test_lgb.Booster.R 42 KB
Newer Older
1
2
3
4
VERBOSITY <- as.integer(
  Sys.getenv("LIGHTGBM_TEST_VERBOSITY", "-1")
)

5
ON_WINDOWS <- .Platform$OS.type == "windows"
6
7
TOLERANCE <- 1e-6

8
9
10
11
12
13
test_that("Booster$finalize() should not fail", {
    X <- as.matrix(as.integer(iris[, "Species"]), ncol = 1L)
    y <- iris[["Sepal.Length"]]
    dtrain <- lgb.Dataset(X, label = y)
    bst <- lgb.train(
        data = dtrain
14
15
16
        , params = list(
            objective = "regression"
        )
17
        , verbose = VERBOSITY
18
19
20
21
22
23
24
25
26
27
28
29
30
31
        , nrounds = 3L
    )
    expect_true(lgb.is.Booster(bst))

    expect_false(lgb.is.null.handle(bst$.__enclos_env__$private$handle))

    bst$finalize()
    expect_true(lgb.is.null.handle(bst$.__enclos_env__$private$handle))

    # calling finalize() a second time shouldn't cause any issues
    bst$finalize()
    expect_true(lgb.is.null.handle(bst$.__enclos_env__$private$handle))
})

32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
test_that("lgb.get.eval.result() should throw an informative error if booster is not an lgb.Booster", {
    bad_inputs <- list(
        matrix(1.0:10.0, 2L, 5L)
        , TRUE
        , c("a", "b")
        , NA
        , 10L
        , lgb.Dataset(
            data = matrix(1.0:10.0, 2L, 5L)
            , params = list()
        )
    )
    for (bad_input in bad_inputs) {
        expect_error({
            lgb.get.eval.result(
                booster = bad_input
                , data_name = "test"
                , eval_name = "l2"
            )
        }, regexp = "Can only use", fixed = TRUE)
    }
})

test_that("lgb.get.eval.result() should throw an informative error for incorrect data_name", {
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    dtrain <- lgb.Dataset(
        agaricus.train$data
        , label = agaricus.train$label
    )
    model <- lgb.train(
        params = list(
            objective = "regression"
            , metric = "l2"
66
67
            , min_data = 1L
            , learning_rate = 1.0
68
            , verbose = VERBOSITY
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
        )
        , data = dtrain
        , nrounds = 5L
        , valids = list(
            "test" = lgb.Dataset.create.valid(
                dtrain
                , agaricus.test$data
                , label = agaricus.test$label
            )
        )
    )
    expect_error({
        eval_results <- lgb.get.eval.result(
            booster = model
            , data_name = "testing"
            , eval_name = "l2"
        )
    }, regexp = "Only the following datasets exist in record evals: [test]", fixed = TRUE)
})

test_that("lgb.get.eval.result() should throw an informative error for incorrect eval_name", {
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    dtrain <- lgb.Dataset(
        agaricus.train$data
        , label = agaricus.train$label
    )
    model <- lgb.train(
        params = list(
            objective = "regression"
            , metric = "l2"
100
101
            , min_data = 1L
            , learning_rate = 1.0
102
            , verbose = VERBOSITY
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
        )
        , data = dtrain
        , nrounds = 5L
        , valids = list(
            "test" = lgb.Dataset.create.valid(
                dtrain
                , agaricus.test$data
                , label = agaricus.test$label
            )
        )
    )
    expect_error({
        eval_results <- lgb.get.eval.result(
            booster = model
            , data_name = "test"
            , eval_name = "l1"
        )
120
    }, regexp = "Only the following eval_names exist for dataset.*\\: \\[l2\\]", fixed = FALSE)
121
})
122
123
124
125
126
127
128
129
130
131

test_that("lgb.load() gives the expected error messages given different incorrect inputs", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    train <- agaricus.train
    test <- agaricus.test
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
132
133
134
135
        , params = list(
            objective = "binary"
            , num_leaves = 4L
            , learning_rate = 1.0
136
            , verbose = VERBOSITY
137
        )
138
139
140
141
142
143
144
145
146
147
148
149
        , nrounds = 2L
    )

    # you have to give model_str or filename
    expect_error({
        lgb.load()
    }, regexp = "either filename or model_str must be given")
    expect_error({
        lgb.load(filename = NULL, model_str = NULL)
    }, regexp = "either filename or model_str must be given")

    # if given, filename should be a string that points to an existing file
150
    model_file <- tempfile(fileext = ".model")
151
    expect_error({
152
        lgb.load(filename = list(model_file))
153
154
155
156
157
158
159
160
161
162
163
164
    }, regexp = "filename should be character")
    file_to_check <- paste0("a.model")
    while (file.exists(file_to_check)) {
        file_to_check <- paste0("a", file_to_check)
    }
    expect_error({
        lgb.load(filename = file_to_check)
    }, regexp = "passed to filename does not exist")

    # if given, model_str should be a string
    expect_error({
        lgb.load(model_str = c(4.0, 5.0, 6.0))
165
    }, regexp = "lgb.load: model_str should be a character/raw vector")
166
167
168

})

169
test_that("Loading a Booster from a text file works", {
170
171
172
173
174
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    train <- agaricus.train
    test <- agaricus.test
175
176
177
178
179
180
181
182
183
184
185
186
187
188
    params <- list(
        num_leaves = 4L
        , boosting = "rf"
        , bagging_fraction = 0.8
        , bagging_freq = 1L
        , boost_from_average = FALSE
        , categorical_feature = c(1L, 2L)
        , interaction_constraints = list(c(1L, 2L), 1L)
        , feature_contri = rep(0.5, ncol(train$data))
        , metric = c("mape", "average_precision")
        , learning_rate = 1.0
        , objective = "binary"
        , verbosity = VERBOSITY
    )
189
190
191
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
192
        , params = params
193
194
195
196
197
        , nrounds = 2L
    )
    expect_true(lgb.is.Booster(bst))

    pred <- predict(bst, test$data)
198
199
    model_file <- tempfile(fileext = ".model")
    lgb.save(bst, model_file)
200
201
202
203
204
205
206

