actuator_unboxing_v2_pi05-absolute

π₀.₅ (Pi05) — Vision-Language-Action model with open-world generalization, from Physical Intelligence. See the π₀.₅ blog post.

Trained with LeDream (a fork of LeRobot).

Training

• Base model: lerobot/pi05_base
• Dataset: DreamMachines/actuator_unboxing_v2
• Step (this revision): 500000
• Batch size: 16
• Total steps target: 500000
• Optimizer: adamw (lr 5e-05)

Available revisions

Every training checkpoint is uploaded as its own git revision; the latest is also on main. Load a specific step with:

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

cfg = PreTrainedConfig.from_pretrained("DreamMachines/actuator_unboxing_v2_pi05-absolute", revision="v002000")
policy = make_policy(cfg)

Available: v025000, v050000, v075000, v100000, v125000, v150000, v175000, v200000, v225000, v250000, v275000, v300000, v325000, v350000, v375000, v400000, v425000, v450000, v475000, v500000

Quick eval on a real DK1

make eval MODEL=DreamMachines/actuator_unboxing_v2_pi05-absolute

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