basic.py 185 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 inspect
6
import json
wxchan's avatar
wxchan committed
7
import warnings
8
from collections import OrderedDict
9
from copy import deepcopy
10
from enum import Enum
11
from functools import wraps
12
from os import SEEK_END, environ
13
14
from os.path import getsize
from pathlib import Path
15
from tempfile import NamedTemporaryFile
16
from typing import TYPE_CHECKING, Any, Callable, Dict, Iterable, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
17
18
19
20

import numpy as np
import scipy.sparse

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

24
25
26
if TYPE_CHECKING:
    from typing import Literal

27
28
29
30
31
32
33
34
35
__all__ = [
    'Booster',
    'Dataset',
    'LGBMDeprecationWarning',
    'LightGBMError',
    'register_logger',
    'Sequence',
]

36
_BoosterHandle = ctypes.c_void_p
37
_DatasetHandle = ctypes.c_void_p
38
39
40
41
_ctypes_int_ptr = Union[
    "ctypes._Pointer[ctypes.c_int32]",
    "ctypes._Pointer[ctypes.c_int64]"
]
42
43
44
45
_ctypes_int_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_int32]]",
    "ctypes.Array[ctypes._Pointer[ctypes.c_int64]]"
]
46
47
48
49
50
51
52
53
_ctypes_float_ptr = Union[
    "ctypes._Pointer[ctypes.c_float]",
    "ctypes._Pointer[ctypes.c_double]"
]
_ctypes_float_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_float]]",
    "ctypes.Array[ctypes._Pointer[ctypes.c_double]]"
]
54
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
55
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
56
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
57
58
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], "Literal['auto']"]
_LGBM_FeatureNameConfiguration = Union[List[str], "Literal['auto']"]
59
60
61
62
63
64
_LGBM_GroupType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
65
66
67
68
_LGBM_PositionType = Union[
    np.ndarray,
    pd_Series
]
69
70
71
72
73
74
75
_LGBM_InitScoreType = Union[
    List[float],
    List[List[float]],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
]
76
77
78
79
80
81
82
83
84
85
86
_LGBM_TrainDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix,
    "Sequence",
    List["Sequence"],
    List[np.ndarray]
]
87
_LGBM_LabelType = Union[
88
89
    List[float],
    List[int],
90
91
92
93
    np.ndarray,
    pd_Series,
    pd_DataFrame
]
94
95
96
97
98
99
100
101
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix
]
102
103
104
105
106
107
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
108
109
110
ZERO_THRESHOLD = 1e-35


111
112
113
114
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


115
def _get_sample_count(total_nrow: int, params: str) -> int:
116
117
118
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
119
        _c_str(params),
120
121
122
123
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
124

125
126
127
128
129
130
class _MissingType(Enum):
    NONE = 'None'
    NAN = 'NaN'
    ZERO = 'Zero'


131
class _DummyLogger:
132
    def info(self, msg: str) -> None:
133
        print(msg)  # noqa: T201
134

135
    def warning(self, msg: str) -> None:
136
137
138
        warnings.warn(msg, stacklevel=3)


139
140
141
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
142
143


144
145
146
147
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


148
149
150
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
151
152
153
154
    """Register custom logger.

    Parameters
    ----------
155
    logger : Any
156
        Custom logger.
157
158
159
160
    info_method_name : str, optional (default="info")
        Method used to log info messages.
    warning_method_name : str, optional (default="warning")
        Method used to log warning messages.
161
    """
162
163
164
165
166
167
    if not _has_method(logger, info_method_name) or not _has_method(logger, warning_method_name):
        raise TypeError(
            f"Logger must provide '{info_method_name}' and '{warning_method_name}' method"
        )

    global _LOGGER, _INFO_METHOD_NAME, _WARNING_METHOD_NAME
168
    _LOGGER = logger
169
170
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
171
172


173
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
174
    """Join log messages from native library which come by chunks."""
175
    msg_normalized: List[str] = []
176
177

    @wraps(func)
178
    def wrapper(msg: str) -> None:
179
180
181
182
183
184
185
186
187
188
189
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


190
def _log_info(msg: str) -> None:
191
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
192
193


194
def _log_warning(msg: str) -> None:
195
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
196
197
198


@_normalize_native_string
199
def _log_native(msg: str) -> None:
200
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
201
202


203
def _log_callback(msg: bytes) -> None:
204
    """Redirect logs from native library into Python."""
205
    _log_native(str(msg.decode('utf-8')))
206
207


208
def _load_lib() -> ctypes.CDLL:
209
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
210
211
212
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
213
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
214
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
215
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
216
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
217
218
    return lib

wxchan's avatar
wxchan committed
219

220
221
222
223
224
225
226
# we don't need lib_lightgbm while building docs
_LIB: ctypes.CDLL
if environ.get('LIGHTGBM_BUILD_DOC', False):
    from unittest.mock import Mock  # isort: skip
    _LIB = Mock(ctypes.CDLL)  # type: ignore
else:
    _LIB = _load_lib()
wxchan's avatar
wxchan committed
227

wxchan's avatar
wxchan committed
228

229
_NUMERIC_TYPES = (int, float, bool)
230
_ArrayLike = Union[List, np.ndarray, pd_Series]
231
232


233
def _safe_call(ret: int) -> None:
234
235
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
236
237
238
    Parameters
    ----------
    ret : int
239
        The return value from C API calls.
wxchan's avatar
wxchan committed
240
241
    """
    if ret != 0:
242
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
243

wxchan's avatar
wxchan committed
244

245
def _is_numeric(obj: Any) -> bool:
246
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
247
248
249
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
250
251
252
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
253
254
        return False

wxchan's avatar
wxchan committed
255

256
def _is_numpy_1d_array(data: Any) -> bool:
257
    """Check whether data is a numpy 1-D array."""
258
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
259

wxchan's avatar
wxchan committed
260

261
def _is_numpy_column_array(data: Any) -> bool:
262
263
264
265
266
267
268
    """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


269
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
270
    """Cast numpy array to given dtype."""
271
272
273
274
275
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


276
def _is_1d_list(data: Any) -> bool:
277
    """Check whether data is a 1-D list."""
278
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
279

wxchan's avatar
wxchan committed
280

281
282
283
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
284
        _is_numpy_1d_array(data)
285
        or _is_numpy_column_array(data)
286
        or _is_1d_list(data)
287
288
289
290
        or isinstance(data, pd_Series)
    )


291
292
def _list_to_1d_numpy(
    data: Any,
293
294
    dtype: "np.typing.DTypeLike",
    name: str
295
) -> np.ndarray:
296
    """Convert data to numpy 1-D array."""
297
    if _is_numpy_1d_array(data):
298
        return _cast_numpy_array_to_dtype(data, dtype)
299
    elif _is_numpy_column_array(data):
300
301
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
302
        return _cast_numpy_array_to_dtype(array, dtype)
303
    elif _is_1d_list(data):
wxchan's avatar
wxchan committed
304
        return np.array(data, dtype=dtype, copy=False)
305
    elif isinstance(data, pd_Series):
306
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
307
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
308
    else:
309
310
        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
311

wxchan's avatar
wxchan committed
312

313
314
315
316
317
318
319
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."""
320
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
321
322
323
324
325
326
327
328
329
330
331


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)
    )


332
333
def _data_to_2d_numpy(
    data: Any,
334
335
    dtype: "np.typing.DTypeLike",
    name: str
336
) -> np.ndarray:
337
338
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
339
        return _cast_numpy_array_to_dtype(data, dtype)
340
341
342
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
343
        _check_for_bad_pandas_dtypes(data.dtypes)
344
        return _cast_numpy_array_to_dtype(data.values, dtype)
345
346
347
348
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


349
def _cfloat32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
350
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
351
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
352
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
353
    else:
354
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
355

Guolin Ke's avatar
Guolin Ke committed
356

357
def _cfloat64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
358
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
359
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
360
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
361
362
363
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
364

365
def _cint32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
366
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
367
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
368
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
369
    else:
370
371
372
        raise RuntimeError('Expected int32 pointer')


373
def _cint64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
374
375
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
376
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
377
378
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
379

wxchan's avatar
wxchan committed
380

381
def _c_str(string: str) -> ctypes.c_char_p:
382
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
383
384
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
385

386
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
387
    """Convert a Python array to C array."""
388
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
389

wxchan's avatar
wxchan committed
390

391
def _json_default_with_numpy(obj: Any) -> Any:
392
393
394
395
396
397
398
399
400
    """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


401
402
403
404
405
406
407
408
def _to_string(x: Union[int, float, str, List]) -> str:
    if isinstance(x, list):
        val_list = ",".join(str(val) for val in x)
        return f"[{val_list}]"
    else:
        return str(x)


409
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
410
    """Convert Python dictionary to string, which is passed to C API."""
411
    if data is None or not data:
wxchan's avatar
wxchan committed
412
413
414
        return ""
    pairs = []
    for key, val in data.items():
415
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
416
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
417
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
418
            pairs.append(f"{key}={val}")
419
        elif val is not None:
420
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
421
    return ' '.join(pairs)
422

wxchan's avatar
wxchan committed
423

424
class _TempFile:
425
426
    """Proxy class to workaround errors on Windows."""

427
428
429
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
430
            self.path = Path(self.name)
431
        return self
wxchan's avatar
wxchan committed
432

433
    def __exit__(self, exc_type, exc_val, exc_tb):
434
435
        if self.path.is_file():
            self.path.unlink()
436

wxchan's avatar
wxchan committed
437

438
class LightGBMError(Exception):
439
440
    """Error thrown by LightGBM."""

441
442
443
    pass


444
445
446
447
448
449
450
451
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
452
453
454
455
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
456
    def _get_all_param_aliases() -> Dict[str, List[str]]:
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_DumpParamAliases(
            ctypes.c_int64(buffer_len),
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
        # if buffer length is not long enough, re-allocate a buffer
        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_DumpParamAliases(
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
474
        return json.loads(
475
            string_buffer.value.decode('utf-8'),
476
            object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
477
        )
478
479

    @classmethod
480
481
482
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
483
484
        ret = set()
        for i in args:
485
            ret.update(cls.get_sorted(i))
486
487
        return ret

488
489
490
491
492
493
    @classmethod
    def get_sorted(cls, name: str) -> List[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
        return cls.aliases.get(name, [name])

494
    @classmethod
495
496
497
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
498
499
500
501
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
502
                    ret.update(aliases)
503
504
505
                    break
        return ret

506

507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
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)

528
529
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
530
531

    # if main_param_name was provided, keep that value and remove all aliases
532
    if main_param_name in params.keys():
533
534
535
        for param in aliases:
            params.pop(param, None)
        return params
536

537
538
539
540
541
    # if main param name was not found, search for an alias
    for param in aliases:
        if param in params.keys():
            params[main_param_name] = params[param]
            break
542

543
544
545
546
547
548
549
    if main_param_name in params.keys():
        for param in aliases:
            params.pop(param, None)
        return params

    # neither of main_param_name, aliases were found
    params[main_param_name] = default_value
550
551
552
553

    return params


554
_MAX_INT32 = (1 << 31) - 1
555

556
"""Macro definition of data type in C API of LightGBM"""
557
558
559
560
_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
561

562
"""Matrix is row major in Python"""
563
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
564

565
"""Macro definition of prediction type in C API of LightGBM"""
566
567
568
569
_C_API_PREDICT_NORMAL = 0
_C_API_PREDICT_RAW_SCORE = 1
_C_API_PREDICT_LEAF_INDEX = 2
_C_API_PREDICT_CONTRIB = 3
wxchan's avatar
wxchan committed
570

571
"""Macro definition of sparse matrix type"""
572
573
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
574

575
"""Macro definition of feature importance type"""
576
577
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
578

579
"""Data type of data field"""
580
581
582
583
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
584
585
    "group": _C_API_DTYPE_INT32,
    "position": _C_API_DTYPE_INT32
586
}
wxchan's avatar
wxchan committed
587

588
"""String name to int feature importance type mapper"""
589
590
591
592
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN
}
593

wxchan's avatar
wxchan committed
594

595
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
596
    """Fix the memory of multi-dimensional sliced object."""
597
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
598
        if not data.flags.c_contiguous:
599
600
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
601
602
603
604
            return np.copy(data)
    return data


605
606
607
def _c_float_array(
    data: np.ndarray
) -> Tuple[_ctypes_float_ptr, int, np.ndarray]:
608
    """Get pointer of float numpy array / list."""
609
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
610
        data = np.array(data, copy=False)
611
    if _is_numpy_1d_array(data):
612
        data = _convert_from_sliced_object(data)
613
        assert data.flags.c_contiguous
614
        ptr_data: _ctypes_float_ptr
wxchan's avatar
wxchan committed
615
616
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
617
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
618
619
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
620
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
621
        else:
622
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
623
    else:
624
        raise TypeError(f"Unknown type({type(data).__name__})")
625
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
626

wxchan's avatar
wxchan committed
627

628
629
630
def _c_int_array(
    data: np.ndarray
) -> Tuple[_ctypes_int_ptr, int, np.ndarray]:
631
    """Get pointer of int numpy array / list."""
632
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
633
        data = np.array(data, copy=False)
634
    if _is_numpy_1d_array(data):
635
        data = _convert_from_sliced_object(data)
636
        assert data.flags.c_contiguous
637
        ptr_data: _ctypes_int_ptr
wxchan's avatar
wxchan committed
638
639
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
640
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
641
642
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
643
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
644
        else:
645
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
646
    else:
647
        raise TypeError(f"Unknown type({type(data).__name__})")
648
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
649

wxchan's avatar
wxchan committed
650

651
def _is_allowed_numpy_dtype(dtype: type) -> bool:
652
    float128 = getattr(np, 'float128', type(None))
653
654
655
656
    return (
        issubclass(dtype, (np.integer, np.floating, np.bool_))
        and not issubclass(dtype, (np.timedelta64, float128))
    )
657
658


659
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
660
661
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
662
        for column_name, pandas_dtype in pandas_dtypes_series.items()
663
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
664
665
666
667
    ]
    if bad_pandas_dtypes:
        raise ValueError('pandas dtypes must be int, float or bool.\n'
                         f'Fields with bad pandas dtypes: {", ".join(bad_pandas_dtypes)}')
668
669


670
def _data_from_pandas(
671
672
673
    data: pd_DataFrame,
    feature_name: _LGBM_FeatureNameConfiguration,
    categorical_feature: _LGBM_CategoricalFeatureConfiguration,
674
    pandas_categorical: Optional[List[List]]
675
676
677
678
679
680
681
682
683
684
685
686
687
) -> Tuple[np.ndarray, List[str], List[str], List[List]]:
    if len(data.shape) != 2 or data.shape[0] < 1:
        raise ValueError('Input data must be 2 dimensional and non empty.')

    # determine feature names
    if feature_name == 'auto':
        feature_name = [str(col) for col in data.columns]

    # determine categorical features
    cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
    cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
    if pandas_categorical is None:  # train dataset
        pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
688
    else:
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
        if len(cat_cols) != len(pandas_categorical):
            raise ValueError('train and valid dataset categorical_feature do not match.')
        for col, category in zip(cat_cols, pandas_categorical):
            if list(data[col].cat.categories) != list(category):
                data[col] = data[col].cat.set_categories(category)
    if len(cat_cols):  # cat_cols is list
        data = data.copy(deep=False)  # not alter origin DataFrame
        data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
    if categorical_feature == 'auto':  # use cat cols from DataFrame
        categorical_feature = cat_cols_not_ordered
    else:  # use cat cols specified by user
        categorical_feature = list(categorical_feature)  # type: ignore[assignment]

    # get numpy representation of the data
    _check_for_bad_pandas_dtypes(data.dtypes)
    df_dtypes = [dtype.type for dtype in data.dtypes]
    df_dtypes.append(np.float32)  # so that the target dtype considers floats
    target_dtype = np.result_type(*df_dtypes)
    try:
        # most common case (no nullable dtypes)
        data = data.to_numpy(dtype=target_dtype, copy=False)
    except TypeError:
        # 1.0 <= pd version < 1.1 and nullable dtypes, least common case
        # raises error because array is casted to type(pd.NA) and there's no na_value argument
        data = data.astype(target_dtype, copy=False).values
    except ValueError:
        # data has nullable dtypes, but we can specify na_value argument and copy will be made
        data = data.to_numpy(dtype=target_dtype, na_value=np.nan)
717
    return data, feature_name, categorical_feature, pandas_categorical
718
719


720
721
722
723
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
    file_name: Optional[Union[str, Path]] = None
) -> str:
724
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
725
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
726
727
728
729
730
731
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


732
733
734
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
    model_str: Optional[str] = None
735
) -> Optional[List[List]]:
736
737
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
738
    if file_name is not None:
739
        max_offset = -getsize(file_name)
740
741
742
743
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
744
                f.seek(offset, SEEK_END)
745
746
747
748
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
749
        last_line = lines[-1].decode('utf-8').strip()
750
        if not last_line.startswith(pandas_key):
751
            last_line = lines[-2].decode('utf-8').strip()
752
    elif model_str is not None:
753
754
755
756
757
758
        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
759
760


761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
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**.

781
782
    .. versionadded:: 3.3.0

783
784
785
786
787
788
789
790
791
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
792
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
793
794
795
796
797
798
799
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
800
                return self._get_one_line(idx)
801
            elif isinstance(idx, slice):
802
803
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
804
                # Only required if using ``Dataset.subset()``.
805
                return np.array([self._get_one_line(i) for i in idx])
806
            else:
807
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
808
809
810

        Parameters
        ----------
811
        idx : int, slice[int], list[int]
812
813
814
815
            Item index.

        Returns
        -------
816
        result : numpy 1-D array or numpy 2-D array
817
            1-D array if idx is int, 2-D array if idx is slice or list.
818
819
820
821
822
823
824
825
826
        """
        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__()")


827
class _InnerPredictor:
828
829
830
831
832
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
833
834
835
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
836
    """
837

838
839
    def __init__(
        self,
840
841
842
843
        booster_handle: _BoosterHandle,
        pandas_categorical: Optional[List[List]],
        pred_parameter: Dict[str, Any],
        manage_handle: bool
844
    ):
845
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
846
847
848

        Parameters
        ----------
849
        booster_handle : object
850
            Handle of Booster.
851
852
853
854
        pandas_categorical : list of list, or None
            If provided, list of categories for ``pandas`` categorical columns.
            Where the ``i``th element of the list contains the categories for the ``i``th categorical feature.
        pred_parameter : dict
855
            Other parameters for the prediction.
856
857
        manage_handle : bool
            If ``True``, free the corresponding Booster on the C++ side when this Python object is deleted.
wxchan's avatar
wxchan committed
858
        """
859
860
861
862
863
864
865
866
        self._handle = booster_handle
        self.__is_manage_handle = manage_handle
        self.pandas_categorical = pandas_categorical
        self.pred_parameter = _param_dict_to_str(pred_parameter)

        out_num_class = ctypes.c_int(0)
        _safe_call(
            _LIB.LGBM_BoosterGetNumClasses(
867
                self._handle,
868
869
870
871
                ctypes.byref(out_num_class)
            )
        )
        self.num_class = out_num_class.value
wxchan's avatar
wxchan committed
872

873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
    @classmethod
    def from_booster(
        cls,
        booster: "Booster",
        pred_parameter: Dict[str, Any]
    ) -> "_InnerPredictor":
        """Initialize an ``_InnerPredictor`` from a ``Booster``.

        Parameters
        ----------
        booster : Booster
            Booster.
        pred_parameter : dict
            Other parameters for the prediction.
        """
        out_cur_iter = ctypes.c_int(0)
        _safe_call(
            _LIB.LGBM_BoosterGetCurrentIteration(
                booster._handle,
                ctypes.byref(out_cur_iter)
            )
        )
        return cls(
            booster_handle=booster._handle,
            pandas_categorical=booster.pandas_categorical,
            pred_parameter=pred_parameter,
            manage_handle=False
        )

    @classmethod
    def from_model_file(
        cls,
        model_file: Union[str, Path],
        pred_parameter: Dict[str, Any]
    ) -> "_InnerPredictor":
        """Initialize an ``_InnerPredictor`` from a text file containing a LightGBM model.

        Parameters
        ----------
        model_file : str or pathlib.Path
            Path to the model file.
        pred_parameter : dict
            Other parameters for the prediction.
        """
        booster_handle = ctypes.c_void_p()
        out_num_iterations = ctypes.c_int(0)
        _safe_call(
            _LIB.LGBM_BoosterCreateFromModelfile(
                _c_str(str(model_file)),
                ctypes.byref(out_num_iterations),
                ctypes.byref(booster_handle)
            )
        )
        return cls(
            booster_handle=booster_handle,
            pandas_categorical=_load_pandas_categorical(file_name=model_file),
            pred_parameter=pred_parameter,
            manage_handle=True
        )
cbecker's avatar
cbecker committed
932

933
    def __del__(self) -> None:
934
935
        try:
            if self.__is_manage_handle:
936
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
937
938
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
939

940
    def __getstate__(self) -> Dict[str, Any]:
941
942
        this = self.__dict__.copy()
        this.pop('handle', None)
943
        this.pop('_handle', None)
944
945
        return this

946
947
    def predict(
        self,
948
        data: _LGBM_PredictDataType,
949
950
951
952
953
954
955
        start_iteration: int = 0,
        num_iteration: int = -1,
        raw_score: bool = False,
        pred_leaf: bool = False,
        pred_contrib: bool = False,
        data_has_header: bool = False,
        validate_features: bool = False
956
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
957
        """Predict logic.
wxchan's avatar
wxchan committed
958
959
960

        Parameters
        ----------
961
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
962
            Data source for prediction.
963
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
964
965
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
966
967
968
969
970
971
972
973
974
975
976
        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.
977
978
979
        validate_features : bool, optional (default=False)
            If True, ensure that the features used to predict match the ones used to train.
            Used only if data is pandas DataFrame.
wxchan's avatar
wxchan committed
980

981
982
            .. versionadded:: 4.0.0

wxchan's avatar
wxchan committed
983
984
        Returns
        -------
985
        result : numpy array, scipy.sparse or list of scipy.sparse
986
            Prediction result.
987
            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
988
        """
wxchan's avatar
wxchan committed
989
        if isinstance(data, Dataset):
990
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
991
992
993
994
995
996
        elif isinstance(data, pd_DataFrame) and validate_features:
            data_names = [str(x) for x in data.columns]
            ptr_names = (ctypes.c_char_p * len(data_names))()
            ptr_names[:] = [x.encode('utf-8') for x in data_names]
            _safe_call(
                _LIB.LGBM_BoosterValidateFeatureNames(
997
                    self._handle,
998
999
1000
1001
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
1002
1003
1004
1005
1006
1007
1008
1009
1010

        if isinstance(data, pd_DataFrame):
            data = _data_from_pandas(
                data=data,
                feature_name="auto",
                categorical_feature="auto",
                pandas_categorical=self.pandas_categorical
            )[0]

1011
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
1012
        if raw_score:
1013
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
1014
        if pred_leaf:
1015
            predict_type = _C_API_PREDICT_LEAF_INDEX
1016
        if pred_contrib:
1017
            predict_type = _C_API_PREDICT_CONTRIB
cbecker's avatar
cbecker committed
1018

1019
        if isinstance(data, (str, Path)):
1020
            with _TempFile() as f:
wxchan's avatar
wxchan committed
1021
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
1022
                    self._handle,
1023
                    _c_str(str(data)),
1024
                    ctypes.c_int(data_has_header),
Guolin Ke's avatar
Guolin Ke committed
1025
                    ctypes.c_int(predict_type),
1026
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1027
                    ctypes.c_int(num_iteration),
1028
1029
                    _c_str(self.pred_parameter),
                    _c_str(f.name)))
1030
1031
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
1032
        elif isinstance(data, scipy.sparse.csr_matrix):
1033
1034
1035
1036
1037
1038
            preds, nrow = self.__pred_for_csr(
                csr=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
Guolin Ke's avatar
Guolin Ke committed
1039
        elif isinstance(data, scipy.sparse.csc_matrix):
1040
1041
1042
1043
1044
1045
            preds, nrow = self.__pred_for_csc(
                csc=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
wxchan's avatar
wxchan committed
1046
        elif isinstance(data, np.ndarray):
1047
1048
1049
1050
1051
1052
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1053
1054
1055
        elif isinstance(data, list):
            try:
                data = np.array(data)
1056
1057
            except BaseException as err:
                raise ValueError('Cannot convert data list to numpy array.') from err
1058
1059
1060
1061
1062
1063
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1064
        elif isinstance(data, dt_DataTable):
1065
1066
1067
1068
1069
1070
            preds, nrow = self.__pred_for_np2d(
                mat=data.to_numpy(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
wxchan's avatar
wxchan committed
1071
1072
        else:
            try:
1073
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
1074
                csr = scipy.sparse.csr_matrix(data)
1075
1076
            except BaseException as err:
                raise TypeError(f'Cannot predict data for type {type(data).__name__}') from err
1077
1078
1079
1080
1081
1082
            preds, nrow = self.__pred_for_csr(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
wxchan's avatar
wxchan committed
1083
1084
        if pred_leaf:
            preds = preds.astype(np.int32)
1085
        is_sparse = isinstance(preds, scipy.sparse.spmatrix) or isinstance(preds, list)
1086
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1087
            if preds.size % nrow == 0:
1088
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1089
            else:
1090
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
1091
1092
        return preds

1093
1094
1095
1096
1097
1098
1099
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
        predict_type: int
    ) -> int:
1100
        """Get size of prediction result."""
1101
        if nrow > _MAX_INT32:
1102
            raise LightGBMError('LightGBM cannot perform prediction for data '
1103
                                f'with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n'
1104
                                'You can split your data into chunks '
1105
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
1106
1107
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
1108
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1109
1110
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
1111
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1112
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
1113
1114
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
1115

1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
    def __inner_predict_np2d(
        self,
        mat: np.ndarray,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
        preds: Optional[np.ndarray]
    ) -> Tuple[np.ndarray, int]:
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
        else:  # change non-float data to float data, need to copy
            data = np.array(mat.reshape(mat.size), dtype=np.float32)
        ptr_data, type_ptr_data, _ = _c_float_array(data)
1129
1130
1131
1132
1133
1134
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
            predict_type=predict_type
        )
1135
1136
1137
1138
1139
1140
        if preds is None:
            preds = np.empty(n_preds, dtype=np.float64)
        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(
1141
            self._handle,
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
            ptr_data,
            ctypes.c_int(type_ptr_data),
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
            ctypes.c_int(predict_type),
            ctypes.c_int(start_iteration),
            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]

    def __pred_for_np2d(
        self,
        mat: np.ndarray,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1164
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1165
        if len(mat.shape) != 2:
1166
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
1167

1168
        nrow = mat.shape[0]
1169
1170
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1171
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1172
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
1173
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1174
            preds = np.empty(sum(n_preds), dtype=np.float64)
1175
1176
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
1177
                # avoid memory consumption by arrays concatenation operations
1178
1179
1180
1181
1182
1183
1184
                self.__inner_predict_np2d(
                    mat=chunk,
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
                    preds=preds[start_idx_pred:end_idx_pred]
                )
1185
            return preds, nrow
wxchan's avatar
wxchan committed
1186
        else:
1187
1188
1189
1190
1191
1192
1193
            return self.__inner_predict_np2d(
                mat=mat,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
                preds=None
            )
wxchan's avatar
wxchan committed
1194

1195
1196
1197
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1198
1199
1200
1201
1202
1203
        out_shape: np.ndarray,
        out_ptr_indptr: "ctypes._Pointer",
        out_ptr_indices: "ctypes._Pointer",
        out_ptr_data: "ctypes._Pointer",
        indptr_type: int,
        data_type: int,
1204
        is_csr: bool
1205
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1206
1207
1208
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1209
        if indptr_type == _C_API_DTYPE_INT32:
1210
            out_indptr = _cint32_array_to_numpy(out_ptr_indptr, indptr_len)
1211
        elif indptr_type == _C_API_DTYPE_INT64:
1212
            out_indptr = _cint64_array_to_numpy(out_ptr_indptr, indptr_len)
1213
1214
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1215
        if data_type == _C_API_DTYPE_FLOAT32:
1216
            out_data = _cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
1217
        elif data_type == _C_API_DTYPE_FLOAT64:
1218
            out_data = _cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
1219
1220
        else:
            raise TypeError("Expected float32 or float64 type for data")
1221
        out_indices = _cint32_array_to_numpy(out_ptr_indices, data_indices_len)
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
        # 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

1250
1251
1252
1253
1254
1255
1256
1257
1258
    def __inner_predict_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
        preds: Optional[np.ndarray]
    ) -> Tuple[np.ndarray, int]:
        nrow = len(csr.indptr) - 1
1259
1260
1261
1262
1263
1264
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1265
1266
1267
1268
1269
        if preds is None:
            preds = np.empty(n_preds, dtype=np.float64)
        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)
wxchan's avatar
wxchan committed
1270

1271
1272
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1273

1274
1275
1276
1277
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

        _safe_call(_LIB.LGBM_BoosterPredictForCSR(
1278
            self._handle,
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
            ptr_indptr,
            ctypes.c_int(type_ptr_indptr),
            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),
            ctypes.c_int(start_iteration),
            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

    def __inner_predict_csr_sparse(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
1303
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1304
1305
1306
1307
        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
1308
        out_ptr_indptr: _ctypes_int_ptr
1309
1310
1311
1312
1313
        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)()
1314
        out_ptr_data: _ctypes_float_ptr
1315
1316
1317
1318
1319
1320
        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)()
        out_shape = np.empty(2, dtype=np.int64)
        _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
1321
            self._handle,
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
            ptr_indptr,
            ctypes.c_int(type_ptr_indptr),
            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),
            ctypes.c_int(start_iteration),
            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(
            cs=csr,
            out_shape=out_shape,
            out_ptr_indptr=out_ptr_indptr,
            out_ptr_indices=out_ptr_indices,
            out_ptr_data=out_ptr_data,
            indptr_type=type_ptr_indptr,
            data_type=type_ptr_data,
            is_csr=True
        )
        nrow = len(csr.indptr) - 1
        return matrices, nrow

1352
1353
1354
1355
1356
1357
1358
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1359
        """Predict for a CSR data."""
1360
        if predict_type == _C_API_PREDICT_CONTRIB:
1361
1362
1363
1364
1365
1366
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1367
        nrow = len(csr.indptr) - 1
1368
1369
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1370
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1371
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1372
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1373
            preds = np.empty(sum(n_preds), dtype=np.float64)
1374
1375
            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:])):
1376
                # avoid memory consumption by arrays concatenation operations
1377
1378
1379
1380
1381
1382
1383
                self.__inner_predict_csr(
                    csr=csr[start_idx:end_idx],
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
                    preds=preds[start_idx_pred:end_idx_pred]
                )
1384
1385
            return preds, nrow
        else:
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
            return self.__inner_predict_csr(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
                preds=None
            )

    def __inner_predict_sparse_csc(
        self,
1396
1397
1398
1399
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
1400
1401
1402
1403
1404
    ):
        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
1405
        out_ptr_indptr: _ctypes_int_ptr
1406
1407
1408
1409
1410
        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)()
1411
        out_ptr_data: _ctypes_float_ptr
1412
1413
1414
1415
1416
1417
        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)()
        out_shape = np.empty(2, dtype=np.int64)
        _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
1418
            self._handle,
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
            ptr_indptr,
            ctypes.c_int(type_ptr_indptr),
            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),
            ctypes.c_int(start_iteration),
            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(
            cs=csc,
            out_shape=out_shape,
            out_ptr_indptr=out_ptr_indptr,
            out_ptr_indices=out_ptr_indices,
            out_ptr_data=out_ptr_data,
            indptr_type=type_ptr_indptr,
            data_type=type_ptr_data,
            is_csr=False
        )
        nrow = csc.shape[0]
        return matrices, nrow
Guolin Ke's avatar
Guolin Ke committed
1448

1449
1450
1451
1452
1453
1454
1455
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1456
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1457
        nrow = csc.shape[0]
1458
        if nrow > _MAX_INT32:
1459
1460
1461
1462
1463
1464
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1465
        if predict_type == _C_API_PREDICT_CONTRIB:
1466
1467
1468
1469
1470
1471
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1472
1473
1474
1475
1476
1477
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1478
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1479
1480
        out_num_preds = ctypes.c_int64(0)

1481
1482
        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
1483

1484
        assert csc.shape[0] <= _MAX_INT32
1485
        csc_indices = csc.indices.astype(np.int32, copy=False)
1486

Guolin Ke's avatar
Guolin Ke committed
1487
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
1488
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1489
            ptr_indptr,
1490
            ctypes.c_int(type_ptr_indptr),
1491
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1492
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1493
1494
1495
1496
1497
            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),
1498
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1499
            ctypes.c_int(num_iteration),
1500
            _c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1501
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1502
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1503
        if n_preds != out_num_preds.value:
1504
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1505
1506
        return preds, nrow

1507
    def current_iteration(self) -> int:
1508
1509
1510
1511
1512
1513
1514
1515
1516
        """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(
1517
            self._handle,
1518
1519
1520
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1521

1522
class Dataset:
wxchan's avatar
wxchan committed
1523
    """Dataset in LightGBM."""
1524

1525
1526
    def __init__(
        self,
1527
        data: _LGBM_TrainDataType,
1528
        label: Optional[_LGBM_LabelType] = None,
1529
        reference: Optional["Dataset"] = None,
1530
1531
1532
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1533
1534
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
1535
        params: Optional[Dict[str, Any]] = None,
1536
1537
        free_raw_data: bool = True,
        position: Optional[_LGBM_PositionType] = None,
1538
    ):
1539
        """Initialize Dataset.
1540

wxchan's avatar
wxchan committed
1541
1542
        Parameters
        ----------
1543
        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
1544
            Data source of Dataset.
1545
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1546
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1547
1548
1549
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1550
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1551
            Weight for each instance. Weights should be non-negative.
1552
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1553
1554
1555
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1556
1557
            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.
1558
        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)
1559
            Init score for Dataset.
1560
        feature_name : list of str, or 'auto', optional (default="auto")
1561
1562
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
1563
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1564
1565
            Categorical features.
            If list of int, interpreted as indices.
1566
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1567
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1568
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1569
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1570
            All negative values in categorical features will be treated as missing values.
1571
            The output cannot be monotonically constrained with respect to a categorical feature.
1572
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1573
        params : dict or None, optional (default=None)
1574
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1575
        free_raw_data : bool, optional (default=True)
1576
            If True, raw data is freed after constructing inner Dataset.
1577
1578
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
wxchan's avatar
wxchan committed
1579
        """
1580
        self._handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1581
1582
1583
1584
1585
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1586
        self.position = position
1587
        self.init_score = init_score
1588
1589
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1590
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1591
        self.free_raw_data = free_raw_data
1592
        self.used_indices: Optional[List[int]] = None
1593
        self._need_slice = True
1594
        self._predictor: Optional[_InnerPredictor] = None
1595
        self.pandas_categorical = None
1596
        self._params_back_up = None
1597
        self.version = 0
1598
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1599

1600
    def __del__(self) -> None:
1601
1602
1603
1604
        try:
            self._free_handle()
        except AttributeError:
            pass
1605

1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
    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.
        """
1623
        param_str = _param_dict_to_str(self.get_params())
1624
1625
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1626
        ptr_data, _, _ = _c_int_array(indices)
1627
1628
1629
1630
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
1631
            _c_str(param_str),
1632
1633
1634
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1635
1636
        assert sample_cnt == actual_sample_cnt.value
        return indices
1637

1638
1639
1640
1641
1642
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
        ref_dataset: _DatasetHandle
    ) -> 'Dataset':
1643
1644
1645
1646
1647
1648
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1649
1650
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1651
1652
1653
1654
1655
1656

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
1657
        self._handle = ctypes.c_void_p()
1658
1659
1660
        _safe_call(_LIB.LGBM_DatasetCreateByReference(
            ref_dataset,
            ctypes.c_int64(total_nrow),
1661
            ctypes.byref(self._handle),
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
        ))
        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
        ----------
1676
        sample_data : list of numpy array
1677
            Sample data for each column.
1678
        sample_indices : list of numpy array
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
            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.
1701
        sample_col_ptr: _ctypes_float_array = (ctypes.POINTER(ctypes.c_double) * ncol)()
1702
1703
        # c type int**
        # each int* points to start of indices for each column
1704
        indices_col_ptr: _ctypes_int_array = (ctypes.POINTER(ctypes.c_int32) * ncol)()
1705
        for i in range(ncol):
1706
1707
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1708
1709

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1710
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1711

1712
        self._handle = ctypes.c_void_p()
1713
        params_str = _param_dict_to_str(self.get_params())
1714
1715
1716
1717
1718
1719
1720
        _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),
1721
            ctypes.c_int64(total_nrow),
1722
            _c_str(params_str),
1723
            ctypes.byref(self._handle),
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
        ))
        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)
1742
        data_ptr, data_type, _ = _c_float_array(data)
1743
1744

        _safe_call(_LIB.LGBM_DatasetPushRows(
1745
            self._handle,
1746
1747
1748
1749
1750
1751
1752
1753
1754
            data_ptr,
            data_type,
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol),
            ctypes.c_int32(self._start_row),
        ))
        self._start_row += nrow
        return self

1755
    def get_params(self) -> Dict[str, Any]:
1756
1757
1758
1759
        """Get the used parameters in the Dataset.

        Returns
        -------
1760
        params : dict
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
            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",
1776
                                                "linear_tree",
1777
1778
1779
1780
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1781
                                                "precise_float_parser",
1782
1783
1784
1785
1786
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1787
1788
        else:
            return {}
1789

1790
    def _free_handle(self) -> "Dataset":
1791
1792
1793
        if self._handle is not None:
            _safe_call(_LIB.LGBM_DatasetFree(self._handle))
            self._handle = None
1794
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
1795
1796
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1797
        return self
wxchan's avatar
wxchan committed
1798

1799
1800
1801
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
1802
        data: _LGBM_TrainDataType,
1803
        used_indices: Optional[Union[List[int], np.ndarray]]
1804
    ) -> "Dataset":
Guolin Ke's avatar
Guolin Ke committed
1805
        data_has_header = False
1806
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
1807
            # check data has header or not
1808
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1809
        num_data = self.num_data()
1810
        if predictor is not None:
1811
1812
1813
1814
1815
            init_score: Union[np.ndarray, scipy.sparse.spmatrix] = predictor.predict(
                data=data,
                raw_score=True,
                data_has_header=data_has_header
            )
1816
            init_score = init_score.ravel()
1817
            if used_indices is not None:
1818
                assert not self._need_slice
1819
                if isinstance(data, (str, Path)):
1820
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1821
                    assert num_data == len(used_indices)
1822
1823
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1824
1825
1826
1827
                            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
1828
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1829
1830
                for i in range(num_data):
                    for j in range(predictor.num_class):
1831
1832
1833
                        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:
1834
            init_score = np.full_like(self.init_score, fill_value=0.0, dtype=np.float64)
1835
1836
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1837
        self.set_init_score(init_score)
1838
        return self
Guolin Ke's avatar
Guolin Ke committed
1839

1840
1841
    def _lazy_init(
        self,
1842
        data: Optional[_LGBM_TrainDataType],
1843
1844
1845
1846
1847
1848
1849
1850
        label: Optional[_LGBM_LabelType],
        reference: Optional["Dataset"],
        weight: Optional[_LGBM_WeightType],
        group: Optional[_LGBM_GroupType],
        init_score: Optional[_LGBM_InitScoreType],
        predictor: Optional[_InnerPredictor],
        feature_name: _LGBM_FeatureNameConfiguration,
        categorical_feature: _LGBM_CategoricalFeatureConfiguration,
1851
1852
        params: Optional[Dict[str, Any]],
        position: Optional[_LGBM_PositionType]
1853
    ) -> "Dataset":
wxchan's avatar
wxchan committed
1854
        if data is None:
1855
            self._handle = None
Nikita Titov's avatar
Nikita Titov committed
1856
            return self
Guolin Ke's avatar
Guolin Ke committed
1857
1858
1859
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1860
1861
1862
1863
1864
1865
1866
        if isinstance(data, pd_DataFrame):
            data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(
                data=data,
                feature_name=feature_name,
                categorical_feature=categorical_feature,
                pandas_categorical=self.pandas_categorical
            )
Guolin Ke's avatar
Guolin Ke committed
1867

1868
        # process for args
wxchan's avatar
wxchan committed
1869
        params = {} if params is None else params
1870
        args_names = inspect.signature(self.__class__._lazy_init).parameters.keys()
1871
        for key in params.keys():
1872
            if key in args_names:
1873
1874
                _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.')
1875
        # get categorical features
1876
        if isinstance(categorical_feature, list):
1877
1878
            categorical_indices = set()
            feature_dict = {}
1879
            if isinstance(feature_name, list):
1880
1881
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
1882
                if isinstance(name, str) and name in feature_dict:
1883
                    categorical_indices.add(feature_dict[name])
1884
                elif isinstance(name, int):
1885
1886
                    categorical_indices.add(name)
                else:
1887
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1888
            if categorical_indices:
1889
1890
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1891
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
1892
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
1893
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1894
                        params.pop(cat_alias, None)
1895
                params['categorical_column'] = sorted(categorical_indices)
1896

1897
        params_str = _param_dict_to_str(params)
1898
        self.params = params
1899
        # process for reference dataset
wxchan's avatar
wxchan committed
1900
        ref_dataset = None
wxchan's avatar
wxchan committed
1901
        if isinstance(reference, Dataset):
1902
            ref_dataset = reference.construct()._handle
wxchan's avatar
wxchan committed
1903
1904
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1905
        # start construct data
1906
        if isinstance(data, (str, Path)):
1907
            self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1908
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1909
1910
                _c_str(str(data)),
                _c_str(params_str),
wxchan's avatar
wxchan committed
1911
                ref_dataset,
1912
                ctypes.byref(self._handle)))
wxchan's avatar
wxchan committed
1913
1914
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
1915
1916
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1917
1918
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1919
1920
1921
1922
1923
1924
1925
1926
1927
        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)
1928
        elif isinstance(data, dt_DataTable):
1929
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1930
1931
1932
1933
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1934
1935
            except BaseException as err:
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}') from err
wxchan's avatar
wxchan committed
1936
1937
1938
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1939
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1940
1941
1942
1943
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1944
1945
        if position is not None:
            self.set_position(position)
1946
1947
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1948
                _log_warning("The init_score will be overridden by the prediction of init_model.")
1949
1950
1951
1952
1953
            self._set_init_score_by_predictor(
                predictor=predictor,
                data=data,
                used_indices=None
            )
1954
1955
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1956
        elif predictor is not None:
1957
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1958
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1959
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1960

1961
1962
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
        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.
1988
        sampled = np.array(list(self._yield_row_from_seqlist(seqs, indices)))
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
        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

2004
2005
2006
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
2007
        ref_dataset: Optional[_DatasetHandle]
2008
    ) -> "Dataset":
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
        """
        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:
2023
            param_str = _param_dict_to_str(self.get_params())
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
            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

2037
2038
2039
2040
2041
2042
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2043
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
2044
2045
2046
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

2047
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2048
2049
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
2050
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
2051
2052
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

2053
        ptr_data, type_ptr_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2054
2055
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2056
            ctypes.c_int(type_ptr_data),
2057
2058
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
2059
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2060
            _c_str(params_str),
wxchan's avatar
wxchan committed
2061
            ref_dataset,
2062
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2063
        return self
wxchan's avatar
wxchan committed
2064

2065
2066
2067
2068
2069
2070
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2071
        """Initialize data from a list of 2-D numpy matrices."""
2072
        ncol = mats[0].shape[1]
2073
        nrow = np.empty((len(mats),), np.int32)
2074
        ptr_data: _ctypes_float_array
2075
2076
2077
2078
2079
2080
        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 = []
2081
        type_ptr_data = -1
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093

        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)
2094
            else:  # change non-float data to float data, need to copy
2095
2096
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2097
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2098
            if type_ptr_data != -1 and chunk_type_ptr_data != type_ptr_data:
2099
2100
2101
2102
2103
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

2104
        self._handle = ctypes.c_void_p()
2105
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
2106
            ctypes.c_int32(len(mats)),
2107
2108
2109
            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)),
2110
            ctypes.c_int32(ncol),
2111
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2112
            _c_str(params_str),
2113
            ref_dataset,
2114
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2115
        return self
2116

2117
2118
2119
2120
2121
2122
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2123
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2124
        if len(csr.indices) != len(csr.data):
2125
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
2126
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2127

2128
2129
        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
2130

2131
        assert csr.shape[1] <= _MAX_INT32
2132
        csr_indices = csr.indices.astype(np.int32, copy=False)
2133

wxchan's avatar
wxchan committed
2134
2135
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2136
            ctypes.c_int(type_ptr_indptr),
2137
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
2138
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2139
2140
2141
2142
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
2143
            _c_str(params_str),
wxchan's avatar
wxchan committed
2144
            ref_dataset,
2145
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2146
        return self
wxchan's avatar
wxchan committed
2147

2148
2149
2150
2151
2152
2153
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2154
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2155
        if len(csc.indices) != len(csc.data):
2156
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
2157
        self._handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2158

2159
2160
        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
2161

2162
        assert csc.shape[0] <= _MAX_INT32
2163
        csc_indices = csc.indices.astype(np.int32, copy=False)
2164

Guolin Ke's avatar
Guolin Ke committed
2165
2166
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2167
            ctypes.c_int(type_ptr_indptr),
2168
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
2169
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2170
2171
2172
2173
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
2174
            _c_str(params_str),
Guolin Ke's avatar
Guolin Ke committed
2175
            ref_dataset,
2176
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2177
        return self
Guolin Ke's avatar
Guolin Ke committed
2178

2179
    @staticmethod
2180
2181
2182
2183
2184
2185
    def _compare_params_for_warning(
        params: Optional[Dict[str, Any]],
        other_params: Optional[Dict[str, Any]],
        ignore_keys: Set[str]
    ) -> bool:
        """Compare two dictionaries with params ignoring some keys.
2186

2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
        It is only for the warning purpose.

        Parameters
        ----------
        params : dict or None
            One dictionary with parameters to compare.
        other_params : dict or None
            Another dictionary with parameters to compare.
        ignore_keys : set
            Keys that should be ignored during comparing two dictionaries.
2197
2198
2199

        Returns
        -------
2200
2201
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
        """
        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

2217
    def construct(self) -> "Dataset":
2218
2219
2220
2221
2222
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2223
            Constructed Dataset object.
2224
        """
2225
        if self._handle is None:
wxchan's avatar
wxchan committed
2226
            if self.reference is not None:
2227
                reference_params = self.reference.get_params()
2228
2229
                params = self.get_params()
                if params != reference_params:
2230
2231
2232
2233
2234
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
2235
                        _log_warning('Overriding the parameters from Reference Dataset.')
2236
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2237
                if self.used_indices is None:
2238
                    # create valid
2239
                    self._lazy_init(data=self.data, label=self.label, reference=self.reference,
2240
                                    weight=self.weight, group=self.group, position=self.position,
2241
                                    init_score=self.init_score, predictor=self._predictor,
2242
                                    feature_name=self.feature_name, categorical_feature='auto', params=self.params)
wxchan's avatar
wxchan committed
2243
                else:
2244
                    # construct subset
2245
                    used_indices = _list_to_1d_numpy(self.used_indices, dtype=np.int32, name='used_indices')
2246
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2247
                    if self.reference.group is not None:
2248
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2249
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
2250
                                                  return_counts=True)
2251
                    self._handle = ctypes.c_void_p()
2252
                    params_str = _param_dict_to_str(self.params)
wxchan's avatar
wxchan committed
2253
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
2254
                        self.reference.construct()._handle,
wxchan's avatar
wxchan committed
2255
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
2256
                        ctypes.c_int32(used_indices.shape[0]),
2257
                        _c_str(params_str),
2258
                        ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
2259
2260
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2261
2262
                    if self.group is not None:
                        self.set_group(self.group)
2263
2264
                    if self.position is not None:
                        self.set_position(self.position)
wxchan's avatar
wxchan committed
2265
2266
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
2267
2268
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
2269
2270
2271
2272
2273
                        self._set_init_score_by_predictor(
                            predictor=self._predictor,
                            data=self.data,
                            used_indices=used_indices
                        )
wxchan's avatar
wxchan committed
2274
            else:
2275
                # create train
2276
                self._lazy_init(data=self.data, label=self.label, reference=None,
2277
2278
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
2279
2280
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature,
                                params=self.params, position=self.position)
wxchan's avatar
wxchan committed
2281
2282
            if self.free_raw_data:
                self.data = None
2283
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2284
        return self
wxchan's avatar
wxchan committed
2285

2286
2287
    def create_valid(
        self,
2288
        data: _LGBM_TrainDataType,
2289
        label: Optional[_LGBM_LabelType] = None,
2290
2291
2292
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2293
2294
        params: Optional[Dict[str, Any]] = None,
        position: Optional[_LGBM_PositionType] = None
2295
    ) -> "Dataset":
2296
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2297
2298
2299

        Parameters
        ----------
2300
        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
2301
            Data source of Dataset.
2302
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2303
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
2304
2305
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
2306
            Weight for each instance. Weights should be non-negative.
2307
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
2308
2309
2310
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2311
2312
            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.
2313
        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)
2314
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2315
        params : dict or None, optional (default=None)
2316
            Other parameters for validation Dataset.
2317
2318
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
2319
2320
2321

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2322
2323
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2324
        """
2325
        ret = Dataset(data, label=label, reference=self,
2326
                      weight=weight, group=group, position=position, init_score=init_score,
2327
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2328
        ret._predictor = self._predictor
2329
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2330
        return ret
wxchan's avatar
wxchan committed
2331

2332
2333
2334
2335
2336
    def subset(
        self,
        used_indices: List[int],
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2337
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2338
2339
2340
2341

        Parameters
        ----------
        used_indices : list of int
2342
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2343
        params : dict or None, optional (default=None)
2344
            These parameters will be passed to Dataset constructor.
2345
2346
2347
2348
2349

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2350
        """
wxchan's avatar
wxchan committed
2351
2352
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
2353
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
2354
2355
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2356
        ret._predictor = self._predictor
2357
        ret.pandas_categorical = self.pandas_categorical
2358
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2359
2360
        return ret

2361
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2362
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2363

2364
2365
2366
2367
2368
        .. 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
2369
2370
        Parameters
        ----------
2371
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2372
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2373
2374
2375
2376
2377

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2378
2379
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
2380
            self.construct()._handle,
2381
            _c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
2382
        return self
wxchan's avatar
wxchan committed
2383

2384
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2385
2386
        if not params:
            return self
2387
        params = deepcopy(params)
2388
2389
2390
2391
2392

        def update():
            if not self.params:
                self.params = params
            else:
2393
                self._params_back_up = deepcopy(self.params)
2394
2395
                self.params.update(params)

2396
        if self._handle is None:
2397
2398
2399
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2400
2401
                _c_str(_param_dict_to_str(self.params)),
                _c_str(_param_dict_to_str(params)))
2402
2403
2404
2405
2406
2407
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2408
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
2409
        return self
wxchan's avatar
wxchan committed
2410

2411
    def _reverse_update_params(self) -> "Dataset":
2412
        if self._handle is None:
2413
2414
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2415
        return self
2416

2417
2418
2419
    def set_field(
        self,
        field_name: str,
2420
        data: Optional[Union[List[List[float]], List[List[int]], List[float], List[int], np.ndarray, pd_Series, pd_DataFrame]]
2421
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2422
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2423
2424
2425

        Parameters
        ----------
2426
        field_name : str
2427
            The field name of the information.
2428
2429
        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
2430
2431
2432
2433
2434

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2435
        """
2436
        if self._handle is None:
2437
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2438
        if data is None:
2439
            # set to None
wxchan's avatar
wxchan committed
2440
            _safe_call(_LIB.LGBM_DatasetSetField(
2441
                self._handle,
2442
                _c_str(field_name),
wxchan's avatar
wxchan committed
2443
                None,
Guolin Ke's avatar
Guolin Ke committed
2444
                ctypes.c_int(0),
2445
                ctypes.c_int(_FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2446
            return self
2447
        dtype: "np.typing.DTypeLike"
2448
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2449
            dtype = np.float64
2450
            if _is_1d_collection(data):
2451
                data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2452
            elif _is_2d_collection(data):
2453
                data = _data_to_2d_numpy(data, dtype=dtype, name=field_name)
2454
2455
2456
2457
2458
2459
2460
                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:
2461
            dtype = np.int32 if (field_name == 'group' or field_name == 'position') else np.float32
2462
            data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2463

2464
        ptr_data: Union[_ctypes_float_ptr, _ctypes_int_ptr]
2465
        if data.dtype == np.float32 or data.dtype == np.float64:
2466
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2467
        elif data.dtype == np.int32:
2468
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2469
        else:
2470
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2471
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2472
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2473
        _safe_call(_LIB.LGBM_DatasetSetField(
2474
            self._handle,
2475
            _c_str(field_name),
wxchan's avatar
wxchan committed
2476
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2477
2478
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2479
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2480
        return self
wxchan's avatar
wxchan committed
2481

2482
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2483
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2484
2485
2486

        Parameters
        ----------
2487
        field_name : str
2488
            The field name of the information.
wxchan's avatar
wxchan committed
2489
2490
2491

        Returns
        -------
2492
        info : numpy array or None
2493
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2494
        """
2495
        if self._handle is None:
2496
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2497
2498
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2499
2500
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
2501
            self._handle,
2502
            _c_str(field_name),
wxchan's avatar
wxchan committed
2503
2504
2505
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
2506
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2507
2508
2509
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2510
        if out_type.value == _C_API_DTYPE_INT32:
2511
            arr = _cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
2512
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2513
            arr = _cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
2514
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2515
            arr = _cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2516
        else:
wxchan's avatar
wxchan committed
2517
            raise TypeError("Unknown type")
2518
2519
2520
2521
2522
2523
        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
2524

2525
2526
    def set_categorical_feature(
        self,
2527
        categorical_feature: _LGBM_CategoricalFeatureConfiguration
2528
    ) -> "Dataset":
2529
        """Set categorical features.
2530
2531
2532

        Parameters
        ----------
2533
        categorical_feature : list of str or int, or 'auto'
2534
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2535
2536
2537
2538
2539

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2540
2541
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2542
            return self
2543
        if self.data is not None:
2544
2545
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2546
                return self._free_handle()
2547
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2548
                return self
2549
            else:
2550
2551
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
2552
                                 f'New categorical_feature is {list(categorical_feature)}')
2553
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2554
                return self._free_handle()
2555
        else:
2556
2557
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2558

2559
2560
2561
2562
    def _set_predictor(
        self,
        predictor: Optional[_InnerPredictor]
    ) -> "Dataset":
2563
2564
2565
2566
        """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
2567
        """
2568
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2569
            return self
2570
2571
2572
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
            if (predictor == self._predictor) and (predictor.current_iteration() == self._predictor.current_iteration()):
                return self
2573
        if self._handle is None:
Guolin Ke's avatar
Guolin Ke committed
2574
            self._predictor = predictor
2575
2576
        elif self.data is not None:
            self._predictor = predictor
2577
2578
2579
2580
2581
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
                used_indices=None
            )
2582
2583
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2584
2585
2586
2587
2588
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.reference.data,
                used_indices=self.used_indices
            )
Guolin Ke's avatar
Guolin Ke committed
2589
        else:
2590
2591
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2592
        return self
Guolin Ke's avatar
Guolin Ke committed
2593

2594
    def set_reference(self, reference: "Dataset") -> "Dataset":
2595
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2596
2597
2598
2599

        Parameters
        ----------
        reference : Dataset
2600
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2601
2602
2603
2604
2605

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2606
        """
2607
2608
2609
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2610
        # we're done if self and reference share a common upstream reference
2611
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2612
            return self
Guolin Ke's avatar
Guolin Ke committed
2613
2614
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2615
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2616
        else:
2617
2618
            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
2619

2620
    def set_feature_name(self, feature_name: _LGBM_FeatureNameConfiguration) -> "Dataset":
2621
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2622
2623
2624

        Parameters
        ----------
2625
        feature_name : list of str
2626
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2627
2628
2629
2630
2631

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2632
        """
2633
2634
        if feature_name != 'auto':
            self.feature_name = feature_name
2635
        if self._handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2636
            if len(feature_name) != self.num_feature():
2637
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2638
            c_feature_name = [_c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2639
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
2640
                self._handle,
2641
                _c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2642
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2643
        return self
Guolin Ke's avatar
Guolin Ke committed
2644

2645
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
2646
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2647
2648
2649

        Parameters
        ----------
2650
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2651
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2652
2653
2654
2655
2656

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2657
2658
        """
        self.label = label
2659
        if self._handle is not None:
2660
2661
2662
2663
            if isinstance(label, pd_DataFrame):
                if len(label.columns) > 1:
                    raise ValueError('DataFrame for label cannot have multiple columns')
                _check_for_bad_pandas_dtypes(label.dtypes)
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
                try:
                    # most common case (no nullable dtypes)
                    label = label.to_numpy(dtype=np.float32, copy=False)
                except TypeError:
                    # 1.0 <= pd version < 1.1 and nullable dtypes, least common case
                    # raises error because array is casted to type(pd.NA) and there's no na_value argument
                    label = label.astype(np.float32, copy=False).values
                except ValueError:
                    # data has nullable dtypes, but we can specify na_value argument and copy will be made
                    label = label.to_numpy(dtype=np.float32, na_value=np.nan)
                label_array = np.ravel(label)
2675
            else:
2676
                label_array = _list_to_1d_numpy(label, dtype=np.float32, name='label')
2677
            self.set_field('label', label_array)
2678
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2679
        return self
Guolin Ke's avatar
Guolin Ke committed
2680

2681
2682
2683
2684
    def set_weight(
        self,
        weight: Optional[_LGBM_WeightType]
    ) -> "Dataset":
2685
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2686
2687
2688

        Parameters
        ----------
2689
        weight : list, numpy 1-D array, pandas Series or None
2690
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2691
2692
2693
2694
2695

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2696
        """
2697
2698
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2699
        self.weight = weight
2700
        if self._handle is not None and weight is not None:
2701
            weight = _list_to_1d_numpy(weight, dtype=np.float32, name='weight')
wxchan's avatar
wxchan committed
2702
            self.set_field('weight', weight)
2703
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2704
        return self
Guolin Ke's avatar
Guolin Ke committed
2705

2706
2707
2708
2709
    def set_init_score(
        self,
        init_score: Optional[_LGBM_InitScoreType]
    ) -> "Dataset":
2710
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2711
2712
2713

        Parameters
        ----------
2714
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2715
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2716
2717
2718
2719
2720

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2721
2722
        """
        self.init_score = init_score
2723
        if self._handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2724
            self.set_field('init_score', init_score)
2725
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2726
        return self
Guolin Ke's avatar
Guolin Ke committed
2727

2728
2729
2730
2731
    def set_group(
        self,
        group: Optional[_LGBM_GroupType]
    ) -> "Dataset":
2732
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2733
2734
2735

        Parameters
        ----------
2736
        group : list, numpy 1-D array, pandas Series or None
2737
2738
2739
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2740
2741
            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
2742
2743
2744
2745
2746

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2747
2748
        """
        self.group = group
2749
        if self._handle is not None and group is not None:
2750
            group = _list_to_1d_numpy(group, dtype=np.int32, name='group')
wxchan's avatar
wxchan committed
2751
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2752
        return self
Guolin Ke's avatar
Guolin Ke committed
2753

2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
    def set_position(
        self,
        position: Optional[_LGBM_PositionType]
    ) -> "Dataset":
        """Set position of Dataset (used for ranking).

        Parameters
        ----------
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.

        Returns
        -------
        self : Dataset
            Dataset with set position.
        """
        self.position = position
        if self._handle is not None and position is not None:
            position = _list_to_1d_numpy(position, dtype=np.int32, name='position')
            self.set_field('position', position)
        return self

2776
    def get_feature_name(self) -> List[str]:
2777
2778
2779
2780
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2781
        feature_names : list of str
2782
2783
            The names of columns (features) in the Dataset.
        """
2784
        if self._handle is None:
2785
2786
2787
2788
2789
            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)
2790
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2791
2792
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
2793
            self._handle,
2794
            ctypes.c_int(num_feature),
2795
            ctypes.byref(tmp_out_len),
2796
            ctypes.c_size_t(reserved_string_buffer_size),
2797
2798
2799
2800
            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")
2801
2802
2803
2804
2805
2806
        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(
2807
                self._handle,
2808
2809
2810
2811
2812
                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))
2813
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2814

2815
    def get_label(self) -> Optional[np.ndarray]:
2816
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2817
2818
2819

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2820
        label : numpy array or None
2821
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2822
        """
2823
        if self.label is None:
wxchan's avatar
wxchan committed
2824
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2825
2826
        return self.label

2827
    def get_weight(self) -> Optional[np.ndarray]:
2828
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2829
2830
2831

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2832
        weight : numpy array or None
2833
            Weight for each data point from the Dataset. Weights should be non-negative.
Guolin Ke's avatar
Guolin Ke committed
2834
        """
2835
        if self.weight is None:
wxchan's avatar
wxchan committed
2836
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2837
2838
        return self.weight

2839
    def get_init_score(self) -> Optional[np.ndarray]:
2840
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2841
2842
2843

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2844
        init_score : numpy array or None
2845
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2846
        """
2847
        if self.init_score is None:
wxchan's avatar
wxchan committed
2848
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2849
2850
        return self.init_score

2851
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
2852
2853
2854
2855
        """Get the raw data of the Dataset.

        Returns
        -------
2856
        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
2857
2858
            Raw data used in the Dataset construction.
        """
2859
        if self._handle is None:
2860
            raise Exception("Cannot get data before construct Dataset")
2861
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
2862
2863
            self.data = self.reference.data
            if self.data is not None:
2864
                if isinstance(self.data, np.ndarray) or isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2865
                    self.data = self.data[self.used_indices, :]
2866
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2867
                    self.data = self.data.iloc[self.used_indices].copy()
2868
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2869
                    self.data = self.data[self.used_indices, :]
2870
2871
2872
                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):
2873
                    self.data = np.array(list(self._yield_row_from_seqlist(self.data, self.used_indices)))
Guolin Ke's avatar
Guolin Ke committed
2874
                else:
2875
2876
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
2877
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
2878
2879
2880
        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.")
2881
2882
        return self.data

2883
    def get_group(self) -> Optional[np.ndarray]:
2884
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2885
2886
2887

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2888
        group : numpy array or None
2889
2890
2891
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2892
2893
            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
2894
        """
2895
        if self.group is None:
wxchan's avatar
wxchan committed
2896
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
2897
2898
            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
2899
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
2900
2901
        return self.group

2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
    def get_position(self) -> Optional[np.ndarray]:
        """Get the position of the Dataset.

        Returns
        -------
        position : numpy 1-D array or None
            Position of items used in unbiased learning-to-rank task.
        """
        if self.position is None:
            self.position = self.get_field('position')
        return self.position

2914
    def num_data(self) -> int:
2915
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2916
2917
2918

        Returns
        -------
2919
2920
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2921
        """
2922
        if self._handle is not None:
2923
            ret = ctypes.c_int(0)
2924
            _safe_call(_LIB.LGBM_DatasetGetNumData(self._handle,
wxchan's avatar
wxchan committed
2925
2926
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2927
        else:
2928
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2929

2930
    def num_feature(self) -> int:
2931
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2932
2933
2934

        Returns
        -------
2935
2936
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2937
        """
2938
        if self._handle is not None:
2939
            ret = ctypes.c_int(0)
2940
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self._handle,
wxchan's avatar
wxchan committed
2941
2942
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2943
        else:
2944
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2945

2946
    def feature_num_bin(self, feature: Union[int, str]) -> int:
2947
2948
        """Get the number of bins for a feature.

2949
2950
        .. versionadded:: 4.0.0

2951
2952
        Parameters
        ----------
2953
2954
        feature : int or str
            Index or name of the feature.
2955
2956
2957
2958
2959
2960

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
2961
        if self._handle is not None:
2962
            if isinstance(feature, str):
2963
2964
2965
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
2966
            ret = ctypes.c_int(0)
2967
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self._handle,
2968
                                                         ctypes.c_int(feature_index),
2969
2970
2971
2972
2973
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

2974
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
2975
2976
2977
2978
2979
        """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.
2980
2981
2982
2983
2984

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2985
2986
2987

        Returns
        -------
2988
2989
2990
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2991
        head = self
2992
        ref_chain: Set[Dataset] = set()
2993
2994
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2995
                ref_chain.add(head)
2996
2997
2998
2999
3000
3001
                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
3002
        return ref_chain
3003

3004
    def add_features_from(self, other: "Dataset") -> "Dataset":
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
        """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.
        """
3019
        if self._handle is None or other._handle is None:
3020
            raise ValueError('Both source and target Datasets must be constructed before adding features')
3021
        _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self._handle, other._handle))
Guolin Ke's avatar
Guolin Ke committed
3022
3023
3024
3025
3026
3027
3028
3029
        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))
3030
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3031
                    self.data = np.hstack((self.data, other.data.toarray()))
3032
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3033
                    self.data = np.hstack((self.data, other.data.values))
3034
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3035
3036
3037
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
3038
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3039
                sparse_format = self.data.getformat()
3040
                if isinstance(other.data, np.ndarray) or isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3041
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
3042
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3043
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
3044
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3045
3046
3047
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
3048
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3049
3050
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
3051
3052
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
3053
                if isinstance(other.data, np.ndarray):
3054
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
3055
                                       axis=1, ignore_index=True)
3056
                elif isinstance(other.data, scipy.sparse.spmatrix):
3057
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
3058
                                       axis=1, ignore_index=True)
3059
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3060
3061
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
3062
3063
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
3064
3065
3066
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
3067
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3068
                if isinstance(other.data, np.ndarray):
3069
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
3070
                elif isinstance(other.data, scipy.sparse.spmatrix):
3071
3072
3073
3074
3075
                    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
3076
3077
3078
3079
3080
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
3081
3082
            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
3083
3084
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
3085
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
3086
        self.feature_name = self.get_feature_name()
3087
3088
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
3089
3090
        self.categorical_feature = "auto"
        self.pandas_categorical = None
3091
3092
        return self

3093
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
3094
3095
3096
3097
3098
3099
        """Save Dataset to a text file.

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

        Parameters
        ----------
3100
        filename : str or pathlib.Path
3101
3102
3103
3104
3105
3106
3107
3108
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
3109
            self.construct()._handle,
3110
            _c_str(str(filename))))
3111
3112
        return self

wxchan's avatar
wxchan committed
3113

3114
3115
3116
3117
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
    Tuple[np.ndarray, np.ndarray]
]
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
        _LGBM_EvalFunctionResultType
    ],
    Callable[
        [np.ndarray, Dataset],
        List[_LGBM_EvalFunctionResultType]
    ]
]
3128
3129


3130
class Booster:
3131
    """Booster in LightGBM."""
3132

3133
3134
3135
3136
3137
3138
3139
    def __init__(
        self,
        params: Optional[Dict[str, Any]] = None,
        train_set: Optional[Dataset] = None,
        model_file: Optional[Union[str, Path]] = None,
        model_str: Optional[str] = None
    ):
3140
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3141
3142
3143

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3144
        params : dict or None, optional (default=None)
3145
3146
3147
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3148
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3149
            Path to the model file.
3150
        model_str : str or None, optional (default=None)
3151
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3152
        """
3153
        self._handle = ctypes.c_void_p()
3154
        self._network = False
wxchan's avatar
wxchan committed
3155
        self.__need_reload_eval_info = True
3156
        self._train_data_name = "training"
3157
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3158
        self.best_iteration = -1
3159
        self.best_score: _LGBM_BoosterBestScoreType = {}
3160
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3161
        if train_set is not None:
3162
            # Training task
wxchan's avatar
wxchan committed
3163
            if not isinstance(train_set, Dataset):
3164
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
            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"]
                )
3199
            # construct booster object
3200
3201
3202
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3203
            params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3204
            _safe_call(_LIB.LGBM_BoosterCreate(
3205
                train_set._handle,
3206
                _c_str(params_str),
3207
                ctypes.byref(self._handle)))
3208
            # save reference to data
wxchan's avatar
wxchan committed
3209
            self.train_set = train_set
3210
3211
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3212
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3213
3214
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
3215
                _safe_call(_LIB.LGBM_BoosterMerge(
3216
3217
                    self._handle,
                    self.__init_predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
3218
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3219
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3220
                self._handle,
wxchan's avatar
wxchan committed
3221
3222
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3223
            # buffer for inner predict
3224
            self.__inner_predict_buffer: List[Optional[np.ndarray]] = [None]
wxchan's avatar
wxchan committed
3225
3226
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3227
            self.pandas_categorical = train_set.pandas_categorical
3228
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3229
        elif model_file is not None:
3230
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3231
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3232
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
3233
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
3234
                ctypes.byref(out_num_iterations),
3235
                ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
3236
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3237
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3238
                self._handle,
wxchan's avatar
wxchan committed
3239
3240
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3241
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3242
3243
3244
            if params:
                _log_warning('Ignoring params argument, using parameters from model file.')
            params = self._get_loaded_param()
3245
        elif model_str is not None:
3246
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3247
        else:
3248
3249
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
3250
        self.params = params
wxchan's avatar
wxchan committed
3251

3252
    def __del__(self) -> None:
3253
        try:
3254
            if self._network:
3255
3256
3257
3258
                self.free_network()
        except AttributeError:
            pass
        try:
3259
3260
            if self._handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
3261
3262
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3263

3264
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3265
3266
        return self.__deepcopy__(None)

3267
    def __deepcopy__(self, _) -> "Booster":
3268
        model_str = self.model_to_string(num_iteration=-1)
3269
        return Booster(model_str=model_str)
wxchan's avatar
wxchan committed
3270

3271
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3272
        this = self.__dict__.copy()
3273
        handle = this['_handle']
wxchan's avatar
wxchan committed
3274
3275
3276
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
3277
            this["_handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3278
3279
        return this

3280
    def __setstate__(self, state: Dict[str, Any]) -> None:
3281
        model_str = state.get('_handle', state.get('handle', None))
3282
        if model_str is not None:
wxchan's avatar
wxchan committed
3283
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3284
            out_num_iterations = ctypes.c_int(0)
3285
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3286
                _c_str(model_str),
3287
3288
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
3289
            state['_handle'] = handle
wxchan's avatar
wxchan committed
3290
3291
        self.__dict__.update(state)

3292
3293
3294
3295
3296
3297
    def _get_loaded_param(self) -> Dict[str, Any]:
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
3298
            self._handle,
3299
3300
3301
3302
3303
3304
3305
3306
3307
            ctypes.c_int64(buffer_len),
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
        # if buffer length is not long enough, re-allocate a buffer
        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_BoosterGetLoadedParam(
3308
                self._handle,
3309
3310
3311
3312
3313
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
        return json.loads(string_buffer.value.decode('utf-8'))

3314
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3315
3316
3317
3318
3319
3320
3321
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
3322
3323
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
3324
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3325
        return self
wxchan's avatar
wxchan committed
3326

3327
    def _free_buffer(self) -> "Booster":
3328
3329
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3330
        return self
3331

3332
3333
3334
3335
3336
3337
3338
    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":
3339
3340
3341
3342
        """Set the network configuration.

        Parameters
        ----------
3343
        machines : list, set or str
3344
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3345
        local_listen_port : int, optional (default=12400)
3346
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3347
        listen_time_out : int, optional (default=120)
3348
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3349
        num_machines : int, optional (default=1)
3350
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3351
3352
3353
3354
3355

        Returns
        -------
        self : Booster
            Booster with set network.
3356
        """
3357
3358
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
3359
        _safe_call(_LIB.LGBM_NetworkInit(_c_str(machines),
3360
3361
3362
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
3363
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3364
        return self
3365

3366
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3367
3368
3369
3370
3371
3372
3373
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3374
        _safe_call(_LIB.LGBM_NetworkFree())
3375
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3376
        return self
3377

3378
    def trees_to_dataframe(self) -> pd_DataFrame:
3379
3380
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3381
3382
3383
3384
        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.
3385
3386
3387
3388
3389
            - ``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.
3390
3391
            - ``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.
3392
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3393
3394
              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.
3395
3396
            - ``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.
3397
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3398
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3399
3400
            - ``count`` : int64, number of records in the training data that fall into this node.

3401
3402
3403
3404
3405
3406
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3407
3408
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
3409
3410
3411
3412

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

3413
        def _is_split_node(tree: Dict[str, Any]) -> bool:
3414
3415
            return 'split_index' in tree.keys()

3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
        def create_node_record(
            tree: Dict[str, Any],
            node_depth: int = 1,
            tree_index: Optional[int] = None,
            feature_names: Optional[List[str]] = None,
            parent_node: Optional[str] = None
        ) -> Dict[str, Any]:

            def _get_node_index(
                tree: Dict[str, Any],
                tree_index: Optional[int]
            ) -> str:
3428
                tree_num = f'{tree_index}-' if tree_index is not None else ''
3429
3430
3431
                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
3432
3433
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
3434

3435
3436
3437
3438
            def _get_split_feature(
                tree: Dict[str, Any],
                feature_names: Optional[List[str]]
            ) -> Optional[str]:
3439
3440
3441
3442
3443
3444
3445
3446
3447
                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

3448
            def _is_single_node_tree(tree: Dict[str, Any]) -> bool:
3449
                return set(tree.keys()) == {'leaf_value'}
3450
3451

            # Create the node record, and populate universal data members
3452
            node: Dict[str, Union[int, str, None]] = OrderedDict()
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
            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

3489
3490
3491
3492
3493
3494
3495
        def tree_dict_to_node_list(
            tree: Dict[str, Any],
            node_depth: int = 1,
            tree_index: Optional[int] = None,
            feature_names: Optional[List[str]] = None,
            parent_node: Optional[str] = None
        ) -> List[Dict[str, Any]]:
3496

3497
            node = create_node_record(tree=tree,
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
                                      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(
3510
                        tree=tree[child],
3511
3512
3513
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
3514
3515
                        parent_node=node['node_index']
                    )
3516
3517
3518
3519
3520
3521
3522
3523
3524
                    # 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']:
3525
            model_list.extend(tree_dict_to_node_list(tree=tree['tree_structure'],
3526
3527
3528
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

3529
        return pd_DataFrame(model_list, columns=model_list[0].keys())
3530

3531
    def set_train_data_name(self, name: str) -> "Booster":
3532
3533
3534
3535
        """Set the name to the training Dataset.

