-
Notifications
You must be signed in to change notification settings - Fork 221
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Only warn for early stop if options.show_trace
#1049
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The decision is Patrick's but just showing my support.
Pinging @pkofod |
I'm wondering if this is the right approach. If you have |
I don’t want to skip all logging information, just the NaNs in the Optim loop. I want all logging info from any user-specified objective, or other optimization warnings, to be displayed. Throwing a warning for NaNs in all optimization loops is a subjective design decision. For some applications where NaNs are frequently observed (like my symbolic regression use cases), it results in a bad user experience with tons of irrelevant warnings being printed. So I’d like to be able to turn it off. |
Okay. I personally think the options struct is way overloaded as it is, but maybe it's better to introduce a |
Sounds good. I didn't want to add yet another option which is why I included it in |
Let me know what your decision is? |
Ping @pkofod. What do you prefer? |
Ping @pkofod. This is holding up SymbolicRegression.jl and PySR from upgrading. |
Okay I still haven't heard any response so I'm just going to submit a second PR with |
Ping @mohamed82008 @pkofod. Sorry for being annoying but I've been waiting for this since August and it is still blocking me from upgrading. |
went with #1058 |
Addresses discussion in #1046 with @mohamed82008
Users of PySR and SymbolicRegression.jl were suddenly getting spammed with warning messages (e.g., MilesCranmer/PySR#416) so I had to hotfix things to use an earlier Optim.jl.
This PR addresses this so that the warning only shows up if the user requested that level of information with
show_trace = true
. Otherwise the warning does not show. But it will still quit early which I agree is the correct behavior.