tug-league-c19
Mjlab-TugOfWar-MicroDuck
A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot.
Run it on a robot
sudo robotctl policy load <slot> nikghodki/microduck-tug-league-c19
The observation normalizer is baked into policy.onnx; feed raw observations.
manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo).
Training
- task_id:
Mjlab-TugOfWar-MicroDuck - repo:
https://github.com/pollen-robotics/microduck-challenges - branch:
duel-pool - commit:
c97a3acbc - checkpoint:
3781 - seed:
19 - base:
mjlab-microduck 0.1.0 @ 1efd853be - started:
2026-11-11T00:22:45Z - exported from a checkout with uncommitted changes
Reproduce
Same code, same uv.lock, same command, same seed. Training it again yields a comparable policy, not the same weights: GPU reinforcement learning is not bit-reproducible across machines.
git clone https://github.com/pollen-robotics/microduck-challenges
cd microduck-challenges
git checkout c97a3acbc
uv sync
uv run train Mjlab-TugOfWar-MicroDuck --env.scene.num-envs 4096 --agent.max-iterations 200 --agent.seed 19 --agent.logger tensorboard --agent.run-name L1-c19 --agent.resume True --agent.load-run 2026-11-10_15-45-18_L1-c18 --agent.load-checkpoint model_3582.pt
- Downloads last month
- 31