SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation | Read Paper on Bytez