Anyone know why AutoRandomForest gets stuck after ...
# mlforecastb
Bersu T
03/26/2025, 9:39 AMAnyone know why AutoRandomForest gets stuck after three trials while other models work fine?
j
jan rathfelder
03/26/2025, 10:39 AMah, you encounter so many problems. Not so sure what is going on here. There is no warning or error message, right? I think the best thing you can do is try to use kind of new versions of all relevant libraries. Optuna, mlforecast, randomforecast. maybe check on pypi how old your versions are
b
Bersu T
03/26/2025, 11:00 AMNope, no warning. It's just stuck. All libraries are updated 🙂
j
jan rathfelder
03/26/2025, 11:03 AMpretty hard to tell what is going on. uninstall and re-install optuna and randomforecast maybe?
jan rathfelder
03/26/2025, 11:04 AMor maybe even mlforecast. beause while mlforecast is super handy in building models, when it comes to debugging what is going on under the hood, it is not so easy, because now you have another layer of complexity (code) on top of optuna and randomforest. maybe some nixtla guys can also jump in here :))
j
José Morales
03/26/2025, 3:43 PMrandom forest is very slow and that's a very expensive combination (900 trees with depth 9), so it's probably fitting, not stuck
j
jan rathfelder
03/26/2025, 3:49 PMah, i didnt check the params. truee RF can be really slow.
b
Bersu T
03/27/2025, 8:49 AMHi, it had been running for 400 minutes after trial 2, so I am not sure whether it is fitting.
Bersu T
03/27/2025, 8:49 AMThe other models do not even take a minute maximum for a trial
j
jan rathfelder
03/27/2025, 8:59 AMi thought about it again: 2 ids, 900 trees and depth=9 should finish fast tbh. 900 trees is not much for RF, depth 9 is a bit expensive, but still should be easy for 2 ids. So there for sure is going on something weird. I mean unless your ids dont have daily data for 1000 years or so...
6 Views