swift VS hummingbird

Because of C/C++ interop, and integration with CMake, you can just add Swift to a Zephyr project and it pretty much Just Works. [The docs](https://github.com/apple/swift/blob/main/docs/EmbeddedSwift/...) should mostly apply to the Zephyr SDK as well.

Fortunately, the Observation framework is part of the Swift 5.9 standard library. We can learn more information by examining its source code.

They do! See https://github.com/apple/swift/blob/main/docs/LibraryEvoluti...

You can also see an example of what a different high level language integration with Swift ABI looks like here: https://github.com/dotnet/designs/blob/main/proposed/swift-i...

So is differentiable Swift a package for Swift or is it part of the Swift standard library? The video says go to swift.org but I can't find any info about differentiable Swift on that site.

Swift's Differentiable Programming Manifesto. Swift has a powerful differentiable programming component, integrated with the compiler.

You can do the same thing the other way around - https://github.com/apple/swift/blob/main/docs/Android.md.

Codable (along with other derived conformances like Equatable, Hashable, and RawRepresentable) is indeed built in to the compiler[0], but unlike Serde, it operates during type-checking on a fully-constructed AST (with access to type information), manipulating the AST to insert code. Because it operates at a later stage of compilation and at a much higher level (with access to type information), the work necessary is significantly less.

With ongoing work for Swift macros, it may eventually be possible to rip this code out of the compiler and rewrite it as a macro, though it would need to be a semantic macro[1] rather a syntactic one, which isn't currently possible in Swift[2].

[0] https://github.com/apple/swift/blob/main/lib/Sema/DerivedCon...

`Int` is a regular struct with a single stored property of type `Builtin.Word` . But the latter is a magical compiler built-in. Source for integer types is generated from this template - https://github.com/apple/swift/blob/9da65ca0a15fdf341649c994b0a77ec3b71f2687/stdlib/public/core/IntegerTypes.swift.gyb

If are interested in just predictions you can try Hummingbird. It is part of the PyTorch ecosystem. We get already trained scikit-learn models and translate them into PyTorch models. From them you can run your model on any hardware support by PyTorch, export it into TVM, ONNX, etc. Performance on hardware acceleration is quite good (orders of magnitude better than scikit-learn is some cases)

I think Rapids AI's cuML tried to go into this direction (essentially scikit-learn on the GPU): https://docs.rapids.ai/api/cuml/stable/api.html#logistic-reg.... For some reason it never took really off though.

Btw., going on a tangent, you might like Hummingbird (https://github.com/microsoft/hummingbird). It allows you trained scikit-learn tree-based models to PyTorch. I watched the SciPy talk last year, and it's a super smart & elegant idea.

ONNX opens an avenue for direct inference using a number of languages and platforms. For example, a model could be run directly on Android to limit data sent to a third party service. ONNX is an exciting development with a lot of promise. Microsoft has also released Hummingbird which enables exporting traditional models (sklearn, decision trees, logistical regression..) to ONNX.

And Python.

For inference, you can convert your models to other formats that support GPU acceleration. See Hummingbird https://github.com/microsoft/hummingbird

The surprising thing is that Hummingbird can be faster than the GPU implementation of LightGBM (and XGBoost) if you use tensor compilers such as TVM. [The paper](https://www.usenix.org/conference/osdi20/presentation/nakandala) describes our findings. We have also open sourced the [benchmark code](https://github.com/microsoft/hummingbird/tree/main/benchmarks) so you try yourself!

I took their sample code from Github and tweaked it to spit out times for each model's prediction, as well as increase the number of rows to 5 million. I used Google's Colab and selected GPU for my hardware accelerator. This gives an option to run code on GPU, not that all computations will happen on the GPU.