Downloads · 30 days
3
38% of all-time downloads
act_carton_middle_labels_5k
stevenzenith/act_carton_middle_labels_5k
act_carton_middle_labels_5k is a robotics model from stevenzenith. Use it for the robotics task on the model card, and read the license before you ship it in a product. It is set up for lerobot.
The last step of the middle-labels fine-tune sweep, not the selected one. ACT, 5,000 steps from actcartonphasec50k.
All-time downloads
8
Public
Parameters
51.6M
206 MB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors206 MB · 100%
At a glance
- Task
- Robotics
- Library
- lerobot
- Access
- Public
- Created
- Sep 2, 2026
- Updated
- Sep 2, 2026
- SHA
- a991c3b2
Model card
From the Hugging Face model README
act_carton_middle_labels_5k
The last step of the middle-labels fine-tune sweep, not the selected one.
ACT, 5,000 steps from
act_carton_phase_c_50k.
Task: dual-view, PressureVision-supervised paper-carton pick with an SO-101.
Use act_carton_middle_labels_2k
instead unless you specifically want the end of the sweep. On held-out middle
episodes 8, 9 and 12, 2k scores body MAE 1.355 against this checkpoint's 1.428,
and closes falsely on 2 of 172 open-phase frames against 3. 2k is the checkpoint
the operator selected as a bounded physical candidate; this one is published so
the sweep is complete and the selection is checkable rather than asserted.
No train_config.json: the training script saves only the policy and its
processors. Recipe and labels are in
training/train_middle_labeled_act.py;
the full comparison table is in training/TRAINING_HANDOFF.md.
- Task
- Robotics
- Library
- lerobot
- Created
- Sep 2, 2026
- Updated
- Sep 2, 2026