How to utilize a pre-established deep reinforcement learning model with AnyLogic as both a training and testing environment?

It is possible to export an AnyLogic model and use a deep reinforcement learning model for training, or alternatively, embed the deep reinforcement learning model into AnyLogic. I have only found information on exporting AnyLogic models and using Java for training with a deep reinforcement learning model. I am unsure whether it is feasible to export an AnyLogic model, train it using Python and PyTorch, and then import it back into AnyLogic for training and testing.