CoreML is such a bad lib compared to Pytorch/Jax and even Tensorflow that im not sure why anyone should use it if they can just use other stuff (i would probably prefer pure MPSGraph for training on device at it allows me to control everything in easy and Tensorflow like way)
I wrote a little service to handle some of the Matrix client-server authentication API.
The first version used Vapor before we had async/await. That was painful at times but doable.
Earlier this year I got some time to rewrite it in modern Swift with async/await. It’s a total game changer. Almost as easy as writing in Python.
Maybe even easier than Python sometimes, because with Swift you have a decent type system watching your back. If it compiles, then you’re probably most of the way there.
That looks nice. I wrote a Swift book last year and I had originally intended on including a server side example but didn’t get to it. Hopefully I will add an example in the next edition.
I have preferred Lisp languages for the last 40 years, but there is something about Swift I like, probably the availability of great libraries like CoreML and good REPL support.
Pre-pandemic I was building a new EV charging kiosk experience on Raspberry Pi industrial hardware using Swift in Docker (actually, Balena). Build times were a little long at the time, due to the RPi CPU speed and Swift’s compiler. But there were a surprising amount of hardware drivers available for things.
I am writing the backend of my side-project with it. Async/Await in Swift is game-changing for safety and productivity. The tooling is only getting better and you can expect greater maturity with the coming of [Swift 6](https://forums.swift.org/t/on-the-road-to-swift-6/32862) likely next year.
We use Vapor3 and the language is ok, but the deployment story (docker/k8s, for us) is a nightmare and has caused us unending drama. Things seem a little better with newer versions of vapor and swift 5.5, but our experiments don’t seem to have gotten very far, and the upgrade story is quite a lot of work.
Highly performant, It runs within docker container. Documentation is highly available.
Challenges: Getting more people to use it and collaborate with it.
SwiftUI has become more trendy, that other frameworks don't get wide relevance.
another library which brings more interesting projects is: CoreML On-Device ML training and inference