Jonathon Misiewicz
Jonathon Misiewicz
A post-DDD project. `normal_modes_write` is currently an option in the `findif` module. Back when all of Psi's hessians were finite difference, this was reasonable. But now, that Psi has RHF/UHF...
According to [the documentation](https://github.com/psi4/psi4/blob/master/psi4/src/psi4/libmints/petitelist.h#L329-L330) of `sotoao()`, the function is supposed to return the inverse of `PetiteList.aotoso()`. This matrix is not guaranteed to be invertible. If the matrix is _not_ invertible,...
Wouldn't a single `std::fill` be better than calling the function consecutively in a loop? If you did that, you could also use a consistent style for the screening i.e. `if...
Right now, our SCF instability following algorithm is the following, very dumb one: * Perform SCF * Check instability * If unstable, rotate in the direction of the unstable vector,...
``` molecule { He } optimize("scf/cc-pvdz") ``` on `1.8a1.dev57` gives ``` Traceback (most recent call last): File "/Users/jonathonmisiewicz/psi4/objdir/stage/bin//psi4", line 345, in exec(content) File "", line 22, in File "/Users/jonathonmisiewicz/psi4/objdir/stage/lib/psi4/driver/driver.py", line...
None of the `CASSCF_*` keywords are [documented](https://forte.readthedocs.io/en/latest/options.html). Those need to be documented. I also recommend a thorough pass that all keywords are documented.
Benchmarking indicates that for transition metals, we see that Molcas can finish an entire CASSCF in the time it takes Psi and L2 to compute integrals. Francesco attributes this to...
Forte doesn't compile for me, even after following the [readme](https://github.com/evangelistalab/forte/blob/master/README.md) instructions for `setup.py`. I compile successfully upon commenting out the declarations and definitions and calls to ``` /// Build OO...
The install instructions in `README.md` don't mention that Boost needs to already be installed before you can install Forte. I had to install that myself (after resolving some conda conflicts)...
The `job_type` keyword at present is too confusing. `job_type = None` actually does things like AVAS, and there's too much overlap with `correlation_solver`. Both are uncomfortably like selecting an electronic...