VLA · aws-gpu

pi0.5

pi0.5 is the VLA policy path for interactive LIBERO/MuJoCo manipulation demos, DROID/Franka action APIs, and future robot policy servers.

LIBERO SpatialBowl between plate and ramekin

Editing this task changes the next rollout. Start a new cloud rollout to run the selected scene and language.

Advanced

Supported LIBERO tasks pair this text with a matching scene, initial state, and success condition.

Runtime Target

First target is the official pi05_libero checkpoint running a closed-loop LIBERO/MuJoCo rollout from a supported Task with video, streaming frames, metrics, action chunks, and logs.

Plan

PRO

Provider

aws-batch

Accelerator

nvidia-l4-or-a10g

Simulator

openpi-libero-mujoco

Embodiment

LIBERO Franka

Placement

cloud-only

Status

Working Now

AWS Batch runs OpenPI pi05_libero in LIBERO/MuJoCo and saves a rollout video artifact. The page rotates through supported LIBERO Spatial tasks and can queue a running task update at replan boundaries. The secondary pi05_droid profile returns raw DROID action tensors for direct model I/O checks.

Ready To Deploy

Scale-to-zero AWS Batch infrastructure, ECR worker image wiring, cloud-session planning, paid gating, and production status polling are deployed.

Runtime Notes

LIBERO/MuJoCo sim is live; DROID/Franka action API is available; Drake/LBM and longer production policy-server serving are later integrations.