Matt Fishman

Also see issue #618 for a proposal on how to do this in a way that is not too confusing in terms of floating point operations and the view behavior....

Note that we could use the notation: ```julia svd(A, .., (k, l)) ``` which is standardized by the package [EllipsisNotation.jl](https://github.com/ChrisRackauckas/EllipsisNotation.jl).

Yes good point, I forgot you can do it that way.

@GTorlai this is a strange one, maybe we can test out an older version of the `rrule` for `apply` that was only written for MPS which should be easier to...

That would be great, thanks!

There is through the `ChainRulesTestUtils.jl` package, but we haven't set it up for `MPS`/`MPO` yet. We have it set up for `rrule`s written just in terms of `ITensor`s but it...

Some follow up thoughts: 1. I created a function `hdf5_type` with definitions like `hdf5_type(::Type{

In the latest commit I added `ITensors.read` which automatically determines the type by `eval`ing the `"type"` attribute, used as follows: ```julia i = Index(2) A = randomITensor(i) B = randomITensor(i)...

Ok, probably the last change. In the latest I added the more compact `ITensors.save`/`ITensors.load` syntax, inspired by `save`/`load` defined in `JLD`/`JLD2`: ```julia i = Index(2) A = randomITensor(i) B =...

In the latest, I added a fallback to the version of `ITensors.read` where no type is specified, so that if there is no `"type"` attribute in the file it is...