LeRobot v0.6.0: Imagine, Evaluate, Improve
LeRobot v0.6.0 introduces three new world model policies—VLA-JEPA, LingBot-VA, and FastWAM—for predicting future states in robots, optimized for inference with zero extra cost. The release includes six new simulation benchmarks, a deployment CLI with DAgger-style corrections, and cloud training on HF Jobs, while enhancing dataset support with depth, language annotations, and faster data loading. The codebase is leaner and feature-complete, offering lightweight installation and support for multi-GPU training and cloud deployment.
More from Huggingface
LFM2.5-VL-3B for Better and Faster Vision Capabilities for the Edge
Thinking of ACE? We Can Do It with Fewer Tokens
Meta is back with Muse Glimmer: local, agentic, multimodal, and open source
