Instructions to use Swenlon/microduck-spin with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Microduck
How to use Swenlon/microduck-spin with Microduck:
# Replace SLOT with the slot specified in the model card (walk, stand, sitstand, ground_pick, kick_left, kick_right, roulade). sudo robotctl policy load SLOT Swenlon/microduck-spin
- Notebooks
- Google Colab
- Kaggle
Commit History
4 days ago
-
Fix local inference timing and add reproducible spin configuration
Verified signatureDocument the required 20 ms actuator delay and add a local launcher and CPU validation results. Model weights and original replay videos remain unchanged.bf84d6c -
Add fast-spinning Microduck policy initialized from a walking base (20,000 updates)
Verified signaturePublish a fast-spinning Microduck policy trained for 20,000 updates, initialized from a standard walking base model. Include the ONNX policy with observation normalization, two original replay videos, model documentation, provenance, and export validation. The original training goal was goose-stepping, but fast-spinning behavior emerged instead. This model is shared as an experimental skill candidate. Export consistency has been verified; safe stopping and deployment on a physical robot remain unvalidated.90fdc99