Agent not learning for maze environment with CRAR

[parthchadha]

I am trying to reproduce the results of the CRAR agent in the maze environment and am observing that the agent's test reward is not improving at all. It stays at about -5 for all the 250 epochs. Can you please point me to the experiment settings that can reproduce the results?

[VinF]

Hi Parth,

Thanks for reaching out. The value for the learning rate of the representation loss in CRAR is set to a different value for that experiment:

deer/deer/learning_algos/CRAR_keras.py

Line 583 in 83bf533

K.set_value(self.diff_s_s_.optimizer.lr, self._lr/5.) # /5. for simple laby or simple catcher; /1. for distrib of laby

It must be set at the value self._lr instead of self._lr/5. Please let me know if you still run into problems.

Apologies for this hard coded value.

[parthchadha]

Hi Vincent,
Thanks for pointing this out. I am able to reproduce the results now!