If dependencies aren't installed yet in a packaged build there may be a blocking load (usually in a black screen) and require a map restart to enable the tensorflow component.
This isn't ideal, consider try/catch importing or checking a json file on bootup and if it's not ready, wait for a signal from the dependency installs to continue with imports.
We could then expose a multicast delegate in the Tensorflow Component that notifies if dependencies are still pending (progress? likely will go 0 to 100 immediately), and when they're fully installed/ready. Devs could use this event to signal to user that there are installations going on and show e.g. a spinning widget.
If dependencies aren't installed yet in a packaged build there may be a blocking load (usually in a black screen) and require a map restart to enable the tensorflow component.
This isn't ideal, consider try/catch importing or checking a json file on bootup and if it's not ready, wait for a signal from the dependency installs to continue with imports.
We could then expose a multicast delegate in the Tensorflow Component that notifies if dependencies are still pending (progress? likely will go 0 to 100 immediately), and when they're fully installed/ready. Devs could use this event to signal to user that there are installations going on and show e.g. a spinning widget.