Skip to main content
LeRobot provides a comprehensive set of examples to help you get started with different features and use cases. These examples demonstrate best practices and common workflows.

​
Dataset Examples

Learn how to work with LeRobot datasets.

​
Loading Datasets

The load_lerobot_dataset.py example demonstrates:
  • Viewing dataset metadata and properties
  • Loading datasets from the Hugging Face Hub
  • Accessing frames by episode number
  • Using timestamp-based frame selection
  • Batch processing with PyTorch DataLoader
from lerobot.datasets.lerobot_dataset import LeRobotDataset

# Load a dataset from the Hub
dataset = LeRobotDataset("lerobot/aloha_mobile_cabinet")

# Access data (automatically handles video decoding)
episode_index = 0
print(f"{dataset[episode_index]['action'].shape=}")
Example location: examples/dataset/load_lerobot_dataset.py

​
Dataset Transformations

The use_dataset_image_transforms.py example shows:
  • Applying image transformations to dataset frames
  • Custom preprocessing pipelines
  • Integration with PyTorch transforms
Example location: examples/dataset/use_dataset_image_transforms.py

​
Dataset Tools

The use_dataset_tools.py example covers:
  • Deleting episodes
  • Splitting datasets by indices/fractions
  • Adding/removing features
  • Merging multiple datasets
Example location: examples/dataset/use_dataset_tools.py

​
Training Examples

​
Training Policies

The train_policy.py example demonstrates how to train a Diffusion Policy on the PushT environment: 
from lerobot.policies.diffusion.configuration_diffusion import DiffusionConfig
from lerobot.policies.diffusion.modeling_diffusion import DiffusionPolicy
from lerobot.datasets.lerobot_dataset import LeRobotDataset

# Load dataset
dataset = LeRobotDataset("lerobot/pusht")

# Create and configure policy
config = DiffusionConfig()
policy = DiffusionPolicy(config, dataset.meta)

# Train the policy
# See full example for training loop details
Example location: examples/training/train_policy.py

​
Streaming Training

The train_with_streaming.py example shows:
  • Training with streaming datasets for large-scale data
  • Memory-efficient loading
  • Distributed training setup
Example location: examples/training/train_with_streaming.py

​
Tutorial Examples

Comprehensive tutorials organized by policy type:

​
ACT Policy

  • Training: examples/tutorial/act/act_training_example.py
  • Inference: examples/tutorial/act/act_using_example.py

​
Diffusion Policy

  • Training: examples/tutorial/diffusion/diffusion_training_example.py
  • Inference: examples/tutorial/diffusion/diffusion_using_example.py

​
VLA Models

  • Pi0: examples/tutorial/pi0/using_pi0_example.py
  • SmolVLA: examples/tutorial/smolvla/using_smolvla_example.py

​
Reinforcement Learning

  • HIL-SERL: examples/tutorial/rl/hilserl_example.py
  • Reward Classifier: examples/tutorial/rl/reward_classifier_example.py

​
Async Inference

  • Policy Server: examples/tutorial/async-inf/policy_server.py
  • Robot Client: examples/tutorial/async-inf/robot_client.py

​
Real-World Use Cases

Mobile Manipulation

Train mobile manipulators using the ALOHA dataset for tasks like cabinet opening and object manipulation.

Bimanual Control

Use ACT policy for bimanual tasks requiring coordinated control of two robot arms.

Vision-Language Tasks

Leverage VLA models like Pi0 or SmolVLA for language-conditioned robotic tasks.

Simulation to Real

Train policies in simulation (PushT, LIBERO) and transfer to real robots.

​
Additional Examples

​
Backward Compatibility

Replay episodes recorded with previous calibration systems: Example location: examples/backward_compatibility/replay.py See the Backward Compatibility guide for migration details.

​
Dataset Porting

Convert datasets from other formats to LeRobotDataset: Example location: examples/port_datasets/

​
Robot-Specific Examples

  • LeKiwi: examples/lekiwi/
  • SO100 to SO100_EE: examples/so100_to_so100_EE/
  • Phone Teleoperation: examples/phone_to_so100/
  • Real-Time Control: examples/rtc/

​
Running Examples

All examples can be run directly from the command line: 
# Navigate to the lerobot directory
cd lerobot

# Run an example
python examples/dataset/load_lerobot_dataset.py
Most examples include configurable parameters. Check the script headers for usage instructions.

​
Community Examples

Explore community-contributed examples and projects:

Build docs developers (and LLMs) love

Get started for freeTalk to us