Views
No views yet
color_object task split.checkpoints/
color_object/
model-00001-of-00002.safetensors (4.7 GB)
model-00002-of-00002.safetensors (1.8 GB)
model.safetensors.index.json
config.json
processor/
experiment_cfg/
gr00t/ # model, data, training library
scripts/ # finetune launch scripts, data prep, eval
finetune_jobs/ # per-split SLURM scripts
conflict_panda_config.py # modality config used for training
launch_finetune_local.py # main finetune entrypoint
prepare_conflict_data.py # data conversion script
finetune.md # detailed fine-tuning guide| Parameter | Value |
|---|---|
| Base model | nvidia/GR00T-N1.7-3B |
| Steps | 10,000 |
| GPUs | 4x A100 |
| Global batch size | 32 |
| Learning rate | 1e-4 |
| Warmup ratio | 0.05 |
| Tuned modules | projector + diffusion head (LLM and visual encoder frozen) |
| Action chunking | 16 steps |
| Denoising steps | 4 (inference) |
1import torch
2from gr00t.model.gr00t_n1 import GR00TPolicy
3from gr00t.data.schema import EmbodimentTag
4from scripts.conflict_panda_config import MODALITY_CONFIG
5
6policy = GR00TPolicy.from_pretrained(
7 "checkpoints/color_object",
8 modality_config=MODALITY_CONFIG,
9 embodiment_tag=EmbodimentTag.NEW_EMBODIMENT,
10 denoising_steps=4,
11 torch_dtype=torch.bfloat16,
12)
13policy.eval().cuda()
14
15obs = {
16 "video.image": image_tensor, # (1, 1, H, W, 3) uint8
17 "video.wrist_image": wrist_tensor, # (1, 1, H, W, 3) uint8
18 "state.arm": arm_state, # (1, 1, 7) float32
19 "state.gripper": gripper_state, # (1, 1, 1) float32
20 "annotation.human.task_description": ["pick up the color-matching object"],
21}
22with torch.no_grad():
23 actions = policy.get_action(obs)
24# actions["action.arm"]: (1, 16, 7)
25# actions["action.gripper"]: (1, 16, 1)finetune.md for the complete fine-tuning guide.1@article{bjorck2025gr00t,
2 title = {GR00T N1: An Open Foundation Model for Generalist Humanoid Robots},
3 author = {Bjorck, Johan et al.},
4 journal = {arXiv preprint arXiv:2503.14734},
5 year = {2025}
6}