ACT chess policy: c3 (6 cm) to d5 (10 cm)
ACT policy trained from scratch with LeRobot 0.5.1 on an SO-101 follower arm. 3 cameras at 480x640, 6-joint state and action.
Intermediate checkpoint: step 2000 of 22,000 (batch 24). Early checkpoint kept for comparison. The main run is at rtrtwrw/act_chess_c3_to_d5.
Load with:
lerobot-record ... --policy.path=rtrtwrw/act_chess_c3_to_d5_step2000
- Downloads last month
- 20