# allenzren/open-pi-zero

Re-implementation of pi0 vision-language-action (VLA) model from Physical Intelligence

Repository: https://github.com/allenzren/open-pi-zero
Canonical: https://ross.abutalabs.com/products/open-pi-zero
Language: Python
License: MIT
License Family: permissive
Last push: 2025-01-31T03:22:27+00:00

## Health v2 (maintenance only)
Score: 25/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 4, release rhythm 40, longevity 45
- inputs: {"age_days": 638, "days_push": 579, "days_rel": 583, "gap_med": 22, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1518, forks 108 (observed 2026-08-28T04:04:57.504862+00:00)

## What it is
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
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-training, robotics
- domain: machine-learning, robotics, deep-learning, artificial-intelligence
- platform: python
- tags: vision-language-action, vla, pi0, flow-matching, paligemma, robot-manipulation, mixture-of-transformers, openx, linux, gpu

## Member repositories
- allenzren/open-pi-zero (main) score 25

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:57.504862+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:32:00.433571+00:00, confidence not recorded.
  - readme: https://github.com/allenzren/open-pi-zero (fetched 2026-08-28T04:04:57.504862+00:00, sha 778bf12e56b7)
- Data as of 2026-08-30T08:39:29.467469+00:00.