You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I tried to play with the re-orientation task by simply running
"python train.py --task=ShadowHandReOrientation --algo=ppo"
without making any further changes. After 6500 iterations, the training ended. But the trained controller could not perform the task well.
The hand only makes small moves (shaking) which could not re-orient the cube.
Is there anything that I'm doing wrong?
Thanks
Zheyu
The text was updated successfully, but these errors were encountered:
First of all, the baselines of all our environments are trained under 2048 environments with 6500 iteration, maybe you can check the number of parallel environments.
But, you're not wrong, we can't actually do it well ourselves too. Even though the setup of our task is completely a two-handed version of IsaacGymEnv's one-handed Re-Orientation, it still doesn't perform well after training. Perhaps the two-handed Re-Orientation is much more difficult than the one-handed Re-Orientation.
In fact, we have always said that there are many challenging tasks in Bi-DexHands that cannot be fully completed, but we still released them. For example, not only ShadowHandReOrientation, but also ShadowHandTwoCatchUnderarm, ShadowHandKettle. We also hope this study to provide a new challenge for the community.
Hi,
I tried to play with the re-orientation task by simply running
"python train.py --task=ShadowHandReOrientation --algo=ppo"
without making any further changes. After 6500 iterations, the training ended. But the trained controller could not perform the task well.
The hand only makes small moves (shaking) which could not re-orient the cube.
Is there anything that I'm doing wrong?
Thanks
Zheyu
The text was updated successfully, but these errors were encountered: