Repository navigation

All pull requests

Labels (9)Milestones (0)New pull request

Open

(35)

Closed

(25)

feat: Windows 10 support
#77billfan123 opened Oct 16, 2026
Windows support for infer_policy.py (termios shims, msvcrt keys, --max-seconds)
#71professorwang opened Oct 4, 2026
docs: Intel GPU training via the mjlab-sycl companion package
#69guang384 opened Oct 3, 2026
sim: a challenge course and a playground to drive in
#68DerekChia opened Oct 3, 2026
fix: keep CPU policy replay compatible with an independent mouth actuator
#60payals opened Sep 26, 2026
Record and publish the recipe of every run; usable from a challenges repo
#57nicolas-rabault opened Sep 25, 2026
Policy export contract
#56pierre-rouanet opened Sep 25, 2026Member
Add headstand routine training tasks
#55zachgarner opened Sep 22, 20266 tasks done
velstand: hand fast-rotating fallen frames to PPO, not the stand expert
#54mina1turintech opened Sep 21, 2026
Add Mjlab-PoliteBow-Flat-MicroDuck episodic task
#53SantiagoGorbea opened Sep 16, 20263 tasks
Cross-stack reward parity: feet_flat sign fix, PR gate, MJX migration scaffold
#52hzzqq opened Sep 15, 2026
Add optional read-only MuJoCo visualization stream
#51microai-lab opened Sep 11, 2026
Approach-and-inspect task (Mjlab-ApproachInspect-*-MicroDuck)
#50mcolonpappaterra opened Sep 11, 2026
Fix windows CUDA installation from pypi defaulting to CPU-only
#49catherinepereira opened Sep 11, 2026
reuse NUM_STEPS_PER_ENV in the RslRlOnPolicyRunnerCfg
#44saikishor opened Sep 7, 2026
docs: note that the viewer needs mjpython on macOS
#43raphael2053 opened Sep 7, 2026
Distill neutral-head Flamingo controller
#42abhinav3 opened Sep 5, 2026
Validate simulator motor targets before updating controls
#41qilinxiaoxiang opened Sep 5, 2026
Jetson AGX Thor (JetPack 7.2): route torch to NVIDIA's SBSA index, keep GB10 on cu129
#39OriNachum opened Sep 4, 2026
feat: add guarded Next-RL training workspace
#37rakeshutekar opened Sep 3, 2026
Add commanded left/right single-leg standing task
#36BingoWon opened Sep 3, 2026
fix: make train --hf-jobs work on Windows
#35lboucher26 opened Sep 3, 2026
feat: re-structured task configs into readable IsaacLab style
#34execbat opened Sep 2, 2026
Train strict grounded MicroDuck backrolls
#32ceddc opened Sep 2, 2026
feat(tasks): Mjlab-Jump-Flat-MicroDuck — phase-driven periodic hop
#28metahubaifeel opened Sep 2, 2026

Navigation Menu

Repository navigation

All pull requests

feat: Windows 10 support

Windows support for infer_policy.py (termios shims, msvcrt keys, --max-seconds)

docs: Intel GPU training via the mjlab-sycl companion package

sim: a challenge course and a playground to drive in

fix: keep CPU policy replay compatible with an independent mouth actuator

Record and publish the recipe of every run; usable from a challenges repo

Policy export contract

Add headstand routine training tasks

velstand: hand fast-rotating fallen frames to PPO, not the stand expert

Add Mjlab-PoliteBow-Flat-MicroDuck episodic task

Cross-stack reward parity: feet_flat sign fix, PR gate, MJX migration scaffold

Add optional read-only MuJoCo visualization stream

Approach-and-inspect task (Mjlab-ApproachInspect-*-MicroDuck)

Fix windows CUDA installation from pypi defaulting to CPU-only

reuse NUM_STEPS_PER_ENV in the RslRlOnPolicyRunnerCfg

docs: note that the viewer needs mjpython on macOS

Distill neutral-head Flamingo controller

Validate simulator motor targets before updating controls

Jetson AGX Thor (JetPack 7.2): route torch to NVIDIA's SBSA index, keep GB10 on cu129

feat: add guarded Next-RL training workspace

Add commanded left/right single-leg standing task

fix: make `train --hf-jobs` work on Windows

feat: re-structured task configs into readable IsaacLab style

Train strict grounded MicroDuck backrolls

feat(tasks): Mjlab-Jump-Flat-MicroDuck — phase-driven periodic hop

Pull requests