basic.py 163 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3
import abc
wxchan's avatar
wxchan committed
4
import ctypes
5
import json
wxchan's avatar
wxchan committed
6
import warnings
7
from collections import OrderedDict
8
from copy import deepcopy
9
10
from functools import wraps
from logging import Logger
11
12
13
from os import SEEK_END
from os.path import getsize
from pathlib import Path
14
from tempfile import NamedTemporaryFile
15
from typing import Any, Callable, Dict, Iterable, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
16
17
18
19

import numpy as np
import scipy.sparse

20
21
from .compat import (PANDAS_INSTALLED, concat, dt_DataTable, is_dtype_sparse, pd_CategoricalDtype, pd_DataFrame,
                     pd_Series)
wxchan's avatar
wxchan committed
22
23
from .libpath import find_lib_path

24
25
26
27
28
29
30
31
32
33
34
35
ZERO_THRESHOLD = 1e-35


def _get_sample_count(total_nrow: int, params: str):
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
        c_str(params),
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
36

37
class _DummyLogger:
38
    def info(self, msg: str) -> None:
39
40
        print(msg)

41
    def warning(self, msg: str) -> None:
42
43
44
        warnings.warn(msg, stacklevel=3)


45
_LOGGER: Union[_DummyLogger, Logger] = _DummyLogger()
46
47


48
def register_logger(logger: Logger) -> None:
49
50
51
52
53
54
55
56
57
58
59
60
61
    """Register custom logger.

    Parameters
    ----------
    logger : logging.Logger
        Custom logger.
    """
    if not isinstance(logger, Logger):
        raise TypeError("Logger should inherit logging.Logger class")
    global _LOGGER
    _LOGGER = logger


62
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
63
    """Join log messages from native library which come by chunks."""
64
    msg_normalized: List[str] = []
65
66

    @wraps(func)
67
    def wrapper(msg: str) -> None:
68
69
70
71
72
73
74
75
76
77
78
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


79
def _log_info(msg: str) -> None:
80
81
82
    _LOGGER.info(msg)


83
def _log_warning(msg: str) -> None:
84
85
86
87
    _LOGGER.warning(msg)


@_normalize_native_string
88
def _log_native(msg: str) -> None:
89
90
91
    _LOGGER.info(msg)


92
def _log_callback(msg: bytes) -> None:
93
    """Redirect logs from native library into Python."""
94
    _log_native(str(msg.decode('utf-8')))
95
96


wxchan's avatar
wxchan committed
97
def _load_lib():
98
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
99
100
    lib_path = find_lib_path()
    if len(lib_path) == 0:
101
        return None
wxchan's avatar
wxchan committed
102
103
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
104
105
106
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
    lib.callback = callback(_log_callback)
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
107
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
108
109
    return lib

wxchan's avatar
wxchan committed
110

wxchan's avatar
wxchan committed
111
112
_LIB = _load_lib()

wxchan's avatar
wxchan committed
113

114
115
116
NUMERIC_TYPES = (int, float, bool)


117
def _safe_call(ret: int) -> None:
118
119
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
120
121
122
    Parameters
    ----------
    ret : int
123
        The return value from C API calls.
wxchan's avatar
wxchan committed
124
125
    """
    if ret != 0:
126
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
127

wxchan's avatar
wxchan committed
128

wxchan's avatar
wxchan committed
129
def is_numeric(obj):
130
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
131
132
133
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
134
135
136
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
137
138
        return False

wxchan's avatar
wxchan committed
139

wxchan's avatar
wxchan committed
140
def is_numpy_1d_array(data):
141
    """Check whether data is a numpy 1-D array."""
142
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
143

wxchan's avatar
wxchan committed
144

145
146
147
148
149
150
151
152
def is_numpy_column_array(data):
    """Check whether data is a column numpy array."""
    if not isinstance(data, np.ndarray):
        return False
    shape = data.shape
    return len(shape) == 2 and shape[1] == 1


153
154
def cast_numpy_array_to_dtype(array, dtype):
    """Cast numpy array to given dtype."""
155
156
157
158
159
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


wxchan's avatar
wxchan committed
160
def is_1d_list(data):
161
162
    """Check whether data is a 1-D list."""
    return isinstance(data, list) and (not data or is_numeric(data[0]))
wxchan's avatar
wxchan committed
163

wxchan's avatar
wxchan committed
164

165
166
167
168
169
170
171
172
173
174
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
        is_numpy_1d_array(data)
        or is_numpy_column_array(data)
        or is_1d_list(data)
        or isinstance(data, pd_Series)
    )


175
def list_to_1d_numpy(data, dtype=np.float32, name='list'):
176
    """Convert data to numpy 1-D array."""
wxchan's avatar
wxchan committed
177
    if is_numpy_1d_array(data):
178
        return cast_numpy_array_to_dtype(data, dtype)
179
180
181
    elif is_numpy_column_array(data):
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
182
        return cast_numpy_array_to_dtype(array, dtype)
wxchan's avatar
wxchan committed
183
184
    elif is_1d_list(data):
        return np.array(data, dtype=dtype, copy=False)
185
    elif isinstance(data, pd_Series):
186
187
        if _get_bad_pandas_dtypes([data.dtypes]):
            raise ValueError('Series.dtypes must be int, float or bool')
188
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
189
    else:
190
191
        raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                        "It should be list, numpy 1-D array or pandas Series")
wxchan's avatar
wxchan committed
192

wxchan's avatar
wxchan committed
193

194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
def _is_numpy_2d_array(data: Any) -> bool:
    """Check whether data is a numpy 2-D array."""
    return isinstance(data, np.ndarray) and len(data.shape) == 2 and data.shape[1] > 1


def _is_2d_list(data: Any) -> bool:
    """Check whether data is a 2-D list."""
    return isinstance(data, list) and len(data) > 0 and is_1d_list(data[0])


def _is_2d_collection(data: Any) -> bool:
    """Check whether data is a 2-D collection."""
    return (
        _is_numpy_2d_array(data)
        or _is_2d_list(data)
        or isinstance(data, pd_DataFrame)
    )


def _data_to_2d_numpy(data: Any, dtype: type = np.float32, name: str = 'list') -> np.ndarray:
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
        return cast_numpy_array_to_dtype(data, dtype)
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
        if _get_bad_pandas_dtypes(data.dtypes):
            raise ValueError('DataFrame.dtypes must be int, float or bool')
        return cast_numpy_array_to_dtype(data.values, dtype)
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


wxchan's avatar
wxchan committed
227
def cfloat32_array_to_numpy(cptr, length):
228
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
229
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
230
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
231
    else:
232
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
233

Guolin Ke's avatar
Guolin Ke committed
234

Guolin Ke's avatar
Guolin Ke committed
235
def cfloat64_array_to_numpy(cptr, length):
236
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
237
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
238
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
239
240
241
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
242

wxchan's avatar
wxchan committed
243
def cint32_array_to_numpy(cptr, length):
244
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
245
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
246
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
247
    else:
248
249
250
251
252
253
        raise RuntimeError('Expected int32 pointer')


def cint64_array_to_numpy(cptr, length):
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
254
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
255
256
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
257

wxchan's avatar
wxchan committed
258

wxchan's avatar
wxchan committed
259
def c_str(string):
260
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
261
262
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
263

wxchan's avatar
wxchan committed
264
def c_array(ctype, values):
265
    """Convert a Python array to C array."""
wxchan's avatar
wxchan committed
266
267
    return (ctype * len(values))(*values)

wxchan's avatar
wxchan committed
268

269
270
271
272
273
274
275
276
277
278
def json_default_with_numpy(obj):
    """Convert numpy classes to JSON serializable objects."""
    if isinstance(obj, (np.integer, np.floating, np.bool_)):
        return obj.item()
    elif isinstance(obj, np.ndarray):
        return obj.tolist()
    else:
        return obj


wxchan's avatar
wxchan committed
279
def param_dict_to_str(data):
280
    """Convert Python dictionary to string, which is passed to C API."""
281
    if data is None or not data:
wxchan's avatar
wxchan committed
282
283
284
        return ""
    pairs = []
    for key, val in data.items():
285
        if isinstance(val, (list, tuple, set)) or is_numpy_1d_array(val):
286
287
            def to_string(x):
                if isinstance(x, list):
288
                    return f"[{','.join(map(str, x))}]"
289
290
                else:
                    return str(x)
291
            pairs.append(f"{key}={','.join(map(to_string, val))}")
292
        elif isinstance(val, (str, Path, NUMERIC_TYPES)) or is_numeric(val):
293
            pairs.append(f"{key}={val}")
294
        elif val is not None:
295
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
296
    return ' '.join(pairs)
297

wxchan's avatar
wxchan committed
298

299
class _TempFile:
300
301
    """Proxy class to workaround errors on Windows."""

302
303
304
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
305
            self.path = Path(self.name)
306
        return self
wxchan's avatar
wxchan committed
307

308
    def __exit__(self, exc_type, exc_val, exc_tb):
309
310
        if self.path.is_file():
            self.path.unlink()
311

wxchan's avatar
wxchan committed
312

313
class LightGBMError(Exception):
314
315
    """Error thrown by LightGBM."""

316
317
318
    pass


319
320
321
322
323
324
325
326
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
327
328
329
    aliases = {"bin_construct_sample_cnt": {"bin_construct_sample_cnt",
                                            "subsample_for_bin"},
               "boosting": {"boosting",
330
331
332
333
334
                            "boosting_type",
                            "boost"},
               "categorical_feature": {"categorical_feature",
                                       "cat_feature",
                                       "categorical_column",
335
336
                                       "cat_column",
                                       "categorical_features"},
337
338
               "data_random_seed": {"data_random_seed",
                                    "data_seed"},
339
340
341
342
               "early_stopping_round": {"early_stopping_round",
                                        "early_stopping_rounds",
                                        "early_stopping",
                                        "n_iter_no_change"},
343
344
345
               "enable_bundle": {"enable_bundle",
                                 "is_enable_bundle",
                                 "bundle"},
346
347
348
349
350
               "eval_at": {"eval_at",
                           "ndcg_eval_at",
                           "ndcg_at",
                           "map_eval_at",
                           "map_at"},
351
352
353
354
355
356
               "group_column": {"group_column",
                                "group",
                                "group_id",
                                "query_column",
                                "query",
                                "query_id"},
357
358
               "header": {"header",
                          "has_header"},
359
360
361
362
363
364
365
366
367
               "ignore_column": {"ignore_column",
                                 "ignore_feature",
                                 "blacklist"},
               "is_enable_sparse": {"is_enable_sparse",
                                    "is_sparse",
                                    "enable_sparse",
                                    "sparse"},
               "label_column": {"label_column",
                                "label"},
Nikita Titov's avatar
Nikita Titov committed
368
369
               "linear_tree": {"linear_tree",
                               "linear_trees"},
370
371
372
               "local_listen_port": {"local_listen_port",
                                     "local_port",
                                     "port"},
373
374
375
               "machines": {"machines",
                            "workers",
                            "nodes"},
376
377
               "max_bin": {"max_bin",
                           "max_bins"},
378
379
380
381
382
383
384
385
386
387
388
389
               "metric": {"metric",
                          "metrics",
                          "metric_types"},
               "num_class": {"num_class",
                             "num_classes"},
               "num_iterations": {"num_iterations",
                                  "num_iteration",
                                  "n_iter",
                                  "num_tree",
                                  "num_trees",
                                  "num_round",
                                  "num_rounds",
390
                                  "nrounds",
391
                                  "num_boost_round",
392
393
                                  "n_estimators",
                                  "max_iter"},
394
395
               "num_machines": {"num_machines",
                                "num_machine"},
396
397
398
399
400
               "num_threads": {"num_threads",
                               "num_thread",
                               "nthread",
                               "nthreads",
                               "n_jobs"},
401
402
403
               "objective": {"objective",
                             "objective_type",
                             "app",
404
405
                             "application",
                             "loss"},
406
407
               "pre_partition": {"pre_partition",
                                 "is_pre_partition"},
408
409
410
411
               "tree_learner": {"tree_learner",
                                "tree",
                                "tree_type",
                                "tree_learner_type"},
412
413
414
               "two_round": {"two_round",
                             "two_round_loading",
                             "use_two_round_loading"},
415
               "verbosity": {"verbosity",
416
417
418
                             "verbose"},
               "weight_column": {"weight_column",
                                 "weight"}}
419
420
421
422
423

    @classmethod
    def get(cls, *args):
        ret = set()
        for i in args:
424
            ret |= cls.aliases.get(i, {i})
425
426
427
        return ret


428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
def _choose_param_value(main_param_name: str, params: Dict[str, Any], default_value: Any) -> Dict[str, Any]:
    """Get a single parameter value, accounting for aliases.

    Parameters
    ----------
    main_param_name : str
        Name of the main parameter to get a value for. One of the keys of ``_ConfigAliases``.
    params : dict
        Dictionary of LightGBM parameters.
    default_value : Any
        Default value to use for the parameter, if none is found in ``params``.

