---
title: "LeRobot v0.4.0: Supercharging OSS Robot Learning"
description: "LeRobot v0.4.0 expands the open-source robot-learning stack with scalable datasets, simulation benchmarks, processing tools, policies, and hardware integrations. LeRobotDataset v3.0 introduces chunked episodes, video streaming, unified Parquet metadata, improved loading and memory efficiency, plus utilities for editing and converting datasets. A modular Processor system separates policy and robot pipelines, while direct Accelerate integration simplifies multi-GPU training; the release also adds LIBERO, Meta-World, PI0, PI0.5, and GR00T N1.5 support. A plugin architecture enables third-party robots, cameras, and teleoperators to be installed without changing the core library, with Reachy 2 and phone teleoperation included, alongside a self-paced Robot Learning Course and hands-on tutorial."
---

# LeRobot v0.4.0: Supercharging OSS Robot Learning

[Hugging Face](https://yomu.fyi/company/hugging-face) · Steven Palma, Michel Aractingi, Pepijn Kooijmans, Caroline Pascal, Jade Choghari, Francesco Capuano, Adil Zouitine, Martino Russi, Thomas Wolf · Oct 24, 2025

**Type:** Announcement

## Summary

LeRobot v0.4.0 expands the open-source robot-learning stack with scalable datasets, simulation benchmarks, processing tools, policies, and hardware integrations. LeRobotDataset v3.0 introduces chunked episodes, video streaming, unified Parquet metadata, improved loading and memory efficiency, plus utilities for editing and converting datasets. A modular Processor system separates policy and robot pipelines, while direct Accelerate integration simplifies multi-GPU training; the release also adds LIBERO, Meta-World, PI0, PI0.5, and GR00T N1.5 support. A plugin architecture enables third-party robots, cameras, and teleoperators to be installed without changing the core library, with Reachy 2 and phone teleoperation included, alongside a self-paced Robot Learning Course and hands-on tutorial.

## Context

LeRobot is being expanded to support the demands of large-scale robot learning, including massive datasets, diverse simulation and evaluation environments, model-to-robot data transformations, distributed training, and easier integration of third-party hardware.

## Approach / What changed

The release overhauls dataset storage with chunked episodes and streaming, adds dataset editing tools and simulation integrations, introduces PolicyProcessorPipeline and RobotProcessorPipeline processing stages, integrates Accelerate for multi-GPU training, adds PI0, PI0.5, and GR00T N1.5 policies, and provides a Python-package plugin system for hardware.

## Takeaways

- LeRobotDataset v3.0 stores episode metadata in unified Parquet files and supports chunked episodes and video streaming for datasets at the OXE level, stated as over 400GB.
- The Processor system uses separate pipelines for batched policy tensors and individual robot observations or actions, handling transformations such as normalization, tokenization, device movement, and action unnormalization.
- The hardware plugin system lets third-party robots, cameras, and teleoperators be distributed as separate Python packages; the release adds Reachy 2 and phone-based teleoperation support.

**Tags:** [Data Pipelines](https://yomu.fyi/topic/data-pipelines), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Open Source](https://yomu.fyi/topic/open-source), [Performance](https://yomu.fyi/topic/performance), [Python](https://yomu.fyi/topic/python)

- Source: [Hugging Face](https://huggingface.co/blog/lerobot-release-v040)
- Source URL: https://huggingface.co/blog/lerobot-release-v040
- Ingested by Yomu: 2026-08-27T15:15:00.413Z

[Read original post](https://huggingface.co/blog/lerobot-release-v040)