imitation-learning-algorithms

Fine-tuned π₀.₅ on 54 own-collected dataset for language-conditioned pick-and-place on a Kinova Gen3 robotic arm. Full pipeline: teleop data collection, LeRobot conversion, LoRA training, and closed-loop deployment with remote inference. A vision-language-action policy that picks up a block and puts it in a bowl.