Daniel Smith

One note here is that JSON is fairly suboptimal for NumPy array encodings as you often need to serialize the raw bytes to ASCII and back which takes quite a...

With msgpack you can skip the base64encode as you can store raw bytes (up to 2GBish) in msgpack directly. See an implementation [here](https://github.com/MolSSI/QCElemental/blob/master/qcelemental/util/serialization.py#L40). We founding avoiding the base64 encoding to...

Right, we just hack a derived class ontop of the BaseModel (see [here](https://github.com/MolSSI/QCElemental/blob/master/qcelemental/models/basemodels.py#L121)). This works fine, but we are always looking to have things a bit more canonical so we...

@mlestep is on this.

Right, but my thought there was that for cases that have poor stability diagonalization might do some wonky things before DIIS kicks in. In these cases you want to use...

I believe the following will help you: https://dgasmith.github.io/opt_einsum/getting_started/sharing_intermediates/

Yes, cache/intermediate aware paths are something we have discussed but have not implemented. It isn't clear that there is a general approach to solve the problem as straightforward approaches become...

I would natively respond that the python overhead and repeated `einsum` calls is the root problem; however, the overall time is larger than expected in the `opt_einsum` route. I wonder...

As a note id really like to separate symmetry and contraction chains from the underlying tensor class. Derived classes can be as complicated as needed, but there should be a...

Both articles lean heavily on the tox assumption, which isn't the case here. It is also worth noting that the citation for the change in cryptography has a pretty long...