SHOW MODEL
Shows useful information about a machine learning model, including its status, the parameters used to create it and the prediction function with its input argument types. You can use the information from the SHOW MODEL to recreate the model. If base tables have changed, running CREATE MODEL with the same SQL statement results in a different model. The information returned by the SHOW MODEL is different for the model owner and a user with the EXECUTE privilege. SHOW MODEL shows different outputs when a model is trained from Amazon Redshift or when the model is a BYOM model.
Syntax
SHOW MODEL ( ALL |
model_name
)
Parameters
- ALL
-
Returns all the models that the user can use and their schemas.
- model_name
-
The name of the model. The model name in a schema must be unique.
Usage notes
The SHOW MODEL command returns the following:
The model name.
The schema where the model was created.
The owner of the model.
The model creation time.
The status of the model, such as READY, TRAINING, or FAILED.
The reason message for a failed model.
The validation error if model has finished training.
The estimated cost needed to derive the model for a non-BYOM approach. Only the owner of the model can view this information.
A list of user-specified parameters and their values, specifically the following:
The specified TARGET column.
The model type, AUTO or XGBoost.
The problem type, such as REGRESSION, BINARY_CLASSIFICATION, MULTICLASS_CLASSIFICATION. This parameter is specific to AUTO.
The name of the Amazon SageMaker training job or the Amazon SageMaker Autopilot job that created the model. You can use this job name to find more information about the model on Amazon SageMaker.
The objective, such as MSE, F1, Accuracy. This parameter is specific to AUTO.
The name of the created function.
The type of inference, local or remote.
The prediction function input arguments.
The prediction function input argument types for models that aren't bring your own model (BYOM).
The return type of the prediction function. This parameter is specific to BYOM.
The name of the Amazon SageMaker endpoint for a BYOM model with remote inference.
The IAM role. Only the owner of the model can see this.
The S3 bucket used. Only the owner of the model can see this.
The Amazon KMS key, if one was provided. Only the owner of the model can see this.
The maximum time that the model can run.
If the model type is not AUTO, then Amazon Redshift also shows the list of hyperparameters provided and their values.
You can also view some of the information provided by SHOW MODEL in other catalog tables, such as pg_proc. Amazon Redshift returns information about the prediction function that is registered in pg_proc catalog table. This information includes the input argument names and their types for the prediction function. Amazon Redshift returns the same information in the SHOW MODEL command.
SELECT * FROM pg_proc WHERE proname ILIKE '%<function_name>%';
Examples
The following example shows the show model output.
SHOW MODEL ALL; Schema Name | Model Name ------------+--------------- public | customer_churn
The owner of the customer_churn can see the following output. A user with only the EXECUTE privilege can't see the IAM role, the Amazon S3 bucket, and the estimated cost of the mode.
SHOW MODEL customer_churn; Key | Value ---------------------------+----------------------------------- Model Name | customer_churn Schema Name | public Owner | 'owner' Creation Time | Sat, 15.01.2000 14:45:20 Model State | READY validation:F1 | 0.855 Estimated Cost | 5.7 | TRAINING DATA: | Table | customer_data Target Column | CHURN | PARAMETERS: | Model Type | auto Problem Type | binary_classification Objective | f1 Function Name | predict_churn Function Parameters | age zip average_daily_spend average_daily_cases Function Parameter Types | int int float float IAM Role | 'iam_role' KMS Key | 'kms_key' Max Runtime | 36000