search

AI4OPT / OPFGenerator

Instance generator for OPF problems
MIT License
2 stars 3 forks source link

Add support for general arithmetic and high-precisions SOCWRConic #47

Closed mtanneau closed 4 months ago

mtanneau commented 4 months ago

Also added default settings in https://github.com/AI4OPT/OPFGenerator/pull/47/commits/bb0bf111e6af0203584f6120b725615725b2d874

klamike commented 4 months ago

Also added default settings in bb0bf11

Nice. Do you want to add more to the ACOPF/Ipopt and DCOPF/Mosek settings too?

mtanneau commented 4 months ago

Do you want to add more to the ACOPF/Ipopt and DCOPF/Mosek settings too?

What we currently have there for DC and AC are the typical settings that we use; I don't have anything smarter to add 🤷‍♂️

mtanneau commented 4 months ago

I cleaned up the HDF5 I/O utilities.

We now have more complete safeguards, and throw more exhaustive warning/error messages. In particular, I improved the checks for Array-typed data, which could silently fail for certain types. I am not 100% sure of what happens under the hood in HDF5, but it looks like some concrete types are automatically converted into a NamedTuple, which would lead to issues when loading the data later.

Our previous code was not catching this when dealing with Array data.

klamike commented 4 months ago

From HDF5.jl docs:

By default Complex numbers are stored as compound types with r and i fields following the h5py convention.

I guess it isn't in the HDF5 spec, but hopefully other implementations also follow this convention.

LGTM!