EEmbodied AI Hub
Resource updatesFeaturedEditor’s pick

LeRobot v0.6.0: Leaner Installs and Refactored RL Stack

Why it mattersDefault installation is now leaner, the built-in GR00T model is upgraded, and the RL stack is fully refactored into swappable components.

This update makes LeRobot leaner and much easier to tinker with, as the default installation no longer bundles heavy training dependencies. The biggest win is the fully refactored reinforcement learning stack, which is now built like swappable Lego blocks so you can swap out a single network without breaking the rest. Plus, the default GR00T embodied AI model gets a smarter N1.7 upgrade with better action generalization.

What changed

  • Leaner installs: pip install now only grabs core features; training and dataset extras must be explicitly requested to avoid bloat and conflicts.
  • Modular RL stack: The reinforcement learning code is fully decoupled into swappable blocks, letting you replace specific networks without touching the rest.
  • Smarter default model: The built-in GR00T embodied AI model upgrades from N1.5 to N1.7, offering better action generalization for robots.
  • Higher PyTorch requirement: Minimum PyTorch version bumped to 2.7 to leverage faster CUDA kernels and improved autograd.
  • Cleaner imports: Internal paths are cleaned up; users should stick to stable public entry points to ensure future compatibility.
changelogLeRobotreleaseprojectv0.6.0GitHubupdate

Source: GitHub Releases · July 6, 2026

Share this article so more people can see it

What changed

  • Leaner installs: pip install now only grabs core features; training and dataset extras must be explicitly requested to avoid bloat and conflicts.
  • Modular RL stack: The reinforcement learning code is fully decoupled into swappable blocks, letting you replace specific networks without touching the rest.
  • Smarter default model: The built-in GR00T embodied AI model upgrades from N1.5 to N1.7, offering better action generalization for robots.
  • Higher PyTorch requirement: Minimum PyTorch version bumped to 2.7 to leverage faster CUDA kernels and improved autograd.
  • Cleaner imports: Internal paths are cleaned up; users should stick to stable public entry points to ensure future compatibility.

Related resources on this hub

Jump to projects, models, or datasets mentioned or closely related.

Discussion

Tell us what you think — comments make stories more useful for builders and founders.

Tell us what you think!

LeRobot v0.6.0: Leaner Installs and Refactored RL Stack

Have an account? Log in to use your display name and avatar.

Email is optional and never shown on the page.

← All insights

More insights

Resource updatesFeaturedGitHub Releases

dm_control 1.0.44: Physics Engine Upgrade, Adhesive Contacts & Conveyor Effects

dm_control 1.0.44 updates the physics engine MuJoCo to 3.11.0 and introduces several new features. You can now simulate adhesive contacts (contacts that pull like magnets) and conveyor/treadmill effects via surface velocity. For rotational joints (e.g., ball joints), control signals now handle rotations beyond 180 degrees more smoothly. Additionally, preparations for future MIMO actuators split actuator counts into inputs, objects, and outputs.

  • Update to MuJoCo 3.11.0, requiring this version
  • Add SO3 transmission and native orientation actuator for rotational control like ball joints
  • Add geom adhesion: contacts that pull via translated friction cones
Read
Resource updatesFeaturedGitHub Releases

Pink v4.3.0: Fix braking-distance calculation for continuous joints

Pink is a Python library for robot kinematics. This release fixes a critical bug where the braking-distance term in AccelerationLimit was incorrect for joints without configuration limits, such as continuous joints. It also updates the list of revolute-joint shortnames supported by ManipulabilityTask, and adds support for Python 3.13 and 3.14. New examples and dependencies are included.

  • Fixed braking-distance calculation in AccelerationLimit for continuous joints
  • Updated the list of revolute-joint shortnames supported by ManipulabilityTask
  • Added support for Python 3.13 and 3.14
Read
Resource updatesFeaturedGitHub Releases

OmniGibson v3.9.1: More Stable Model Downloads and Performance Tuning

OmniGibson is a 3D environment platform for robot simulation and interaction. v3.9.1 improves model downloading from Hugging Face for more stable resource access; adds PyTorch thread count configuration for performance tuning; fixes the R1Pro robot base velocity frame issue; and updates demo documentation.

  • Updated Hugging Face model download method for improved stability
  • Added constant options for PyTorch thread count to control parallel computing resources
  • Edited sponsor information in index.md
Read