This repository contains Google Deepmind's RT-1-X code, with an added fine tuning pipeline to further train the model on data from new embodiments. The fine tuning code can be found in colabs/Fine_Tuning.py.
This code was developed in the course of my Master's Thesis with the title "Bringing the RT-1-X Foundation Model for Robotic Control to New Embodiments" where I explored the potential of adapting the model to a UMI robotic arm at the Intelligent Robotics Lab of University of Amsterdam. Along with this, a ROS2 node for running inference with actual robots was developed.