Series

Unifolm Vla G1 Series

The "Unifolm Vla G1 Series" series has 5 parts — read them in order from part 1.

humanoidvlawhole-bodyPart 1

unifolm-vla + Unitree G1 (Post 1): WBC+VLA system architecture — from data to real hardware

Overview of the 3-repo architecture — unifolm-vla (VLA), xr_teleoperate (data collection), unitree_rl_gym (locomotion) — and how they connect into a whole-body control pipeline for Unitree G1 on real hardware.

5/31/20266 min read

Tutorial

unifolm-vla + Unitree G1 (Post 2): collecting data with xr_teleoperate + Meta Quest 3

humanoidvladata-collectionPart 2

humanoid

unifolm-vla + Unitree G1 (Post 2): collecting data with xr_teleoperate + Meta Quest 3

Step-by-step guide to setting up xr_teleoperate with Meta Quest 3, connecting G1, recording teleoperation demos, and collecting high-quality data for VLA training on real hardware.

6/2/20267 min read

Tutorial

unifolm-vla + Unitree G1 (Post 3): data pipeline — JSON → LeRobot → HDF5 → RLDS

humanoidvladata-pipelinePart 3

humanoid

unifolm-vla + Unitree G1 (Post 3): data pipeline — JSON → LeRobot → HDF5 → RLDS

3-step guide to converting teleoperation data: from JSON (xr_teleoperate) to LeRobot V2.1, to HDF5, to RLDS format ready for unifolm-vla training — with dataset verification at each step.

6/4/20265 min read

Tutorial

unifolm-vla + Unitree G1 (Post 4): fine-tuning from Qwen2.5-VL-7B — 8-GPU and single-GPU LoRA

humanoidvlafine-tuningPart 4

humanoid

unifolm-vla + Unitree G1 (Post 4): fine-tuning from Qwen2.5-VL-7B — 8-GPU and single-GPU LoRA

Fine-tune unifolm-vla from the public Qwen2.5-VL-7B-Instruct checkpoint (since Unifolm-VLM-0 is not public) — covering the official 8-GPU DeepSpeed approach and a single-GPU QLoRA workaround for beginners.

6/6/20267 min read

Tutorial

unifolm-vla + Unitree G1 (Post 5): deploy inference server, SSH tunnel, and parallel locomotion

humanoidvladeployPart 5

humanoid

unifolm-vla + Unitree G1 (Post 5): deploy inference server, SSH tunnel, and parallel locomotion

Final series post: start the FastAPI inference server, connect G1 via SSH tunnel, send action commands, run arm VLA and locomotion simultaneously — with safety checklist and debug guide for common real-hardware failures.

6/7/20267 min read