bat-cha/microduck-batduck

batduck

Mjlab-Sprint2m-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> bat-cha/microduck-batduck

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).

A remix of HannesVonEssen/microduck-running.

Training

• task_id: Mjlab-Sprint2m-MicroDuck
• repo: https://github.com/bat-cha/microduck-challenges.git
• branch: sprint-race
• commit: 033ca0ef7
• checkpoint: 500
• seed: 1
• base: mjlab-microduck 0.1.0 @ 8d0dbc94c
• started: 2026-10-21T20:49:31Z

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/bat-cha/microduck-challenges.git
cd microduck-challenges
git checkout 033ca0ef7
uv sync
uv run train Mjlab-Sprint2m-MicroDuck --env.scene.num-envs 4096 --agent.max-iterations 900 --agent.seed 1 --agent.logger tensorboard --agent.run-name race-ws-arena --agent.resume True --agent.load-run warmstart_hannes --agent.load-checkpoint model_0.pt --agent.save-interval 100