Add int64 reductions

[klokare]

Ran into an issue with the sum function when using int64.

the following runs ok

import mlx.core as mx
y = mx.ones((70), dtype=mx.int32)
mx.sum(y)

the following crashes every time on my MacBook Pro M3 Pro / 18GB

import mlx.core as mx
y = mx.ones((70), dtype=mx.int64)
mx.sum(y)

with error message "libc++abi: terminating due to uncaught exception of type std::runtime_error: [Reduce::eval_gpu] Does not support int64"

can you confirm this is intended behaviour? or if it something that is planned for a future version?

[awni]

Yes, int64 is not yet supported in our reductions, that's expected. It's definitely possible to add, but not the highest priority at the moment. Happy to take a contribution for this though if someone is interested in adding it!

[vj-krish]

@awni we can close this now.

[awni]

Thanks @vj-krish !!

[angeloskath]

Fixed in #569.