Modelpublic

alexhegit/so101-simstudio-lab01-pnp-act

sourceHugging Faceapache-2.0updated 26d agoView on Hugging Face

0likes14downloads

Model Card

SO-101 Lab01 Pick-and-Place — ACT

ACT policy fine-tuned on expert demonstrations collected and validated with [SO-101 SimStudio](https://github.com/rocPAI-Forge/so101-simstudio) (MuJoCo sim2sim, leader-arm teleop).

Training data: alexhegit/so101-simstudio-lab01-pnp

Documentation

Resource	Link
SimStudio repo	rocPAI-Forge/so101-simstudio
Lab 01 walkthrough (record → train → eval)	labs/lab01_pnp/lab01_pnp.md

For dataset layout, training commands, eval configs, and reference metrics, see Lab 01 in the repo above.

Quick load

python

from lerobot.policies.act.modeling_act import ACTPolicy

policy = ACTPolicy.from_pretrained("alexhegit/so101-simstudio-lab01-pnp-act")

The 15-D ACT reference (same 50K schedule; policy trained on pos+vel+ee) lives in this repo.

For joint-position-only ACT (6-D state, aligned with real SO-101 IL / sim2real), use alexhegit/so101-simstudio-lab01-pnp-act-state6.