Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

write script to bench the max control freq when using PPO, SAC etc #55

Open
4 tasks
Tracked by #51
Armandpl opened this issue Jan 17, 2024 · 1 comment
Open
4 tasks
Tracked by #51

Comments

@Armandpl
Copy link
Owner

Armandpl commented Jan 17, 2024

  • write wrapper that keeps track of the last time step was called
    • and also makes the action zero as we actually don't want to actuate the robot
  • configure the script with hydra
    • we want to be able to select the algorithm
    • i think we would use the files under algos, this way we can do things like bench sac vs sac but updating every episode instead of every step
  • setup and run a sweep on my machine
    • then run it on my mac
    • on Pierre's computer
    • on the rig
    • find a way to display all that in a wandb report
  • bench the velocity filter
    • run the bench with/without the velocity filter. i dont know if it takes a lot of time/compute or not. if it is slow maybe we need a faster one. or maybe we need to use recurrent algos that don't need the speed as an observation
This was referenced Jan 17, 2024
@Armandpl
Copy link
Owner Author

  • should we try benching SBX too? seems like UTD ratio is important and feel like my laptop will be too weak for that so Jax on mps might be better?
  • look at SAC when doing episodic training
  • look at SAC when doing one update per step, two updates per step etc

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant