Several computer vision use cases, for example, in autonomous driving, have very low latency constraints. To meet these requirements, model developers may need to deploy models at the edge (i.e., to eliminate the time it takes to gather data, send it to the cloud for processing, and make predictions). D2Go, which is built on top of Detectron2, PyTorch Mobile, and TorchVision, is designed to enable ML practitioners to train and deploy deep learning object detection and image classification models on mobile devices and other hardware. Specifically, D2Go makes it easy to create FBnet models, which are optimized for model devices.