Any decent ML library should have support for GPU nowadays. Or is this just a proof of concept to practice a bit with Rust?

I am not that used to Rust yet but I wonder, is it easy to write generic (templated?) code which works both on GPU and on CPU?

Is there any GPU (CUDA) based ML lib in Rust? I searched a bit and found only this (and related): https://internals.rust-lang.org/t/single-source-gpu-support/...

Or maybe this, although this is only indirect: https://github.com/arrayfire/arrayfire-rust

Also, next topics for ML libs: multi-GPU, distributed computing, ...

There is Leaf[1], the Hacker's Machine Learning Framework, which has GPU (CUDA, OpenCL) support for Machine Learning in Rust.

James linked to it in the community section of his post, at the end.

[1]: https://github.com/autumnai/leaf

Was looking at leaf yesterday, looks good. How good is openCL perf with Leaf compared to CUDA?

We would love to test that and release the performance on Deep Learning Benchmarks[1] or convnet-benchmarks[2]. But so far no convolution/NN-related OpenCl kernels are linked[3], although it might be quite easy, if the kernels exist.

[1]: https://github.com/autumnai/deep-learning-benchmarks

[2]: https://github.com/soumith/convnet-benchmarks

[3]: https://github.com/autumnai/collenchyma-nn

You could do a lot of ML without GPU support. Especially if you are dealing with sparse data (like text).

This is an interesting read of ML in Rust, but it jumps straight into the modeling part. One question I have is how does one get a good idea of what the data is (e.g. data visualization and simple summary statistics) in Rust? Or is this something that you do first in another language before using Rust for the ML part?

This is nice, thanks, will make for a great weekend read and study.