File size: 1,488 Bytes
55e9d6b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 | ---
license: apache-2.0
tags:
- robotics
- inverse-dynamics-model
- idm
- omx
- gr00t
---
# 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](https://github.com/NVIDIA-Omniverse/GR00T-Dreams).
|