Transform is available as a standalone library.
The tft module documentation is the only module that is relevant to TFX users.
The tft_beam module is relevant only when using Transform as a standalone library. Typically, a TFX user constructs a preprocessing_fn, and the rest of the
Transform library calls are made by the Transform component.
You can also use the Apache Beam MLTransform class to preprocess data
for training and inference. The MLTransform class wraps multiple TFX data
processing transforms in one class. For more information, see
Preprocess data with MLTransform
in the Apache Beam documentation.