fbsh96/so101-sock-ball-act-handcam-100eps-mi300x-b16-20000steps

SO101 Sock/Ball Pick-Place ACT Hand-Camera Checkpoint

This is a LeRobot ACT checkpoint trained on fbsh96/so101_sock_ball_pick_place_formal_100ep.

Training Summary

• —Policy: ACT
• —Dataset: fbsh96/so101_sock_ball_pick_place_formal_100ep
• —Episodes: 100
• —Frames: 91,600
• —Device: AMD Instinct MI300X via ROCm/PyTorch
• —Steps: 20,000
• —Batch size: 16
• —Input features: observation.state and observation.images.hand_cam
• —Action dimension: 12
• —Final logged loss: ~0.074

Important Note

This checkpoint intentionally uses only observation.images.hand_cam. The source dataset also contains front_cam, but the raw 720p front camera path was not used in this training run because it was significantly slower without preprocessing/downsampling.