OnBoard is an experimental tool introduced in the Vitis AI 3.5. release. Its main objective is to enhance the analysis and debugging process for WeGO users by visualizing the models' inference flow. Built as an extension of the TensorBoard web server, OnBoard writes events with simple function calls. It can visualize the images of inference results, models' graph structure before and after WeGO transformation, and DPU platform-related information. As of Vitis AI 3.5, OnBoard is available for PyTorch and TensorFlow 2.
For OnBoard examples and more information, see Vitis AI GitHub .