Hello,
I've used this package with some success to model hundreds of time series so thanks for the effort !
I'd like to use it in production but I've run into some issues related to the size of the model. One model currently takes 300mb with the maximum compression using joblib ! This would take more than 150Gb to save all the models which isn't tractable.
Do you have any tricks to reduce the disk footprint ?
Hello, I've used this package with some success to model hundreds of time series so thanks for the effort ! I'd like to use it in production but I've run into some issues related to the size of the model. One model currently takes 300mb with the maximum compression using joblib ! This would take more than 150Gb to save all the models which isn't tractable.
Do you have any tricks to reduce the disk footprint ?