Open FBruzzesi opened 1 year ago
I'm open to it, but the maintainers of the project are spread a bit thin (I blame the baby). So it might take a while before we're able to get back to you.
Thank you Vincent for getting back to me. I can barely imagine the life with a baby 😁. As I needed that with a certain urgency I started to work towards a completely independent solution. I will calmly try to add the feature here as well!
In the current implementation of
TimeGapSplit
the argumentvalidation_duration
is used as stride as well.This leads to the impossibility of having overlapping validation periods. What about the possibility of adding a stride parameter to indicate the time shift for the training period?
The concrete use case is as follows: I would like to validate a model that retrains every week but forecast for multiple weeks ahead.
Happy to give it a try in the next following days if there is interest in adding this feature.