Edit

Make predictions with Machine Learning extension for Azure Data Studio (Preview)

  • Article

Learn how to use the Machine Learning extension for Azure Data Studio to make predictions with an ONNX model in your database. The extension will generate a T-SQL script using PREDICT to make predictions on the dataset stored in your table with a model that is previously imported, resides in a local file, or from Azure Machine Learning.

Important

Make predictions with the Machine Learning extension currently only supports Machine Learning Services in Azure SQL Managed Instance and Azure SQL Edge with ONNX.

Prerequisites

Make predictions from ONNX model

Follow the steps below to use an ONNX model to make predictions.

  1. Select Make predictions.

  2. If you're asked to install onnxruntime, mlflow, and mlflow-dbstore, select Yes.

  3. Choose where your model is located and select Next. You can use:

    • Imported models. Choose this to use a model that is already stored in your database. Choose the Model database and Model table where your model is located, select the model you want to use, and select Next.
    • File upload. Choose this to use a model from a file. Select the model file under Source files and select Next.
    • Azure Machine Learning. Choose this to use a model from Azure Machine Learning. First, Sign in to Azure. Then select your Azure account, Azure subscription, Azure resource group, and Azure ML workspace. Select the model you want to use and select Next.

  4. Map the source data to your model.

    • Select the Source database and Source table containing the data set for which you want to apply the prediction.
    • Map the columns under Model Input mapping and Model output. The extension will automatically map columns that have the same name and data type.

  5. Select Predict.

Azure Data Studio will create a new T-SQL query with the PREDICT, which you can use to make predictions on your data.

Next steps