# AI use cases for Lerobot (huggingface/lerobot)

Repo: [huggingface/lerobot](https://github.com/huggingface/lerobot) | Use cases: 3 | Page: https://dedupedai.com/use-cases/huggingface-lerobot/

LeRobot is Hugging Face's PyTorch library of models, datasets and tools for real-world robotics. It gives a hardware-agnostic Robot interface from low-cost arms like the SO-100 and SO-101 up to humanoids, the LeRobotDataset format (Parquet plus MP4) hosted on the Hugging Face Hub, and state-of-the-art policies for imitation learning, reinforcement learning and vision-language-action models, with CLI commands to record, train, evaluate and roll out policies.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Train a robot policy on a Hugging Face dataset: https://dedupedai.com/use-cases/huggingface-lerobot/train-a-robot-policy-on-a-hub-dataset/
- Benchmark a vision-language-action policy in simulation: https://dedupedai.com/use-cases/huggingface-lerobot/evaluate-a-vla-policy-on-libero-with-lerobot/
- Load and explore a robot dataset from the Hub: https://dedupedai.com/use-cases/huggingface-lerobot/load-and-explore-a-lerobot-dataset/

## How to train a robot policy on a Hugging Face dataset with Lerobot

Repo: [huggingface/lerobot](https://github.com/huggingface/lerobot) | Difficulty: Intermediate | Time: half a day | Basis: From the README

Train an ACT policy on a public LeRobot dataset with one lerobot-train command, no robot needed.

**Who it is for:** A machine learning engineer curious about robot learning who does not own hardware yet.

**The problem:** Robotics research code is often tied to one lab's robot and data format.

**The outcome:** A trained ACT policy checkpoint from the lerobot/aloha_mobile_cabinet dataset.

### Steps

1. Install with `pip install lerobot` and check the setup with lerobot-info
2. Add training dependencies with the 'Lerobot[training]' extra
3. Train with `lerobot-train` using `--policy.type=act`, `--dataset.repo_id=lerobot/aloha_mobile_cabinet` and `--policy.push_to_hub=false`
4. Try other policies from the README table, such as Diffusion or SmolVLA, by changing policy.type
5. Check the Compute Hardware Guide for GPU, RAM and training time per policy

### Prompt to paste into your agent

```text
Using huggingface/lerobot, install the training extra and run lerobot-train with --policy.type=act on --dataset.repo_id=lerobot/aloha_mobile_cabinet with --policy.push_to_hub=false. Report training loss and where the checkpoint was saved.
```

### Caveat

This card is written from the huggingface/lerobot README, not from a test run by Deduped. Before you start, make sure you have Python and pip. Check the repo for newer instructions before you rely on these steps.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/huggingface-lerobot/train-a-robot-policy-on-a-hub-dataset/

## How to benchmark a vision-language-action policy in simulation with Lerobot

Repo: [huggingface/lerobot](https://github.com/huggingface/lerobot) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README

Evaluate a pretrained Pi0 policy on the LIBERO benchmark with lerobot-eval.

**Who it is for:** A researcher comparing VLA models before trying one on a real arm.

**The problem:** Testing policies on real robots is slow and risky without a simulation baseline.

**The outcome:** Success results over 10 LIBERO episodes for a published policy.

### Steps

1. Install with `pip install lerobot`
2. Run `lerobot-eval` with `--policy.path=lerobot/pi0_libero_finetuned`, `--env.type=libero`, `--env.task=libero_object` and `--eval.n_episodes=10`
3. Swap in another policy path or benchmark such as MetaWorld to compare
4. Use lerobot-rollout, per the Policy Deployment Documentation, once you want to run a trained policy on a real robot
5. Publish your own environment through EnvHub if you build a new benchmark

### Prompt to paste into your agent

```text
With huggingface/lerobot, run lerobot-eval on --policy.path=lerobot/pi0_libero_finetuned with --env.type=libero --env.task=libero_object --eval.n_episodes=10, then summarise the success rate.
```

### Caveat

Simulation success does not guarantee the same behaviour on a physical robot.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/huggingface-lerobot/evaluate-a-vla-policy-on-libero-with-lerobot/

## How to load and explore a robot dataset from the Hub with Lerobot

Repo: [huggingface/lerobot](https://github.com/huggingface/lerobot) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README

Load a LeRobotDataset in Python and inspect synchronized video and action data episode by episode.

**Who it is for:** A student or researcher who wants real robot demonstration data to study or reuse.

**The problem:** Robot datasets come in incompatible formats and are hard to stream and inspect.

**The outcome:** A dataset loaded from the Hub with video decoding handled for you, ready to edit, split or merge.

### Steps

1. Install the dataset extra with `pip install 'lerobot[dataset]'`
2. Load a dataset in Python with LeRobotDataset("lerobot/aloha_mobile_cabinet"), as in the README
3. Index an episode and print its action shape; video decoding is automatic
4. Use the dataset tools to delete episodes, split by indices or fractions, add or remove features, or merge datasets
5. Browse thousands more datasets on the Hugging Face Hub under Lerobot

### Prompt to paste into your agent

```text
Using huggingface/lerobot, load the dataset lerobot/aloha_mobile_cabinet with LeRobotDataset, print the features and the action shape of episode 0, and split it into train and validation by fraction.
```

### Caveat

This card is written from the huggingface/lerobot README, not from a test run by Deduped. Before you start, make sure you have Python and pip. Check the repo for newer instructions before you rely on these steps.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/huggingface-lerobot/load-and-explore-a-lerobot-dataset/