    Returns
    -------
    params : dict
        A ``params`` dict with exactly one value for ``main_param_name``, and all aliases ``main_param_name`` removed.
        If both ``main_param_name`` and one or more aliases for it are found, the value of ``main_param_name`` will be preferred.
    """
    # avoid side effects on passed-in parameters
    params = deepcopy(params)

    # find a value, and remove other aliases with .pop()
    # prefer the value of 'main_param_name' if it exists, otherwise search the aliases
    found_value = None
    if main_param_name in params.keys():
        found_value = params[main_param_name]

    for param in _ConfigAliases.get(main_param_name):
        val = params.pop(param, None)
        if found_value is None and val is not None:
            found_value = val

    if found_value is not None:
        params[main_param_name] = found_value
    else:
        params[main_param_name] = default_value

    return params


468
469
MAX_INT32 = (1 << 31) - 1

470
"""Macro definition of data type in C API of LightGBM"""
wxchan's avatar
wxchan committed
471
472
473
474
C_API_DTYPE_FLOAT32 = 0
C_API_DTYPE_FLOAT64 = 1
C_API_DTYPE_INT32 = 2
C_API_DTYPE_INT64 = 3
Guolin Ke's avatar
Guolin Ke committed
475

476
"""Matrix is row major in Python"""
wxchan's avatar
wxchan committed
477
478
C_API_IS_ROW_MAJOR = 1

479
"""Macro definition of prediction type in C API of LightGBM"""
wxchan's avatar
wxchan committed
480
481
482
C_API_PREDICT_NORMAL = 0
C_API_PREDICT_RAW_SCORE = 1
C_API_PREDICT_LEAF_INDEX = 2
483
C_API_PREDICT_CONTRIB = 3
wxchan's avatar
wxchan committed
484

485
486
487
488
"""Macro definition of sparse matrix type"""
C_API_MATRIX_TYPE_CSR = 0
C_API_MATRIX_TYPE_CSC = 1

489
490
491
492
"""Macro definition of feature importance type"""
C_API_FEATURE_IMPORTANCE_SPLIT = 0
C_API_FEATURE_IMPORTANCE_GAIN = 1

493
"""Data type of data field"""
wxchan's avatar
wxchan committed
494
495
FIELD_TYPE_MAPPER = {"label": C_API_DTYPE_FLOAT32,
                     "weight": C_API_DTYPE_FLOAT32,
Guolin Ke's avatar
Guolin Ke committed
496
                     "init_score": C_API_DTYPE_FLOAT64,
497
                     "group": C_API_DTYPE_INT32}
wxchan's avatar
wxchan committed
498

499
500
501
502
"""String name to int feature importance type mapper"""
FEATURE_IMPORTANCE_TYPE_MAPPER = {"split": C_API_FEATURE_IMPORTANCE_SPLIT,
                                  "gain": C_API_FEATURE_IMPORTANCE_GAIN}

wxchan's avatar
wxchan committed
503

504
def convert_from_sliced_object(data):
505
    """Fix the memory of multi-dimensional sliced object."""
506
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
507
        if not data.flags.c_contiguous:
508
509
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
510
511
512
513
            return np.copy(data)
    return data


wxchan's avatar
wxchan committed
514
def c_float_array(data):
515
    """Get pointer of float numpy array / list."""
wxchan's avatar
wxchan committed
516
517
518
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
519
520
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
521
522
523
524
525
526
527
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
            type_data = C_API_DTYPE_FLOAT32
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
            type_data = C_API_DTYPE_FLOAT64
        else:
528
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
529
    else:
530
        raise TypeError(f"Unknown type({type(data).__name__})")
531
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
532

wxchan's avatar
wxchan committed
533

wxchan's avatar
wxchan committed
534
def c_int_array(data):
535
    """Get pointer of int numpy array / list."""
wxchan's avatar
wxchan committed
536
537
538
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
539
540
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
541
542
543
544
545
546
547
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
            type_data = C_API_DTYPE_INT32
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
            type_data = C_API_DTYPE_INT64
        else:
548
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
549
    else:
550
        raise TypeError(f"Unknown type({type(data).__name__})")
551
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
552

wxchan's avatar
wxchan committed
553

554
555
556
557
558
559
560
561
562
563
564
def _get_bad_pandas_dtypes(dtypes):
    pandas_dtype_mapper = {'int8': 'int', 'int16': 'int', 'int32': 'int',
                           'int64': 'int', 'uint8': 'int', 'uint16': 'int',
                           'uint32': 'int', 'uint64': 'int', 'bool': 'int',
                           'float16': 'float', 'float32': 'float', 'float64': 'float'}
    bad_indices = [i for i, dtype in enumerate(dtypes) if (dtype.name not in pandas_dtype_mapper
                                                           and (not is_dtype_sparse(dtype)
                                                                or dtype.subtype.name not in pandas_dtype_mapper))]
    return bad_indices


565
def _data_from_pandas(data, feature_name, categorical_feature, pandas_categorical):
566
    if isinstance(data, pd_DataFrame):
567
568
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
569
570
        if feature_name == 'auto' or feature_name is None:
            data = data.rename(columns=str)
571
        cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
572
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
573
574
575
576
577
        if pandas_categorical is None:  # train dataset
            pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
        else:
            if len(cat_cols) != len(pandas_categorical):
                raise ValueError('train and valid dataset categorical_feature do not match.')
578
            for col, category in zip(cat_cols, pandas_categorical):
579
580
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
581
        if len(cat_cols):  # cat_cols is list
582
            data = data.copy()  # not alter origin DataFrame
583
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
584
585
586
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
587
            if categorical_feature == 'auto':  # use cat cols from DataFrame
588
                categorical_feature = cat_cols_not_ordered
589
590
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
591
592
        if feature_name == 'auto':
            feature_name = list(data.columns)
593
594
        bad_indices = _get_bad_pandas_dtypes(data.dtypes)
        if bad_indices:
595
            bad_index_cols_str = ', '.join(data.columns[bad_indices])
596
            raise ValueError("DataFrame.dtypes for data must be int, float or bool.\n"
597
                             "Did not expect the data types in the following fields: "
598
                             f"{bad_index_cols_str}")
599
600
601
        data = data.values
        if data.dtype != np.float32 and data.dtype != np.float64:
            data = data.astype(np.float32)
602
603
604
605
606
607
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
608
609
610


def _label_from_pandas(label):
611
    if isinstance(label, pd_DataFrame):
612
613
        if len(label.columns) > 1:
            raise ValueError('DataFrame for label cannot have multiple columns')
614
        if _get_bad_pandas_dtypes(label.dtypes):
615
            raise ValueError('DataFrame.dtypes for label must be int, float or bool')
616
        label = np.ravel(label.values.astype(np.float32, copy=False))
617
618
619
    return label


620
def _dump_pandas_categorical(pandas_categorical, file_name=None):
621
622
    categorical_json = json.dumps(pandas_categorical, default=json_default_with_numpy)
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
623
624
625
626
627
628
629
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


def _load_pandas_categorical(file_name=None, model_str=None):
630
631
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
632
    if file_name is not None:
633
        max_offset = -getsize(file_name)
634
635
636
637
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
638
                f.seek(offset, SEEK_END)
639
640
641
642
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
643
        last_line = lines[-1].decode('utf-8').strip()
644
        if not last_line.startswith(pandas_key):
645
            last_line = lines[-2].decode('utf-8').strip()
646
    elif model_str is not None:
647
648
649
650
651
652
        idx = model_str.rfind('\n', 0, offset)
        last_line = model_str[idx:].strip()
    if last_line.startswith(pandas_key):
        return json.loads(last_line[len(pandas_key):])
    else:
        return None
653
654


655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
class Sequence(abc.ABC):
    """
    Generic data access interface.

    Object should support the following operations:

    .. code-block::

        # Get total row number.
        >>> len(seq)
        # Random access by row index. Used for data sampling.
        >>> seq[10]
        # Range data access. Used to read data in batch when constructing Dataset.
        >>> seq[0:100]
        # Optionally specify batch_size to control range data read size.
        >>> seq.batch_size

    - With random access, **data sampling does not need to go through all data**.
    - With range data access, there's **no need to read all data into memory thus reduce memory usage**.

675
676
    .. versionadded:: 3.3.0

677
678
679
680
681
682
683
684
685
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

    batch_size = 4096  # Defaults to read 4K rows in each batch.

    @abc.abstractmethod
686
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
687
688
689
690
691
692
693
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
694
                return self._get_one_line(idx)
695
            elif isinstance(idx, slice):
696
697
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
698
                # Only required if using ``Dataset.subset()``.
699
                return np.array([self._get_one_line(i) for i in idx])
700
            else:
701
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
702
703
704

        Parameters
        ----------
705
        idx : int, slice[int], list[int]
706
707
708
709
710
            Item index.

        Returns
        -------
        result : numpy 1-D array, numpy 2-D array
711
            1-D array if idx is int, 2-D array if idx is slice or list.
712
713
714
715
716
717
718
719
720
        """
        raise NotImplementedError("Sub-classes of lightgbm.Sequence must implement __getitem__()")

    @abc.abstractmethod
    def __len__(self) -> int:
        """Return row count of this sequence."""
        raise NotImplementedError("Sub-classes of lightgbm.Sequence must implement __len__()")


721
class _InnerPredictor:
722
723
724
725
726
    """_InnerPredictor of LightGBM.

    Not exposed to user.
    Used only for prediction, usually used for continued training.

Nikita Titov's avatar
Nikita Titov committed
727
728
729
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
730
    """
731

732
    def __init__(self, model_file=None, booster_handle=None, pred_parameter=None):
733
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
734
735
736

        Parameters
        ----------
737
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
738
            Path to the model file.
739
740
741
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
742
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
743
744
745
746
747
        """
        self.handle = ctypes.c_void_p()
        self.__is_manage_handle = True
        if model_file is not None:
            """Prediction task"""
Guolin Ke's avatar
Guolin Ke committed
748
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
749
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
750
                c_str(str(model_file)),
wxchan's avatar
wxchan committed
751
752
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
753
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
754
755
756
757
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
758
            self.num_total_iteration = out_num_iterations.value
759
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
760
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
761
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
762
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
763
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
764
765
766
767
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
768
            self.num_total_iteration = self.current_iteration()
769
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
770
        else:
771
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
772

773
774
        pred_parameter = {} if pred_parameter is None else pred_parameter
        self.pred_parameter = param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
775

wxchan's avatar
wxchan committed
776
    def __del__(self):
777
778
779
780
781
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
782

783
784
785
786
787
    def __getstate__(self):
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

788
    def predict(self, data, start_iteration=0, num_iteration=-1,
789
                raw_score=False, pred_leaf=False, pred_contrib=False, data_has_header=False,
wxchan's avatar
wxchan committed
790
                is_reshape=True):
791
        """Predict logic.
wxchan's avatar
wxchan committed
792
793
794

        Parameters
        ----------
795
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
796
            Data source for prediction.
797
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
798
799
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
800
801
802
803
804
805
806
807
808
809
810
811
812
        num_iteration : int, optional (default=-1)
            Iteration used for prediction.
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
        data_has_header : bool, optional (default=False)
            Whether data has header.
            Used only for txt data.
        is_reshape : bool, optional (default=True)
            Whether to reshape to (nrow, ncol).
wxchan's avatar
wxchan committed
813
814
815

        Returns
        -------
816
        result : numpy array, scipy.sparse or list of scipy.sparse
817
            Prediction result.
818
            Can be sparse or a list of sparse objects (each element represents predictions for one class) for feature contributions (when ``pred_contrib=True``).
wxchan's avatar
wxchan committed
819
        """
wxchan's avatar
wxchan committed
820
        if isinstance(data, Dataset):
821
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
822
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
wxchan's avatar
wxchan committed
823
824
825
826
827
        predict_type = C_API_PREDICT_NORMAL
        if raw_score:
            predict_type = C_API_PREDICT_RAW_SCORE
        if pred_leaf:
            predict_type = C_API_PREDICT_LEAF_INDEX
828
829
        if pred_contrib:
            predict_type = C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
830
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
831

832
        if isinstance(data, (str, Path)):
833
            with _TempFile() as f:
wxchan's avatar
wxchan committed
834
835
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
836
                    c_str(str(data)),
Guolin Ke's avatar
Guolin Ke committed
837
838
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
839
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
840
                    ctypes.c_int(num_iteration),
841
                    c_str(self.pred_parameter),
wxchan's avatar
wxchan committed
842
                    c_str(f.name)))
843
844
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
845
        elif isinstance(data, scipy.sparse.csr_matrix):
846
            preds, nrow = self.__pred_for_csr(data, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
847
        elif isinstance(data, scipy.sparse.csc_matrix):
848
            preds, nrow = self.__pred_for_csc(data, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
849
        elif isinstance(data, np.ndarray):
850
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
851
852
853
        elif isinstance(data, list):
            try:
                data = np.array(data)
854
            except BaseException:
855
                raise ValueError('Cannot convert data list to numpy array.')
856
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
857
        elif isinstance(data, dt_DataTable):
858
            preds, nrow = self.__pred_for_np2d(data.to_numpy(), start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
859
860
        else:
            try:
861
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
862
                csr = scipy.sparse.csr_matrix(data)
863
            except BaseException:
864
                raise TypeError(f'Cannot predict data for type {type(data).__name__}')
865
            preds, nrow = self.__pred_for_csr(csr, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
866
867
        if pred_leaf:
            preds = preds.astype(np.int32)
868
869
        is_sparse = scipy.sparse.issparse(preds) or isinstance(preds, list)
        if is_reshape and not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
870
            if preds.size % nrow == 0:
871
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
872
            else:
873
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
874
875
        return preds

876
    def __get_num_preds(self, start_iteration, num_iteration, nrow, predict_type):
877
        """Get size of prediction result."""
878
        if nrow > MAX_INT32:
879
            raise LightGBMError('LightGBM cannot perform prediction for data '
880
                                f'with number of rows greater than MAX_INT32 ({MAX_INT32}).\n'
881
                                'You can split your data into chunks '
882
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
883
884
885
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
886
887
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
888
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
889
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
890
891
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
892

893
    def __pred_for_np2d(self, mat, start_iteration, num_iteration, predict_type):
894
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
895
        if len(mat.shape) != 2:
896
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
897

898
        def inner_predict(mat, start_iteration, num_iteration, predict_type, preds=None):
899
900
            if mat.dtype == np.float32 or mat.dtype == np.float64:
                data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
901
            else:  # change non-float data to float data, need to copy
902
903
                data = np.array(mat.reshape(mat.size), dtype=np.float32)
            ptr_data, type_ptr_data, _ = c_float_array(data)
904
            n_preds = self.__get_num_preds(start_iteration, num_iteration, mat.shape[0], predict_type)
905
            if preds is None:
906
                preds = np.empty(n_preds, dtype=np.float64)
907
908
909
910
911
912
913
            elif len(preds.shape) != 1 or len(preds) != n_preds:
                raise ValueError("Wrong length of pre-allocated predict array")
            out_num_preds = ctypes.c_int64(0)
            _safe_call(_LIB.LGBM_BoosterPredictForMat(
                self.handle,
                ptr_data,
                ctypes.c_int(type_ptr_data),
914
915
                ctypes.c_int32(mat.shape[0]),
                ctypes.c_int32(mat.shape[1]),
916
917
                ctypes.c_int(C_API_IS_ROW_MAJOR),
                ctypes.c_int(predict_type),
918
                ctypes.c_int(start_iteration),
919
920
921
922
923
924
925
926
927
928
929
930
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.byref(out_num_preds),
                preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
            if n_preds != out_num_preds.value:
                raise ValueError("Wrong length for predict results")
            return preds, mat.shape[0]

        nrow = mat.shape[0]
        if nrow > MAX_INT32:
            sections = np.arange(start=MAX_INT32, stop=nrow, step=MAX_INT32)
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
931
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
932
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
933
            preds = np.empty(sum(n_preds), dtype=np.float64)
934
935
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
936
                # avoid memory consumption by arrays concatenation operations
937
                inner_predict(chunk, start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
938
            return preds, nrow
wxchan's avatar
wxchan committed
939
        else:
940
            return inner_predict(mat, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
941

942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
    def __create_sparse_native(self, cs, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                               indptr_type, data_type, is_csr=True):
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
        if indptr_type == C_API_DTYPE_INT32:
            out_indptr = cint32_array_to_numpy(out_ptr_indptr, indptr_len)
        elif indptr_type == C_API_DTYPE_INT64:
            out_indptr = cint64_array_to_numpy(out_ptr_indptr, indptr_len)
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
        if data_type == C_API_DTYPE_FLOAT32:
            out_data = cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
        elif data_type == C_API_DTYPE_FLOAT64:
            out_data = cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
        else:
            raise TypeError("Expected float32 or float64 type for data")
        out_indices = cint32_array_to_numpy(out_ptr_indices, data_indices_len)
        # break up indptr based on number of rows (note more than one matrix in multiclass case)
        per_class_indptr_shape = cs.indptr.shape[0]
        # for CSC there is extra column added
        if not is_csr:
            per_class_indptr_shape += 1
        out_indptr_arrays = np.split(out_indptr, out_indptr.shape[0] / per_class_indptr_shape)
        # reformat output into a csr or csc matrix or list of csr or csc matrices
        cs_output_matrices = []
        offset = 0
        for cs_indptr in out_indptr_arrays:
            matrix_indptr_len = cs_indptr[cs_indptr.shape[0] - 1]
            cs_indices = out_indices[offset + cs_indptr[0]:offset + matrix_indptr_len]
            cs_data = out_data[offset + cs_indptr[0]:offset + matrix_indptr_len]
            offset += matrix_indptr_len
            # same shape as input csr or csc matrix except extra column for expected value
            cs_shape = [cs.shape[0], cs.shape[1] + 1]
            # note: make sure we copy data as it will be deallocated next
            if is_csr:
                cs_output_matrices.append(scipy.sparse.csr_matrix((cs_data, cs_indices, cs_indptr), cs_shape))
            else:
                cs_output_matrices.append(scipy.sparse.csc_matrix((cs_data, cs_indices, cs_indptr), cs_shape))
        # free the temporary native indptr, indices, and data
        _safe_call(_LIB.LGBM_BoosterFreePredictSparse(out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                      ctypes.c_int(indptr_type), ctypes.c_int(data_type)))
        if len(cs_output_matrices) == 1:
            return cs_output_matrices[0]
        return cs_output_matrices

988
    def __pred_for_csr(self, csr, start_iteration, num_iteration, predict_type):
989
        """Predict for a CSR data."""
990
        def inner_predict(csr, start_iteration, num_iteration, predict_type, preds=None):
991
            nrow = len(csr.indptr) - 1
992
            n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
993
            if preds is None:
994
                preds = np.empty(n_preds, dtype=np.float64)
995
996
997
998
999
1000
1001
            elif len(preds.shape) != 1 or len(preds) != n_preds:
                raise ValueError("Wrong length of pre-allocated predict array")
            out_num_preds = ctypes.c_int64(0)

            ptr_indptr, type_ptr_indptr, __ = c_int_array(csr.indptr)
            ptr_data, type_ptr_data, _ = c_float_array(csr.data)

1002
            assert csr.shape[1] <= MAX_INT32
1003
            csr_indices = csr.indices.astype(np.int32, copy=False)
1004

1005
1006
1007
            _safe_call(_LIB.LGBM_BoosterPredictForCSR(
                self.handle,
                ptr_indptr,
1008
                ctypes.c_int(type_ptr_indptr),
1009
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
1010
1011
1012
1013
1014
1015
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csr.indptr)),
                ctypes.c_int64(len(csr.data)),
                ctypes.c_int64(csr.shape[1]),
                ctypes.c_int(predict_type),
1016
                ctypes.c_int(start_iteration),
1017
1018
1019
1020
1021
1022
1023
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.byref(out_num_preds),
                preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
            if n_preds != out_num_preds.value:
                raise ValueError("Wrong length for predict results")
            return preds, nrow
wxchan's avatar
wxchan committed
1024

1025
        def inner_predict_sparse(csr, start_iteration, num_iteration, predict_type):
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
            ptr_indptr, type_ptr_indptr, __ = c_int_array(csr.indptr)
            ptr_data, type_ptr_data, _ = c_float_array(csr.data)
            csr_indices = csr.indices.astype(np.int32, copy=False)
            matrix_type = C_API_MATRIX_TYPE_CSR
            if type_ptr_indptr == C_API_DTYPE_INT32:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int32)()
            else:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int64)()
            out_ptr_indices = ctypes.POINTER(ctypes.c_int32)()
            if type_ptr_data == C_API_DTYPE_FLOAT32:
                out_ptr_data = ctypes.POINTER(ctypes.c_float)()
            else:
                out_ptr_data = ctypes.POINTER(ctypes.c_double)()
1039
            out_shape = np.empty(2, dtype=np.int64)
1040
1041
1042
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
1043
                ctypes.c_int(type_ptr_indptr),
1044
1045
1046
1047
1048
1049
1050
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csr.indptr)),
                ctypes.c_int64(len(csr.data)),
                ctypes.c_int64(csr.shape[1]),
                ctypes.c_int(predict_type),
1051
                ctypes.c_int(start_iteration),
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.c_int(matrix_type),
                out_shape.ctypes.data_as(ctypes.POINTER(ctypes.c_int64)),
                ctypes.byref(out_ptr_indptr),
                ctypes.byref(out_ptr_indices),
                ctypes.byref(out_ptr_data)))
            matrices = self.__create_sparse_native(csr, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                   type_ptr_indptr, type_ptr_data, is_csr=True)
            nrow = len(csr.indptr) - 1
            return matrices, nrow

        if predict_type == C_API_PREDICT_CONTRIB:
1065
            return inner_predict_sparse(csr, start_iteration, num_iteration, predict_type)
1066
1067
1068
1069
        nrow = len(csr.indptr) - 1
        if nrow > MAX_INT32:
            sections = [0] + list(np.arange(start=MAX_INT32, stop=nrow, step=MAX_INT32)) + [nrow]
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1070
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1071
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1072
            preds = np.empty(sum(n_preds), dtype=np.float64)
1073
1074
            for (start_idx, end_idx), (start_idx_pred, end_idx_pred) in zip(zip(sections, sections[1:]),
                                                                            zip(n_preds_sections, n_preds_sections[1:])):
1075
                # avoid memory consumption by arrays concatenation operations
1076
                inner_predict(csr[start_idx:end_idx], start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
1077
1078
            return preds, nrow
        else:
1079
            return inner_predict(csr, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
1080

1081
    def __pred_for_csc(self, csc, start_iteration, num_iteration, predict_type):
1082
        """Predict for a CSC data."""
1083
        def inner_predict_sparse(csc, start_iteration, num_iteration, predict_type):
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
            ptr_indptr, type_ptr_indptr, __ = c_int_array(csc.indptr)
            ptr_data, type_ptr_data, _ = c_float_array(csc.data)
            csc_indices = csc.indices.astype(np.int32, copy=False)
            matrix_type = C_API_MATRIX_TYPE_CSC
            if type_ptr_indptr == C_API_DTYPE_INT32:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int32)()
            else:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int64)()
            out_ptr_indices = ctypes.POINTER(ctypes.c_int32)()
            if type_ptr_data == C_API_DTYPE_FLOAT32:
                out_ptr_data = ctypes.POINTER(ctypes.c_float)()
            else:
                out_ptr_data = ctypes.POINTER(ctypes.c_double)()
1097
            out_shape = np.empty(2, dtype=np.int64)
1098
1099
1100
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
1101
                ctypes.c_int(type_ptr_indptr),
1102
1103
1104
1105
1106
1107
1108
                csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csc.indptr)),
                ctypes.c_int64(len(csc.data)),
                ctypes.c_int64(csc.shape[0]),
                ctypes.c_int(predict_type),
1109
                ctypes.c_int(start_iteration),
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.c_int(matrix_type),
                out_shape.ctypes.data_as(ctypes.POINTER(ctypes.c_int64)),
                ctypes.byref(out_ptr_indptr),
                ctypes.byref(out_ptr_indices),
                ctypes.byref(out_ptr_data)))
            matrices = self.__create_sparse_native(csc, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                   type_ptr_indptr, type_ptr_data, is_csr=False)
            nrow = csc.shape[0]
            return matrices, nrow

Guolin Ke's avatar
Guolin Ke committed
1122
        nrow = csc.shape[0]
1123
        if nrow > MAX_INT32:
1124
            return self.__pred_for_csr(csc.tocsr(), start_iteration, num_iteration, predict_type)
1125
        if predict_type == C_API_PREDICT_CONTRIB:
1126
1127
            return inner_predict_sparse(csc, start_iteration, num_iteration, predict_type)
        n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
1128
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1129
1130
        out_num_preds = ctypes.c_int64(0)

1131
1132
        ptr_indptr, type_ptr_indptr, __ = c_int_array(csc.indptr)
        ptr_data, type_ptr_data, _ = c_float_array(csc.data)
Guolin Ke's avatar
Guolin Ke committed
1133

1134
        assert csc.shape[0] <= MAX_INT32
1135
        csc_indices = csc.indices.astype(np.int32, copy=False)
1136

Guolin Ke's avatar
Guolin Ke committed
1137
1138
1139
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
1140
            ctypes.c_int(type_ptr_indptr),
1141
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1142
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1143
1144
1145
1146
1147
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
            ctypes.c_int(predict_type),
1148
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1149
            ctypes.c_int(num_iteration),
1150
            c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1151
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1152
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1153
        if n_preds != out_num_preds.value:
1154
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1155
1156
        return preds, nrow

1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
    def current_iteration(self):
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
        out_cur_iter = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1171

1172
class Dataset:
wxchan's avatar
wxchan committed
1173
    """Dataset in LightGBM."""
1174

1175
    def __init__(self, data, label=None, reference=None,
1176
                 weight=None, group=None, init_score=None, silent='warn',
1177
                 feature_name='auto', categorical_feature='auto', params=None,
wxchan's avatar
wxchan committed
1178
                 free_raw_data=True):
1179
        """Initialize Dataset.
1180

wxchan's avatar
wxchan committed
1181
1182
        Parameters
        ----------
1183
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence or list of numpy array
wxchan's avatar
wxchan committed
1184
            Data source of Dataset.
1185
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1186
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1187
1188
1189
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1190
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
1191
            Weight for each instance.
1192
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1193
1194
1195
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1196
1197
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
1198
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None, optional (default=None)
1199
            Init score for Dataset.
1200
1201
        silent : bool, optional (default=False)
            Whether to print messages during construction.
1202
        feature_name : list of str, or 'auto', optional (default="auto")
1203
1204
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
1205
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1206
1207
            Categorical features.
            If list of int, interpreted as indices.
1208
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1209
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1210
            All values in categorical features should be less than int32 max value (2147483647).
1211
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1212
            All negative values in categorical features will be treated as missing values.
1213
            The output cannot be monotonically constrained with respect to a categorical feature.
Nikita Titov's avatar
Nikita Titov committed
1214
        params : dict or None, optional (default=None)
1215
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1216
        free_raw_data : bool, optional (default=True)
1217
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
1218
        """
wxchan's avatar
wxchan committed
1219
1220
1221
1222
1223
1224
        self.handle = None
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1225
        self.init_score = init_score
wxchan's avatar
wxchan committed
1226
1227
        self.silent = silent
        self.feature_name = feature_name
1228
        self.categorical_feature = categorical_feature
1229
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1230
1231
        self.free_raw_data = free_raw_data
        self.used_indices = None
1232
        self.need_slice = True
wxchan's avatar
wxchan committed
1233
        self._predictor = None
1234
        self.pandas_categorical = None
1235
        self.params_back_up = None
1236
1237
        self.feature_penalty = None
        self.monotone_constraints = None
1238
        self.version = 0
1239
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1240
1241

    def __del__(self):
1242
1243
1244
1245
        try:
            self._free_handle()
        except AttributeError:
            pass
1246

1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
    def _create_sample_indices(self, total_nrow: int) -> np.ndarray:
        """Get an array of randomly chosen indices from this ``Dataset``.

        Indices are sampled without replacement.

        Parameters
        ----------
        total_nrow : int
            Total number of rows to sample from.
            If this value is greater than the value of parameter ``bin_construct_sample_cnt``, only ``bin_construct_sample_cnt`` indices will be used.
            If Dataset has multiple input data, this should be the sum of rows of every file.

        Returns
        -------
        indices : numpy array
            Indices for sampled data.
        """
        param_str = param_dict_to_str(self.get_params())
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
        ptr_data, _, _ = c_int_array(indices)
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
            c_str(param_str),
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1276
1277
        assert sample_cnt == actual_sample_cnt.value
        return indices
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312

    def _init_from_ref_dataset(self, total_nrow: int, ref_dataset: 'Dataset') -> 'Dataset':
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
        ref_dataset : Dataset
            Reference dataset to extract meta from.

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
        self.handle = ctypes.c_void_p()
        _safe_call(_LIB.LGBM_DatasetCreateByReference(
            ref_dataset,
            ctypes.c_int64(total_nrow),
            ctypes.byref(self.handle),
        ))
        return self

    def _init_from_sample(
        self,
        sample_data: List[np.ndarray],
        sample_indices: List[np.ndarray],
        sample_cnt: int,
        total_nrow: int,
    ) -> "Dataset":
        """Create Dataset from sampled data structures.

        Parameters
        ----------
1313
        sample_data : list of numpy array
1314
            Sample data for each column.
1315
        sample_indices : list of numpy array
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
            Sample data row index for each column.
        sample_cnt : int
            Number of samples.
        total_nrow : int
            Total number of rows for all input files.

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
        ncol = len(sample_indices)
        assert len(sample_data) == ncol, "#sample data column != #column indices"

        for i in range(ncol):
            if sample_data[i].dtype != np.double:
                raise ValueError(f"sample_data[{i}] type {sample_data[i].dtype} is not double")
            if sample_indices[i].dtype != np.int32:
                raise ValueError(f"sample_indices[{i}] type {sample_indices[i].dtype} is not int32")

        # c type: double**
        # each double* element points to start of each column of sample data.
        sample_col_ptr = (ctypes.POINTER(ctypes.c_double) * ncol)()
        # c type int**
        # each int* points to start of indices for each column
        indices_col_ptr = (ctypes.POINTER(ctypes.c_int32) * ncol)()
        for i in range(ncol):
            sample_col_ptr[i] = c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = c_int_array(sample_indices[i])[0]

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
        num_per_col_ptr, _, _ = c_int_array(num_per_col)

        self.handle = ctypes.c_void_p()
        params_str = param_dict_to_str(self.get_params())
        _safe_call(_LIB.LGBM_DatasetCreateFromSampledColumn(
            ctypes.cast(sample_col_ptr, ctypes.POINTER(ctypes.POINTER(ctypes.c_double))),
            ctypes.cast(indices_col_ptr, ctypes.POINTER(ctypes.POINTER(ctypes.c_int32))),
            ctypes.c_int32(ncol),
            num_per_col_ptr,
            ctypes.c_int32(sample_cnt),
            ctypes.c_int32(total_nrow),
            c_str(params_str),
            ctypes.byref(self.handle),
        ))
        return self

    def _push_rows(self, data: np.ndarray) -> 'Dataset':
        """Add rows to Dataset.

        Parameters
        ----------
        data : numpy 1-D array
            New data to add to the Dataset.

