Filippo Vicentini

@inailuig has recently (sucesfully) experimented with plugging MPI inside of a CPU device... He can say more than I can. However, what we would love is for a way to...

Over at @netket we'd love this to compute kinetic energies of quantum states, like @dpfau does. I've got a private branch where I use @y1xiaoc 's code internally, and we're...

So it turns out even more people are interested in this! https://github.com/microsoft/folx

@mattjj what's the status on this? Do you plan on merging it?

From my point of view, aka that of a maintainer of [a library](https://github.com/netket/netket) that could benefit from such fast laplacian calculations, but with no desire/bandwith to work directly on them......

Hi, I think this would be a great addition to jax! I have two minor comments: - When we implemented [batch_vmap](https://netket.readthedocs.io/en/stable/api/_generated/jax/netket.jax.vmap_chunked.html#netket.jax.vmap_chunked) in netket a few years ago, we soon found...

I fully support this effort! This would remarkably simplify setup for several academic users. I've had to fight with setting up jax.distributed in weird clusters several times and this seems...

Plum only dispatches positional arguments, and pipes through keyword arguments. To all effects, what plum is doing is interpreting your function definitions as ```python @dispatch def f(x: int, y: float,...

This should be throughly tested before merging. In limited testing on my laptop (though apple throttles aggressively, so should test on a cluster) I've seen no changes to slight performance...