Remote Profiling

DL Workbench can collect performance data not only on the machine on which you run it, but also on other machines in your local network. This helps when you cannot run the DL Workbench on a machine due to security or network issues or because it is impossible to install Docker. If this is the case, run the DL Workbench on another machine and collect performance data on a remote machine in your local network.

When connected to a remote machine, you can currently use a limited set of DL Workbench features:

Single and group inference Yes
(HDDL plugin is not supported)
INT8 calibration No
Winograd algorithmic tuning No
Accuracy measurements No
Performance comparison between models on local and remote machinesYes
Deployment package creation No

Follow the steps below to profile your model on a remote target:

  1. Set up the target machine
  2. Register the remote target in the DL Workbench
  3. Profile on the remote machine