Examples¶
Getting-started notebooks¶
New to Strands Robots? The examples/notebooks/
folder is a click-and-run series that runs end-to-end in simulation - no
hardware, no GPU, no Hugging Face credentials.
| Notebook | What it shows |
|---|---|
01_getting_started.ipynb |
Robot("so100"), run a policy, read joint state, create_policy(). |
02_record_and_stream.ipynb |
Record a LeRobotDataset, then stream it back with stream_dataset(). |
03_record_train_deploy.ipynb |
The full loop: record, train an ACT policy on CPU, export, and load it back. |
04_discover_lerobot.ipynb |
Discover the LeRobot API with use_lerobot: list robots, policies, teleoperators, cameras, and inspect any class. |
05_streaming_data_loop.ipynb |
The streaming data loop: record, render, stream back, train, and load in one notebook (the optional Storage Bucket sync needs LeRobot >= 0.6.1). |
uv pip install "strands-robots[sim-mujoco,lerobot]" jupyterlab
jupyter lab # open examples/notebooks/
Scripts¶
Browse examples/:
| File | What it does |
|---|---|
cosmos3_sim_rollout.py |
Cosmos 3 sim rollout: spawn SO-100, connect to Cosmos 3 server, run episodes, save LeRobot v3 recording. |
molmoact2_so101_pickplace.py |
SO-101 pick-and-place via LerobotLocalPolicy with norm_tag / image_keys / inference_action_mode. Requires hardware + GPU. |
mesh_acl_example.json5 |
Mesh ACL config: per-peer allow/deny rules for Zenoh mesh. |
git clone https://github.com/strands-labs/robots
cd robots
uv pip install -e ".[all]"
python examples/vla/cosmos3_sim_rollout.py # needs cosmos3-service + server on :8000
python examples/vla/molmoact2_so101_pickplace.py # requires hardware
# requires GPU
See also¶
- Quickstart - minimal starter.
- Cosmos3Policy - Cosmos 3 details.
- LerobotLocalPolicy - MolmoAct2 and other local models.