Train AI Robots Without Writing Code! (Introducing LeLab)

2026-06-03 · Source: HuggingFace · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Novice, short

Summary

Leloup is a new graphical user interface for the Robot Library, designed to simplify the process of training AI robots without requiring code. It allows users to teleoperate robots, configure hardware including calibration and camera integration, and collect high-quality datasets. The platform supports training models both on local machines and leveraging powerful GPUs via Hugging Face Jobs, offering options for policy selection, training steps, and batch size, with parallel training capabilities. Users can also test models on robots while training is ongoing, observing performance improvements from early checkpoints (e.g., 3,000 steps) to final policies (e.g., 30,000 steps). A typical task, like putting a pen into a holder, involved recording 50 episodes in about 15 minutes, emphasizing smooth and consistent movements for optimal model performance.

Key takeaway

For AI Engineers or Robotics Students aiming to develop robot policies without extensive coding, Leloup offers a streamlined workflow. You can quickly configure robots, collect high-quality datasets by recording 30-50 smooth episodes, and train models efficiently using local hardware or Hugging Face GPUs. This approach allows you to rapidly iterate on robot behaviors, testing model improvements from early checkpoints to final deployment, significantly accelerating development cycles for robotic tasks.

Key insights

Leloup simplifies robot AI training via a GUI, enabling teleoperation, data collection, and model deployment without coding.

Principles

• High-quality data is crucial for robot model performance.
• Consistent, smooth movements improve dataset quality.
• Iterative testing during training shows model improvement.

Method

Configure robots, teleoperate to practice tasks, record 30-50 smooth episodes for a dataset, train models locally or on Hugging Face GPUs, then test and deploy policies on robots.

In practice

• Record 50 episodes for a task in about 15 minutes.
• Use Hugging Face Jobs for powerful GPU training.
• Test model checkpoints on robots during training.

Topics

• Robot Library
• Leloup GUI
• AI Robot Training
• Data Collection
• Hugging Face Jobs
• Policy Deployment

Best for: Robotics Engineer, AI Engineer, AI Student

Related on AIssential

• Edge-to-Cloud Robotics with AMD ROCm: From Data Collection to Real-Time Inference — AMD's edge-to-cloud robotics solution integrates LeRobot, VLA models, and Imitation Learning on AMD hardware for efficient robot training and deployment.
• Everyone Can Build a Robot: Open Source Embodied AI With Seeed Studio | NVIDIA AI Podcast Ep. 300 — Robot training has evolved from complex programming to intuitive, physical guidance, with cloud-based training and JSON deployment.
• An Embodied Simulation Platform, Benchmark, and Data-Efficient Augmentation Framework for Wet-Lab Robotics — Pipette offers a simulation platform and data augmentation to significantly improve wet-lab robot learning efficiency and accessibility.
• Open-Source Software Is Starting to Help Robots Think — Open-source AI is rapidly lowering barriers to advanced robotics development, accelerating innovation and accessibility.
• We Hired 10 Humans to Train a Robot to Fold Shirts — Robots can learn complex manipulation tasks like t-shirt folding through teleoperated data collection.
• How Kodiak trains the brain behind 28 driverless trucks — Kodiak's GigaFusionNet uses a multi-stage training pipeline and an AI Flywheel for continuous improvement in autonomous trucking.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by HuggingFace.