        Returns
        -------
        self : Dataset
            Dataset object.
        """
        nrow, ncol = data.shape
        data = data.reshape(data.size)
        data_ptr, data_type, _ = c_float_array(data)

        _safe_call(_LIB.LGBM_DatasetPushRows(
            self.handle,
            data_ptr,
            data_type,
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol),
            ctypes.c_int32(self._start_row),
        ))
        self._start_row += nrow
        return self

1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
    def get_params(self):
        """Get the used parameters in the Dataset.

        Returns
        -------
        params : dict or None
            The used parameters in this Dataset object.
        """
        if self.params is not None:
            # no min_data, nthreads and verbose in this function
            dataset_params = _ConfigAliases.get("bin_construct_sample_cnt",
                                                "categorical_feature",
                                                "data_random_seed",
                                                "enable_bundle",
                                                "feature_pre_filter",
                                                "forcedbins_filename",
                                                "group_column",
                                                "header",
                                                "ignore_column",
                                                "is_enable_sparse",
                                                "label_column",
1412
                                                "linear_tree",
1413
1414
1415
1416
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1417
                                                "precise_float_parser",
1418
1419
1420
1421
1422
1423
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}

1424
    def _free_handle(self):
1425
        if self.handle is not None:
1426
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1427
            self.handle = None
Guolin Ke's avatar
Guolin Ke committed
1428
1429
1430
        self.need_slice = True
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1431
        return self
wxchan's avatar
wxchan committed
1432

Guolin Ke's avatar
Guolin Ke committed
1433
1434
    def _set_init_score_by_predictor(self, predictor, data, used_indices=None):
        data_has_header = False
1435
        if isinstance(data, (str, Path)):
Guolin Ke's avatar
Guolin Ke committed
1436
            # check data has header or not
1437
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1438
        num_data = self.num_data()
1439
1440
1441
1442
1443
1444
1445
        if predictor is not None:
            init_score = predictor.predict(data,
                                           raw_score=True,
                                           data_has_header=data_has_header,
                                           is_reshape=False)
            if used_indices is not None:
                assert not self.need_slice
1446
                if isinstance(data, (str, Path)):
1447
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1448
                    assert num_data == len(used_indices)
1449
1450
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1451
1452
1453
1454
                            sub_init_score[i * predictor.num_class + j] = init_score[used_indices[i] * predictor.num_class + j]
                    init_score = sub_init_score
            if predictor.num_class > 1:
                # need to regroup init_score
1455
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1456
1457
                for i in range(num_data):
                    for j in range(predictor.num_class):
1458
1459
1460
                        new_init_score[j * num_data + i] = init_score[i * predictor.num_class + j]
                init_score = new_init_score
        elif self.init_score is not None:
1461
            init_score = np.zeros(self.init_score.shape, dtype=np.float64)
1462
1463
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1464
1465
        self.set_init_score(init_score)

1466
    def _lazy_init(self, data, label=None, reference=None,
1467
                   weight=None, group=None, init_score=None, predictor=None,
wxchan's avatar
wxchan committed
1468
                   silent=False, feature_name='auto',
1469
                   categorical_feature='auto', params=None):
wxchan's avatar
wxchan committed
1470
1471
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1472
            return self
Guolin Ke's avatar
Guolin Ke committed
1473
1474
1475
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1476
1477
1478
1479
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
wxchan's avatar
wxchan committed
1480
        label = _label_from_pandas(label)
Guolin Ke's avatar
Guolin Ke committed
1481

1482
        # process for args
wxchan's avatar
wxchan committed
1483
        params = {} if params is None else params
1484
1485
1486
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1487
        for key in params.keys():
1488
            if key in args_names:
1489
1490
                _log_warning(f'{key} keyword has been found in `params` and will be ignored.\n'
                             f'Please use {key} argument of the Dataset constructor to pass this parameter.')
1491
        # user can set verbose with params, it has higher priority
1492
1493
1494
1495
1496
        if silent != "warn":
            _log_warning("'silent' argument is deprecated and will be removed in a future release of LightGBM. "
                         "Pass 'verbose' parameter via 'params' instead.")
        else:
            silent = False
1497
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
1498
            params["verbose"] = -1
1499
        # get categorical features
1500
1501
1502
1503
1504
1505
        if categorical_feature is not None:
            categorical_indices = set()
            feature_dict = {}
            if feature_name is not None:
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
1506
                if isinstance(name, str) and name in feature_dict:
1507
                    categorical_indices.add(feature_dict[name])
1508
                elif isinstance(name, int):
1509
1510
                    categorical_indices.add(name)
                else:
1511
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1512
            if categorical_indices:
1513
1514
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1515
1516
1517
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
                        if not(isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1518
                        params.pop(cat_alias, None)
1519
                params['categorical_column'] = sorted(categorical_indices)
1520

wxchan's avatar
wxchan committed
1521
        params_str = param_dict_to_str(params)
1522
        self.params = params
1523
        # process for reference dataset
wxchan's avatar
wxchan committed
1524
        ref_dataset = None
wxchan's avatar
wxchan committed
1525
        if isinstance(reference, Dataset):
1526
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1527
1528
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1529
        # start construct data
1530
        if isinstance(data, (str, Path)):
wxchan's avatar
wxchan committed
1531
1532
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1533
                c_str(str(data)),
wxchan's avatar
wxchan committed
1534
1535
1536
1537
1538
                c_str(params_str),
                ref_dataset,
                ctypes.byref(self.handle)))
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
1539
1540
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1541
1542
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1543
1544
1545
1546
1547
1548
1549
1550
1551
        elif isinstance(data, list) and len(data) > 0:
            if all(isinstance(x, np.ndarray) for x in data):
                self.__init_from_list_np2d(data, params_str, ref_dataset)
            elif all(isinstance(x, Sequence) for x in data):
                self.__init_from_seqs(data, ref_dataset)
            else:
                raise TypeError('Data list can only be of ndarray or Sequence')
        elif isinstance(data, Sequence):
            self.__init_from_seqs([data], ref_dataset)
1552
        elif isinstance(data, dt_DataTable):
1553
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1554
1555
1556
1557
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1558
            except BaseException:
1559
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}')
wxchan's avatar
wxchan committed
1560
1561
1562
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1563
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1564
1565
1566
1567
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1568
1569
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1570
                _log_warning("The init_score will be overridden by the prediction of init_model.")
Guolin Ke's avatar
Guolin Ke committed
1571
            self._set_init_score_by_predictor(predictor, data)
1572
1573
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1574
        elif predictor is not None:
1575
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1576
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1577
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1578

1579
1580
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
        offset = 0
        seq_id = 0
        seq = seqs[seq_id]
        for row_id in indices:
            assert row_id >= offset, "sample indices are expected to be monotonic"
            while row_id >= offset + len(seq):
                offset += len(seq)
                seq_id += 1
                seq = seqs[seq_id]
            id_in_seq = row_id - offset
            row = seq[id_in_seq]
            yield row if row.flags['OWNDATA'] else row.copy()

    def __sample(self, seqs: List[Sequence], total_nrow: int) -> Tuple[List[np.ndarray], List[np.ndarray]]:
        """Sample data from seqs.

        Mimics behavior in c_api.cpp:LGBM_DatasetCreateFromMats()

        Returns
        -------
            sampled_rows, sampled_row_indices
        """
        indices = self._create_sample_indices(total_nrow)

        # Select sampled rows, transpose to column order.
1606
        sampled = np.array([row for row in self._yield_row_from_seqlist(seqs, indices)])
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
        sampled = sampled.T

        filtered = []
        filtered_idx = []
        sampled_row_range = np.arange(len(indices), dtype=np.int32)
        for col in sampled:
            col_predicate = (np.abs(col) > ZERO_THRESHOLD) | np.isnan(col)
            filtered_col = col[col_predicate]
            filtered_row_idx = sampled_row_range[col_predicate]

            filtered.append(filtered_col)
            filtered_idx.append(filtered_row_idx)

        return filtered, filtered_idx

    def __init_from_seqs(self, seqs: List[Sequence], ref_dataset: Optional['Dataset'] = None):
        """
        Initialize data from list of Sequence objects.

        Sequence: Generic Data Access Object
            Supports random access and access by batch if properly defined by user

        Data scheme uniformity are trusted, not checked
        """
        total_nrow = sum(len(seq) for seq in seqs)

        # create validation dataset from ref_dataset
        if ref_dataset is not None:
            self._init_from_ref_dataset(total_nrow, ref_dataset)
        else:
            param_str = param_dict_to_str(self.get_params())
            sample_cnt = _get_sample_count(total_nrow, param_str)

            sample_data, col_indices = self.__sample(seqs, total_nrow)
            self._init_from_sample(sample_data, col_indices, sample_cnt, total_nrow)

        for seq in seqs:
            nrow = len(seq)
            batch_size = getattr(seq, 'batch_size', None) or Sequence.batch_size
            for start in range(0, nrow, batch_size):
                end = min(start + batch_size, nrow)
                self._push_rows(seq[start:end])
        return self

wxchan's avatar
wxchan committed
1651
    def __init_from_np2d(self, mat, params_str, ref_dataset):
1652
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1653
1654
1655
1656
1657
1658
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

        self.handle = ctypes.c_void_p()
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
1659
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1660
1661
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1662
        ptr_data, type_ptr_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1663
1664
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1665
            ctypes.c_int(type_ptr_data),
1666
1667
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
Guolin Ke's avatar
Guolin Ke committed
1668
            ctypes.c_int(C_API_IS_ROW_MAJOR),
wxchan's avatar
wxchan committed
1669
1670
1671
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1672
        return self
wxchan's avatar
wxchan committed
1673

1674
    def __init_from_list_np2d(self, mats, params_str, ref_dataset):
1675
        """Initialize data from a list of 2-D numpy matrices."""
1676
        ncol = mats[0].shape[1]
1677
        nrow = np.empty((len(mats),), np.int32)
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
        if mats[0].dtype == np.float64:
            ptr_data = (ctypes.POINTER(ctypes.c_double) * len(mats))()
        else:
            ptr_data = (ctypes.POINTER(ctypes.c_float) * len(mats))()

        holders = []
        type_ptr_data = None

        for i, mat in enumerate(mats):
            if len(mat.shape) != 2:
                raise ValueError('Input numpy.ndarray must be 2 dimensional')

            if mat.shape[1] != ncol:
                raise ValueError('Input arrays must have same number of columns')

            nrow[i] = mat.shape[0]

            if mat.dtype == np.float32 or mat.dtype == np.float64:
                mats[i] = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
1697
            else:  # change non-float data to float data, need to copy
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

            chunk_ptr_data, chunk_type_ptr_data, holder = c_float_array(mats[i])
            if type_ptr_data is not None and chunk_type_ptr_data != type_ptr_data:
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

        self.handle = ctypes.c_void_p()
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
1709
            ctypes.c_int32(len(mats)),
1710
1711
1712
            ctypes.cast(ptr_data, ctypes.POINTER(ctypes.POINTER(ctypes.c_double))),
            ctypes.c_int(type_ptr_data),
            nrow.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
1713
            ctypes.c_int32(ncol),
1714
1715
1716
1717
            ctypes.c_int(C_API_IS_ROW_MAJOR),
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1718
        return self
1719

wxchan's avatar
wxchan committed
1720
    def __init_from_csr(self, csr, params_str, ref_dataset):
1721
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
1722
        if len(csr.indices) != len(csr.data):
1723
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
wxchan's avatar
wxchan committed
1724
1725
        self.handle = ctypes.c_void_p()

1726
1727
        ptr_indptr, type_ptr_indptr, __ = c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = c_float_array(csr.data)
wxchan's avatar
wxchan committed
1728

1729
        assert csr.shape[1] <= MAX_INT32
1730
        csr_indices = csr.indices.astype(np.int32, copy=False)
1731

wxchan's avatar
wxchan committed
1732
1733
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1734
            ctypes.c_int(type_ptr_indptr),
1735
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
1736
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1737
1738
1739
1740
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
wxchan's avatar
wxchan committed
1741
1742
1743
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1744
        return self
wxchan's avatar
wxchan committed
1745

Guolin Ke's avatar
Guolin Ke committed
1746
    def __init_from_csc(self, csc, params_str, ref_dataset):
1747
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
1748
        if len(csc.indices) != len(csc.data):
1749
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
Guolin Ke's avatar
Guolin Ke committed
1750
1751
        self.handle = ctypes.c_void_p()

1752
1753
        ptr_indptr, type_ptr_indptr, __ = c_int_array(csc.indptr)
        ptr_data, type_ptr_data, _ = c_float_array(csc.data)
Guolin Ke's avatar
Guolin Ke committed
1754

1755
        assert csc.shape[0] <= MAX_INT32
1756
        csc_indices = csc.indices.astype(np.int32, copy=False)
1757

Guolin Ke's avatar
Guolin Ke committed
1758
1759
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1760
            ctypes.c_int(type_ptr_indptr),
1761
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1762
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1763
1764
1765
1766
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
Guolin Ke's avatar
Guolin Ke committed
1767
1768
1769
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1770
        return self
Guolin Ke's avatar
Guolin Ke committed
1771

1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
    @staticmethod
    def _compare_params_for_warning(params, other_params):
        """Compare params.

        It is only for the warning purpose. Thus some keys are ignored.

        Returns
        -------
        compare_result: bool
          If they are equal, return True; Otherwise, return False.
        """
        ignore_keys = _ConfigAliases.get("categorical_feature")
        if params is None:
            params = {}
        if other_params is None:
            other_params = {}
        for k in other_params:
            if k not in ignore_keys:
                if k not in params or params[k] != other_params[k]:
                    return False
        for k in params:
            if k not in ignore_keys:
                if k not in other_params or params[k] != other_params[k]:
                    return False
        return True

wxchan's avatar
wxchan committed
1798
    def construct(self):
1799
1800
1801
1802
1803
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
1804
            Constructed Dataset object.
1805
        """
1806
        if self.handle is None:
wxchan's avatar
wxchan committed
1807
            if self.reference is not None:
1808
                reference_params = self.reference.get_params()
1809
1810
1811
1812
                params = self.get_params()
                if params != reference_params:
                    if self._compare_params_for_warning(params, reference_params) is False:
                        _log_warning('Overriding the parameters from Reference Dataset.')
1813
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
1814
                if self.used_indices is None:
1815
                    # create valid
1816
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
1817
1818
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
1819
                                    silent=self.silent, feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
1820
                else:
1821
                    # construct subset
wxchan's avatar
wxchan committed
1822
                    used_indices = list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
1823
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
1824
                    if self.reference.group is not None:
1825
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
1826
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
1827
                                                  return_counts=True)
1828
                    self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1829
1830
                    params_str = param_dict_to_str(self.params)
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
1831
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
1832
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
1833
                        ctypes.c_int32(used_indices.shape[0]),
wxchan's avatar
wxchan committed
1834
1835
                        c_str(params_str),
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
1836
1837
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
1838
1839
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
1840
1841
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
1842
1843
1844
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
                        self._set_init_score_by_predictor(self._predictor, self.data, used_indices)
wxchan's avatar
wxchan committed
1845
            else:
1846
                # create train
1847
                self._lazy_init(self.data, label=self.label,
1848
1849
1850
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
                                silent=self.silent, feature_name=self.feature_name,
1851
                                categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
1852
1853
1854
            if self.free_raw_data:
                self.data = None
        return self
wxchan's avatar
wxchan committed
1855

wxchan's avatar
wxchan committed
1856
    def create_valid(self, data, label=None, weight=None, group=None,
1857
                     init_score=None, silent='warn', params=None):
1858
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
1859
1860
1861

        Parameters
        ----------
1862
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence or list of numpy array
wxchan's avatar
wxchan committed
1863
            Data source of Dataset.
1864
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1865
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1866
1867
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
1868
            Weight for each instance.
1869
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1870
1871
1872
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1873
1874
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
1875
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None, optional (default=None)
1876
            Init score for Dataset.
1877
1878
        silent : bool, optional (default=False)
            Whether to print messages during construction.
Nikita Titov's avatar
Nikita Titov committed
1879
        params : dict or None, optional (default=None)
1880
            Other parameters for validation Dataset.
1881
1882
1883

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1884
1885
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
1886
        """
1887
        ret = Dataset(data, label=label, reference=self,
1888
1889
                      weight=weight, group=group, init_score=init_score,
                      silent=silent, params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1890
        ret._predictor = self._predictor
1891
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
1892
        return ret
wxchan's avatar
wxchan committed
1893

wxchan's avatar
wxchan committed
1894
    def subset(self, used_indices, params=None):
1895
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
1896
1897
1898
1899

        Parameters
        ----------
        used_indices : list of int
1900
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
1901
        params : dict or None, optional (default=None)
1902
            These parameters will be passed to Dataset constructor.
1903
1904
1905
1906
1907

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
1908
        """
wxchan's avatar
wxchan committed
1909
1910
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
1911
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
1912
1913
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1914
        ret._predictor = self._predictor
1915
        ret.pandas_categorical = self.pandas_categorical
1916
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
1917
1918
1919
        return ret

    def save_binary(self, filename):
1920
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
1921

1922
1923
1924
1925
1926
        .. note::

            Please note that `init_score` is not saved in binary file.
            If you need it, please set it again after loading Dataset.

wxchan's avatar
wxchan committed
1927
1928
        Parameters
        ----------
1929
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
1930
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
1931
1932
1933
1934
1935

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
1936
1937
1938
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
1939
            c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
1940
        return self
wxchan's avatar
wxchan committed
1941
1942

    def _update_params(self, params):
1943
1944
        if not params:
            return self
1945
        params = deepcopy(params)
1946
1947
1948
1949
1950

        def update():
            if not self.params:
                self.params = params
            else:
1951
                self.params_back_up = deepcopy(self.params)
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
                self.params.update(params)

        if self.handle is None:
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
                c_str(param_dict_to_str(self.params)),
                c_str(param_dict_to_str(params)))
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
1966
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
1967
        return self
wxchan's avatar
wxchan committed
1968

1969
    def _reverse_update_params(self):
1970
        if self.handle is None:
1971
            self.params = deepcopy(self.params_back_up)
1972
            self.params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
1973
        return self
1974

wxchan's avatar
wxchan committed
1975
    def set_field(self, field_name, data):
wxchan's avatar
wxchan committed
1976
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
1977
1978
1979

        Parameters
        ----------
1980
        field_name : str
1981
            The field name of the information.
1982
1983
        data : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
            The data to be set.
Nikita Titov's avatar
Nikita Titov committed
1984
1985
1986
1987
1988

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
1989
        """
1990
        if self.handle is None:
1991
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
1992
        if data is None:
1993
            # set to None
wxchan's avatar
wxchan committed
1994
1995
1996
1997
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
                c_str(field_name),
                None,
Guolin Ke's avatar
Guolin Ke committed
1998
1999
                ctypes.c_int(0),
                ctypes.c_int(FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2000
            return self
2001
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2002
            dtype = np.float64
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
            if _is_1d_collection(data):
                data = list_to_1d_numpy(data, dtype, name=field_name)
            elif _is_2d_collection(data):
                data = _data_to_2d_numpy(data, dtype, name=field_name)
                data = data.ravel(order='F')
            else:
                raise TypeError(
                    'init_score must be list, numpy 1-D array or pandas Series.\n'
                    'In multiclass classification init_score can also be a list of lists, numpy 2-D array or pandas DataFrame.'
                )
        else:
            dtype = np.int32 if field_name == 'group' else np.float32
            data = list_to_1d_numpy(data, dtype, name=field_name)

2017
2018
        if data.dtype == np.float32 or data.dtype == np.float64:
            ptr_data, type_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
2019
        elif data.dtype == np.int32:
2020
            ptr_data, type_data, _ = c_int_array(data)
wxchan's avatar
wxchan committed
2021
        else:
2022
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
wxchan's avatar
wxchan committed
2023
        if type_data != FIELD_TYPE_MAPPER[field_name]:
2024
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2025
2026
2027
2028
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
            c_str(field_name),
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2029
2030
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2031
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2032
        return self
wxchan's avatar
wxchan committed
2033

wxchan's avatar
wxchan committed
2034
2035
    def get_field(self, field_name):
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2036
2037
2038

        Parameters
        ----------
2039
        field_name : str
2040
            The field name of the information.
wxchan's avatar
wxchan committed
2041
2042
2043

        Returns
        -------
2044
        info : numpy array or None
2045
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2046
        """
2047
        if self.handle is None:
2048
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2049
2050
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
            self.handle,
            c_str(field_name),
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
        if out_type.value != FIELD_TYPE_MAPPER[field_name]:
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
        if out_type.value == C_API_DTYPE_INT32:
2063
            arr = cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
wxchan's avatar
wxchan committed
2064
        elif out_type.value == C_API_DTYPE_FLOAT32:
2065
            arr = cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
Guolin Ke's avatar
Guolin Ke committed
2066
        elif out_type.value == C_API_DTYPE_FLOAT64:
2067
            arr = cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2068
        else:
wxchan's avatar
wxchan committed
2069
            raise TypeError("Unknown type")
2070
2071
2072
2073
2074
2075
        if field_name == 'init_score':
            num_data = self.num_data()
            num_classes = arr.size // num_data
            if num_classes > 1:
                arr = arr.reshape((num_data, num_classes), order='F')
        return arr
Guolin Ke's avatar
Guolin Ke committed
2076

2077
    def set_categorical_feature(self, categorical_feature):
2078
        """Set categorical features.
2079
2080
2081

        Parameters
        ----------
2082
        categorical_feature : list of int or str
2083
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2084
2085
2086
2087
2088

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2089
2090
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2091
            return self
2092
        if self.data is not None:
2093
2094
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2095
                return self._free_handle()
2096
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2097
                return self
2098
            else:
2099
2100
2101
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
                                 f'New categorical_feature is {sorted(list(categorical_feature))}')
2102
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2103
                return self._free_handle()
2104
        else:
2105
2106
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2107

Guolin Ke's avatar
Guolin Ke committed
2108
    def _set_predictor(self, predictor):
2109
2110
2111
2112
        """Set predictor for continued training.

        It is not recommended for user to call this function.
        Please use init_model argument in engine.train() or engine.cv() instead.
Guolin Ke's avatar
Guolin Ke committed
2113
        """
2114
        if predictor is self._predictor and (predictor is None or predictor.current_iteration() == self._predictor.current_iteration()):
Nikita Titov's avatar
Nikita Titov committed
2115
            return self
2116
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
2117
            self._predictor = predictor
2118
2119
2120
2121
2122
2123
        elif self.data is not None:
            self._predictor = predictor
            self._set_init_score_by_predictor(self._predictor, self.data)
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
            self._set_init_score_by_predictor(self._predictor, self.reference.data, self.used_indices)
Guolin Ke's avatar
Guolin Ke committed
2124
        else:
2125
2126
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2127
        return self
Guolin Ke's avatar
Guolin Ke committed
2128
2129

    def set_reference(self, reference):
2130
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2131
2132
2133
2134

        Parameters
        ----------
        reference : Dataset
2135
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2136
2137
2138
2139
2140

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2141
        """
2142
2143
2144
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2145
        # we're done if self and reference share a common upstream reference
2146
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2147
            return self
Guolin Ke's avatar
Guolin Ke committed
2148
2149
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2150
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2151
        else:
2152
2153
            raise LightGBMError("Cannot set reference after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
Guolin Ke's avatar
Guolin Ke committed
2154
2155

    def set_feature_name(self, feature_name):
2156
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2157
2158
2159

        Parameters
        ----------
2160
        feature_name : list of str
2161
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2162
2163
2164
2165
2166

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2167
        """
2168
2169
        if feature_name != 'auto':
            self.feature_name = feature_name
2170
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2171
            if len(feature_name) != self.num_feature():
2172
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2173
            c_feature_name = [c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2174
2175
2176
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
                c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2177
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2178
        return self
Guolin Ke's avatar
Guolin Ke committed
2179
2180

    def set_label(self, label):
2181
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2182
2183
2184

        Parameters
        ----------
2185
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2186
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2187
2188
2189
2190
2191

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2192
2193
        """
        self.label = label
2194
        if self.handle is not None:
2195
            label = list_to_1d_numpy(_label_from_pandas(label), name='label')
wxchan's avatar
wxchan committed
2196
            self.set_field('label', label)
2197
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2198
        return self
Guolin Ke's avatar
Guolin Ke committed
2199
2200

    def set_weight(self, weight):
2201
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2202
2203
2204

        Parameters
        ----------
2205
        weight : list, numpy 1-D array, pandas Series or None
2206
            Weight to be set for each data point.
Nikita Titov's avatar
Nikita Titov committed
2207
2208
2209
2210
2211

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2212
        """
2213
2214
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2215
        self.weight = weight
2216
        if self.handle is not None and weight is not None:
wxchan's avatar
wxchan committed
2217
2218
            weight = list_to_1d_numpy(weight, name='weight')
            self.set_field('weight', weight)
2219
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2220
        return self
Guolin Ke's avatar
Guolin Ke committed
2221
2222

    def set_init_score(self, init_score):
2223
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2224
2225
2226

        Parameters
        ----------
2227
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2228
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2229
2230
2231
2232
2233

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2234
2235
        """
        self.init_score = init_score
2236
        if self.handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2237
            self.set_field('init_score', init_score)
2238
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2239
        return self
Guolin Ke's avatar
Guolin Ke committed
2240
2241

    def set_group(self, group):
2242
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2243
2244
2245

        Parameters
        ----------
2246
        group : list, numpy 1-D array, pandas Series or None
2247
2248
2249
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2250
2251
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
Nikita Titov's avatar
Nikita Titov committed
2252
2253
2254
2255
2256

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2257
2258
        """
        self.group = group
2259
        if self.handle is not None and group is not None:
wxchan's avatar
wxchan committed
2260
2261
            group = list_to_1d_numpy(group, np.int32, name='group')
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2262
        return self
Guolin Ke's avatar
Guolin Ke committed
2263

2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
    def get_feature_name(self):
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
        feature_names : list
            The names of columns (features) in the Dataset.
        """
        if self.handle is None:
            raise LightGBMError("Cannot get feature_name before construct dataset")
        num_feature = self.num_feature()
        tmp_out_len = ctypes.c_int(0)
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
2278
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2279
2280
2281
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
2282
            ctypes.c_int(num_feature),
2283
            ctypes.byref(tmp_out_len),
2284
            ctypes.c_size_t(reserved_string_buffer_size),
2285
2286
2287
2288
            ctypes.byref(required_string_buffer_size),
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
                self.handle,
                ctypes.c_int(num_feature),
                ctypes.byref(tmp_out_len),
                ctypes.c_size_t(actual_string_buffer_size),
                ctypes.byref(required_string_buffer_size),
                ptr_string_buffers))
2301
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2302

Guolin Ke's avatar
Guolin Ke committed
2303
    def get_label(self):
2304
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2305
2306
2307

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2308
        label : numpy array or None
2309
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2310
        """
2311
        if self.label is None:
wxchan's avatar
wxchan committed
2312
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2313
2314
2315
        return self.label

    def get_weight(self):
2316
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2317
2318
2319

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2320
        weight : numpy array or None
2321
            Weight for each data point from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2322
        """
2323
        if self.weight is None:
wxchan's avatar
wxchan committed
2324
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2325
2326
2327
        return self.weight

    def get_init_score(self):
2328
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2329
2330
2331

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2332
        init_score : numpy array or None
2333
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2334
        """
2335
        if self.init_score is None:
wxchan's avatar
wxchan committed
2336
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2337
2338
        return self.init_score

2339
2340
2341
2342
2343
    def get_data(self):
        """Get the raw data of the Dataset.

        Returns
        -------
2344
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence or list of numpy array or None
2345
2346
2347
2348
            Raw data used in the Dataset construction.
        """
        if self.handle is None:
            raise Exception("Cannot get data before construct Dataset")
Guolin Ke's avatar
Guolin Ke committed
2349
2350
2351
2352
2353
        if self.need_slice and self.used_indices is not None and self.reference is not None:
            self.data = self.reference.data
            if self.data is not None:
                if isinstance(self.data, np.ndarray) or scipy.sparse.issparse(self.data):
                    self.data = self.data[self.used_indices, :]
2354
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2355
                    self.data = self.data.iloc[self.used_indices].copy()
2356
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2357
                    self.data = self.data[self.used_indices, :]
2358
2359
2360
2361
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
                elif isinstance(self.data, list) and len(self.data) > 0 and all(isinstance(x, Sequence) for x in self.data):
                    self.data = np.array([row for row in self._yield_row_from_seqlist(self.data, self.used_indices)])
Guolin Ke's avatar
Guolin Ke committed
2362
                else:
2363
2364
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
2365
            self.need_slice = False
Guolin Ke's avatar
Guolin Ke committed
2366
2367
2368
        if self.data is None:
            raise LightGBMError("Cannot call `get_data` after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2369
2370
        return self.data

Guolin Ke's avatar
Guolin Ke committed
2371
    def get_group(self):
2372
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2373
2374
2375

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2376
        group : numpy array or None
2377
2378
2379
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2380
2381
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
Guolin Ke's avatar
Guolin Ke committed
2382
        """
2383
        if self.group is None:
wxchan's avatar
wxchan committed
2384
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
2385
2386
            if self.group is not None:
                # group data from LightGBM is boundaries data, need to convert to group size
Nikita Titov's avatar
Nikita Titov committed
2387
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
2388
2389
2390
        return self.group

    def num_data(self):
2391
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2392
2393
2394

        Returns
        -------
2395
2396
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2397
        """
2398
        if self.handle is not None:
2399
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2400
2401
2402
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2403
        else:
2404
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2405
2406

    def num_feature(self):
2407
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2408
2409
2410

        Returns
        -------
2411
2412
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2413
        """
2414
        if self.handle is not None:
2415
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2416
2417
2418
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2419
        else:
2420
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2421

2422
    def get_ref_chain(self, ref_limit=100):
2423
2424
2425
2426
2427
        """Get a chain of Dataset objects.

        Starts with r, then goes to r.reference (if exists),
        then to r.reference.reference, etc.
        until we hit ``ref_limit`` or a reference loop.
2428
2429
2430
2431
2432

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2433
2434
2435

        Returns
        -------
2436
2437
2438
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2439
        head = self
2440
        ref_chain = set()
2441
2442
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2443
                ref_chain.add(head)
2444
2445
2446
2447
2448
2449
                if (head.reference is not None) and (head.reference not in ref_chain):
                    head = head.reference
                else:
                    break
            else:
                break
Nikita Titov's avatar
Nikita Titov committed
2450
        return ref_chain
2451

2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
    def add_features_from(self, other):
        """Add features from other Dataset to the current Dataset.

        Both Datasets must be constructed before calling this method.

        Parameters
        ----------
        other : Dataset
            The Dataset to take features from.

        Returns
        -------
        self : Dataset
            Dataset with the new features added.
        """
        if self.handle is None or other.handle is None:
            raise ValueError('Both source and target Datasets must be constructed before adding features')
        _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self.handle, other.handle))
Guolin Ke's avatar
Guolin Ke committed
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
        was_none = self.data is None
        old_self_data_type = type(self.data).__name__
        if other.data is None:
            self.data = None
        elif self.data is not None:
            if isinstance(self.data, np.ndarray):
                if isinstance(other.data, np.ndarray):
                    self.data = np.hstack((self.data, other.data))
                elif scipy.sparse.issparse(other.data):
                    self.data = np.hstack((self.data, other.data.toarray()))
2480
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2481
                    self.data = np.hstack((self.data, other.data.values))
2482
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2483
2484
2485
2486
2487
2488
2489
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
            elif scipy.sparse.issparse(self.data):
                sparse_format = self.data.getformat()
                if isinstance(other.data, np.ndarray) or scipy.sparse.issparse(other.data):
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
2490
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2491
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
2492
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2493
2494
2495
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
2496
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2497
2498
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
2499
2500
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
2501
                if isinstance(other.data, np.ndarray):
2502
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
2503
2504
                                       axis=1, ignore_index=True)
                elif scipy.sparse.issparse(other.data):
2505
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
2506
                                       axis=1, ignore_index=True)
2507
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2508
2509
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
2510
2511
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
2512
2513
2514
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
2515
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2516
                if isinstance(other.data, np.ndarray):
2517
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
Guolin Ke's avatar
Guolin Ke committed
2518
                elif scipy.sparse.issparse(other.data):
2519
2520
2521
2522
2523
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data.toarray())))
                elif isinstance(other.data, pd_DataFrame):
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data.values)))
                elif isinstance(other.data, dt_DataTable):
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data.to_numpy())))
Guolin Ke's avatar
Guolin Ke committed
2524
2525
2526
2527
2528
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
2529
2530
            err_msg = (f"Cannot add features from {type(other.data).__name__} type of raw data to "
                       f"{old_self_data_type} type of raw data.\n")
Guolin Ke's avatar
Guolin Ke committed
2531
2532
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
2533
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
2534
        self.feature_name = self.get_feature_name()
2535
2536
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
2537
2538
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2539
2540
        return self

2541
    def _dump_text(self, filename):
2542
2543
2544
2545
2546
2547
        """Save Dataset to a text file.

        This format cannot be loaded back in by LightGBM, but is useful for debugging purposes.

        Parameters
        ----------
2548
        filename : str or pathlib.Path
2549
2550
2551
2552
2553
2554
2555
2556
2557
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
            self.construct().handle,
2558
            c_str(str(filename))))
2559
2560
        return self

wxchan's avatar
wxchan committed
2561

2562
class Booster:
2563
    """Booster in LightGBM."""
2564

2565
    def __init__(self, params=None, train_set=None, model_file=None, model_str=None, silent='warn'):
2566
        """Initialize the Booster.
wxchan's avatar
wxchan committed
2567
2568
2569

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2570
        params : dict or None, optional (default=None)
2571
2572
2573
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
2574
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
2575
            Path to the model file.
2576
        model_str : str or None, optional (default=None)
2577
            Model will be loaded from this string.
2578
2579
        silent : bool, optional (default=False)
            Whether to print messages during construction.
wxchan's avatar
wxchan committed
2580
        """
2581
        self.handle = None
2582
        self.network = False
wxchan's avatar
wxchan committed
2583
        self.__need_reload_eval_info = True
2584
        self._train_data_name = "training"
wxchan's avatar
wxchan committed
2585
        self.__attr = {}
2586
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
2587
        self.best_iteration = -1
wxchan's avatar
wxchan committed
2588
        self.best_score = {}
2589
        params = {} if params is None else deepcopy(params)
2590
        # user can set verbose with params, it has higher priority
2591
2592
2593
2594
2595
        if silent != 'warn':
            _log_warning("'silent' argument is deprecated and will be removed in a future release of LightGBM. "
                         "Pass 'verbose' parameter via 'params' instead.")
        else:
            silent = False
2596
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
2597
            params["verbose"] = -1
wxchan's avatar
wxchan committed
2598
        if train_set is not None:
2599
            # Training task
wxchan's avatar
wxchan committed
2600
            if not isinstance(train_set, Dataset):
2601
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
            params = _choose_param_value(
                main_param_name="machines",
                params=params,
                default_value=None
            )
            # if "machines" is given, assume user wants to do distributed learning, and set up network
            if params["machines"] is None:
                params.pop("machines", None)
            else:
                machines = params["machines"]
                if isinstance(machines, str):
                    num_machines_from_machine_list = len(machines.split(','))
                elif isinstance(machines, (list, set)):
                    num_machines_from_machine_list = len(machines)
                    machines = ','.join(machines)
                else:
                    raise ValueError("Invalid machines in params.")

                params = _choose_param_value(
                    main_param_name="num_machines",
                    params=params,
                    default_value=num_machines_from_machine_list
                )
                params = _choose_param_value(
                    main_param_name="local_listen_port",
                    params=params,
                    default_value=12400
                )
                self.set_network(
                    machines=machines,
                    local_listen_port=params["local_listen_port"],
                    listen_time_out=params.get("time_out", 120),
                    num_machines=params["num_machines"]
                )
2636
            # construct booster object
2637
2638
2639
2640
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
            params_str = param_dict_to_str(params)
2641
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2642
            _safe_call(_LIB.LGBM_BoosterCreate(
2643
                train_set.handle,
wxchan's avatar
wxchan committed
2644
2645
                c_str(params_str),
                ctypes.byref(self.handle)))
2646
            # save reference to data
wxchan's avatar
wxchan committed
2647
2648
2649
2650
            self.train_set = train_set
            self.valid_sets = []
            self.name_valid_sets = []
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
2651
2652
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
2653
2654
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
2655
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
2656
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2657
2658
2659
2660
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2661
            # buffer for inner predict
wxchan's avatar
wxchan committed
2662
2663
2664
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
2665
            self.pandas_categorical = train_set.pandas_categorical
2666
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
2667
        elif model_file is not None:
2668
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
2669
            out_num_iterations = ctypes.c_int(0)
2670
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2671
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
2672
                c_str(str(model_file)),
wxchan's avatar
wxchan committed
2673
2674
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2675
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2676
2677
2678
2679
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2680
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
2681
        elif model_str is not None:
2682
            self.model_from_string(model_str, verbose="_silent_false")
wxchan's avatar
wxchan committed
2683
        else:
2684
2685
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
2686
        self.params = params
wxchan's avatar
wxchan committed
2687
2688

    def __del__(self):
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
        try:
            if self.network:
                self.free_network()
        except AttributeError:
            pass
        try:
            if self.handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
2699

wxchan's avatar
wxchan committed
2700
2701
2702
2703
    def __copy__(self):
        return self.__deepcopy__(None)

    def __deepcopy__(self, _):
2704
        model_str = self.model_to_string(num_iteration=-1)
2705
        booster = Booster(model_str=model_str)
2706
        return booster
wxchan's avatar
wxchan committed
2707
2708
2709
2710
2711
2712
2713

    def __getstate__(self):
        this = self.__dict__.copy()
        handle = this['handle']
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
2714
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
2715
2716
2717
        return this

    def __setstate__(self, state):
2718
2719
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
2720
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2721
            out_num_iterations = ctypes.c_int(0)
2722
2723
2724
2725
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
                c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
2726
2727
2728
            state['handle'] = handle
        self.__dict__.update(state)

wxchan's avatar
wxchan committed
2729
    def free_dataset(self):
Nikita Titov's avatar
Nikita Titov committed
2730
2731
2732
2733
2734
2735
2736
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
2737
2738
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
2739
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
2740
        return self
wxchan's avatar
wxchan committed
2741

2742
2743
2744
    def _free_buffer(self):
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
2745
        return self
2746

2747
2748
2749
2750
2751
2752
2753
    def set_network(
        self,
        machines: Union[List[str], Set[str], str],
        local_listen_port: int = 12400,
        listen_time_out: int = 120,
        num_machines: int = 1
    ) -> "Booster":
2754
2755
2756
2757
        """Set the network configuration.

        Parameters
        ----------
2758
        machines : list, set or str
2759
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
2760
        local_listen_port : int, optional (default=12400)
2761
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
2762
        listen_time_out : int, optional (default=120)
2763
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
2764
        num_machines : int, optional (default=1)
2765
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
2766
2767
2768
2769
2770

        Returns
        -------
        self : Booster
            Booster with set network.
2771
        """
2772
2773
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
2774
2775
2776
2777
2778
        _safe_call(_LIB.LGBM_NetworkInit(c_str(machines),
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
        self.network = True
Nikita Titov's avatar
Nikita Titov committed
2779
        return self
2780
2781

    def free_network(self):
Nikita Titov's avatar
Nikita Titov committed
2782
2783
2784
2785
2786
2787
2788
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
2789
2790
        _safe_call(_LIB.LGBM_NetworkFree())
        self.network = False
Nikita Titov's avatar
Nikita Titov committed
2791
        return self
2792

2793
2794
2795
    def trees_to_dataframe(self):
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

2796
2797
2798
2799
        The returned DataFrame has the following columns.

            - ``tree_index`` : int64, which tree a node belongs to. 0-based, so a value of ``6``, for example, means "this node is in the 7th tree".
            - ``node_depth`` : int64, how far a node is from the root of the tree. The root node has a value of ``1``, its direct children are ``2``, etc.
2800
2801
2802
2803
2804
            - ``node_index`` : str, unique identifier for a node.
            - ``left_child`` : str, ``node_index`` of the child node to the left of a split. ``None`` for leaf nodes.
            - ``right_child`` : str, ``node_index`` of the child node to the right of a split. ``None`` for leaf nodes.
            - ``parent_index`` : str, ``node_index`` of this node's parent. ``None`` for the root node.
            - ``split_feature`` : str, name of the feature used for splitting. ``None`` for leaf nodes.
2805
2806
            - ``split_gain`` : float64, gain from adding this split to the tree. ``NaN`` for leaf nodes.
            - ``threshold`` : float64, value of the feature used to decide which side of the split a record will go down. ``NaN`` for leaf nodes.
2807
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
2808
2809
              For example, ``split_feature = "Column_10", threshold = 15, decision_type = "<="`` means that
              records where ``Column_10 <= 15`` follow the left side of the split, otherwise follows the right side of the split. ``None`` for leaf nodes.
2810
2811
            - ``missing_direction`` : str, split direction that missing values should go to. ``None`` for leaf nodes.
            - ``missing_type`` : str, describes what types of values are treated as missing.
2812
2813
2814
2815
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
            - ``weight`` : float64 or int64, sum of hessian (second-order derivative of objective), summed over observations that fall in this node.
            - ``count`` : int64, number of records in the training data that fall into this node.

2816
2817
2818
2819
2820
2821
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
2822
2823
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834

        if self.num_trees() == 0:
            raise LightGBMError('There are no trees in this Booster and thus nothing to parse')

        def _is_split_node(tree):
            return 'split_index' in tree.keys()

        def create_node_record(tree, node_depth=1, tree_index=None,
                               feature_names=None, parent_node=None):

            def _get_node_index(tree, tree_index):
2835
                tree_num = f'{tree_index}-' if tree_index is not None else ''
2836
2837
2838
                is_split = _is_split_node(tree)
                node_type = 'S' if is_split else 'L'
                # if a single node tree it won't have `leaf_index` so return 0
2839
2840
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852

            def _get_split_feature(tree, feature_names):
                if _is_split_node(tree):
                    if feature_names is not None:
                        feature_name = feature_names[tree['split_feature']]
                    else:
                        feature_name = tree['split_feature']
                else:
                    feature_name = None
                return feature_name

            def _is_single_node_tree(tree):
2853
                return set(tree.keys()) == {'leaf_value'}
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926

            # Create the node record, and populate universal data members
            node = OrderedDict()
            node['tree_index'] = tree_index
            node['node_depth'] = node_depth
            node['node_index'] = _get_node_index(tree, tree_index)
            node['left_child'] = None
            node['right_child'] = None
            node['parent_index'] = parent_node
            node['split_feature'] = _get_split_feature(tree, feature_names)
            node['split_gain'] = None
            node['threshold'] = None
            node['decision_type'] = None
            node['missing_direction'] = None
            node['missing_type'] = None
            node['value'] = None
            node['weight'] = None
            node['count'] = None

            # Update values to reflect node type (leaf or split)
            if _is_split_node(tree):
                node['left_child'] = _get_node_index(tree['left_child'], tree_index)
                node['right_child'] = _get_node_index(tree['right_child'], tree_index)
                node['split_gain'] = tree['split_gain']
                node['threshold'] = tree['threshold']
                node['decision_type'] = tree['decision_type']
                node['missing_direction'] = 'left' if tree['default_left'] else 'right'
                node['missing_type'] = tree['missing_type']
                node['value'] = tree['internal_value']
                node['weight'] = tree['internal_weight']
                node['count'] = tree['internal_count']
            else:
                node['value'] = tree['leaf_value']
                if not _is_single_node_tree(tree):
                    node['weight'] = tree['leaf_weight']
                    node['count'] = tree['leaf_count']

            return node

        def tree_dict_to_node_list(tree, node_depth=1, tree_index=None,
                                   feature_names=None, parent_node=None):

            node = create_node_record(tree,
                                      node_depth=node_depth,
                                      tree_index=tree_index,
                                      feature_names=feature_names,
                                      parent_node=parent_node)

            res = [node]

            if _is_split_node(tree):
                # traverse the next level of the tree
                children = ['left_child', 'right_child']
                for child in children:
                    subtree_list = tree_dict_to_node_list(
                        tree[child],
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
                        parent_node=node['node_index'])
                    # In tree format, "subtree_list" is a list of node records (dicts),
                    # and we add node to the list.
                    res.extend(subtree_list)
            return res

        model_dict = self.dump_model()
        feature_names = model_dict['feature_names']
        model_list = []
        for tree in model_dict['tree_info']:
            model_list.extend(tree_dict_to_node_list(tree['tree_structure'],
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

2927
        return pd_DataFrame(model_list, columns=model_list[0].keys())
2928

wxchan's avatar
wxchan committed
2929
    def set_train_data_name(self, name):
2930
2931
2932
2933
        """Set the name to the training Dataset.