    # finalize the booster and destroy it so you know we aren't cheating
    bst$finalize()
    expect_null(bst$.__enclos_env__$private$handle)
    rm(bst)

    bst2 <- lgb.load(
207
        filename = model_file
208
209
210
    )
    pred2 <- predict(bst2, test$data)
    expect_identical(pred, pred2)
211
212
213

    # check that the parameters are loaded correctly
    expect_equal(bst2$params[names(params)], params)
214
215
})

216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
test_that("boosters with linear models at leaves can be written to text file and re-loaded successfully", {
    X <- matrix(rnorm(100L), ncol = 1L)
    labels <- 2L * X + runif(nrow(X), 0L, 0.1)
    dtrain <- lgb.Dataset(
        data = X
        , label = labels
    )

    params <- list(
        objective = "regression"
        , verbose = -1L
        , metric = "mse"
        , seed = 0L
        , num_leaves = 2L
    )

    bst <- lgb.train(
        data = dtrain
        , nrounds = 10L
        , params = params
236
        , verbose = VERBOSITY
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
    )
    expect_true(lgb.is.Booster(bst))

    # save predictions, then write the model to a file and destroy it in R
    preds <- predict(bst, X)
    model_file <- tempfile(fileext = ".model")
    lgb.save(bst, model_file)
    bst$finalize()
    expect_null(bst$.__enclos_env__$private$handle)
    rm(bst)

    # load the booster and make predictions...should be the same
    bst2 <- lgb.load(
        filename = model_file
    )
252
253
    preds2 <- predict(bst2, X)
    expect_identical(preds, preds2)
254
255
256
})


257
258
259
260
261
262
263
264
265
test_that("Loading a Booster from a string works", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    train <- agaricus.train
    test <- agaricus.test
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
266
267
268
269
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
270
            , verbose = VERBOSITY
271
        )
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
        , nrounds = 2L
    )
    expect_true(lgb.is.Booster(bst))

    pred <- predict(bst, test$data)
    model_string <- bst$save_model_to_string()

    # finalize the booster and destroy it so you know we aren't cheating
    bst$finalize()
    expect_null(bst$.__enclos_env__$private$handle)
    rm(bst)

    bst2 <- lgb.load(
        model_str = model_string
    )
    pred2 <- predict(bst2, test$data)
    expect_identical(pred, pred2)
})

291
292
293
294
295
296
297
test_that("Saving a large model to string should work", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    train <- agaricus.train
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
298
299
300
301
302
        , params = list(
            num_leaves = 100L
            , learning_rate = 0.01
            , objective = "binary"
        )
303
        , nrounds = 500L
304
        , verbose = VERBOSITY
305
306
307
    )

    pred <- predict(bst, train$data)
308
309
    pred_leaf_indx <- predict(bst, train$data, type = "leaf")
    pred_raw_score <- predict(bst, train$data, type = "raw")
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
    model_string <- bst$save_model_to_string()

    # make sure this test is still producing a model bigger than the default
    # buffer size used in LGBM_BoosterSaveModelToString_R
    expect_gt(nchar(model_string), 1024L * 1024L)

    # finalize the booster and destroy it so you know we aren't cheating
    bst$finalize()
    expect_null(bst$.__enclos_env__$private$handle)
    rm(bst)

    # make sure a new model can be created from this string, and that it
    # produces expected results
    bst2 <- lgb.load(
        model_str = model_string
    )
    pred2 <- predict(bst2, train$data)
327
328
    pred2_leaf_indx <- predict(bst2, train$data, type = "leaf")
    pred2_raw_score <- predict(bst2, train$data, type = "raw")
329
330
331
332
333
334
335
336
337
338
339
340
    expect_identical(pred, pred2)
    expect_identical(pred_leaf_indx, pred2_leaf_indx)
    expect_identical(pred_raw_score, pred2_raw_score)
})

test_that("Saving a large model to JSON should work", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    train <- agaricus.train
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
341
342
343
344
345
        , params = list(
            num_leaves = 100L
            , learning_rate = 0.01
            , objective = "binary"
        )
346
        , nrounds = 200L
347
        , verbose = VERBOSITY
348
349
350
351
352
353
354
355
356
357
358
359
360
    )

    model_json <- bst$dump_model()

    # make sure this test is still producing a model bigger than the default
    # buffer size used in LGBM_BoosterDumpModel_R
    expect_gt(nchar(model_json), 1024L * 1024L)

    # check that it is valid JSON that looks like a LightGBM model
    model_list <- jsonlite::fromJSON(model_json)
    expect_equal(model_list[["objective"]], "binary sigmoid:1")
})

361
362
363
364
365
366
367
368
369
test_that("If a string and a file are both passed to lgb.load() the file is used model_str is totally ignored", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    train <- agaricus.train
    test <- agaricus.test
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
370
371
372
373
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
374
            , verbose = VERBOSITY
375
        )
376
377
378
379
380
        , nrounds = 2L
    )
    expect_true(lgb.is.Booster(bst))

    pred <- predict(bst, test$data)
381
382
    model_file <- tempfile(fileext = ".model")
    lgb.save(bst, model_file)
383
384
385
386
387
388
389

