allenzren/open-pi-zero

Health v2 · maintenance only

Adoption not part of the score

1518 stars · 108 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

An open-source re-implementation of the pi0 vision-language-action (VLA) model from Physical Intelligence, built on a pre-trained PaliGemma VLM with a separate action expert trained via flow matching. It includes training scripts, released checkpoints on Bridge and Fractal datasets, and evaluation support via a SimplerEnv fork.

Use cases

• reimplement the pi0 VLA model
• train a vision-language-action model for robot manipulation
• fine-tune PaliGemma with an action expert
• evaluate robot policies in SimplerEnv
• experiment with flow matching for action chunk generation
• run VLA checkpoints on bridge and fractal datasets

When to choose

• you want an open, readable implementation of pi0 for research
• you need to train or fine-tune a VLA model on OXE-style datasets like Bridge or Fractal
• you want to evaluate robot manipulation policies in simulation

When to avoid

• you need a production-supported robotics stack with official vendor support
• you need real-robot deployment tooling out of the box
• you want a fully trained generalist policy rather than a research codebase

Facets

library · maturity active

machine-learning deep-learning llm-training robotics machine-learning robotics deep-learning artificial-intelligence python vision-language-action vla pi0 flow-matching paligemma robot-manipulation mixture-of-transformers openx linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="allenzren/open-pi-zero")