-
Notifications
You must be signed in to change notification settings - Fork 71
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
[MLForecast] Add the possibility to pass custom parameters to the fit function #322
Comments
Hey @helderPereira22, thanks for using mlforecast. I believe you can achieve that following this guide. Please let us know if that doesn't work for you. |
Hello @jmoralez! I think it works! But do I lose any of the Nixtla capabilities by doing this? |
All we do with the trained models is call predict, so it should work exactly in the same way. |
Hello @jmoralez, For instance, adopting the custom training approach you recommended means I wouldn't have the ability to employ Conformal Prediction for creating prediction intervals. Therefore, it seems I would be sacrificing this feature by not applying the .fit method provided by Nixtla. Could you take this into account? |
The intervals are created by performing cross validation, so if you were to set the early stopping rounds you'd end up with a potentially different number of iterations in each fold. My suggestion is to run it once (maybe with all of your data), check which iteration was the best and then fix that value in the catboost constructor and use the regular MLForecast.fit, that way when computing the intervals each fold will use the same number of iterations. |
This issue has been automatically closed because it has been awaiting a response for too long. When you have time to to work with the maintainers to resolve this issue, please post a new comment and it will be re-opened. If the issue has been locked for editing by the time you return to it, please open a new issue and reference this one. |
Description
Enable the ability to pass custom parameters (for instance with kwargs) for the fit function of the model
Use case
For instance, to use the Overfitting detector of the Catboost algorithm (where we pass 'early_stopping_rounds' to the fit function)
The text was updated successfully, but these errors were encountered: