External Model

Note

Sample formats can be downloaded from here txt

class ExternalModelFactsElements(facts_client: FactsClientAdapter)

Bases: object

save_external_model_asset(model_identifier: str, name: str, description: str = None, model_details: ModelDetails = None, schemas: ExternalModelSchemas = None, training_data_reference: TrainingDataReference = None, deployment_details: DeploymentDetails = None, model_entry_props: ModelEntryProps = None, catalog_id: str = None) → ModelAssetUtilities

Warning

Parameter(s) model_entry_props deprecated. Use save_external_model_asset().add_tracking_model_usecase() to create/link to model usecase instead. This param(s) will be removed in a future release.;

Save External model assets in catalog and (Optional) link to model usecase. By default external model is goig to save in Platform Asset Catalog ( PAC ), if user wants to save it to different catalog user has to pass catalog_id parameter.

Parameters:

model_identifier (str) – Identifier specific to ML providers (i.e., Azure ML service: service_id, AWS Sagemaker:model_name)
name (str) – Name of the model
description (str) – (Optional) description of the model
model_details (ModelDetails) – (Optional) Model details. Supported only after CP4D >= 4.7.0
schemas (ExternalModelSchemas) – (Optional) Input and Output schema of the model
training_data_reference (TrainingDataReference) – (Optional) Training data schema
deployment_details (DeploymentDetails) – (Optional) Model deployment details
model_entry_props (ModelEntryProps) – (Optional) Properties about model usecase and model usecase catalog.
catalog_id (str) – (Optional) catalog id as external model can be saved in catalog itslef..

Return type:

ModelAssetUtilities

If using external models with manual log option, initiate client as:

from ibm_aigov_facts_client import AIGovFactsClient
client= AIGovFactsClient(api_key=API_KEY,experiment_name="external",enable_autolog=False,external_model=True)

If using external models with Autolog, initiate client as:

from ibm_aigov_facts_client import AIGovFactsClient
client= AIGovFactsClient(api_key=API_KEY,experiment_name="external",external_model=True)

If using external models with no tracing, initiate client as:

from ibm_aigov_facts_client import AIGovFactsClient
client= AIGovFactsClient(api_key=API_KEY,external_model=True,disable_tracing=True)

If using Cloud pak for Data:

creds=CloudPakforDataConfig(service_url="<HOST URL>",
                            username="<username>",
                            password="<password>")

client = AIGovFactsClient(experiment_name=<experiment_name>,external_model=True,cloud_pak_for_data_configs=creds)

Payload example by supported external providers:

Azure ML Service:

from ibm_aigov_facts_client.supporting_classes.factsheet_utils import DeploymentDetails,TrainingDataReference,ExternalModelSchemas

external_schemas=ExternalModelSchemas(input=input_schema,output=output_schema)
trainingdataref=TrainingDataReference(schema=training_ref)
deployment=DeploymentDetails(identifier=<service_url in Azure>,name="deploymentname",deployment_type="online",scoring_endpoint="test/score")

client.external_model_facts.save_external_model_asset(model_identifier=<service_id in Azure>
                                                            ,name=<model_name>
                                                            ,model_details=<model_stub_details>
                                                            ,deployment_details=deployment
                                                            ,schemas=external_schemas
                                                            ,training_data_reference=tdataref)

client.external_model_facts.save_external_model_asset(model_identifier=<service_id in Azure>
                                                            ,name=<model_name>
                                                            ,model_details=<model_stub_details>
                                                            ,deployment_details=deployment
                                                            ,schemas=external_schemas
                                                            ,training_data_reference=tdataref,
                                                            ,catalog_id=<catalog_id>) Different catalog_id other than Platform Asset Catalog (PAC)

AWS Sagemaker:

external_schemas=ExternalModelSchemas(input=input_schema,output=output_schema)
trainingdataref=TrainingDataReference(schema=training_ref)
deployment=DeploymentDetails(identifier=<endpoint_name in Sagemaker>,name="deploymentname",deployment_type="online",scoring_endpoint="test/score")

client.external_model_facts.save_external_model_asset(model_identifier=<model_name in Sagemaker>
                                                            ,name=<model_name>
                                                            ,model_details=<model_stub_details>
                                                            ,deployment_details=deployment
                                                            ,schemas=external_schemas
                                                            ,training_data_reference=tdataref)


client.external_model_facts.save_external_model_asset(model_identifier=<model_name in Sagemaker>
                                                            ,name=<model_name>
                                                            ,model_details=<model_stub_details>
                                                            ,deployment_details=deployment
                                                            ,schemas=external_schemas
                                                            ,training_data_reference=tdataref,
                                                            ,catalog_id=<catalog_id>) Different catalog_id other than Platform Asset Catalog (PAC)

NOTE:

If you are are using Watson OpenScale to monitor this external model the evaluation results will automatically become available in the external model.

To enable that automatic sync of evaluation results for Sagemaker model make sure to use the Sagemaker endpoint name when creating the external model in the notebook
To enable that for Azure ML model make sure to use the scoring URL.

Example format: https://southcentralus.modelmanagement.azureml.net/api/subscriptions/{az_subscription_id}/resourceGroups/{az_resource_group}/ providers/Microsoft.MachineLearningServices/workspaces/{az_workspace_name}/services/{az_service_name}?api-version=2018-03-01-preview

model usecase props example, IBM Cloud and CPD:

>>> from ibm_aigov_facts_client.supporting_classes.factsheet_utils import ModelEntryProps,DeploymentDetails,TrainingDataReference,ExternalModelSchemas

Older way:

For new model usecase:

>>> props=ModelEntryProps(
            model_entry_catalog_id=<catalog_id>,
            model_entry_name=<name>,
            model_entry_desc=<description>
            )

For linking to existing model usecase:

>>> props=ModelEntryProps(
            model_entry_catalog_id=<catalog_id>,
            model_entry_id=<model_entry_id to link>
            )

>>> client.external_model_facts.save_external_model_asset(model_identifier=<model_name in Sagemaker>
                                                                ,name=<model_name>
                                                                ,model_details=<model_stub_details>
                                                                ,deployment_details=deployment
                                                                ,schemas=external_schemas
                                                                ,training_data_reference=tdataref
                                                                ,model_entry_props= props)

Current and go forward suggested way:

external_model=client.external_model_facts.save_external_model_asset(model_identifier=<service_id in Azure>
                                                    ,name=<model_name>
                                                    ,model_details=<model_stub_details>
                                                    ,deployment_details=deployment
                                                    ,schemas=external_schemas
                                                    ,training_data_reference=tdataref)

Create and link to new model usecase:

>>> external_model.add_tracking_model_usecase(model_usecase_name=<entry name>, model_usecase_catalog_id=<catalog id>)

Link to existing model usecase:

>>> external_model.add_tracking_model_usecase(model_usecase_id=<model_usecase_id>, model_usecase_catalog_id=<catalog id>)

To remove model usecase:

>>> external_model.remove_tracking_model_usecase()

unregister_model_entry(asset_id, catalog_id)

Warning

ibm_aigov_facts_client.factsheet.external_modelfacts_utility.ExternalModelFactsElements.unregister_model_entry is deprecated. This method will be removed in a future release.Use save_external_model_asset().remove_tracking_model_usecase() instead.

Unregister WKC Model usecase

Parameters:

asset_id (str) – WKC model usecase id
catalog_id (str) – Catalog ID where asset is stored

Example for IBM Cloud or CPD:

>>> client.external_model_facts.unregister_model_entry(asset_id=<model asset id>,catalog_id=<catalog_id>)

list_model_entries(catalog_id=None) → list

Warning

ibm_aigov_facts_client.factsheet.external_modelfacts_utility.ExternalModelFactsElements.list_model_entries is deprecated. This method will be removed in a future release.Use client.assets.list_model_usecases() instead.

Returns all WKC Model usecase assets for a catalog

Parameters:: catalog_id (str) – (Optional) Catalog ID where you want to register model, if None list from all catalogs
Returns:: All WKC Model usecase assets for a catalog
Return type:: list

Example:

>>> client.external_model_facts.list_model_entries()
>>> client.external_model_facts.list_model_entries(catalog_id=<catalog_id>)

Deployments Methods