R/TrainResult.R

In mlr-org/mlrng: Machine Learning in R - Next Generation

#' @include Result.R
#' @title TrainResult
#' @format [R6Class()] object
#'
#' @description
#' A [R6::R6Class()] containing data of a [train()].
#'
#' @field task ([Task()]): Task used to fit the model.
#' @field learner ([Learner()]): Learner used to fit the model.
#' @field wrapped.model [any]: Result of the model fit as returned by third party packages.
#' @field train (`integer`): Indices of training data used to fit the model.
#' @field train.log ([TrainLog()]): Logging information from model fit.
#' @field train.success (`logical(1)`): Was the training sucessfull.
#'  Depending on the settings of `mlrng.continue.on.train.error` this can still be a valid model, but it uses a dummy fallback learner.
TrainResult = R6Class("TrainResult",
  inherit = Result,
  cloneable = FALSE,
  public = list(
    initialize = function(task, learner, rmodel, train.set, train.log) {
      super$initialize(data.table(
        task = list(assertTask(task)),
        learner = list(assertLearner(learner)),
        rmodel = list(rmodel),
        train.set = list(assertIndexSet(train.set, for.task = task)),
        train.log = list(assertR6(train.log, "TrainLog"))
      ))
    },

    print = function(...) {
      gcat("Training result of {self$learner$id} on {self$task$id}.")
      gcat("Training took {pretty_sec(self$train.log$train.time)}.")
      if (!self$train.success)
        gcat("Training failed with error {stri_peek(self$train.log$errors[[1]]$message)}.
              Model will output constant predictions from dummy learner.")
      if (getOption("mlrng.debug", FALSE))
        cat("\n", format(self), "\n")
    }
  ),

  active = list(
    task = function() self$data$task[[1L]],
    learner = function() self$data$learner[[1L]],
    rmodel = function() self$data$rmodel[[1L]],
    train.set = function() self$data$train.set[[1L]],
    train.log = function() self$data$train.log[[1L]],
    train.success = function() self$data$train.log[[1L]]$n.errors == 0L
  )
)