Create a template/demo app that runs on GPU and is accessed via SSH port-forwarding or web url. This involves (1) initializing a model on GPU within the app outside the process_fn(), (2) using the --listen flag when launching the app, and (3) creating an SSH tunnel. This will help showcase HARP's main advantage (running on hosted compute).
Create a template/demo app that runs on GPU and is accessed via SSH port-forwarding or web url. This involves (1) initializing a model on GPU within the app outside the
process_fn(), (2) using the--listenflag when launching the app, and (3) creating an SSH tunnel. This will help showcase HARP's main advantage (running on hosted compute).