---
library_name: openpi
license: apache-2.0
base_model: physical-intelligence/pi0.5
tags:
  - robotics
  - openpi
  - pi0.5
  - lora
  - piper
---

# π0.5 Piper Take-Cup-Off-Rack LoRA

OpenPI/JAX π0.5 LoRA checkpoint trained for 5,000 steps on
`take_off_cup_short16_dual14d_pi05_rgb224_v21` (16 episodes, 9,065 frames).

The exact language prompt is:

`take the cup off the rack and place it on the plate`

## Runtime contract

- RGB inputs: top, left wrist, and right wrist, letterboxed to 224×224.
- State/action: 14-D `[L_j1..j6, L_gripper_m, R_j1..j6, R_gripper_m]`; the 12 joint targets are trained as state-relative deltas and both grippers remain absolute.
- Recorded action supervision contains absolute joint-angle targets in radians.
- Grippers are meters and are clipped to `[0, 0.08]` at the policy boundary.
- Action horizon: 30.
- OpenPI config: `pi05_piper_take_cup_lora`.

Normalization statistics under `assets/` and the bundled OpenPI overlay are
required. The validation report checks checkpoint loading and inference pipeline
integrity; it is not a real-robot success-rate evaluation.