ML - Data transfer when running a model in Vertica?

An ML model is a catalog object backed by several files on Vertica's internal DFS. When an appropriate predict function is called to use a model for in-database scoring (prediction), its files are read from the DFS (storage layer) into memory, and then this information is employed for calculating the predicted values. The input data is fed to the function the same way it is fed into any other function in database.