[Question] Early stopping in RSLRL

[Dafodilrat]

Question

I wanted to know if a feature for early stopping exsists in RSLRL environment since currently I end up wasting a lot of iterations.

[garylvov]

Early stopping in training is a work in progress with the Isaac Lab Ray integration: #3276

The above example is for RL games, but once the feature is functional it can be applied to RSL RL with a custom configuration.

This could also be totally possible with RSL RL out of the box, but I'm not aware of such functionality.

[Mayankm96]

This seems specific to RSL-RL runner and not Isaac Lab itself. I suggest opening an issue over there instead: https://github.com/leggedrobotics/rsl_rl