Downloads · 30 days
12
28% of all-time downloads
RooibosT/gr00t-n1.7-g1-dex1-bct-relarm-aug-30hz-h40
gr00t-n1.7-g1-dex1-bct-relarm-aug-30hz-h40 is a robotics model from RooibosT. Use it for the robotics task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Action-expert fine-tune of nvidia/GR00T-N1.7-3B for a Unitree G1 with Dex1 grippers assembling a children's table. Selected by open-loop scan over 6 checkpoints, not by evalloss (see Checkpoint selection below).
Downloads · 30 days
12
28% of all-time downloads
All-time downloads
43
Public
Parameters
3.1B
12.6 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors12.6 GB · 100%
From the Hugging Face model README
Action-expert fine-tune of nvidia/GR00T-N1.7-3B for a Unitree G1 with Dex1 grippers
assembling a children's table. Selected by open-loop scan over 6 checkpoints, not by
eval_loss (see Checkpoint selection below).
| Modality | Keys | Notes |
|---|---|---|
| video (3) | cam_head, cam_left_wrist, cam_right_wrist | 4th head camera was ablated and dropped — it changed arm accuracy by 0.00% |
| state (46) | legs(12) waist(3) left_arm(7) right_arm(7) left_gripper(1) right_gripper(1) base_gravity(3) left_eef(6) right_eef(6) | see State conventions |
| language | annotation.human.task_description | 5 subtasks |
Chunk of 40 steps (1.33 s at 30 Hz) over 19 dims:
| Key | Dims | Representation |
|---|---|---|
waist | 3 | ABSOLUTE |
left_arm, right_arm | 7 + 7 | RELATIVE (delta from current state) |
left_gripper, right_gripper | 1 + 1 | ABSOLUTE |
Legs are observed but not commanded — locomotion stays with the whole-body controller.
base_gravity — gravity direction in the pelvis frame, from the IMU quaternion (w,x,y,z):
-[2(xz - wy), 2(yz + wx), 1 - 2(x² + y²)]. Yaw-invariant, unlike the raw quaternion
whose w/z components are ~all session heading (yaw std 109° vs roll 0.6°).left_eef / right_eef — wrist pose recomputed by forward kinematics from the arm joints
in the same row, not the recorded end-effector state (which lagged the joints by
10 frames / 333 ms). Convention: g1_body29_hand14.urdf, waist joints zeroed (torso frame),
wrist_yaw origin translated +0.05 m along local x, orientation as extrinsic xyz Euler.A mismatch in these conventions silently corrupts 15 of 46 input dims. state_dropout=0.2
makes the model robust to missing state, not to wrong state.
| Executed steps | Re-inference period | Arm MAE | Gripper MAE |
|---|---|---|---|
| 5 | 0.17 s | 1.20° | 0.080 |
| 8 | 0.27 s | 1.50° | 0.091 |
| 16 | 0.53 s | 2.25° | 0.122 |
| 40 (full chunk) | 1.33 s | 3.98° | 0.198 |
Error grows with horizon, so the full-chunk figure is not the deployment figure — execute the head of the chunk and re-infer. A hold-current-pose baseline gives 1.13° at h1 and 5.72° over the full chunk. For scale, the arms move 10.2° over a 40-step chunk.
| Data | 1,249 train clips / 500,038 frames @ 30 fps, teleop stalls removed, subtasks rebalanced |
| Trainable | action expert only (1.62B of 3.14B); Cosmos-Reason2 backbone frozen |
| Schedule | 15,000 steps, effective batch 192, cosine, lr 1e-4, warmup 0.05, wd 1e-5 |
| Regularization | state_dropout 0.2 |
| Hardware | 3x A100 80GB, DDP with bf16 gradient communication, 12h22m |
eval_loss is the flow-matching regression objective at random noise levels; across five
runs it rose after ~7.5k steps while open-loop action accuracy kept improving. It is not a
model-selection signal here. Checkpoints were instead scanned with a fixed-seed open-loop
evaluation on the held-out split.
A longer schedule was tried and rejected: 30,000 steps at effective batch 256 saturated by ~20k and finished slightly worse than this model (full-chunk MSE 0.0299 vs 0.0291, gripper 0.204 vs 0.198), so roughly 5-6 epochs is enough for this dataset.