        Parameters
        ----------
2934
        name : str
Nikita Titov's avatar
Nikita Titov committed
2935
2936
2937
2938
2939
2940
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
2941
        """
2942
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
2943
        return self
wxchan's avatar
wxchan committed
2944
2945

    def add_valid(self, data, name):
2946
        """Add validation data.
wxchan's avatar
wxchan committed
2947
2948
2949
2950

        Parameters
        ----------
        data : Dataset
2951
            Validation data.
2952
        name : str
2953
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
2954
2955
2956
2957
2958

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
2959
        """
Guolin Ke's avatar
Guolin Ke committed
2960
        if not isinstance(data, Dataset):
2961
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2962
        if data._predictor is not self.__init_predictor:
2963
2964
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2965
2966
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
2967
            data.construct().handle))
wxchan's avatar
wxchan committed
2968
2969
2970
2971
2972
        self.valid_sets.append(data)
        self.name_valid_sets.append(name)
        self.__num_dataset += 1
        self.__inner_predict_buffer.append(None)
        self.__is_predicted_cur_iter.append(False)
Nikita Titov's avatar
Nikita Titov committed
2973
        return self
wxchan's avatar
wxchan committed
2974
2975

    def reset_parameter(self, params):
2976
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
2977
2978
2979
2980

        Parameters
        ----------
        params : dict
2981
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
2982
2983
2984
2985
2986

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
2987
2988
2989
2990
2991
2992
        """
        params_str = param_dict_to_str(params)
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
                self.handle,
                c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
2993
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
2994
        return self
wxchan's avatar
wxchan committed
2995
2996

    def update(self, train_set=None, fobj=None):
Nikita Titov's avatar
Nikita Titov committed
2997
        """Update Booster for one iteration.
2998

wxchan's avatar
wxchan committed
2999
3000
        Parameters
        ----------
3001
3002
3003
3004
        train_set : Dataset or None, optional (default=None)
            Training data.
            If None, last training data is used.
        fobj : callable or None, optional (default=None)
wxchan's avatar
wxchan committed
3005
            Customized objective function.
3006
3007
3008
3009
3010
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

                preds : list or numpy 1-D array
                    The predicted values.
3011
3012
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3013
3014
3015
                train_data : Dataset
                    The training dataset.
                grad : list or numpy 1-D array
3016
3017
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3018
                hess : list or numpy 1-D array
3019
3020
                    The value of the second order derivative (Hessian) of the loss
                    with respect to the elements of preds for each sample point.
wxchan's avatar
wxchan committed
3021

3022
3023
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is score[j * num_data + i]
3024
3025
            and you should group grad and hess in this way as well.

wxchan's avatar
wxchan committed
3026
3027
        Returns
        -------
3028
3029
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3030
        """
3031
        # need reset training data
3032
3033
3034
3035
3036
3037
        if train_set is None and self.train_set_version != self.train_set.version:
            train_set = self.train_set
            is_the_same_train_set = False
        else:
            is_the_same_train_set = train_set is self.train_set and self.train_set_version == train_set.version
        if train_set is not None and not is_the_same_train_set:
Guolin Ke's avatar
Guolin Ke committed
3038
            if not isinstance(train_set, Dataset):
3039
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3040
            if train_set._predictor is not self.__init_predictor:
3041
3042
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3043
3044
3045
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
3046
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
3047
            self.__inner_predict_buffer[0] = None
3048
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3049
3050
        is_finished = ctypes.c_int(0)
        if fobj is None:
3051
            if self.__set_objective_to_none:
3052
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3053
3054
3055
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
3056
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3057
3058
            return is_finished.value == 1
        else:
3059
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3060
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3061
3062
3063
3064
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

    def __boost(self, grad, hess):
3065
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3066

Nikita Titov's avatar
Nikita Titov committed
3067
3068
        .. note::

3069
3070
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
Nikita Titov's avatar
Nikita Titov committed
3071
3072
3073
            For multi-class task, the score is group by class_id first, then group by row_id.
            If you want to get i-th row score in j-th class, the access way is score[j * num_data + i]
            and you should group grad and hess in this way as well.
3074

wxchan's avatar
wxchan committed
3075
3076
        Parameters
        ----------
3077
        grad : list or numpy 1-D array
3078
3079
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3080
        hess : list or numpy 1-D array
3081
3082
            The value of the second order derivative (Hessian) of the loss
            with respect to the elements of score for each sample point.
wxchan's avatar
wxchan committed
3083
3084
3085

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3086
3087
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3088
        """
3089
3090
        grad = list_to_1d_numpy(grad, name='gradient')
        hess = list_to_1d_numpy(hess, name='hessian')
3091
3092
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3093
        if len(grad) != len(hess):
3094
            raise ValueError(f"Lengths of gradient({len(grad)}) and hessian({len(hess)}) don't match")
wxchan's avatar
wxchan committed
3095
3096
3097
3098
3099
3100
        is_finished = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterUpdateOneIterCustom(
            self.handle,
            grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            ctypes.byref(is_finished)))
3101
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3102
3103
3104
        return is_finished.value == 1

    def rollback_one_iter(self):
Nikita Titov's avatar
Nikita Titov committed
3105
3106
3107
3108
3109
3110
3111
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3112
3113
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
3114
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3115
        return self
wxchan's avatar
wxchan committed
3116
3117

    def current_iteration(self):
3118
3119
3120
3121
3122
3123
3124
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3125
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3126
3127
3128
3129
3130
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
    def num_model_per_iteration(self):
        """Get number of models per iteration.

        Returns
        -------
        model_per_iter : int
            The number of models per iteration.
        """
        model_per_iter = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterNumModelPerIteration(
            self.handle,
            ctypes.byref(model_per_iter)))
        return model_per_iter.value

    def num_trees(self):
        """Get number of weak sub-models.

        Returns
        -------
        num_trees : int
            The number of weak sub-models.
        """
        num_trees = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterNumberOfTotalModel(
            self.handle,
            ctypes.byref(num_trees)))
        return num_trees.value

3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
    def upper_bound(self):
        """Get upper bound value of a model.

        Returns
        -------
        upper_bound : double
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

    def lower_bound(self):
        """Get lower bound value of a model.

        Returns
        -------
        lower_bound : double
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

wxchan's avatar
wxchan committed
3187
    def eval(self, data, name, feval=None):
3188
        """Evaluate for data.
wxchan's avatar
wxchan committed
3189
3190
3191

        Parameters
        ----------
3192
3193
        data : Dataset
            Data for the evaluating.
3194
        name : str
3195
3196
            Name of the data.
        feval : callable or None, optional (default=None)
3197
            Customized evaluation function.
3198
            Should accept two parameters: preds, eval_data,
3199
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3200
3201
3202

                preds : list or numpy 1-D array
                    The predicted values.
3203
3204
                    If ``fobj`` is specified, predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3205
3206
                eval_data : Dataset
                    The evaluation dataset.
3207
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3208
                    The name of evaluation function (without whitespace).
3209
3210
3211
3212
3213
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

3214
3215
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
3216

wxchan's avatar
wxchan committed
3217
3218
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3219
        result : list
3220
            List with evaluation results.
wxchan's avatar
wxchan committed
3221
        """
Guolin Ke's avatar
Guolin Ke committed
3222
3223
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3224
3225
3226
3227
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3228
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3229
3230
3231
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3232
        # need to push new valid data
wxchan's avatar
wxchan committed
3233
3234
3235
3236
3237
3238
3239
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

    def eval_train(self, feval=None):
3240
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3241
3242
3243

        Parameters
        ----------
3244
        feval : callable or None, optional (default=None)
3245
            Customized evaluation function.
3246
3247
            Should accept two parameters: preds, train_data,
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3248
3249
3250

                preds : list or numpy 1-D array
                    The predicted values.
3251
3252
                    If ``fobj`` is specified, predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3253
3254
                train_data : Dataset
                    The training dataset.
3255
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3256
                    The name of evaluation function (without whitespace).
3257
3258
3259
3260
3261
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

3262
3263
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
wxchan's avatar
wxchan committed
3264
3265
3266

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3267
        result : list
3268
            List with evaluation results.
wxchan's avatar
wxchan committed
3269
        """
3270
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
3271
3272

    def eval_valid(self, feval=None):
3273
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
3274
3275
3276

        Parameters
        ----------
3277
        feval : callable or None, optional (default=None)
3278
            Customized evaluation function.
3279
            Should accept two parameters: preds, valid_data,
3280
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3281
3282
3283

                preds : list or numpy 1-D array
                    The predicted values.
3284
3285
                    If ``fobj`` is specified, predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3286
3287
                valid_data : Dataset
                    The validation dataset.
3288
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3289
                    The name of evaluation function (without whitespace).
3290
3291
3292
3293
3294
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

3295
3296
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
wxchan's avatar
wxchan committed
3297
3298
3299

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3300
        result : list
3301
            List with evaluation results.
wxchan's avatar
wxchan committed
3302
        """
3303
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
3304
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
3305

3306
    def save_model(self, filename, num_iteration=None, start_iteration=0, importance_type='split'):
3307
        """Save Booster to file.
wxchan's avatar
wxchan committed
3308
3309
3310

        Parameters
        ----------
3311
        filename : str or pathlib.Path
3312
            Filename to save Booster.
3313
3314
3315
3316
        num_iteration : int or None, optional (default=None)
            Index of the iteration that should be saved.
            If None, if the best iteration exists, it is saved; otherwise, all iterations are saved.
            If <= 0, all iterations are saved.
Nikita Titov's avatar
Nikita Titov committed
3317
        start_iteration : int, optional (default=0)
3318
            Start index of the iteration that should be saved.
3319
        importance_type : str, optional (default="split")
3320
3321
3322
            What type of feature importance should be saved.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
Nikita Titov's avatar
Nikita Titov committed
3323
3324
3325
3326
3327

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
3328
        """
3329
        if num_iteration is None:
3330
            num_iteration = self.best_iteration
3331
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3332
3333
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
3334
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3335
            ctypes.c_int(num_iteration),
3336
            ctypes.c_int(importance_type_int),
3337
            c_str(str(filename))))
3338
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
3339
        return self
wxchan's avatar
wxchan committed
3340

3341
    def shuffle_models(self, start_iteration=0, end_iteration=-1):
3342
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
3343

3344
3345
3346
        Parameters
        ----------
        start_iteration : int, optional (default=0)
3347
            The first iteration that will be shuffled.
3348
3349
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
3350
            If <= 0, means the last available iteration.
3351

Nikita Titov's avatar
Nikita Titov committed
3352
3353
3354
3355
        Returns
        -------
        self : Booster
            Booster with shuffled models.
3356
        """
3357
3358
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
3359
3360
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
3361
        return self
3362

3363
    def model_from_string(self, model_str, verbose='warn'):
3364
3365
3366
3367
        """Load Booster from a string.

        Parameters
        ----------
3368
        model_str : str
3369
            Model will be loaded from this string.
Nikita Titov's avatar
Nikita Titov committed
3370
3371
        verbose : bool, optional (default=True)
            Whether to print messages while loading model.
3372
3373
3374

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3375
        self : Booster
3376
3377
            Loaded Booster object.
        """
3378
3379
3380
3381
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
3382
3383
3384
3385
3386
3387
3388
3389
3390
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
            c_str(model_str),
            ctypes.byref(out_num_iterations),
            ctypes.byref(self.handle)))
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
            self.handle,
            ctypes.byref(out_num_class)))
3391
3392
3393
3394
        if verbose in {'warn', '_silent_false'}:
            verbose = verbose == 'warn'
        else:
            _log_warning("'verbose' argument is deprecated and will be removed in a future release of LightGBM.")
3395
        if verbose:
3396
            _log_info(f'Finished loading model, total used {int(out_num_iterations.value)} iterations')
3397
        self.__num_class = out_num_class.value
3398
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
3399
3400
        return self

3401
    def model_to_string(self, num_iteration=None, start_iteration=0, importance_type='split'):
3402
        """Save Booster to string.
3403

3404
3405
3406
3407
3408
3409
        Parameters
        ----------
        num_iteration : int or None, optional (default=None)
            Index of the iteration that should be saved.
            If None, if the best iteration exists, it is saved; otherwise, all iterations are saved.
            If <= 0, all iterations are saved.
Nikita Titov's avatar
Nikita Titov committed
3410
        start_iteration : int, optional (default=0)
3411
            Start index of the iteration that should be saved.
3412
        importance_type : str, optional (default="split")
3413
3414
3415
            What type of feature importance should be saved.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
3416
3417
3418

        Returns
        -------
3419
        str_repr : str
3420
3421
            String representation of Booster.
        """
3422
        if num_iteration is None:
3423
            num_iteration = self.best_iteration
3424
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3425
        buffer_len = 1 << 20
3426
        tmp_out_len = ctypes.c_int64(0)
3427
3428
3429
3430
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
            self.handle,
3431
            ctypes.c_int(start_iteration),
3432
            ctypes.c_int(num_iteration),
3433
            ctypes.c_int(importance_type_int),
3434
            ctypes.c_int64(buffer_len),
3435
3436
3437
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
3438
        # if buffer length is not long enough, re-allocate a buffer
3439
3440
3441
3442
3443
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _safe_call(_LIB.LGBM_BoosterSaveModelToString(
                self.handle,
3444
                ctypes.c_int(start_iteration),
3445
                ctypes.c_int(num_iteration),
3446
                ctypes.c_int(importance_type_int),
3447
                ctypes.c_int64(actual_len),
3448
3449
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
3450
        ret = string_buffer.value.decode('utf-8')
3451
3452
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
3453

3454
    def dump_model(self, num_iteration=None, start_iteration=0, importance_type='split', object_hook=None):
Nikita Titov's avatar
Nikita Titov committed
3455
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
3456