    # finalize the booster and destroy it so you know we aren't cheating
    bst$finalize()
    expect_null(bst$.__enclos_env__$private$handle)
    rm(bst)

    bst2 <- lgb.load(
390
        filename = model_file
391
392
393
394
395
        , model_str = 4.0
    )
    pred2 <- predict(bst2, test$data)
    expect_identical(pred, pred2)
})
396
397
398
399
400
401
402
403
404
405
406
407

test_that("Creating a Booster from a Dataset should work", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    dtrain <- lgb.Dataset(
        agaricus.train$data
        , label = agaricus.train$label
    )
    bst <- Booster$new(
        params = list(
            objective = "binary"
408
            , verbose = VERBOSITY
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
        ),
        train_set = dtrain
    )
    expect_true(lgb.is.Booster(bst))
    expect_equal(bst$current_iter(), 0L)
    expect_true(is.na(bst$best_score))
    expect_true(all(bst$predict(agaricus.train$data) == 0.5))
})

test_that("Creating a Booster from a Dataset with an existing predictor should work", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    nrounds <- 2L
    bst <- lightgbm(
        data = as.matrix(agaricus.train$data)
        , label = agaricus.train$label
425
426
427
428
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
429
            , verbose = VERBOSITY
430
        )
431
432
433
434
435
436
437
438
439
440
        , nrounds = nrounds
    )
    data(agaricus.test, package = "lightgbm")
    dtest <- Dataset$new(
        data = agaricus.test$data
        , label = agaricus.test$label
        , predictor = bst$to_predictor()
    )
    bst_from_ds <- Booster$new(
        train_set = dtest
441
442
443
        , params = list(
            verbose = VERBOSITY
        )
444
445
446
447
    )
    expect_true(lgb.is.Booster(bst))
    expect_equal(bst$current_iter(), nrounds)
    expect_equal(bst$eval_train()[[1L]][["value"]], 0.1115352)
448
    expect_true(lgb.is.Booster(bst_from_ds))
449
    expect_equal(bst_from_ds$current_iter(), nrounds)
450
    expect_equal(bst_from_ds$eval_train()[[1L]][["value"]], 5.65704892)
451
452
    dumped_model <- jsonlite::fromJSON(bst$dump_model())
})
453

454
455
456
457
458
459
460
461
462
463
464
test_that("Booster$eval() should work on a Dataset stored in a binary file", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    train <- agaricus.train
    dtrain <- lgb.Dataset(train$data, label = train$label)

    bst <- lgb.train(
        params = list(
            objective = "regression"
            , metric = "l2"
            , num_leaves = 4L
465
            , verbose = VERBOSITY
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
        )
        , data = dtrain
        , nrounds = 2L
    )

    data(agaricus.test, package = "lightgbm")
    test <- agaricus.test
    dtest <- lgb.Dataset.create.valid(
        dataset = dtrain
        , data = test$data
        , label = test$label
    )
    dtest$construct()

    eval_in_mem <- bst$eval(
        data = dtest
        , name = "test"
    )

    test_file <- tempfile(pattern = "lgb.Dataset_")
    lgb.Dataset.save(
        dataset = dtest
        , fname = test_file
    )
    rm(dtest)

    eval_from_file <- bst$eval(
        data = lgb.Dataset(
            data = test_file
495
            , params = list(verbose = VERBOSITY)
496
497
498
499
500
        )$construct()
        , name = "test"
    )

    expect_true(abs(eval_in_mem[[1L]][["value"]] - 0.1744423) < TOLERANCE)
501
    # refer to https://github.com/microsoft/LightGBM/issues/4680
502
503
504
505
506
    if (isTRUE(ON_WINDOWS)) {
      expect_equal(eval_in_mem, eval_from_file)
    } else {
      expect_identical(eval_in_mem, eval_from_file)
    }
507
508
})

509
510
511
512
513
514
515
516
517
518
test_that("Booster$rollback_one_iter() should work as expected", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    data(agaricus.test, package = "lightgbm")
    train <- agaricus.train
    test <- agaricus.test
    nrounds <- 5L
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
519
520
521
522
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
523
            , verbose = VERBOSITY
524
        )
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
        , nrounds = nrounds
    )
    expect_equal(bst$current_iter(), nrounds)
    expect_true(lgb.is.Booster(bst))
    logloss <- bst$eval_train()[[1L]][["value"]]
    expect_equal(logloss, 0.01904786)

    x <- bst$rollback_one_iter()

    # rollback_one_iter() should return a booster and modify the original
    # booster in place
    expect_true(lgb.is.Booster(x))
    expect_equal(bst$current_iter(), nrounds - 1L)

    # score should now come from the model as of 4 iterations
    logloss <- bst$eval_train()[[1L]][["value"]]
    expect_equal(logloss, 0.027915146)
})
543
544
545
546
547
548
549
550
551
552

test_that("Booster$update() passing a train_set works as expected", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    nrounds <- 2L

    # train with 2 rounds and then update
    bst <- lightgbm(
        data = as.matrix(agaricus.train$data)
        , label = agaricus.train$label
553
554
555
556
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
557
            , verbose = VERBOSITY
558
        )
559
560
561
562
563
564
565
566
        , nrounds = nrounds
    )
    expect_true(lgb.is.Booster(bst))
    expect_equal(bst$current_iter(), nrounds)
    bst$update(
        train_set = Dataset$new(
            data = agaricus.train$data
            , label = agaricus.train$label
567
            , params = list(verbose = VERBOSITY)
568
569
570
571
572
        )
    )
    expect_true(lgb.is.Booster(bst))
    expect_equal(bst$current_iter(), nrounds + 1L)

573
    # train with 3 rounds directly
574
575
576
    bst2 <- lightgbm(
        data = as.matrix(agaricus.train$data)
        , label = agaricus.train$label
577
578
579
580
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
581
            , verbose = VERBOSITY
582
        )
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
        , nrounds = nrounds +  1L
    )
    expect_true(lgb.is.Booster(bst2))
    expect_equal(bst2$current_iter(), nrounds +  1L)

