Dear All, I would like to ask, what happened to XLA.jl. When I checked the repository, it says it was archived. Is the project of building the Julia-to-XLA interface dead? Best, Jan

Not sure, and while I/you can still install XLA under FluxML this other XLA.jl is not achieved and I could also install (not both projects at the same time, I don’t think possible nor needed, at least in this case, to install projects with same name): pkg> add https://github.com/JuliaTPU/XLA.jl I …

The general answer is the same, but there is some nuance depending on what you’re thinking of when you say “Julia-to-XLA interface”: A way to run Julia code + models on TPUs. This would require wrapping libtpu and updating JuliaTPU/XLA.jl to support a recent version of Julia. The latter may requir…

Thank you for your reply. I was thinking about point 2, for my type of problem (basically solving a system of functional equations), TPU doesn’t make much sense. However, being a poor macroeconomist, I will probably stick to JAX. :sweat_smile: I was originally working in Julia, however, a year ago, …

[image] Honza9723: TPU doesn’t make much sense. However, being a poor macroeconomist, I will probably stick to JAX. [image] Why is JAX so fast? · google/jax · Discussion #11078 I was writing a notebook to illustrate the performance trade-offs of array-oriented programming and impe…

As @jling mentioned, it’s not hard to beat many types of kernel fusion JAX does. However, it is also true that doing so often gives up AD or GPU support. Given that you mention nested AD as a selling point, can you elaborate on some of your macroeconomic use cases which use it? I’d also be curious …

Thank you for your interest! Small disclaimer: what I work on is really a niche application, even within my field. First, to those nested derivatives: Sometimes, I work on continuous time models, where the functional equation to solve is a second-order HJB equation coupled with an associated syst…

Related issues [image] Julia's Broadcast vs Jax's vmap Internals & Design My (high-level) understanding of jax.vmap is that it automatically vectorizes vectorizes a function along a specified axis of its input by introducing an “abstract” axis and compiling the code as th…

FWIW you are likely a more professional user than I, someone who doesn’t write Julia for work/research :wink: . Thanks for providing such a detailed post. I think examples like this are important to understand where the ecosystem and possibly the language are still lacking. Not just on a technical l…

[image] ToucheSir: I assume the intersection of people who want to use XLA and/or TPUs, want to use Julia, and would be willing to take on this project is basically the empty set. I would be very happy to be proven wrong on that :slight_smile: Apologies for reviving an older thread but I’d ce…

I stumble upon this thread by chance and just want to say that I’m in a very similar situation myself. I invested quite some time learning Julia with the idea that I could write efficient code without worrying about broadcasting. But I soon discovered that it did not work out with Zygote, and I’m no…

What happened to XLA.jl

Specific Domains Machine Learning

jar1 October 2, 2022, 10:39pm 8

Related issues

Topic		Replies	Views
Thoughts on JAX vs Julia Community jax	21	12086	October 25, 2022
State of machine learning in Julia Machine Learning	60	66906	August 26, 2022
Where does Julia provide the biggest benefits over other ML frameworks for research? Machine Learning	34	10804	September 16, 2019
Deep learning in Julia Machine Learning	35	12918	April 22, 2024
Why is Python, not Julia, still used for most state-of-the-art AI research? Offtopic knet , flux , machine-learning , mlj , sciml	65	11861	August 14, 2021

What happened to XLA.jl

Related topics