| tags: | |
| - microduck | |
| - robotics | |
| - reinforcement-learning | |
| - onnx | |
| library_name: microduck | |
| pipeline_tag: robotics | |
| # backroll | |
| Mjlab-GotATalent-MicroDuck | |
| A **perpetual** policy for the [microduck](https://github.com/pollen-robotics/microduck) (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot. | |
| ## Run it on a robot | |
| ```bash | |
| sudo robotctl policy load <slot> dashi/microduck-backroll | |
| ``` | |
| 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-GotATalent-MicroDuck` | |
| - **repo**: `https://github.com/bxshi/microduck-challenges.git` | |
| - **branch**: `HEAD` | |
| - **commit**: `a15317f40` | |
| - **checkpoint**: `1750` | |
| - **seed**: `1` | |
| - **base**: `mjlab-microduck 0.1.0 @ 8d0dbc94c` | |
| - **started**: `2026-10-22T07:55:17Z` | |
| ## 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. | |
| ```bash | |
| git clone https://github.com/bxshi/microduck-challenges.git | |
| cd microduck-challenges | |
| git checkout a15317f40 | |
| uv sync | |
| uv run train Mjlab-GotATalent-MicroDuck --env.scene.num-envs 4096 --agent.max-iterations 2500 --agent.seed 1 --agent.logger tensorboard --agent.run-name roll12c --agent.resume True --agent.load-run 2026-10-22_07-14-05_roll12 --agent.load-checkpoint model_500.pt | |
| ``` | |