Closed MaxvdKolk closed 3 years ago
@MaxvdKolk Yes, this is certainly one of the requirements (with view on replication of results) and the argument that currently results are hard to re-produce. Let's discuss the format in the meeting.
Yes, maybe some bullets to discuss
Isn't this more something that users can do themselves? I don't think it's necessary for us to add this as a feature. Non-default arguments can easily be recognized and saved manually. What is the main purpose of this feature? If it is just for adding to a paper, it might be a little overkill.
I was thinking more from a convenience point of view. Personally I would like a feature that could just report on the settings that were used, or read the settings from a file versus a main script.
It can be something very simple, just one routine for a (sub)solver that prints its dictionary of settings to a file, maybe a couple lines of code. However, if this is something we rather leave to the user, that's fine with me.
I'll just close it for now and not bother with implementing it yet. Lets see once more core features of the package are in place what we do with this.
As extension of #15, we might want to provide support to write the complete configuration of an optimisation problem to disk. We could opt to write to a dictionary, or pickle-like formats, that contain the used constants, (sub)solver types, convergence criteria etc.
Preferably it is human-readable (so no binary blob), such that it can be included easily with papers?