from lerobot.policies.factory import make_policy
from lerobot.configs.policies import PreTrainedConfig

cfg = PreTrainedConfig.from_pretrained("DreamMachines/pi05_sweep_v2_01_freeze_01_full", revision="v002000")
policy = make_policy(cfg)

Available: v010000, v020000, v030000, v040000, v050000, v060000, v070000, v080000, v090000, v100000

Quick eval on a real DK1

Base model: <a href="https://huggingface.co/lerobot/pi05_base" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors"> lerobot/pi05_base
Dataset: <a href="https://huggingface.co/datasets/DreamMachines/actuator_unboxing_20h_345" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors"> DreamMachines/actuator_unboxing_20h_345
Step (this revision): 100000
Batch size: 32
Total steps target: 210000
Optimizer: adamw (lr 5e-05 )

make eval MODEL=DreamMachines/pi05_sweep_v2_01_freeze_01_full

See docs/source/policy_pi05_README.md for π₀.₅-specific notes.