Matt Fishman

This is pretty close to being ready, in the latest I've implemented taking the naive tensor product of two graded axes (for abelian symmetries, which was all that was implemented...

@ogauthe let's discuss this issue around `dual` offline. I was planning to address that issue in a future PR, and merge this PR as-is as a starting point for future...

Thanks for the report @ArtemStrashko. We'll look into it. We are in the process of rewriting and simplifying a lot of the NDTensors library so it's good to keep track...

Thanks for the report @csmaczny. Basically I don't think we ever had support for `Diag` tensors on GPU, so I'm not surprised this doesn't work. A workaround should be to...

We have been discussing a general buffer system which can help automate pre-allocating data to decrease allocations in tensor operations (@kmp5VT was testing out a prototype in #1011), while being...

Thanks for the report @ArtemStrashko, I agree we should be handling that promotion better using information from the buffer tensor.

Maybe not quite what you're hoping for, but you can do this: ```julia julia> (OpSum() + ("Z", 1)) * (OpSum() + ("Z", 2) + ("Z", 3)) prod( sum( 1.0 Z(1,)...

This is a bit more involved then I expected, and ultimately would be made much easier by making `BlockSparse` and `DiagBlockSparse` more generic by sharing more code. I think we...

Thanks for the report @ArtemStrashko. I think you mean `l2 = x -> lognorm(x)` in the example you show? I was thinking of handling this by defining custom `rrule`s for...

I would have assumed this was already defined, but should be easy enough to add. Note that is just special syntax for `ITensors.contract!`, i.e. `A .+= B .* C` would...