IBM Watson Machine Learning

You can use IBM Watson Machine Learning to perform payload logging, feedback logging, and to measure performance accuracy, runtime bias detection, drift detection, explainability, and auto-debias function in IBM Watson OpenScale.

IBM Watson OpenScale fully supports the following IBM Watson Machine Learning frameworks:

Table 1. Framework support details

¹To learn more about AutoAI, see AutoAI implementation details. For models where the training data is in Cloud Object Storage, there is no support for fairness attributes of type boolean. However, if the training data is in Db2, Watson OpenScale supports fairness attributes that are boolean type. When using the AutoAI option, Watson OpenScale does not support models where the data type of the model prediction is binary. You must change such models so that the data type of their prediction is a string data type.

²Fairness and drift metrics are not supported for unstructured (image or text) data types.

³Keras support does not include support for fairness.

⁴Explainability is supported if your model / framework outputs prediction probabilities.

⁵For XGBoost binary and multiple class models, you must update the model to return prediction probability in the form of numerical values for binary models and a list of probabilities per class for multi-class models. Support for the XGBoost framework has the following limitations for classification problems: For binary classification, Watson OpenScale supports the binary:logistic logistic regression function with an output as a probability of True. For multiclass classification, Watson OpenScale supports the multi:softprob function where the result contains the predicted probability of each data point belonging to each class.

XGBoost subscriptions limitation for native XGBoost models

With the latest version WML , the xgboost_0.82 framework is deprecated. The supported framework is xgboost_0.90, which must be used in combination with Python 3.7. To enable and persist XGBoost version 0.90 and Python 3.7, you must patch subscriptions by running the following command:

PATCH /v2/subscriptions/<subscription_id>

For example, the following command provides the full URL and subscription ID:

PATCH https://namespace.apps.server.company.com/openscale/00000000-0000-0000-0000-000000000000/v2/subscriptions/18bbe3f9-c550-44e7-b948-675bafcb1c63

The expected payload is returned:

[{
    "op": "replace",
    "path": "\/asset\/runtime_environment",
    "value": "xgboost_0.90"
}]

After the subscription is successfully patched and scoring is done, the output data schema will have right modeling role set:

{
	"metadata": {
		"modeling_role": "class_probability"
	},
	"name": "prediction",
	"nullable": true,
	"type": "double"
}, {
	"metadata": {},
	"name": "probability",
	"nullable": true,
	"type": "double"
}, {
	"metadata": {
		"modeling_role": "prediction-probability"
	},
	"name": "prediction_probability",
	"nullable": true,
	"type": "double"
}, {
	"metadata": {
		"modeling_role": "prediction"
	},
	"name": "scoring_prediction",
	"nullable": true,
	"type": "integer"
}

AutoAI models that use the other field type fail

If the training_data_schema contains some fields with the data type other, Watson OpenScale fails to create the subscription. The other data type is not supported by Watson OpenScale. Perform the following steps to suppress the training_data_schema as a way to work around this limitation and create a subscription for the AutoAI deployment.

1. Create subscription using SDK without setting the following attributes in the AssetPropertiesRequest objects

   • AssetPropertiesRequest.feature_fields
   • AssetPropertiesRequest.categorical_fields
   • AssetPropertiesRequest.training_data_schema
   • AssetPropertiesRequest.input_data_schema
   • AssetPropertiesRequest.output_data_schema

2. Perform payload logging by scoring the deployment against IBM Watson Machine Learning either through Watson OpenScale UI payload logging sample code or through the IBM Watson Machine Learning Client.

   This internally sets all the subscription related schema information.

AutoAI models and training data

AutoAI automatically prepares data, applies algorithms, or estimators, and builds model pipelines best suited for your data and use case. Watson OpenScale requires access to the training data to analyze the model.

Because Watson OpenScale cannot detect the training data location for an AutoAI model like it can for a regular model, you must explicitly provide the needed details to access the training data location:

• For the online path, where you manually configuring monitors, you must provide the database details from which training data can be accessed .
• For the custom notebook path, where you upload training data distribution, you can use the JSON file that is produced by running the notebook.

For more information, see Provide model details and Numeric/categorical data.

Specifying an IBM Watson Machine Learning service instance

Your first step in the Watson OpenScale tool is to specify an IBM Watson Machine Learning instance. Your Machine Learning instance is where you store your AI models and deployments.

Prerequisites

You should have provisioned an IBM Watson Machine Learning instance in the same account or cluster where the Watson OpenScale service instance is present. If you have provisioned a IBM Watson Machine Learning instance in some other account or cluster, then you will not be able to configure that instance with automatic payload logging with Watson OpenScale.

Connect your Machine Learning service instance

Watson OpenScale connects to AI models and deployments in an IBM Watson Machine Learning instance. To connect your service to Watson OpenScale, go to the Configure tab, add a machine learning provider, and click the Edit icon. Provide a name and description and whether this is a Pre-production or Production environment type. If you have an instance of IBM Watson Machine Learning, Watson OpenScale detects it, along with the configuration information.

Next steps

You are now ready to select deployed models and configure your monitors. Watson OpenScale lists your deployed models on the Insights dashboard where you can click the Add to dashboard button. Select the deployments you want to monitor and click Configure.

For more information, see Configure monitors.