You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I implemented the first improvement in commit 75cbcb9.
Even though I understand that your last table is more compact (and similar to what IPOPT does), I would argue against it. Abbreviations in the column names create ambiguity and make it difficult to read (although I know what they are!) I'd like it to still be human-readable.
That said, I'm happy to unify on a fixed scientific precision and squeeze some of the wider columns.
The people that look at the log are advanced users who know how Uno works under the hood
Advanced users will understand abbreviations
The log is mostly useful for debugging support questios. People will ask "Why did Uno fail?" and ignore the fact that the log has 2000 iterations and then hits an iteration limit...
It's easier to visually grep trends in the log if there is less vertical whitespace
I still like 80 character wide terminals...
I mostly look at the e+XY, e.g., Is the barrier decreasing by >1 OOM in each iteration at the end? What is the magnitude of the residuals/obj, does it ever go above e+08 or less than e-04. That sort of stuff.
The default is
To my eyes, an improvement is
and then you could unify on a fixed scientific precision (I didn't do the stat column):
The text was updated successfully, but these errors were encountered: