Questions about using TFF with tfdf.keras.RandomForestModel

[drakstik]

I've been trying to use TFF for a simple tfdf.keras.RandomForestModel model. I've followed the MNIST example, however I think I need to create a custom FL algorithm for such a model. I have tried this using the build your own custom algorithm example (https://www.tensorflow.org/federated/tutorials/building_your_own_federated_learning_algorithm#preparing_the_input_data). In the example the use a Sequential model for the custom algorithm.

I am having issues implementing a Sequential model for a tfdf.keras.RandomForestModel. How would I go about building my own custom model for a simple tfdf.keras.RandomForestModel that is trained in a Federal Learning manner? Should I be building a custom model? If so do I need to make it Sequential?

Any help would be super appreciated!