Skip to content

Training in mjlab #10

Description

@1YI-DING

Hi FlashSAC team, thanks for the great work! :)

I'm trying to integrate FlashSAC into mjlab for a humanoid velocity-tracking task. Reward terms are ported from the PPO recipe, and I've already rebalanced several weights following the guidance in Table 14 of the paper.
Across multiple runs, the policy still converges to a degenerate behavior: the robot stands almost still and only tilts its torso slightly toward the velocity command, instead of walking.

A few questions:

  1. Is this a known failure mode? Any specific reward terms that typically need the most attention beyond what's listed in the paper?
  2. Is there a recommended reward template for mjlab specifically ?

Happy to share my reward config and training curves. Thanks!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions