Difference in result between test and train_test

[wissamsleiman]

Thank you for the great work!
My issue is as follow:

When training the presslight agent, the training testing results are as follows:

When only testing this trained model by changing load_model to True in base.yml and the train_model to False in presslight.yml, I obtain the following:

I don't understand where this difference is coming from and would appreciate any help or clarification in case I am missing something. Thank you in advance

edit: I changed the pressure from difference in queues to difference in densities, but can't find where to change it for the testing function too

[Red-Pheonix]

If you changed the reward/observer function in the agent themselves, then no need. It will work for both training and testing.

Regarding the load_model, I remember it not working properly for me. I need to check again what exactly was the problem. Will let you know.