Jonatan Kłosko

@jalberto interesting, the build itself doesn't require an actual GPU, so the Docker build should be reproducible. What kind of errors are you getting?

@jalberto ah yeah, the first error is because I removed the file and forget to update, I've just fixed on main. The build error is very confusing, I was suspecting...

@jalberto I've just run `build/build.sh rocm` on a fresh AWS amd64 instance with Ubuntu 20.04 and it run without failure. I'm wondering if the issue could be that you build...

@Eiji7 you can try the new release and use ROCm 6.0, see https://github.com/elixir-nx/xla/issues/82#issuecomment-2124230058.

Yeah, it looks like latest XLA requires 6.0+, so I think this ship has sailed on this side. I don't think there's anything else we can do for 5.7, so...

I've just realised that `:other` is reached for decimal numbers (as shown in [this spec](https://github.com/svenfuchs/rails-i18n/blob/master/spec/unit/pluralization_spec.rb#L591-L596)), but with that in mind [this](https://github.com/svenfuchs/rails-i18n/blob/master/rails/locale/pl.yml#L69) (and others below) seems wrong to me; looks like...

@josevalim the nested map is not a problem, it's other Nx.Containers (currently tuples), so it may make sense to restrict Axon parameters to tensors.

Sidenote: sharding is a nice-to-have, but with https://github.com/elixir-nx/safetensors/pull/8 we should be able to write all parameters into a single file efficiently.

With #344 the main motivation (excessive memory usage) is addressed, so this is less of a priority. It would still reduce some time overhead necessary for transforming the params. Either...

Hey @michalwarda! The repository uses the BERT, so Bumblebee supports it, however we don't have any serving that fits cross-encoder. So currently you could do this: ```elixir {:ok, model_info} =...