Julia is looking more and more like a viable alternative to Python's data science stack.
For small things, interop with R is good, so one can defer things to R packages and get access to a great set of functionality.
What do you mean by that? AFAICT for the most part Julia is already like the best that Jax could ever hope to be — since there is language level support for JIT compilation, gradients, etc — given how well libraries compose without having to pick different incompatible subsets of the language.
Now I agree that Julia, with its JIT, is well positioned to do all this. But I don’t think it does.
I’m curious to get perspectives on what remains to be done to solve the missing piece, and what are the biggest challenges to overcome (assuming one gets access to the appropriate low level APIs on top of the hardware — like OpenAI’s Triton).
I would really prefer to use Julia, I actually dislike Python, but it's unrealistic to implement large architectures in Flux. It is buggy, lots of gradient calculations are unreliable.
See some discussion here for instance: https://kidger.site/thoughts/jax-vs-julia