Open muyu11 opened 2 weeks ago
To record additional information returned in the step() method, you can edit on_policy_marl.py by adding the following line after line 427 (version 1.2.3):
step_info["Train-Results/Other-Info"] = {"env-%d" % i: info[i]["Other-Info"]}
This will allow you to capture and log the desired information.
I'm training a customize mappo environment, now I can visualize the rewards during my training, but I wanna record other information in a dict "self.infos" and can be visualized in tensorboard, self.infos will be returned in the method "step", so how can I realize that?
Thanks