    # model with 2 rounds + 1 update should be identical to 3 rounds
    expect_equal(bst2$eval_train()[[1L]][["value"]], 0.04806585)
    expect_equal(bst$eval_train()[[1L]][["value"]], bst2$eval_train()[[1L]][["value"]])
})

test_that("Booster$update() throws an informative error if you provide a non-Dataset to update()", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    nrounds <- 2L

    # train with 2 rounds and then update
    bst <- lightgbm(
        data = as.matrix(agaricus.train$data)
        , label = agaricus.train$label
602
603
604
605
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
606
            , verbose = VERBOSITY
607
        )
608
609
610
611
612
613
614
615
        , nrounds = nrounds
    )
    expect_error({
        bst$update(
            train_set = data.frame(x = rnorm(10L))
        )
    }, regexp = "lgb.Booster.update: Only can use lgb.Dataset", fixed = TRUE)
})
616

617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
test_that("Booster should store parameters and Booster$reset_parameter() should update them", {
    data(agaricus.train, package = "lightgbm")
    dtrain <- lgb.Dataset(
        agaricus.train$data
        , label = agaricus.train$label
    )
    # testing that this works for some cases that could break it:
    #    - multiple metrics
    #    - using "metric", "boosting", "num_class" in params
    params <- list(
        objective = "multiclass"
        , max_depth = 4L
        , bagging_fraction = 0.8
        , metric = c("multi_logloss", "multi_error")
        , boosting = "gbdt"
        , num_class = 5L
633
        , verbose = VERBOSITY
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
    )
    bst <- Booster$new(
        params = params
        , train_set = dtrain
    )
    expect_identical(bst$params, params)

    params[["bagging_fraction"]] <- 0.9
    ret_bst <- bst$reset_parameter(params = params)
    expect_identical(ret_bst$params, params)
    expect_identical(bst$params, params)
})

test_that("Booster$params should include dataset params, before and after Booster$reset_parameter()", {
    data(agaricus.train, package = "lightgbm")
    dtrain <- lgb.Dataset(
        agaricus.train$data
        , label = agaricus.train$label
        , params = list(
            max_bin = 17L
        )
    )
    params <- list(
        objective = "binary"
        , max_depth = 4L
        , bagging_fraction = 0.8
660
        , verbose = VERBOSITY
661
662
663
664
665
666
667
668
669
670
671
    )
    bst <- Booster$new(
        params = params
        , train_set = dtrain
    )
    expect_identical(
        bst$params
        , list(
            objective = "binary"
            , max_depth = 4L
            , bagging_fraction = 0.8
672
            , verbose = VERBOSITY
673
674
675
676
677
678
679
680
681
682
            , max_bin = 17L
        )
    )

    params[["bagging_fraction"]] <- 0.9
    ret_bst <- bst$reset_parameter(params = params)
    expected_params <- list(
        objective = "binary"
        , max_depth = 4L
        , bagging_fraction = 0.9
683
        , verbose = VERBOSITY
684
685
686
687
688
689
        , max_bin = 17L
    )
    expect_identical(ret_bst$params, expected_params)
    expect_identical(bst$params, expected_params)
})

690
691
692
693
694
695
696
test_that("Saving a model with different feature importance types works", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    train <- agaricus.train
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
697
698
699
700
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
701
            , verbose = VERBOSITY
702
        )
703
704
705
706
707
        , nrounds = 2L
    )
    expect_true(lgb.is.Booster(bst))

    .feat_importance_from_string <- function(model_string) {
708
        file_lines <- strsplit(model_string, "\n", fixed = TRUE)[[1L]]
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
        start_indx <- which(grepl("^feature_importances\\:$", file_lines)) + 1L
        blank_line_indices <- which(file_lines == "")
        end_indx <- blank_line_indices[blank_line_indices > start_indx][1L] - 1L
        importances <- file_lines[start_indx: end_indx]
        return(importances)
    }

    GAIN_IMPORTANCE <- 1L
    model_string <- bst$save_model_to_string(feature_importance_type = GAIN_IMPORTANCE)
    expect_equal(
        .feat_importance_from_string(model_string)
        , c(
            "odor=none=4010"
            , "stalk-root=club=1163"
            , "stalk-root=rooted=573"
            , "stalk-surface-above-ring=silky=450"
            , "spore-print-color=green=397"
            , "gill-color=buff=281"
        )
    )

    SPLIT_IMPORTANCE <- 0L
    model_string <- bst$save_model_to_string(feature_importance_type = SPLIT_IMPORTANCE)
    expect_equal(
        .feat_importance_from_string(model_string)
        , c(
            "odor=none=1"
            , "gill-color=buff=1"
            , "stalk-root=club=1"
            , "stalk-root=rooted=1"
            , "stalk-surface-above-ring=silky=1"
            , "spore-print-color=green=1"
        )
    )
743
744
745
746
747
748
749
750
751
})

test_that("Saving a model with unknown importance type fails", {
    set.seed(708L)
    data(agaricus.train, package = "lightgbm")
    train <- agaricus.train
    bst <- lightgbm(
        data = as.matrix(train$data)
        , label = train$label
752
753
754
755
        , params = list(
            num_leaves = 4L
            , learning_rate = 1.0
            , objective = "binary"
756
            , verbose = VERBOSITY
757
        )
758
759
760
        , nrounds = 2L
    )
    expect_true(lgb.is.Booster(bst))
761
762
763

    UNSUPPORTED_IMPORTANCE <- 2L
    expect_error({
764
765
766
767
768
        capture.output({
          model_string <- bst$save_model_to_string(
            feature_importance_type = UNSUPPORTED_IMPORTANCE
          )
        }, type = "message")
769
770
    }, "Unknown importance type")
})
771

772

773
.params_from_model_string <- function(model_str) {
774
    file_lines <- strsplit(model_str, "\n", fixed = TRUE)[[1L]]
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
    start_indx <- which(grepl("^parameters\\:$", file_lines)) + 1L
    blank_line_indices <- which(file_lines == "")
    end_indx <- blank_line_indices[blank_line_indices > start_indx][1L] - 1L
    params <- file_lines[start_indx: end_indx]
    return(params)
}

test_that("all parameters are stored correctly with save_model_to_string()", {
    dtrain <- lgb.Dataset(
        data = matrix(rnorm(500L), nrow = 100L)
        , label = rnorm(100L)
    )
    nrounds <- 4L
    bst <- lgb.train(
        params = list(
            objective = "regression"
            , metric = "l2"
        )
        , data = dtrain
        , nrounds = nrounds
795
        , verbose = VERBOSITY
796
797
798
799
800
801
    )

    model_str <- bst$save_model_to_string()
    params_in_file <- .params_from_model_string(model_str = model_str)

    # parameters should match what was passed from the R package
802
    expect_equal(sum(startsWith(params_in_file, "[metric:")), 1L)
803
804
    expect_equal(sum(params_in_file == "[metric: l2]"), 1L)

805
    expect_equal(sum(startsWith(params_in_file, "[num_iterations:")), 1L)
806
807
    expect_equal(sum(params_in_file == "[num_iterations: 4]"), 1L)

808
    expect_equal(sum(startsWith(params_in_file, "[objective:")), 1L)
809
810
    expect_equal(sum(params_in_file == "[objective: regression]"), 1L)

811
    expect_equal(sum(startsWith(params_in_file, "[verbosity:")), 1L)
812
    expect_equal(sum(params_in_file == sprintf("[verbosity: %i]", VERBOSITY)), 1L)
813
814

    # early stopping should be off by default
815
    expect_equal(sum(startsWith(params_in_file, "[early_stopping_round:")), 1L)
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
    expect_equal(sum(params_in_file == "[early_stopping_round: 0]"), 1L)
})

test_that("early_stopping, num_iterations are stored correctly in model string even with aliases", {
    dtrain <- lgb.Dataset(
        data = matrix(rnorm(500L), nrow = 100L)
        , label = rnorm(100L)
    )
    dvalid <- lgb.Dataset(
        data = matrix(rnorm(500L), nrow = 100L)
        , label = rnorm(100L)
    )

    # num_iterations values (all different)
    num_iterations <- 4L
    num_boost_round <- 2L
    n_iter <- 3L
    nrounds_kwarg <- 6L

    # early_stopping_round values (all different)
    early_stopping_round <- 2L
    early_stopping_round_kwarg <- 3L
    n_iter_no_change <- 4L

    params <- list(
        objective = "regression"
        , metric = "l2"
        , num_boost_round = num_boost_round
        , num_iterations = num_iterations
        , n_iter = n_iter
        , early_stopping_round = early_stopping_round
        , n_iter_no_change = n_iter_no_change
    )

    bst <- lgb.train(
        params = params
        , data = dtrain
        , nrounds = nrounds_kwarg
        , early_stopping_rounds = early_stopping_round_kwarg
        , valids = list(
            "random_valid" = dvalid
        )
858
        , verbose = VERBOSITY
859
860
861
862
863
864
865
    )

    model_str <- bst$save_model_to_string()
    params_in_file <- .params_from_model_string(model_str = model_str)

    # parameters should match what was passed from the R package, and the "main" (non-alias)
    # params values in `params` should be preferred to keyword argumentts or aliases
866
    expect_equal(sum(startsWith(params_in_file, "[num_iterations:")), 1L)
867
    expect_equal(sum(params_in_file == sprintf("[num_iterations: %s]", num_iterations)), 1L)
868
    expect_equal(sum(startsWith(params_in_file, "[early_stopping_round:")), 1L)
869
870
871
    expect_equal(sum(params_in_file == sprintf("[early_stopping_round: %s]", early_stopping_round)), 1L)

    # none of the aliases shouold have been written to the model file
872
873
874
    expect_equal(sum(startsWith(params_in_file, "[num_boost_round:")), 0L)
    expect_equal(sum(startsWith(params_in_file, "[n_iter:")), 0L)
    expect_equal(sum(startsWith(params_in_file, "[n_iter_no_change:")), 0L)
875
876
877

})

878
879
880
881
882
883
884
885
886
887
888
test_that("Booster: method calls Booster with a null handle should raise an informative error and not segfault", {
    data(agaricus.train, package = "lightgbm")
    train <- agaricus.train
    dtrain <- lgb.Dataset(train$data, label = train$label)
    bst <- lgb.train(
        params = list(
            objective = "regression"
            , metric = "l2"
            , num_leaves = 8L
        )
        , data = dtrain
889
        , verbose = VERBOSITY
890
891
892
893
        , nrounds = 5L
        , valids = list(
            train = dtrain
        )
894
        , serializable = FALSE
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
    )
    tmp_file <- tempfile(fileext = ".rds")
    saveRDS(bst, tmp_file)
    rm(bst)
    bst <- readRDS(tmp_file)
    .expect_booster_error <- function(object) {
        error_regexp <- "Attempting to use a Booster which no longer exists"
        expect_error(object, regexp = error_regexp)
    }
    .expect_booster_error({
        bst$current_iter()
    })
    .expect_booster_error({
        bst$dump_model()
    })
    .expect_booster_error({
        bst$eval(data = dtrain, name = "valid")
    })
    .expect_booster_error({
        bst$eval_train()
    })
    .expect_booster_error({
        bst$lower_bound()
    })
    .expect_booster_error({
        bst$predict(data = train$data[seq_len(5L), ])
    })
    .expect_booster_error({
        bst$reset_parameter(params = list(learning_rate = 0.123))
    })
    .expect_booster_error({
        bst$rollback_one_iter()
    })
    .expect_booster_error({
929
        bst$save_raw()
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
    })
    .expect_booster_error({
        bst$save_model(filename = tempfile(fileext = ".model"))
    })
    .expect_booster_error({
        bst$save_model_to_string()
    })
    .expect_booster_error({
        bst$update()
    })
    .expect_booster_error({
        bst$upper_bound()
    })
    predictor <- bst$to_predictor()
    .expect_booster_error({
        predictor$current_iter()
    })
    .expect_booster_error({
        predictor$predict(data = train$data[seq_len(5L), ])
    })
})

test_that("Booster$new() using a Dataset with a null handle should raise an informative error and not segfault", {
    data(agaricus.train, package = "lightgbm")
    train <- agaricus.train
    dtrain <- lgb.Dataset(train$data, label = train$label)
    dtrain$construct()
    tmp_file <- tempfile(fileext = ".bin")
    saveRDS(dtrain, tmp_file)
    rm(dtrain)
    dtrain <- readRDS(tmp_file)
    expect_error({
962
963
964
965
966
967
        bst <- Booster$new(
            train_set = dtrain
            , params = list(
                verbose = VERBOSITY
            )
        )
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
    }, regexp = "Attempting to create a Dataset without any raw data")
})

test_that("Booster$new() raises informative errors for malformed inputs", {
  data(agaricus.train, package = "lightgbm")
  train <- agaricus.train
  dtrain <- lgb.Dataset(train$data, label = train$label)

  # no inputs
  expect_error({
    Booster$new()
  }, regexp = "lgb.Booster: Need at least either training dataset, model file, or model_str")

  # unrecognized objective
  expect_error({
983
984
985
986
987
988
    capture.output({
      Booster$new(
        params = list(objective = "not_a_real_objective")
        , train_set = dtrain
      )
    }, type = "message")
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
  }, regexp = "Unknown objective type name: not_a_real_objective")

  # train_set is not a Dataset
  expect_error({
    Booster$new(
      train_set = data.table::data.table(rnorm(1L:10L))
    )
  }, regexp = "lgb.Booster: Can only use lgb.Dataset as training data")

  # model file isn't a string
  expect_error({
    Booster$new(
      modelfile = list()
    )
  }, regexp = "lgb.Booster: Can only use a string as model file path")

  # model file doesn't exist
  expect_error({
1007
1008
1009
1010
1011
1012
    capture.output({
      Booster$new(
        params = list()
        , modelfile = "file-that-does-not-exist.model"
      )
    }, type = "message")
1013
1014
1015
1016
1017
1018
1019
1020
1021
  }, regexp = "Could not open file-that-does-not-exist.model")

  # model file doesn't contain a valid LightGBM model
  model_file <- tempfile(fileext = ".model")
  writeLines(
    text = c("make", "good", "predictions")
    , con = model_file
  )
  expect_error({
1022
1023
1024
1025
1026
1027
    capture.output({
      Booster$new(
        params = list()
        , modelfile = model_file
      )
    }, type = "message")
1028
1029
1030
1031
  }, regexp = "Unknown model format or submodel type in model file")

  # malformed model string
  expect_error({
1032
1033
1034
1035
1036
1037
    capture.output({
      Booster$new(
        params = list()
        , model_str = "a\nb\n"
      )
    }, type = "message")
1038
1039
1040
1041
1042
1043
1044
1045
  }, regexp = "Model file doesn't specify the number of classes")

  # model string isn't character or raw
  expect_error({
    Booster$new(
      model_str = numeric()
    )
  }, regexp = "lgb.Booster: Can only use a character/raw vector as model_str")
1046
1047
})

1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
# this is almost identical to the test above it, but for lgb.cv(). A lot of code
# is duplicated between lgb.train() and lgb.cv(), and this will catch cases where
# one is updated and the other isn't
test_that("lgb.cv() correctly handles passing through params to the model file", {
    dtrain <- lgb.Dataset(
        data = matrix(rnorm(500L), nrow = 100L)
        , label = rnorm(100L)
    )

    # num_iterations values (all different)
    num_iterations <- 4L
    num_boost_round <- 2L
    n_iter <- 3L
    nrounds_kwarg <- 6L

    # early_stopping_round values (all different)
    early_stopping_round <- 2L
    early_stopping_round_kwarg <- 3L
    n_iter_no_change <- 4L

    params <- list(
        objective = "regression"
        , metric = "l2"
        , num_boost_round = num_boost_round
        , num_iterations = num_iterations
        , n_iter = n_iter
        , early_stopping_round = early_stopping_round
        , n_iter_no_change = n_iter_no_change
1076
        , verbose = VERBOSITY
1077
1078
1079
1080
1081
1082
1083
1084
    )

    cv_bst <- lgb.cv(
        params = params
        , data = dtrain
        , nrounds = nrounds_kwarg
        , early_stopping_rounds = early_stopping_round_kwarg
        , nfold = 3L
1085
        , verbose = VERBOSITY
1086
1087
1088
1089
1090
1091
1092
1093
    )

    for (bst in cv_bst$boosters) {
        model_str <- bst[["booster"]]$save_model_to_string()
        params_in_file <- .params_from_model_string(model_str = model_str)

        # parameters should match what was passed from the R package, and the "main" (non-alias)
        # params values in `params` should be preferred to keyword argumentts or aliases
1094
        expect_equal(sum(startsWith(params_in_file, "[num_iterations:")), 1L)
1095
        expect_equal(sum(params_in_file == sprintf("[num_iterations: %s]", num_iterations)), 1L)
1096
        expect_equal(sum(startsWith(params_in_file, "[early_stopping_round:")), 1L)
1097
1098
1099
        expect_equal(sum(params_in_file == sprintf("[early_stopping_round: %s]", early_stopping_round)), 1L)

