Calling of predict method in DRL_predict

[rhelpacc]

Hi - I noticed that the DRL_predict which is used when we take the trained model to trade out of sample calls the model.predict method without specifying deterministic=True. The default value of this argument in stable_baselines for model.predict is False. This means that at out of sample, we sample out of the policy distribution. I am wondering if you could provide rationale for this sampling as opposed to just use the distribution mode? Thank you very much.