Can ACT policy support pushT task?

[Kimho666]

I want to train the ACT policy with pushT dataset, but the evaluation accuracy is only 0%.

Here is my yaml
act_pusht.txt
And my training command is
''
python lerobot/scripts/train.py
hydra.run.dir=outputs/train/2025_1_3_1654_act_pusht
hydra.job.name=act_pusht
policy=act_pusht
policy.use_vae=true
env=pusht
env.task=PushT-v0
dataset_repo_id=lerobot/pusht
training.offline_steps=50000
training.save_freq=25000
training.eval_freq=5000
eval.n_episodes=50
wandb.enable=false
device=cuda
''