IDM-OMX-v3-50K

NVIDIA GR00T IDM (Inverse Dynamics Model) fine-tuned for ROBOTIS OMX-F robot.

Model Details

• Architecture: SiGLIP2(768d) → VL Transformer(4L) → DiT(8L, 1024d)
• Parameters: 612M
• Embodiment: OMX (slot 30)
• Action Space: 6-DOF delta joint angles (radians), 30 FPS
• Action Horizon: 16 steps

Training

• Base Model: IDM-OMX-v2 (20K steps checkpoint)
• Dataset: 558 episodes, 257,466 frames (OMX combined + SO-101 megamix FK→IK converted)
• Steps: 50,000 (6.21 epochs)
• Batch Size: 16 (effective 32 with gradient accumulation)
• Learning Rate: 1e-4 (cosine schedule)
• Duration: 5h 08min on RTX 5090
• Final Train Loss: 0.2644

Evaluation

Metric	Baseline (10K)	v2 (20K)	v3 (50K)
Normalized MAE	0.6092	0.5582	0.3551 (-36.4%)
Forward Loss	0.3942	0.3453	0.2517 (-27.1%)

Per-Joint MAE (degrees)

Joint	v2	v3	Change
shoulder_pan	82.12	71.76	-12.6%
shoulder_lift	83.01	72.74	-12.4%
elbow_flex	65.65	30.12	-54.1%
wrist_flex	141.40	96.43	-31.8%
wrist_roll	67.87	56.43	-16.9%
gripper	6.50	4.21	-35.2%

Usage

Pseudo-label Extraction

Citation

Based on NVIDIA GR00T IDM.