How is success rate computed in the paper?

[CoffeePot1206]

Hi,

Your work is really helpful for my start on RL training! But I have a question about how to compute the "Average Episode Success Rate" reported in your paper, using your codebase. I have found in class AllegroHandDynamicHandover (in dynamic_handover/dexteroushandenvs/tasks/allegro_hand_dynamic_handover.py), there is a member variable self.successes. But it is true only when goal_dist<=0. And I also outputed the self.successes and found it was all False during the whole training process. There must be something I omitted.

Thanks in advance for your reply!