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
At the time of this writing, the current latest stable CPython version is 3.13. However, the SMAC PyPI package only lists support for versions up to 3.10. For my and possibly others' use cases, it would be nice to be able to use SMAC with more recent versions for a few reasons like:
support for newer language features, and not having to either give them up or maintain separate code versions
performance gains in the target code for optimization (here, an alternative would be to run the target code with a different interpreter, which can be achieved by using the CLI)
easier deployment on various OSs
Now, I can see there's a note in #1116 about this, so I assume the problem is probably generated by compatibility with pyrfr via swig (that issue says "(see Eddies comment Feb 23rd in SMAC3)", but I couldn't find them :) ), so I guess the work is more complex, as I imagine it would probably entail either making, adding and maintaining some new pre-compiled binaries for rfr/pyrfr, or dropping it all together and replacing it, which in itself is a bag of snakes, because that's what benchmarking the performance of machine learning algorithm implementations usually is. I just wanted to leave this issue out here to show there's some interest for it. So no pressure, and thanks for your hard work :).
The text was updated successfully, but these errors were encountered:
Hi Bogdan,
thanks for your suggestion.
We are looking into supporting newer python versions.
I sometimes run SMAC in python 3.12 without any problems.
We are also in the process of replacing the random forest, however this might take a while until we properly benchmarked the new version.
Best
At the time of this writing, the current latest stable CPython version is 3.13. However, the SMAC PyPI package only lists support for versions up to 3.10. For my and possibly others' use cases, it would be nice to be able to use SMAC with more recent versions for a few reasons like:
Now, I can see there's a note in #1116 about this, so I assume the problem is probably generated by compatibility with
pyrfr
viaswig
(that issue says "(see Eddies comment Feb 23rd in SMAC3)", but I couldn't find them :) ), so I guess the work is more complex, as I imagine it would probably entail either making, adding and maintaining some new pre-compiled binaries forrfr
/pyrfr
, or dropping it all together and replacing it, which in itself is a bag of snakes, because that's what benchmarking the performance of machine learning algorithm implementations usually is. I just wanted to leave this issue out here to show there's some interest for it. So no pressure, and thanks for your hard work :).The text was updated successfully, but these errors were encountered: