Giuseppe Carleo

yes this seems correct now. By the way, I recall that I was not using ```choice``` at the time because it wasn't supported by numba

Yes @ShHsLin the idea is that in v3.0 we are removing all C++ code, to make everything more easily interoperable with autograd engines (pytorch, jax, tensorflow). Basically, now all high-level...

No problem @ShHsLin and thanks for taking the time to do this! Things have changed quite a bit in this release 3.0, so it's completely understandable if you are a...

Hi @twesterhout, sure it would be valuable to have (automatic?) benchmarks in NetKet, also to make sure that some of the new features are not slowing down the existing ones....

Alright @twesterhout , that sounds like a plan! Indeed google benchmark seems the best option out there. I still don't quite get why the scripts running the benchmarks shouldn't be...

Even if we have moved past the C++ phase, this issue is still quite current. We have some primitive benchmarks in ```netket/Benchmarks``` but it would be nice to study a...

Hi, yes thanks this is something important to do especially in light of the new release. I would prioritize work on the small benchmarks maybe for the moment? also, I...

Yes, nice interface

I don't have strong opinions about this, it seems that the behavior is roughly the same anyways? (maybe marginally better, but I guess that at large alpha there might more...

If there is no known optimal way of initializing rbm weights(even though I recall some tricks in Bengio's paper?) I think this should be first explored systematically, potentially in a...