mcMotionChallengeGitHub ↗

MODEL / CHECKPOINT

Galbot / Tsinghua & collaborators

Model

The original 216 Transformer checkpoint with sixteen chronological observation frames and the author reference preprocessing.

Organization
Galbot / Tsinghua & collaborators
Type
Dense
Continuous-action Transformer conditioned on sixteen136D observations; this is a separate checkpoint from the264 MLP.
Version
Transformer 216 · 82313c76
author-released Transformer checkpoint
Robot support
Unitree G1 · 29 DoF
Author source revision
9f9e7b74ecadb532abbb34b6a779d87191a9bbb6

Checkpoint artifacts

Author release
GalaxyGeneralRobotics/Humanoid-GPT ↗
Artifact type
Author-released checkpoint
Original file
pns_wo_priv216_trans.onnx ↗

Original author Transformer 216 ONNX without conversion or weight changes. SHA256 and author Git blob 82ed6796c50b53d3fc260741dbb676b09ac0084f verified.

pns_wo_priv216_trans.onnx uses 16 chronological frames of 136 observations, as required by the released model input shape.

Download links deliver the adapter files. The release links and verification notes distinguish author files from conversions; SHA256 identifies the bytes used.

Policy checkpoint
Download ↗
82313c76869ca796395abf80ee4288cced424d25087e74e32f9ec739ff59c853

Inference interface

The adapter declares its control clock and reference window below. Source and configuration describe the observation and action transforms.

Observation requirements

Observation rules ↗

Model-specific handling

Author infer_onnx_real hardware route: fixed synthetic root [0,0,0.78], author reference FK, EMA and 16-frame history. No measured robot position or position integration.

Exact checkpoint & evidence ↗
Control interval
20 ms
Reference history
0 ms
Reference preview
20 ms
Output interface
joint-pd-feedforward
Configuration
policy.json

Evaluations