[R-package] lightgbm() cannot evaluate metrics on validation sets

[jameslamb]

Evaluation on validation sets using the function lightgbm() is currently broken. The documentation suggests that you should be able to pass valids to this function similar to the way it is passed to lgb.train().

I discovered this working on #2912. As long as this bug exists, early stopping with the lightgbm() function is also broken.

Environment info

Operating System: macOS Mojave 10.14.6

CPU/GPU model: 2.3 GHz Intel Core i5

C++/Python/R version: R 3.6.1

LightGBM version or commit hash: 4aa888c

Error message

Running the code in "Reproducible examples" yields the following error:

Error in lgb.train(params = params, data = dtrain, nrounds = nrounds, :
formal argument "valids" matched by multiple actual arguments

Reproducible examples

data(agaricus.train, package = "lightgbm")
data(agaricus.test, package = "lightgbm")
train <- agaricus.train
test <- agaricus.test
dvalid1 <- lgb.Dataset(
    data = train$data
    , labels = train$label
  )
  dvalid2 <- lgb.Dataset(
    data = train$data
    , labels = train$label
  )
  nrounds <- 10L
  bst <- lightgbm(
    data = train$data
    , label = train$label
    , num_leaves = 5L
    , nrounds = nrounds
    , objective = "binary"
    , metric = "binary_error"
    , valids = list(
      "valid1" = dvalid1
      , "valid2" = dvalid2
    )
  )