[R-package] init_model should accept booster

[sebffischer]

Summary

It would be great if one could pass the lgb.Booster directly as init_model and not have to save it to disk first.

[jameslamb]

Thanks for your interest in LightGBM!

It is possible to pass an lgb.Booster model into lgb.train() as init_model without saving it to disk.

The following minimal example works on the latest released version (v3.3.2) and on the latest development version (44fe591).

library(lightgbm)

data(agaricus.train, package = "lightgbm")

dtrain <- lgb.Dataset(
    agaricus.train$data
    , label = agaricus.train$label
    , free_raw_data = FALSE
)

initial_model <- lightgbm::lgb.train(
    data = dtrain
    , params = list(
        objective = "binary"
        , num_iterations = 10L
    )
)

new_model <- lightgbm::lgb.train(
    init_model = initial_model
    , data = dtrain
    , params = list(
        objective = "binary"
        , num_iterations = 17
    )
)
new_model$current_iter()
# 27

In the future, when you open issues here please provide the information asked for in the issue template. Including:

• version of LightGBM + how you installed it
• version of R
• minimal, reproducible example demonstrating specifically what you are asking about.

[sebffischer]

Thanks, what threw me off here were the docs (version 3.2.2)

init_model: path of model file of ‘lgb.Booster’ object, will continue
          training from this model

[jameslamb]

ha! that should say "OR" lgb.Booster not "OF" .

Would you like to open a pull request fixing that in the documentation?

[sebffischer]

Sure, will do it tomorrow :)

[github-actions]

This issue has been automatically locked since there has not been any recent activity since it was closed.
To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues
including a reference to this.