶ Gets update status on the endpoint.
Returns: update_status (dict of str to {None, bool, float, int, str, list, dict})
-
delete
()¶ Deletes this endpoint.