Repository navigation
[Algorithm] Add MicroDuck PPO MJLab example - #4183
Merged
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/4183
Note: Links to docs will display an error until the docs builds have been completed. This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
github-actions
Bot
added
new algo
vmoens
force-pushed
the
codex/microduck-ppo-example
branch
from
August 31, 2026 07:58
69076d8 to
331e5a8
Compare
github-actions
Bot
added
Documentation
Group MicroDuck examples by robot rather than by simulator backend so the MJLab and backend-neutral MuJoCo scripts live side by side. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
mjlab reports episode reward terms, curriculum levels and termination counts through the extras["log"] mapping returned by reset and step. The wrapper discarded it, so training scripts had to subclass _reset and _step to recover those metrics. Add a log_extras option that accumulates the numeric entries and a pop_logged_extras method returning their means since the previous call. MJLabEnv and MJLabEnvConfig forward the option. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Provide the KL-targeted learning-rate schedule popularized by rsl_rl as a reusable utility: shrink the learning rate when the measured policy KL exceeds twice the target, grow it when the KL falls below half the target, and clamp every parameter group to a configurable range. The kl_approx output of ClipPPOLoss can be passed to step() directly. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Use MJLabWrapper(log_extras=True) instead of a private subclass, replace the hand-rolled adaptive learning-rate schedule with KLAdaptiveLR, fill the replay buffer once per iteration instead of once per epoch, return the environment and its normalizer from make_env, and log through torchrl_logger. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
vmoens
force-pushed
the
codex/microduck-ppo-example
branch
from
September 2, 2026 19:58
b65177f to
cad787a
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.