Hello everyone. I have a question. I would like to work on a cross validation in which only certain weeks are being tested, The process could be: 1. Train up to week1 test 1 2. Reuse the model and add up to week2 data and test on week 2 3. and continue. I know in neuralforecast I can just set

use_init_models

to

false

and call the cross_validation with the different datasets. Is there a native solution in MlForecast to do so? My data is rather massive and I am running the experiments on a very expensive machine, so a full retrain does not seems as a solution. Thanks

Hey. You can set

refit=False

that'll just train the model at the first fold