slvDev
Featured
esp32-ai
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
robot_futuring_rl
robot_futuring_rl is an open-source full-stack system for closed-loop real-robot learning, built on OpenPI's π-series models and adding VLA fine-tuning, tri-state progress annotation, two-stage RL Token training, PICO teleoperation, and human-in-the-loop correction data collection.
Stars
28
7-day growth
No data
Forks
4
Open issues
0
License
Apache-2.0
Last updated
2026-07-31
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It provides a complete, modular closed-loop workflow from robot data collection and human correction labeling to advantage-weighted training and policy deployment, with independent components for models, trainers, annotation, ROS2 teleoperation, and robot protocol adaptation.
Who it is for
- Robotics researchers and engineers working on VLA training and real-robot deployment
- Imitation learning and reinforcement learning practitioners exploring progress-based reward signals
- Developers building human-in-the-loop data collection and teleoperation pipelines
- Teams integrating new robot arms or mobile manipulators into learning systems
Use cases
- Fine-tuning or deploying π₀ / π₀-FAST / π₀.₅ VLA policies
- Annotating robot episodes with -1/0/1/2 progress states for reward generation
- Collecting human correction data through the HIL PICO WebUI and LeRobot v2.1 format
- Training and serving RL Token Stage 2 actors with replay and TD3+BC
Strengths
- Modular design: model training, labeler, teleoperation bridge, and robot protocol adapters can be used independently
- YAML-based robot configuration decouples state/action dimensions, ROS topics, cameras, and inference parameters from core code
- Tri-state labeler with ready-data validation rules and fixed reward mapping
- RL Token Stage 2 cache/replay pipeline binds configs, labels, and checkpoint hashes to prevent inconsistent reuse
Considerations
- The repository does not distribute training data, model checkpoints, or XRoboToolkit SDK/PC Service binaries
- Real-robot PICO and HIL operation depends on ROS2, robot message packages, camera topics, mode services, and robot-side safety controllers
- Configuration-level soft limits are not a substitute for a robot safety controller
README quick start
安装
Related repositories
Similar projects matched by category, topics, and programming language.
gavamedia
Featured
deltafin
Deltafin is a research project that runs the 2.8-trillion-parameter Mixture-of-Experts model Kimi K3 on a single Apple Silicon Mac (e.g., M1 Max with 64 GB) at about 16 seconds per token, using exact, reproducible inference with local or streaming expert loading.
jamesob
Featured
local-llm
A comprehensive guide for building and configuring a high-end local machine to run state-of-the-art LLMs, with detailed hardware choices, BIOS tuning, and Docker-based model serving.