3457
3458
        Parameters
        ----------
3459
3460
3461
3462
        num_iteration : int or None, optional (default=None)
            Index of the iteration that should be dumped.
            If None, if the best iteration exists, it is dumped; otherwise, all iterations are dumped.
            If <= 0, all iterations are dumped.
Nikita Titov's avatar
Nikita Titov committed
3463
        start_iteration : int, optional (default=0)
3464
            Start index of the iteration that should be dumped.
3465
        importance_type : str, optional (default="split")
3466
3467
3468
            What type of feature importance should be dumped.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
3469
3470
3471
3472
3473
3474
3475
3476
3477
        object_hook : callable or None, optional (default=None)
            If not None, ``object_hook`` is a function called while parsing the json
            string returned by the C API. It may be used to alter the json, to store
            specific values while building the json structure. It avoids
            walking through the structure again. It saves a significant amount
            of time if the number of trees is huge.
            Signature is ``def object_hook(node: dict) -> dict``.
            None is equivalent to ``lambda node: node``.
            See documentation of ``json.loads()`` for further details.
3478

wxchan's avatar
wxchan committed
3479
3480
        Returns
        -------
3481
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
3482
            JSON format of Booster.
wxchan's avatar
wxchan committed
3483
        """
3484
        if num_iteration is None:
3485
            num_iteration = self.best_iteration
3486
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3487
        buffer_len = 1 << 20
3488
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
3489
3490
3491
3492
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterDumpModel(
            self.handle,
3493
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3494
            ctypes.c_int(num_iteration),
3495
            ctypes.c_int(importance_type_int),
3496
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
3497
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3498
            ptr_string_buffer))
wxchan's avatar
wxchan committed
3499
        actual_len = tmp_out_len.value
3500
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
3501
3502
3503
3504
3505
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _safe_call(_LIB.LGBM_BoosterDumpModel(
                self.handle,
3506
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3507
                ctypes.c_int(num_iteration),
3508
                ctypes.c_int(importance_type_int),
3509
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
3510
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3511
                ptr_string_buffer))
3512
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
3513
3514
3515
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
                                                          default=json_default_with_numpy))
        return ret
wxchan's avatar
wxchan committed
3516

3517
    def predict(self, data, start_iteration=0, num_iteration=None,
3518
                raw_score=False, pred_leaf=False, pred_contrib=False,
3519
                data_has_header=False, is_reshape=True, **kwargs):
3520
        """Make a prediction.
wxchan's avatar
wxchan committed
3521
3522
3523

        Parameters
        ----------
3524
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
3525
            Data source for prediction.
3526
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3527
        start_iteration : int, optional (default=0)
3528
            Start index of the iteration to predict.
3529
            If <= 0, starts from the first iteration.
3530
        num_iteration : int or None, optional (default=None)
3531
3532
3533
3534
            Total number of iterations used in the prediction.
            If None, if the best iteration exists and start_iteration <= 0, the best iteration is used;
            otherwise, all iterations from ``start_iteration`` are used (no limits).
            If <= 0, all iterations from ``start_iteration`` are used (no limits).
3535
3536
3537
3538
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
3539
3540
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
3541

Nikita Titov's avatar
Nikita Titov committed
3542
3543
3544
3545
3546
3547
3548
            .. note::

                If you want to get more explanations for your model's predictions using SHAP values,
                like SHAP interaction values,
                you can install the shap package (https://github.com/slundberg/shap).
                Note that unlike the shap package, with ``pred_contrib`` we return a matrix with an extra
                column, where the last column is the expected value.
3549

3550
3551
        data_has_header : bool, optional (default=False)
            Whether the data has header.
3552
            Used only if data is str.
3553
3554
        is_reshape : bool, optional (default=True)
            If True, result is reshaped to [nrow, ncol].
3555
3556
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
3557
3558
3559

        Returns
        -------
3560
        result : numpy array, scipy.sparse or list of scipy.sparse
3561
            Prediction result.
3562
            Can be sparse or a list of sparse objects (each element represents predictions for one class) for feature contributions (when ``pred_contrib=True``).
wxchan's avatar
wxchan committed
3563
        """
3564
        predictor = self._to_predictor(deepcopy(kwargs))
3565
        if num_iteration is None:
3566
            if start_iteration <= 0:
3567
3568
3569
3570
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
        return predictor.predict(data, start_iteration, num_iteration,
3571
3572
                                 raw_score, pred_leaf, pred_contrib,
                                 data_has_header, is_reshape)
wxchan's avatar
wxchan committed
3573

3574
    def refit(self, data, label, decay_rate=0.9, **kwargs):
Guolin Ke's avatar
Guolin Ke committed
3575
3576
3577
3578
        """Refit the existing Booster by new data.

        Parameters
        ----------
3579
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
Guolin Ke's avatar
Guolin Ke committed
3580
            Data source for refit.
3581
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3582
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
3583
3584
            Label for refit.
        decay_rate : float, optional (default=0.9)
3585
3586
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
3587
3588
        **kwargs
            Other parameters for refit.
3589
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
3590
3591
3592
3593
3594
3595

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
3596
3597
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
3598
        predictor = self._to_predictor(deepcopy(kwargs))
3599
        leaf_preds = predictor.predict(data, -1, pred_leaf=True)
3600
        nrow, ncol = leaf_preds.shape
3601
        out_is_linear = ctypes.c_int(0)
3602
3603
3604
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
3605
3606
3607
3608
3609
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
3610
        new_params["linear_tree"] = bool(out_is_linear.value)
3611
        train_set = Dataset(data, label, silent=True, params=new_params)
3612
        new_params['refit_decay_rate'] = decay_rate
3613
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
3614
3615
3616
3617
3618
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
3619
        ptr_data, _, _ = c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
3620
3621
3622
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
3623
3624
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
3625
3626
        new_booster.network = self.network
        new_booster.__attr = self.__attr.copy()
Guolin Ke's avatar
Guolin Ke committed
3627
3628
        return new_booster

3629
    def get_leaf_output(self, tree_id, leaf_id):
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
        """Get the output of a leaf.

        Parameters
        ----------
        tree_id : int
            The index of the tree.
        leaf_id : int
            The index of the leaf in the tree.

        Returns
        -------
        result : float
            The output of the leaf.
        """
3644
3645
3646
3647
3648
3649
3650
3651
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
            self.handle,
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

3652
    def _to_predictor(self, pred_parameter=None):
3653
        """Convert to predictor."""
3654
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
3655
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
3656
3657
        return predictor

3658
    def num_feature(self):
3659
3660
3661
3662
3663
3664
3665
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
3666
3667
3668
3669
3670
3671
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
            self.handle,
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

wxchan's avatar
wxchan committed
3672
    def feature_name(self):
3673
        """Get names of features.
wxchan's avatar
wxchan committed
3674
3675
3676

        Returns
        -------
3677
3678
        result : list
            List with names of features.
wxchan's avatar
wxchan committed
3679
        """
3680
        num_feature = self.num_feature()
3681
        # Get name of features
wxchan's avatar
wxchan committed
3682
        tmp_out_len = ctypes.c_int(0)
3683
3684
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
3685
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
3686
3687
3688
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
3689
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
3690
            ctypes.byref(tmp_out_len),
3691
            ctypes.c_size_t(reserved_string_buffer_size),
3692
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3693
3694
3695
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
                self.handle,
                ctypes.c_int(num_feature),
                ctypes.byref(tmp_out_len),
                ctypes.c_size_t(actual_string_buffer_size),
                ctypes.byref(required_string_buffer_size),
                ptr_string_buffers))
3708
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
3709

3710
    def feature_importance(self, importance_type='split', iteration=None):
3711
        """Get feature importances.
3712

3713
3714
        Parameters
        ----------
3715
        importance_type : str, optional (default="split")
3716
3717
3718
            How the importance is calculated.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
3719
3720
3721
3722
        iteration : int or None, optional (default=None)
            Limit number of iterations in the feature importance calculation.
            If None, if the best iteration exists, it is used; otherwise, all trees are used.
            If <= 0, all trees are used (no limits).
3723

3724
3725
        Returns
        -------
3726
3727
        result : numpy array
            Array with feature importances.
3728
        """
3729
3730
        if iteration is None:
            iteration = self.best_iteration
3731
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3732
        result = np.empty(self.num_feature(), dtype=np.float64)
3733
3734
3735
3736
3737
3738
        _safe_call(_LIB.LGBM_BoosterFeatureImportance(
            self.handle,
            ctypes.c_int(iteration),
            ctypes.c_int(importance_type_int),
            result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
        if importance_type_int == 0:
3739
            return result.astype(np.int32)
3740
3741
        else:
            return result
3742

3743
3744
3745
3746
3747
    def get_split_value_histogram(self, feature, bins=None, xgboost_style=False):
        """Get split value histogram for the specified feature.

        Parameters
        ----------
3748
        feature : int or str
3749
3750
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
3751
            If str, interpreted as name.
3752

Nikita Titov's avatar
Nikita Titov committed
3753
3754
3755
            .. warning::

                Categorical features are not supported.
3756

3757
        bins : int, str or None, optional (default=None)
3758
3759
3760
            The maximum number of bins.
            If None, or int and > number of unique split values and ``xgboost_style=True``,
            the number of bins equals number of unique split values.
3761
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
        xgboost_style : bool, optional (default=False)
            Whether the returned result should be in the same form as it is in XGBoost.
            If False, the returned value is tuple of 2 numpy arrays as it is in ``numpy.histogram()`` function.
            If True, the returned value is matrix, in which the first column is the right edges of non-empty bins
            and the second one is the histogram values.

        Returns
        -------
        result_tuple : tuple of 2 numpy arrays
            If ``xgboost_style=False``, the values of the histogram of used splitting values for the specified feature
            and the bin edges.
        result_array_like : numpy array or pandas DataFrame (if pandas is installed)
            If ``xgboost_style=True``, the histogram of used splitting values for the specified feature.
        """
        def add(root):
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
3779
                if feature_names is not None and isinstance(feature, str):
3780
3781
3782
3783
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
3784
                    if isinstance(root['threshold'], str):
3785
3786
3787
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
        values = []
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

3798
        if bins is None or isinstance(bins, int) and xgboost_style:
3799
3800
3801
3802
3803
3804
3805
            n_unique = len(np.unique(values))
            bins = max(min(n_unique, bins) if bins is not None else n_unique, 1)
        hist, bin_edges = np.histogram(values, bins=bins)
        if xgboost_style:
            ret = np.column_stack((bin_edges[1:], hist))
            ret = ret[ret[:, 1] > 0]
            if PANDAS_INSTALLED:
3806
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
3807
3808
3809
3810
3811
            else:
                return ret
        else:
            return hist, bin_edges

wxchan's avatar
wxchan committed
3812
    def __inner_eval(self, data_name, data_idx, feval=None):
3813
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
3814
        if data_idx >= self.__num_dataset:
3815
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3816
3817
3818
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
3819
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
3820
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3821
3822
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3823
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3824
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3825
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
3826
            if tmp_out_len.value != self.__num_inner_eval:
3827
                raise ValueError("Wrong length of eval results")
3828
            for i in range(self.__num_inner_eval):
3829
3830
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
3831
3832
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
3833
3834
3835
3836
3837
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
3838
3839
3840
3841
3842
3843
3844
3845
3846
            for eval_function in feval:
                if eval_function is None:
                    continue
                feval_ret = eval_function(self.__inner_predict(data_idx), cur_data)
                if isinstance(feval_ret, list):
                    for eval_name, val, is_higher_better in feval_ret:
                        ret.append((data_name, eval_name, val, is_higher_better))
                else:
                    eval_name, val, is_higher_better = feval_ret
wxchan's avatar
wxchan committed
3847
3848
3849
3850
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

    def __inner_predict(self, data_idx):
3851
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
3852
        if data_idx >= self.__num_dataset:
3853
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3854
3855
3856
3857
3858
        if self.__inner_predict_buffer[data_idx] is None:
            if data_idx == 0:
                n_preds = self.train_set.num_data() * self.__num_class
            else:
                n_preds = self.valid_sets[data_idx - 1].num_data() * self.__num_class
3859
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
3860
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
3861
3862
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
3863
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
3864
3865
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3866
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3867
3868
3869
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
3870
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
3871
3872
3873
3874
            self.__is_predicted_cur_iter[data_idx] = True
        return self.__inner_predict_buffer[data_idx]

    def __get_eval_info(self):
3875
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
3876
3877
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
3878
            out_num_eval = ctypes.c_int(0)
3879
            # Get num of inner evals
wxchan's avatar
wxchan committed
3880
3881
3882
3883
3884
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
                self.handle,
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
3885
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
3886
                tmp_out_len = ctypes.c_int(0)
3887
3888
3889
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
3890
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
3891
                ]
wxchan's avatar
wxchan committed
3892
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
3893
3894
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
3895
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
3896
                    ctypes.byref(tmp_out_len),
3897
                    ctypes.c_size_t(reserved_string_buffer_size),
3898
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3899
3900
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
3901
                    raise ValueError("Length of eval names doesn't equal with num_evals")
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
                actual_string_buffer_size = required_string_buffer_size.value
                # if buffer length is not long enough, reallocate buffers
                if reserved_string_buffer_size < actual_string_buffer_size:
                    string_buffers = [
                        ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(self.__num_inner_eval)
                    ]
                    ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
                    _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                        self.handle,
                        ctypes.c_int(self.__num_inner_eval),
                        ctypes.byref(tmp_out_len),
                        ctypes.c_size_t(actual_string_buffer_size),
                        ctypes.byref(required_string_buffer_size),
                        ptr_string_buffers))
                self.__name_inner_eval = [
                    string_buffers[i].value.decode('utf-8') for i in range(self.__num_inner_eval)
                ]
                self.__higher_better_inner_eval = [
                    name.startswith(('auc', 'ndcg@', 'map@', 'average_precision')) for name in self.__name_inner_eval
                ]
3922

wxchan's avatar
wxchan committed
3923
    def attr(self, key):
3924
        """Get attribute string from the Booster.
wxchan's avatar
wxchan committed
3925
3926
3927

        Parameters
        ----------
3928
        key : str
3929
            The name of the attribute.
wxchan's avatar
wxchan committed
3930
3931
3932

        Returns
        -------
3933
        value : str or None
3934
            The attribute value.
Nikita Titov's avatar
Nikita Titov committed
3935
            Returns None if attribute does not exist.
wxchan's avatar
wxchan committed
3936
        """
3937
        return self.__attr.get(key, None)
wxchan's avatar
wxchan committed
3938
3939

    def set_attr(self, **kwargs):
3940
        """Set attributes to the Booster.
wxchan's avatar
wxchan committed
3941
3942
3943
3944

        Parameters
        ----------
        **kwargs
3945
3946
            The attributes to set.
            Setting a value to None deletes an attribute.
Nikita Titov's avatar
Nikita Titov committed
3947
3948
3949
3950

        Returns
        -------
        self : Booster
3951
            Booster with set attributes.
wxchan's avatar
wxchan committed
3952
3953
3954
        """
        for key, value in kwargs.items():
            if value is not None:
3955
                if not isinstance(value, str):
Nikita Titov's avatar
Nikita Titov committed
3956
                    raise ValueError("Only string values are accepted")
wxchan's avatar
wxchan committed
3957
3958
3959
                self.__attr[key] = value
            else:
                self.__attr.pop(key, None)
Nikita Titov's avatar
Nikita Titov committed
3960
        return self