Use Snowflake machine learning models in a Snowflake Native App¶

Overview of using Snowpark ML in a Snowflake Native App¶

Snowflake ML is an integrated set of capabilities for end-to-end machine learning in a single platform on top of your governed data. You can this functionality within a Snowflake Native App.

The Snowflake Native App Framework supports the following use cases:

• Providers include a training algorithm in the app, but the trained model is not included. Providers include the source code for the model, for example linear regression or logistical regression, in the app.

  After the app is installed, training occurs on data in the consumer account, for example by calling the model’s fit() method.

  For more information, see Create, train and use a Snowflake ML model in an app.

• Providers share data with the consumer and include a training algorithm in the app. After installation, the app trains the model based on data in the consumer account that has been shared with the app

  For more information, see Create, train and use a Snowflake ML model in an app.

• Providers train a model based on data in their account and include these models in the app. When the app is installed, consumers can use the model directly, for example by calling the model’s :predict() method.

  For more information, see Include a trained model in an app.

Limitations when using Snowflake ML in an app¶

The following limitations apply when using Snowflake ML in an app:

• Only models based on warehouses are currently supported.

• Providers must use the Snowflake Model Registry to share models with consumers. Snowpark ML functions like fit() store results in a temporary stage which is not supported for Snowflake Native Apps.

• There are limitations on machine learning algorithms that are runnable in a Snowpark sandbox within a warehouse. More complex machine learning frameworks like TensorFlow or PyTorch are not runnable in these sandboxes.

• Training performed on a provider’s dataset may not yield a model sufficiently effective for a consumer’s data. Training a model on consumer data may provide better results.

Calling Snowflake Cortex functions from an app¶

To call a Snowflake Cortex function from an app, consumers must first grant the CORTEX_USER database role to the app as shown in the following example:

GRANT DATABASE ROLE SNOWFLAKE.CORTEX_USER TO APPLICATION my_app;

The CORTEX_USER database role in the SNOWFLAKE database includes the privileges that allow users to call Snowflake Cortex LLM functions. See Snowflake Cortex AISQL (including LLM functions) for more information.

After consumers this role to the app, the app can call Snowflake Cortex functions as shown in the following example:

SELECT SNOWFLAKE.CORTEX.TRANSLATE('La plateforme unique de Snowflake élimine les silos de données!','fr','en');