Downloads · 30 days
0
GR00T-N1.7-PegInsertion-v7-18D-H10-30K
HanLinqi/GR00T-N1.7-PegInsertion-v7-18D-H10-30K
GR00T-N1.7-PegInsertion-v7-18D-H10-30K is a robotics model from HanLinqi. Use it for the robotics task on the model card, and read the license before you ship it in a product.
Final no-force and force-token GR00T N1.7 checkpoints trained on the public ForceBenchmark PegInsertion v7 18D planning dataset.
Access
Public
Updated Sep 14, 2026
Repo size
13.9 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors13.9 GB · 100%
At a glance
- Task
- Robotics
- Access
- Public
- Created
- Sep 14, 2026
- Updated
- Sep 14, 2026
- SHA
- 4fcde78a
Base models
Model card
From the Hugging Face model README
GR00T N1.7 PegInsertion v7 — 18D planning
Final no-force and force-token GR00T N1.7 checkpoints trained on the public ForceBenchmark PegInsertion v7 18D planning dataset.
Checkpoints
noforce/checkpoint-30000/: RGB + 18D qpos/qvel, without force tokens.force/checkpoint-30000/: RGB + 18D qpos/qvel + causal 6D wrench history.
Both runs completed 30,000 optimizer steps. Each checkpoint includes the model shards, configuration, processor, statistics, and training metadata.
Training contract
- Dataset:
HanLinqi/ForceBenchmark,planning_data/forcebenchmarkpandapeginsertion_v7 - Dataset revision:
9bd7daa0354caaf526987c16b3c8fa54e9069dc5 - Dataset H5 SHA-256:
f96248dc9cc4dda55eda0799926922800329d9eb6a65c4f64d84ca6e985a40fb - 1,000 successful demonstrations; 194,047 frames
- State: qpos9 + qvel9 (18D); action: 7D; wrench: 6D
- Action horizon: 10; global batch size: 32; seed: 42
- GPUs: noforce 0–3; force 4–7
- Force history: length 4, stride 1, one token, latent dimension 256, MLP encoder
- Evaluation was not run for this training batch.
- Task
- Robotics
- Created
- Sep 14, 2026
- Updated
- Sep 14, 2026