You can use the Python transformation to define simple or complex transformation functionality. You can also use the Python transformation to implement machine learning. For example, you can load a pre-trained model through a resource file and use the model to classify input data or to create predictions.

To create a Python transformation, you write the following types of Python code snippets:

Pre-partition Python code that runs one time before it processes any input rows.
Main Python code that runs when the transformation receives an input row.
Post-partition Python code that runs after the transformation processes all input rows.

You can't use the Python transformation with a Graviton-enabled cluster. For more information about Graviton-enabled clusters, see the Administrator help.