FAQ
Do I need a GPU?
Section titled “Do I need a GPU?”The Jetson TX2 has one built in. Inference runs on its integrated GPU through CUDA. If CUDA cannot start, the service falls back to the CPU and keeps running, with higher latency. See Hardware Setup.
Can I run without internet?
Section titled “Can I run without internet?”Yes. You install and update the box from an offline release bundle. The box needs no internet access, no registry access, and no image builds. See Offline Bundle Install.
Which model formats are supported?
Section titled “Which model formats are supported?”ONNX only. The dashboard rejects any other file type at upload. Export models with ONNX opset 15 or lower, because the ONNX Runtime build for the TX2 does not load newer opsets. See Preparing Models with modelctl.
How is data secured?
Section titled “How is data secured?”Every user signs in, and admin-only actions are limited to admin accounts. The inference service listens on the box’s loopback address only, so your network cannot reach it. To serve the dashboard over HTTPS, set AIBOARD_TLS_ENABLED=1. See Security and Environment Variables.
How do I update?
Section titled “How do I update?”Copy the newer release bundle to the box and run sudo ./update.sh. The update snapshots the database first, then waits for the stack to report healthy. If it does not, the update stops and prints how to roll back. See Offline Bundle Install.