Action Chunking Transformer Policy trained for the pick and place blue blocks into the circle with SO-ARM100.
We strong recommend the environment needs to be the same as the video. I use the camera of Macbook Air M2, Also The model was inference by Macbook Air M2 16GB.
The model was trained using
LeRobot's training script and with the
masato-ka/so100_grasp_lego dataset, using this command:
1python lerobot/scripts/train.py \
2 --dataset.repo_id='masato-ka/so100_lego_sort' \
3 --policy.type=act \
4 --output_dir=outputs/train/act_so100_lego_sort \
5 --job_name=act_so100_lego_sort \
6 --policy.device=cuda \
7 --wandb.enable=true
Actually, this model was stopped traning at 60k steps.
The training curves may be found at
https://wandb.ai/masato-ka-personal/lerobot/runs/u67z9fk8. The current model corresponds to the checkpoint at 60k steps.
This took about 2h to train on an Nvida L4.
