weewooo
Reward sprint speed beyond target
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> weerawhin/microduck-weewooo
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-Sprint2m-MicroDuck - repo:
https://github.com/NIRUN-Weerawit/microduck-challenges - branch:
experiment/aten - commit:
8fa205b11 - checkpoint:
2000 - seed:
1 - base:
mjlab-microduck 0.1.0 @ 8d0dbc94c - started:
2026-10-14T11:10:45Z
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/NIRUN-Weerawit/microduck-challenges
cd microduck-challenges
git checkout 8fa205b11
uv sync
uv run train Mjlab-Sprint2m-MicroDuck --env.scene.num-envs 8196 --agent.max-iterations 3000 --agent.seed 1 --agent.logger tensorboard --agent.run-name run-1014-2010
- Downloads last month
- 13