-
Notifications
You must be signed in to change notification settings - Fork 167
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
[CSE 276F Submission] PlaceSphere-v1 #379
Conversation
Work looks good. Just resolve conflicts with main and see my comments. Can confirm RL works as well. |
Conflicts resolved. I re-structured a bit and hopefully I didn't make their RollBall docs wrong. |
Still missing fixes for the 2 reviewed parts. see comments above |
I can't see any comment or required change in the code... Could you specify where they are? |
Okay now the 2 parts are fixed |
The task involves grasping a sphere and placing that on top of a little bin. The task is solvable via ppo in 5000_0000 epochs.
Training command: python ppo.py --env_id="PlaceSphere-v1" --num_envs=1024 --update_epochs=8 --num_minibatches=32 --total_timesteps=50_000_000
Evaluation command: python ppo.py --env_id="PlaceSphere-v1" --evaluate --checkpoint=/path_to_final_ckpt.pt --num_eval_envs=1 --num-eval-steps=1000 --seed=2
Evaluation results:
eval_success_rate=0.8478260869565217