arunos728/gr00t-n16-rby1-assembletissue-chunk30-2gpu-b64-30k

GR00T N1.6 — RB-Y1 assemble-tissue

RLWRLD RB-Y1 bimanual platform finetune of GR00T N1.6 3B. 30,000 steps on 2x H100, 3h 58m.

Training setup

base model	nvidia/GR00T-N1.6-3B
embodiment tag	new_embodiment
steps	30,000
GPUs	2x H100 80GB
global batch size	64 (32 per GPU)
learning rate	1e-4, cosine, warmup ratio 0.05
weight decay	1e-5
action chunk	30 (1.0 s at 30 fps)
action representation	absolute joint targets

Robot

RB-Y1 + Wuji hand (rby1m_wujihand2). 54-D absolute joint targets, grouped as:

right_arm_joints  [0:7]     right_hand_joints [7:27]
left_arm_joints  [27:34]    left_hand_joints [34:54]
action uses the same four groups as *_command_joints

No head joints, and the hands are 20-D each. State and action group identically here -- worth stating because the sibling OpenArm/RH56F1 checkpoints do not (there the action side splits head from arm while the state side folds them together). Do not carry that caveat across.

Cameras are a stereo ego pair, ego_left / ego_right, 320x180.

Data

35,613 frames / 50 episodes at 30 fps, single task "Put the tissue roll onto the holder." Internal dataset, not public.

A companion run of the same task on the OpenArm/RH56F1 platform (28-D) exists with the same chunk, batch and step count, so the two robots can be read against each other.

54-D needs no special handling on the GR00T side: the base model's max_state_dim and max_action_dim are both 128.

What is in here

The final export only: model-0000{1,2}-of-00002.safetensors, config.json, processor/, experiment_cfg/. Optimizer state and the 10000 / 20000 checkpoints are not included.

Evaluation notes

• Do not reimplement the preprocessing. letterbox -> SmallestMaxSize(256) -> center crop(0.95) -> SmallestMaxSize(256) is baked into the bundled processor. processor.eval() reproduces training; hand-rolling it is how eval silently diverges.
• Camera order is ego_left then ego_right.
• Feed the native 320x180. Pre-resizing to 256x256 gets letterboxed a second time.
• Use the bundled processor/statistics.json for normalization.
• Do not request chunks longer than 30.