pi05_piper_stacking_v3_latency - step 9999

Chef-style latency-aware cube-stacking policy. Same data as axiboai/piper_stacking_v3 (155 eps), trained from pi05_base with a training-time action-target shift of 5 frames (obs[t] -> action[t+5:t+5+H], 176ms exec @30fps). Clean A/B vs pi05_piper_stacking_v3. Checkpoint: step 9999 / 10000 (final 9999 to follow). Prompt: stack red cube on blue cube

import pathlib
from huggingface_hub import snapshot_download
from openpi.training import config as _config
from openpi.policies import policy_config
ckpt = pathlib.Path(snapshot_download("axiboai/pi05_piper_stacking_v3_latency"))
policy = policy_config.create_trained_policy(_config.get_config("pi05_piper_stacking_v3_latency"), ckpt)