You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

chip_pick_g1_act — ACT policies for Unitree G1 chip pick

Three ACT policy variants for the lab's chip-pick task. Trained on real-robot demonstrations recorded with Han's teleoperation rig, converted to LeRobot datasets, and trained with lerobot_train.py on a single 5090.

Subfolder	Policy I/O	`--repo_id` to pass at eval time
`abs_ee/`	Absolute EE pose (xyz + 6D rotation) + 12 hand DoFs	`aleksantari/chip_pick_eepose`
`rel_ee/`	Relative EE pose (chunk-start anchor) + 12 hand DoFs	`aleksantari/chip_pick_eepose` (same dataset as abs — rel anchoring happens inside the policy preprocessor)
`mask/`	Absolute EE pose + `cam_head_mask` segmentation	`aleksantari/chip_pick_eepose_mask`

⚠️ There is no chip_pick_eepose_rel dataset. Both abs_ee/ and rel_ee/ use the same aleksantari/chip_pick_eepose repo. The relative-pose transform is a preprocessor step inside the policy, not a dataset variant.

Each variant has snapshots step_060000, step_085000, step_100000. All ACT, chunk_size=100, n_action_steps=100, ResNet-18 backbone, kl_weight=10.0.

Loading a checkpoint

Absolute EE pose

huggingface-cli download aleksantari/chip_pick_g1_act \
    --include "abs_ee/step_100000/*" --local-dir ./ckpts

python -m unitree_lerobot.eval_robot.eval_g1 \
    --policy.path=./ckpts/abs_ee/step_100000 \
    --repo_id=aleksantari/chip_pick_eepose \
    --arm=G1_29 --ee=inspire_ftp --image_source=teleimager

Relative EE pose (same dataset as abs)

huggingface-cli download aleksantari/chip_pick_g1_act \
    --include "rel_ee/step_100000/*" --local-dir ./ckpts

python -m unitree_lerobot.eval_robot.eval_g1 \
    --policy.path=./ckpts/rel_ee/step_100000 \
    --repo_id=aleksantari/chip_pick_eepose \
    --arm=G1_29 --ee=inspire_ftp --image_source=teleimager

Mask-augmented (needs the SAM3 streamer)

huggingface-cli download aleksantari/chip_pick_g1_act \
    --include "mask/step_100000/*" --local-dir ./ckpts

python -m unitree_lerobot.eval_robot.eval_g1 \
    --policy.path=./ckpts/mask/step_100000 \
    --repo_id=aleksantari/chip_pick_eepose_mask \
    --arm=G1_29 --ee=inspire_ftp --image_source=teleimager

See docs/sam3_mask_eval.md in the source repo for env setup (SAM3 install) and the runtime click-prompt flow.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support