kwargs from the lightgbm predict are not available with the LightGBMLSS.predict

[EchoDel]

For model governance reason I want to be able to run tree by tree using the start_iteration and num_iteration arguments to see if there is a number of over fitted trees towards the end of the training.

Would it be possible to add in the **kwargs to the predict all the way through to the lightgbm.predict.

[StatMixedML]

@EchoDel Thanks for raising this issue.

I would rather suggest to add an additional function that allows to return the trained lgb.Booster so that you can use it for further analysis. What do you think?

[EchoDel]

I'm looking for the parameters after the transformations and the probabilities so to avoid duplicating the code in the distribution class I would prefer either specifying the kwargs or being able to pass the raw predictions in to the distribution class to apply the transformations.

[StatMixedML]

Yeah, you are probably correct. Would be good to have the kwargs available. Not sure though when I find the time to look into it.

[EchoDel]

So I can work through it and submit and PR which you can merge after the tests have been added but thinking about if there is a better way to do this.
Does it make sense to change the DistributionClass.predict_dist to take a predictions argument instead of the booster and test set, then moving the booster.predict into the LightGBMLSS.predict?
Again I can work through this and submit a PR if you are happy with this change.

[StatMixedML]

Sorry was busy with the recent releases. Yes makes sense, thank you. Happy if you create a first draft via a PR