        # none of the aliases shouold have been written to the model file
1100
1101
1102
        expect_equal(sum(startsWith(params_in_file, "[num_boost_round:")), 0L)
        expect_equal(sum(startsWith(params_in_file, "[n_iter:")), 0L)
        expect_equal(sum(startsWith(params_in_file, "[n_iter_no_change:")), 0L)
1103
1104
1105
    }

})
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119

test_that("params (including dataset params) should be stored in .rds file for Booster", {
    data(agaricus.train, package = "lightgbm")
    dtrain <- lgb.Dataset(
        agaricus.train$data
        , label = agaricus.train$label
        , params = list(
            max_bin = 17L
        )
    )
    params <- list(
        objective = "binary"
        , max_depth = 4L
        , bagging_fraction = 0.8
1120
        , verbose = VERBOSITY
1121
1122
1123
1124
1125
1126
    )
    bst <- Booster$new(
        params = params
        , train_set = dtrain
    )
    bst_file <- tempfile(fileext = ".rds")
1127
    expect_warning(saveRDS.lgb.Booster(bst, file = bst_file))
1128

1129
    expect_warning(bst_from_file <- readRDS.lgb.Booster(file = bst_file))
1130
1131
1132
1133
1134
1135
    expect_identical(
        bst_from_file$params
        , list(
            objective = "binary"
            , max_depth = 4L
            , bagging_fraction = 0.8
1136
            , verbose = VERBOSITY
1137
1138
1139
1140
            , max_bin = 17L
        )
    )
})
1141

1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
test_that("params (including dataset params) should be stored in .rds file for Booster", {
    data(agaricus.train, package = "lightgbm")
    dtrain <- lgb.Dataset(
        agaricus.train$data
        , label = agaricus.train$label
        , params = list(
            max_bin = 17L
        )
    )
    params <- list(
        objective = "binary"
        , max_depth = 4L
        , bagging_fraction = 0.8
1155
        , verbose = VERBOSITY
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
    )
    bst <- Booster$new(
        params = params
        , train_set = dtrain
    )
    bst_file <- tempfile(fileext = ".rds")
    saveRDS(bst, file = bst_file)

    bst_from_file <- readRDS(file = bst_file)
    expect_identical(
        bst_from_file$params
        , list(
            objective = "binary"
            , max_depth = 4L
            , bagging_fraction = 0.8
1171
            , verbose = VERBOSITY
1172
1173
1174
1175
1176
1177
1178
            , max_bin = 17L
        )
    )
})

test_that("Handle is automatically restored when calling predict", {
    data(agaricus.train, package = "lightgbm")
1179
1180
1181
1182
1183
1184
1185
1186
1187
    bst <- lightgbm(
        agaricus.train$data
        , agaricus.train$label
        , nrounds = 5L
        , obj = "binary"
        , params = list(
            verbose = VERBOSITY
        )
    )
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
    bst_file <- tempfile(fileext = ".rds")
    saveRDS(bst, file = bst_file)

    bst_from_file <- readRDS(file = bst_file)

    pred_before <- predict(bst, agaricus.train$data)
    pred_after <- predict(bst_from_file, agaricus.train$data)
    expect_equal(pred_before, pred_after)
})

test_that("boosters with linear models at leaves work with saveRDS.lgb.Booster and readRDS.lgb.Booster", {
    X <- matrix(rnorm(100L), ncol = 1L)
    labels <- 2L * X + runif(nrow(X), 0L, 0.1)
    dtrain <- lgb.Dataset(
        data = X
        , label = labels
    )

    params <- list(
        objective = "regression"
1208
        , verbose = VERBOSITY
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
        , metric = "mse"
        , seed = 0L
        , num_leaves = 2L
    )

    bst <- lgb.train(
        data = dtrain
        , nrounds = 10L
        , params = params
    )
    expect_true(lgb.is.Booster(bst))

    # save predictions, then write the model to a file and destroy it in R
    preds <- predict(bst, X)
    model_file <- tempfile(fileext = ".rds")
    expect_warning(saveRDS.lgb.Booster(bst, file = model_file))
    bst$finalize()
    expect_null(bst$.__enclos_env__$private$handle)
    rm(bst)

    # load the booster and make predictions...should be the same
1230
1231
1232
    expect_warning({
        bst2 <- readRDS.lgb.Booster(file = model_file)
    })
1233
1234
1235
1236
    preds2 <- predict(bst2, X)
    expect_identical(preds, preds2)
})

1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
test_that("boosters with linear models at leaves can be written to RDS and re-loaded successfully", {
    X <- matrix(rnorm(100L), ncol = 1L)
    labels <- 2L * X + runif(nrow(X), 0L, 0.1)
    dtrain <- lgb.Dataset(
        data = X
        , label = labels
    )

    params <- list(
        objective = "regression"
1247
        , verbose = VERBOSITY
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
        , metric = "mse"
        , seed = 0L
        , num_leaves = 2L
    )

    bst <- lgb.train(
        data = dtrain
        , nrounds = 10L
        , params = params
    )
    expect_true(lgb.is.Booster(bst))

    # save predictions, then write the model to a file and destroy it in R
    preds <- predict(bst, X)
    model_file <- tempfile(fileext = ".rds")
1263
    saveRDS(bst, file = model_file)
1264
1265
1266
1267
1268
    bst$finalize()
    expect_null(bst$.__enclos_env__$private$handle)
    rm(bst)

