Skip to content
1 minute readNew

Model Profiling connects to your own AITune server

Model Profiling now runs on the AITune server your workspace connects on the Compute page. Before, it could only use a server set up by Last Price, which did not exist, so the page always said AITune was not connected.

The Model Profiling page in dark mode saying AITune is not connected, with a Connect AITune button

Model Profiling needs an NVIDIA AITune server on a GPU machine. Until now the page could only talk to one server shared by the whole deployment, and none was set up, so there was no way to use it.

  • Connect your own AITune server on the Compute page, under Inference: its https address and, if it has one, its key. Last Price checks that it answers before saving it.
  • Model Profiling then shows that server's system stats, loads its reports and starts profiling runs on it.
  • Each workspace uses its own server; switching workspace switches server.
  • Only public https addresses are accepted, and error pages from your server are summarized rather than shown as they are.