Closed gianscarpe closed 1 year ago
Hi @gianscarpe, are you trying to run the random agent with a discrete action space as suggested in the snippet from the README that I posted below. Overall, its very difficult to say why the code is not able to use the discrete_waypoint_controller without the steps to reproduce the error and a stack trace.
Hi @ykarmesh, My fault I didn't specified all the details. I'm not trying with the random agent. I'm trying to run a discrete goalfollower from habitat-baselines and I need to customize my code (which predicts discrete actions) to work in the challenge environment. I tried to modify "velocitycontrol" action space in the hydra config with either v0, v1, and discrete_waypoint_controller, but none of those worked. Can you help me figure this out?
A few questions to understand the error more: 1) Are you using a branch based of the challenge-2023 tag? 2) Do you have the hydra configs belonging to the different actions inside the config folder? 3) What is the exact error that you are getting?
Thank you! 2 solved my issue
Hi! I'm trying running a policy with discrete actions, but I couldn't figure out the action_space_config to use, as
discrete_waypoint_controller
does not exist. Can you point me the correct action space to use?