host the container
R specifically? 🧵👇🏽 (View Tweet)
Note: Thread
You need to serialize your model to disk in some format. {butcher} will make it as small as possible
https://t.co/objuvt81iw. After you’ve saved it to disk you need to make it callable with all your preprocessing etc. (View Tweet)
I recommend making a my_predict() type function.
Then use this function to create a plumber API https://t.co/cgCDOsYF1Q (View Tweet)
This plumber API is all you really need to put it into a docker container. But the problem is that its single threaded. To make it parallel use {valve}
https://t.co/PNgTnX9KLT (View Tweet)
ALL THAT SAID?!!!! Use {vetiver} for creating the plumber API for you. https://t.co/GnFwMuL5ko
vetiver creates a plumber API. So just use that plumber API with Valve for “free” multiprocess scaling.
Lastly!! (View Tweet)
You need to “deploy” it. In AWS you need to publish the container to your registry, then deploy it via EC2 or ECS for scaling. (View Tweet)