    # load the booster and make predictions...should be the same
1269
    bst2 <- readRDS(file = model_file)
1270
1271
    preds2 <- predict(bst2, X)
    expect_identical(preds, preds2)
1272
})
1273
1274
1275
1276
1277
1278
1279
1280
1281

test_that("Booster's print, show, and summary work correctly", {
    .have_same_handle <- function(model, other_model) {
       expect_equal(
         model$.__enclos_env__$private$handle
         , other_model$.__enclos_env__$private$handle
       )
    }

1282
    .has_expected_content_for_fitted_model <- function(printed_txt) {
1283
1284
      expect_true(any(startsWith(printed_txt, "LightGBM Model")))
      expect_true(any(startsWith(printed_txt, "Fitted to dataset")))
1285
1286
1287
1288
    }

    .has_expected_content_for_finalized_model <- function(printed_txt) {
      expect_true(any(grepl("^LightGBM Model$", printed_txt)))
1289
      expect_true(any(grepl("Booster handle is invalid", printed_txt, fixed = TRUE)))
1290
1291
    }

1292
1293
    .check_methods_work <- function(model) {

1294
1295
1296
1297
1298
1299
        #--- should work for fitted models --- #

        # print()
        log_txt <- capture.output({
          ret <- print(model)
        })
1300
        .have_same_handle(ret, model)
1301
1302
1303
1304
1305
1306
        .has_expected_content_for_fitted_model(log_txt)

        # show()
        log_txt <- capture.output({
          ret <- show(model)
        })
1307
        expect_null(ret)
1308
1309
1310
1311
1312
1313
        .has_expected_content_for_fitted_model(log_txt)

        # summary()
        log_text <- capture.output({
          ret <- summary(model)
        })
1314
        .have_same_handle(ret, model)
1315
        .has_expected_content_for_fitted_model(log_txt)
1316

1317
        #--- should not fail for finalized models ---#
1318
        model$finalize()
1319
1320
1321
1322
1323
1324
1325
1326

        # print()
        log_txt <- capture.output({
          ret <- print(model)
        })
        .has_expected_content_for_finalized_model(log_txt)

        # show()
1327
        .have_same_handle(ret, model)
1328
1329
1330
        log_txt <- capture.output({
          ret <- show(model)
        })
1331
        expect_null(ret)
1332
1333
1334
1335
1336
1337
        .has_expected_content_for_finalized_model(log_txt)

        # summary()
        log_txt <- capture.output({
          ret <- summary(model)
        })
1338
        .have_same_handle(ret, model)
1339
        .has_expected_content_for_finalized_model(log_txt)
1340
1341
1342
1343
    }

    data("mtcars")
    model <- lgb.train(
1344
1345
1346
1347
        params = list(
          objective = "regression"
          , min_data_in_leaf = 1L
        )
1348
1349
        , data = lgb.Dataset(
            as.matrix(mtcars[, -1L])
1350
1351
1352
1353
1354
            , label = mtcars$mpg
            , params = list(
              min_data_in_bin = 1L
            )
        )
1355
        , verbose = VERBOSITY
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
        , nrounds = 5L
    )
    .check_methods_work(model)

    data("iris")
    model <- lgb.train(
        params = list(objective = "multiclass", num_class = 3L)
        , data = lgb.Dataset(
            as.matrix(iris[, -5L])
            , label = as.numeric(factor(iris$Species)) - 1.0
        )
1367
        , verbose = VERBOSITY
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
        , nrounds = 5L
    )
    .check_methods_work(model)


    # with custom objective
    .logregobj <- function(preds, dtrain) {
        labels <- get_field(dtrain, "label")
        preds <- 1.0 / (1.0 + exp(-preds))
        grad <- preds - labels
        hess <- preds * (1.0 - preds)
        return(list(grad = grad, hess = hess))
    }

    .evalerror <- function(preds, dtrain) {
        labels <- get_field(dtrain, "label")
        preds <- 1.0 / (1.0 + exp(-preds))
        err <- as.numeric(sum(labels != (preds > 0.5))) / length(labels)
        return(list(
            name = "error"
            , value = err
            , higher_better = FALSE
        ))
    }

    model <- lgb.train(
        data = lgb.Dataset(
            as.matrix(iris[, -5L])
            , label = as.numeric(iris$Species == "virginica")
        )
        , obj = .logregobj
        , eval = .evalerror
1400
        , verbose = VERBOSITY
1401
1402
1403
1404
1405
1406
1407
1408
1409
        , nrounds = 5L
    )

    .check_methods_work(model)
})

test_that("LGBM_BoosterGetNumFeature_R returns correct outputs", {
    data("mtcars")
    model <- lgb.train(
1410
1411
1412
1413
        params = list(
          objective = "regression"
          , min_data_in_leaf = 1L
        )
1414
1415
        , data = lgb.Dataset(
            as.matrix(mtcars[, -1L])
1416
1417
1418
1419
1420
            , label = mtcars$mpg
            , params = list(
              min_data_in_bin = 1L
            )
        )
1421
        , verbose = VERBOSITY
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
        , nrounds = 5L
    )
    ncols <- .Call(LGBM_BoosterGetNumFeature_R, model$.__enclos_env__$private$handle)
    expect_equal(ncols, ncol(mtcars) - 1L)

    data("iris")
    model <- lgb.train(
        params = list(objective = "multiclass", num_class = 3L)
        , data = lgb.Dataset(
            as.matrix(iris[, -5L])
            , label = as.numeric(factor(iris$Species)) - 1.0
        )
1434
        , verbose = VERBOSITY
1435
1436
1437
1438
1439
        , nrounds = 5L
    )
    ncols <- .Call(LGBM_BoosterGetNumFeature_R, model$.__enclos_env__$private$handle)
    expect_equal(ncols, ncol(iris) - 1L)
})