[R-package] session crashes when serializing and deserializing model

[verajosemanuel]

I trained a model in R using LightGBM (treesnip) and saved using common saveRDS command.

To my surprise, when trying to predict some data after reading the saved model, readRDS seems to work but script&predict is crashing.

Read this issue #4208 regarding some similar issue but I am not sure that is the cause.

is there any safe procedure to save my model fot later use in R?

regards

[jameslamb]

It is hard to say for sure since you haven't provided a reproducible example showing exactly what you tried, but based on the information you've provided, #4208 probably describe problems you're facing using saveRDS() / readRDS(). I recommend subscribing to #4208 to be sure you're notified once that issue is resolved.

Until then, you can use the functions lightgbm::saveRDS.lgb.Booster() and lightgbm::readRDS.lgb.Booster() to save a {lightgbm} model object.

You can try these by using the following code:

library(lightgbm)

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

dtrain <- lightgbm::lgb.Dataset(
    agaricus.train$data
    , label = agaricus.train$label
)
bst <- lightgbm::lgb.train(
    params = list(
        "learning_rate" = 0.05
    )
    , data = dtrain
    , nrounds = 10L
    , obj = "binary"
)

model_file <- "test-model.rds"
saveRDS.lgb.Booster(
    object = bst
    , file = model_file
)

rm(list = ls())

After running that code, restart your R session and re-load the model. You should see that you're able to successfully predict on new data.

library(lightgbm)

model_file <- "test-model.rds"

data(agaricus.test, package = "lightgbm")
bst <- lightgbm::readRDS.lgb.Booster(
    file = model_file
)

preds <- predict(bst, data = agaricus.test$data)
print(preds)

For more, see:

• ?lightgbm::readRDS.lgb.Booster
• unit tests on model serialization and deserialization:

  LightGBM/R-package/tests/testthat/test_lgb.Booster.R

  Line 831 in 53ffba7

  context("saveRDS.lgb.Booster() and readRDS.lgb.Booster()")

NOTE: I'm not familiar with how {treesnip} works. If it returns its own object which wraps a {lightgbm} Booster, then until #4208 is resolved you will have to extract the Booster from that {treesnip}-specific object to use the example code above.

[verajosemanuel]

that's correct! just the issue I was having. Thanks for the workaround! Sadly, the code you provided crashed my R environment

[jameslamb]

Sadly, the code you provided crashed my R environment

How did you install {lightgbm} and what version of it are you using? As I mentioned in #4007 (comment), there have been several stability fixes to the R package that have not yet been released.

If you need further help, please provide a reproducible example.

[no-response]

This issue has been automatically closed because it has been awaiting a response for too long. When you have time to to work with the maintainers to resolve this issue, please post a new comment and it will be re-opened. If the issue has been locked for editing by the time you return to it, please open a new issue and reference this one. Thank you for taking the time to improve LightGBM!

[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.