opentau.policies.pi07_paligemma.low_level
Low-level component for π07 (PaliGemma variant).
This module implements the low-level component of the π07 hierarchical architecture. It uses a SpaceTime SigLIP video encoder, processes temporal state sequences (one continuous token per timestep), and supports optional subtask response, subgoal image, and metadata conditioning. Action generation combines flow matching (continuous actions via an action expert) with FAST discrete token prediction (through the VLM backbone with Knowledge Insulation).
Modules
Configuration module for the PI05 Policy. |
|
π05: A Vision-Language-Action Flow Model for General Robot Control |