Abstract: This paper presents the design of a framework for loading a pre-trained model in PyTorch on embedded devices to run local inference. Currently, TensorFlow Lite is the most widely used ...