ACT β€” C5b_pusht_chunk20

ACT trained with LeRobot (ba83e4c, official ACT hyperparameters: batch 8, lr 1e-5, chunk 20, 100k steps) on one RTX 5090.

lerobot-eval --policy.path=LUC1O/act-5090-C5b_pusht_chunk20 --env.type=pusht --env.task=PushT-v0 --eval.n_episodes=50 --eval.use_async_envs=false
Downloads last month
23
Safetensors
Model size
51.6M params
Tensor type
F32
Β·
Video Preview
loading

Dataset used to train LUC1O/act-5090-C5b_pusht_chunk20