smolvla 开源项目推荐（88 个）

datawhalechina /every-embodied

仅需Python基础，从0构建自己的具身智能机器人；从0逐步构建VLA/OpenVLA/SmolVLA/Pi0，深入理解具身智能

FastCrest /tether

FastCrest Tether：开源软件的边缘到云AI部署CLI。优化、验证、部署到Jetson、RTX、Apple Silicon、AMD上。混合边缘-云推理与奇偶校验证书。

JereoZero /so101-real

实机部署so101流程和踩坑全记录，使用lerobot框架so101真机。在本地 ubuntu22+5070显卡 搭建环境，部署，录数据，训练smolvla，推理全流程。

nik658 /UR10e_pi0_lerobot

This repository implements pi0, pi0fast and smolvla models on sim and real robots using Lerobot interface

alexantaluo0 /DeepScholar

Automated agentic research loop for vision-language-action (VLA) robot policy self-improvement

Sa74ll /smolvla-so100-finetune

该项目在 GitHub 上没有填写描述

noteflowai /robot-reel

Review Physical AI with original evidence: compare VLM explanations against recorded robot outcomes, replay SmolVLA failures, and inspect Newton/Genesis simulations offline.

Hashtag-Robotics /so101-tic-tac-toe

End-to-end SO-101 tic-tac-toe: printable hardware, local UI, LeRobot dataset, SmolVLA policy and guarded Strands agent

provael /provael

Provael — red-team open Vision-Language-Action (VLA) robot policies in simulation and report an Attack Success Rate (ASR). Prove it. Prevail.

vishal-naveen /TactileVLA-Edge

Vision-language-action stack for contact-rich robot manipulation on sub-$300 hardware. SO-101 + LeRobot; ACT 18/20 and SmolVLA 16/20, both generalizing to a held-out workspace cell.

BlackMirean /FlowVLA-RL

Online GRPO for SmolVLA's flow-matching action expert on LIBERO: marginal-preserving SDE retrofit, locked eval protocol, seed-replicated +6.3pp on LIBERO-Object, failure-boundary analysis. Fits in 24 GB VRAM.

2437buaa /smolvla-libero-repro

在 RTX 3060 6GB 上复现 SmolVLA × LIBERO，并开展低显存 LoRA 微调、多随机种子评测与遗忘控制实验。Low-VRAM SmolVLA × LIBERO reproduction with LoRA adaptation, fixed multi-seed evaluation, and forgetting controls.

ajmcastro /langgraph-vla-agent

Agentic planning for Vision-Language-Action robotics — orchestrating SmolVLA/LeRobot policies with LangGraph for multi-step robot manipulation.

dariusss04 /so101-vla-benchmarking

Manual real‑world benchmarking of ACT/SmolVLA on a physical SO101 robot using LeRobot.

BugLoverNaN /vla-portfolio

Reproducing SmolVLA-450M on LIBERO benchmark with cross-embodiment experiments | LeRobot

paumontagut /per-group-mse-vla

Per-Group Error, Not Total MSE. Fine-tuning VLAs on the Toyota HSR. ICRA 2026 Workshop From Data to Decisions, 1st place out of 36.

Mr-Moody /SmolVLA-Testing

Unified pipeline for collecting, labeling, converting, and training robot demonstration data with SmolVLA, Pi0, Pi0.5, or ACT policies - extended with Qwen3-VL offline phase annotation, a runtime FSM, and three phase-conditioned policy variants.

ahmedsleem109 /WorkshopRobot

Language-commanded mobile manipulation in MuJoCo: a Go2 quadruped with a Z1 arm fetches a named tool on one 6 GB GPU. 125/125 grasps, 86% end-to-end, and a VLA negative result diagnosed to 18/20.

zhaoyufei122 /piper-smolvla

Real-world robot learning with AgileX Piper: SpaceMouse demonstrations, SmolVLA fine-tuning, and wrist-camera policy execution.

ShreyaVijaykumar /Edgebot-Reflex

A heterogeneous dual-model robot navigation system pairing Falcon Mamba (SSM) as a fast reflex layer with SmolVLA (VLA) as a deliberate planning layer, arbitrated by a priority controller, fully simulatable in Wokwi on ESP32. as a part of learning project.