Idea here is that currently the model train runs after the mode.predict in tensorflow.js.
As an improvement we don't need to recall the train model every 5min... we could do it only every 60min or even more. So it would be a good enhancement to split this in a function that will only be executed every N model.predict loops.
That will not affect the model predict and we would be able able to significantly reduce the aws CPU average usage.
Idea here is that currently the model train runs after the mode.predict in tensorflow.js.
As an improvement we don't need to recall the train model every 5min... we could do it only every 60min or even more. So it would be a good enhancement to split this in a function that will only be executed every N model.predict loops.
That will not affect the model predict and we would be able able to significantly reduce the aws CPU average usage.