DreamMachines/arena2036_insertion_pi05_2op

arena2036_insertion/pi05_2op

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/arena2036_insertion_2op_224x224
• Step (this revision): 30000
• Batch size: 32
• Total steps target: 30000
• Optimizer: adamw (lr 0.0001)

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/arena2036_insertion_pi05_2op", revision="v002000")
policy = make_policy(cfg)

Available: v005000, v010000, v015000, v020000, v025000, v030000

Quick eval on a real DK1

make eval MODEL=DreamMachines/arena2036_insertion_pi05_2op

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