wafngar

Very common in ML research these days - claim novelty / cite prior work in an obfuscated way and so on.

But they have built a fully “independent” pipeline. Deepseek and others probably trained in gpt4, o1 or whatever data.

Why not use torch.compile()?

They train with ERA5 and observations.

Should be a relevant reference: https://arxiv.org/abs/2111.13587

Adaptive Fourier Neural Operators: Efficient Token Mixers for Transformers John Guibas, Morteza Mardani, Zongyi Li, Andrew Tao, Anima Anandkumar, Bryan Catanzaro

Brussels is quite small. Probably less than Washington.

Depends. PyTorch on the other hand has a large user base and well defined and tested api. So should be doable; and is already progressing and rapid speed..

Probably unfair as a reaction to the unfair statements in the blog…

PyTorch is developed by multiple companies / stake holders while jax is google only with internal tooling they don’t share with the world. This alone is a major reason not to use jax. Also I think it is more the other way around: with torch.compile the main advantage of jax is disappearing.

AIFS is transformer based (Graphcast is pure GNN) so different architecture and is already running operationally, see:

https://www.ecmwf.int/en/about/media-centre/aifs-blog/2024/i...

For AI you have to support PyTorch. That works well now on the big AMD GPUs. Consumer gpu capabilities do not matter here.

US has no bureaucrats? All the EU member states, no bureaucrats? How many projects in member states cost the tax payer billions? Btw the EU budget is tiny in comparison.

The UK would do much better had they not chosen Brexit. Basically everyone here admits that (bank of England etc) apart from some zealot politicians. And the health system has collapsed. The problems are not solely caused by Brexit but it exacerbates nearly all of them.

What concrete advantage would it have for you?

sure, my point was more that it uses pytorch -> also lots of c++ :-)

But isn’t that PyTorch with a few extra steps?