Justin Turney
Justin Turney
Sounds good to me. There should be `test_*.py` in the `lib` directory that can test the Python side of things.
Cool. I'll take a look at it.
Nice! One of our grad students has been playing with TBLIS, too, with very impressive performance benefits.
There are not many places internal to Psi4 that need to really know the ordering of our orbitals. So perhaps us changing to match default libint2 wouldn't be too big...
This seems like a neat ability.
The input file should be fort.56 Memory should be handled here: https://github.com/psi4/psi4/blob/8418efe67717940cda6e8446814d9689e9c303e8/psi4/src/psi4/mrcc/mrcc.cc#L932 Threading should be handled here: https://github.com/psi4/psi4/blob/8418efe67717940cda6e8446814d9689e9c303e8/psi4/driver/procrouting/proc.py#L4994 It's been a long time since any of these interfaces were managed....
Are there any tests that can be added to help ensure things are working properly? I think I found an example of using HIP CPU on GHA. Might be useful...
This test failed on Azure: `dlpnomp2-3/test_input.py`. An input keyword wasn't understood. Can you look into it? @andyj10224
Yeah that's fine with me. I used h5cpp because it did some stuff automatically for me. Reducing the number of dependencies einsums has is always good. We just need to...
Rebased. I think ciflow/inductor and ciflow/xpu labels need to be added back.