Create a trainer, like misty.py, ketchum.py or trainer.py, that can handle both markov kernel and RNN models (and ideally any new type of model).
Ideas:
Might require offloading some functions, like which data retrieval function to use, to the model definition.
Also requires adding the handling of soft_margin to the trainer.
Create a trainer, like misty.py, ketchum.py or trainer.py, that can handle both markov kernel and RNN models (and ideally any new type of model). Ideas:
Also requires adding the handling of soft_margin to the trainer.