wee-climb
First run
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-wee-climb
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-HillClimb1m5-MicroDuck - repo:
https://github.com/NIRUN-Weerawit/microduck-challenges - branch:
experiment/aten - commit:
079ed7298 - checkpoint:
2250 - seed:
42 - base:
mjlab-microduck 0.1.0 @ 8d0dbc94c - started:
2026-10-14T06:05:06Z
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 079ed7298
uv sync
uv run train Mjlab-HillClimb1m5-MicroDuck --env.scene.num-envs 2048 --agent.max-iterations 3000 --agent.seed 42 --agent.logger tensorboard --agent.run-name hill-progress-seed42
- Downloads last month
- 23