        Parameters
        ----------
3536
        name : str
Nikita Titov's avatar
Nikita Titov committed
3537
3538
3539
3540
3541
3542
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
3543
        """
3544
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
3545
        return self
wxchan's avatar
wxchan committed
3546

3547
    def add_valid(self, data: Dataset, name: str) -> "Booster":
3548
        """Add validation data.
wxchan's avatar
wxchan committed
3549
3550
3551
3552

        Parameters
        ----------
        data : Dataset
3553
            Validation data.
3554
        name : str
3555
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
3556
3557
3558
3559
3560

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
3561
        """
Guolin Ke's avatar
Guolin Ke committed
3562
        if not isinstance(data, Dataset):
3563
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3564
        if data._predictor is not self.__init_predictor:
3565
3566
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3567
        _safe_call(_LIB.LGBM_BoosterAddValidData(
3568
3569
            self._handle,
            data.construct()._handle))
wxchan's avatar
wxchan committed
3570
3571
3572
3573
3574
        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
3575
        return self
wxchan's avatar
wxchan committed
3576

3577
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
3578
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
3579
3580
3581
3582

        Parameters
        ----------
        params : dict
3583
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
3584
3585
3586
3587
3588

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
3589
        """
3590
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3591
3592
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
3593
                self._handle,
3594
                _c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
3595
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
3596
        return self
wxchan's avatar
wxchan committed
3597

3598
3599
3600
3601
3602
    def update(
        self,
        train_set: Optional[Dataset] = None,
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
3603
        """Update Booster for one iteration.
3604

wxchan's avatar
wxchan committed
3605
3606
        Parameters
        ----------
3607
3608
3609
3610
        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
3611
            Customized objective function.
3612
3613
3614
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3615
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3616
                    The predicted values.
3617
3618
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3619
3620
                train_data : Dataset
                    The training dataset.
3621
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3622
3623
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3624
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3625
3626
                    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
3627

3628
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3629
            and grad and hess should be returned in the same format.
3630

wxchan's avatar
wxchan committed
3631
3632
        Returns
        -------
3633
3634
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3635
        """
3636
        # need reset training data
3637
3638
3639
3640
3641
3642
        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
3643
            if not isinstance(train_set, Dataset):
3644
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3645
            if train_set._predictor is not self.__init_predictor:
3646
3647
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3648
3649
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
3650
3651
                self._handle,
                self.train_set.construct()._handle))
wxchan's avatar
wxchan committed
3652
            self.__inner_predict_buffer[0] = None
3653
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3654
3655
        is_finished = ctypes.c_int(0)
        if fobj is None:
3656
            if self.__set_objective_to_none:
3657
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3658
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
3659
                self._handle,
wxchan's avatar
wxchan committed
3660
                ctypes.byref(is_finished)))
3661
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3662
3663
            return is_finished.value == 1
        else:
3664
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3665
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3666
3667
3668
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

3669
3670
3671
3672
3673
    def __boost(
        self,
        grad: np.ndarray,
        hess: np.ndarray
    ) -> bool:
3674
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3675

Nikita Titov's avatar
Nikita Titov committed
3676
3677
        .. note::

3678
3679
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3680
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3681
            and grad and hess should be returned in the same format.
3682

wxchan's avatar
wxchan committed
3683
3684
        Parameters
        ----------
3685
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3686
3687
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3688
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3689
3690
            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
3691
3692
3693

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3694
3695
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3696
        """
3697
3698
3699
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3700
3701
        grad = _list_to_1d_numpy(grad, dtype=np.float32, name='gradient')
        hess = _list_to_1d_numpy(hess, dtype=np.float32, name='hessian')
3702
3703
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3704
        if len(grad) != len(hess):
3705
3706
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3707
        if len(grad) != num_train_data * self.__num_class:
3708
3709
3710
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3711
                f"number of models per one iteration ({self.__num_class})"
3712
            )
wxchan's avatar
wxchan committed
3713
3714
        is_finished = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterUpdateOneIterCustom(
3715
            self._handle,
wxchan's avatar
wxchan committed
3716
3717
3718
            grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            ctypes.byref(is_finished)))
3719
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3720
3721
        return is_finished.value == 1

3722
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3723
3724
3725
3726
3727
3728
3729
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3730
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
3731
            self._handle))
3732
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3733
        return self
wxchan's avatar
wxchan committed
3734

3735
    def current_iteration(self) -> int:
3736
3737
3738
3739
3740
3741
3742
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3743
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3744
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
3745
            self._handle,
wxchan's avatar
wxchan committed
3746
3747
3748
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3749
    def num_model_per_iteration(self) -> int:
3750
3751
3752
3753
3754
3755
3756
3757
3758
        """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(
3759
            self._handle,
3760
3761
3762
            ctypes.byref(model_per_iter)))
        return model_per_iter.value

3763
    def num_trees(self) -> int:
3764
3765
3766
3767
3768
3769
3770
3771
3772
        """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(
3773
            self._handle,
3774
3775
3776
            ctypes.byref(num_trees)))
        return num_trees.value

3777
    def upper_bound(self) -> float:
3778
3779
3780
3781
        """Get upper bound value of a model.

        Returns
        -------
3782
        upper_bound : float
3783
3784
3785
3786
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
3787
            self._handle,
3788
3789
3790
            ctypes.byref(ret)))
        return ret.value

3791
    def lower_bound(self) -> float:
3792
3793
3794
3795
        """Get lower bound value of a model.

        Returns
        -------
3796
        lower_bound : float
3797
3798
3799
3800
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
3801
            self._handle,
3802
3803
3804
            ctypes.byref(ret)))
        return ret.value

3805
3806
3807
3808
3809
3810
    def eval(
        self,
        data: Dataset,
        name: str,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3811
        """Evaluate for data.
wxchan's avatar
wxchan committed
3812
3813
3814

        Parameters
        ----------
3815
3816
        data : Dataset
            Data for the evaluating.
3817
        name : str
3818
            Name of the data.
3819
        feval : callable, list of callable, or None, optional (default=None)
3820
            Customized evaluation function.
3821
            Each evaluation function should accept two parameters: preds, eval_data,
3822
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3823

3824
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3825
                    The predicted values.
3826
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3827
                    If custom objective function is used, predicted values are returned before any transformation,
3828
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3829
                eval_data : Dataset
3830
                    A ``Dataset`` to evaluate.
3831
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3832
                    The name of evaluation function (without whitespace).
3833
3834
3835
3836
3837
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

wxchan's avatar
wxchan committed
3838
3839
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3840
        result : list
3841
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3842
        """
Guolin Ke's avatar
Guolin Ke committed
3843
3844
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3845
3846
3847
3848
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3849
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3850
3851
3852
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3853
        # need to push new valid data
wxchan's avatar
wxchan committed
3854
3855
3856
3857
3858
3859
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

3860
3861
3862
3863
    def eval_train(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3864
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3865
3866
3867

        Parameters
        ----------
3868
        feval : callable, list of callable, or None, optional (default=None)
3869
            Customized evaluation function.
3870
            Each evaluation function should accept two parameters: preds, eval_data,
3871
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3872

3873
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3874
                    The predicted values.
3875
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3876
                    If custom objective function is used, predicted values are returned before any transformation,
3877
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
Akshita Dixit's avatar
Akshita Dixit committed
3878
                eval_data : Dataset
3879
                    The training dataset.
3880
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3881
                    The name of evaluation function (without whitespace).
3882
3883
3884
3885
3886
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

wxchan's avatar
wxchan committed
3887
3888
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3889
        result : list
3890
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3891
        """
3892
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
3893

3894
3895
3896
3897
    def eval_valid(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3898
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
3899
3900
3901

        Parameters
        ----------
3902
        feval : callable, list of callable, or None, optional (default=None)
3903
            Customized evaluation function.
3904
            Each evaluation function should accept two parameters: preds, eval_data,
3905
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3906

3907
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3908
                    The predicted values.
3909
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3910
                    If custom objective function is used, predicted values are returned before any transformation,
3911
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
Akshita Dixit's avatar
Akshita Dixit committed
3912
                eval_data : Dataset
3913
                    The validation dataset.
3914
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3915
                    The name of evaluation function (without whitespace).
3916
3917
3918
3919
3920
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

wxchan's avatar
wxchan committed
3921
3922
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3923
        result : list
3924
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3925
        """
3926
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
3927
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
3928

3929
3930
3931
3932
3933
3934
3935
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> "Booster":
3936
        """Save Booster to file.
wxchan's avatar
wxchan committed
3937
3938
3939

        Parameters
        ----------
3940
        filename : str or pathlib.Path
3941
            Filename to save Booster.
3942
3943
3944
3945
        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
3946
        start_iteration : int, optional (default=0)
3947
            Start index of the iteration that should be saved.
3948
        importance_type : str, optional (default="split")
3949
3950
3951
            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
3952
3953
3954
3955
3956

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
3957
        """
3958
        if num_iteration is None:
3959
            num_iteration = self.best_iteration
3960
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3961
        _safe_call(_LIB.LGBM_BoosterSaveModel(
3962
            self._handle,
3963
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3964
            ctypes.c_int(num_iteration),
3965
            ctypes.c_int(importance_type_int),
3966
            _c_str(str(filename))))
3967
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
3968
        return self
wxchan's avatar
wxchan committed
3969

3970
3971
3972
3973
3974
    def shuffle_models(
        self,
        start_iteration: int = 0,
        end_iteration: int = -1
    ) -> "Booster":
3975
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
3976

3977
3978
3979
        Parameters
        ----------
        start_iteration : int, optional (default=0)
3980
            The first iteration that will be shuffled.
3981
3982
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
3983
            If <= 0, means the last available iteration.
3984

Nikita Titov's avatar
Nikita Titov committed
3985
3986
3987
3988
        Returns
        -------
        self : Booster
            Booster with shuffled models.
3989
        """
3990
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
3991
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
3992
3993
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
3994
        return self
3995

3996
    def model_from_string(self, model_str: str) -> "Booster":
3997
3998
3999
4000
        """Load Booster from a string.

        Parameters
        ----------
4001
        model_str : str
4002
4003
4004
4005
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4006
        self : Booster
4007
4008
            Loaded Booster object.
        """
4009
4010
4011
        # ensure that existing Booster is freed before replacing it
        # with a new one createdfrom file
        _safe_call(_LIB.LGBM_BoosterFree(self._handle))
4012
        self._free_buffer()
4013
        self._handle = ctypes.c_void_p()
4014
4015
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
4016
            _c_str(model_str),
4017
            ctypes.byref(out_num_iterations),
4018
            ctypes.byref(self._handle)))
4019
4020
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
4021
            self._handle,
4022
4023
            ctypes.byref(out_num_class)))
        self.__num_class = out_num_class.value
4024
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
4025
4026
        return self

4027
4028
4029
4030
4031
4032
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> str:
4033
        """Save Booster to string.
4034

4035
4036
4037
4038
4039
4040
        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
4041
        start_iteration : int, optional (default=0)
4042
            Start index of the iteration that should be saved.
4043
        importance_type : str, optional (default="split")
4044
4045
4046
            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.
4047
4048
4049

        Returns
        -------
4050
        str_repr : str
4051
4052
            String representation of Booster.
        """
4053
        if num_iteration is None:
4054
            num_iteration = self.best_iteration
4055
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4056
        buffer_len = 1 << 20
4057
        tmp_out_len = ctypes.c_int64(0)
4058
4059
4060
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
4061
            self._handle,
4062
            ctypes.c_int(start_iteration),
4063
            ctypes.c_int(num_iteration),
4064
            ctypes.c_int(importance_type_int),
4065
            ctypes.c_int64(buffer_len),
4066
4067
4068
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
4069
        # if buffer length is not long enough, re-allocate a buffer
4070
4071
4072
4073
        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(
4074
                self._handle,
4075
                ctypes.c_int(start_iteration),
4076
                ctypes.c_int(num_iteration),
4077
                ctypes.c_int(importance_type_int),
4078
                ctypes.c_int64(actual_len),
4079
4080
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
4081
        ret = string_buffer.value.decode('utf-8')
4082
4083
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
4084

4085
4086
4087
4088
4089
4090
4091
    def dump_model(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split',
        object_hook: Optional[Callable[[Dict[str, Any]], Dict[str, Any]]] = None
    ) -> Dict[str, Any]:
Nikita Titov's avatar
Nikita Titov committed
4092
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
4093

4094
4095
        Parameters
        ----------
4096
4097
4098
4099
        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
4100
        start_iteration : int, optional (default=0)
4101
            Start index of the iteration that should be dumped.
4102
        importance_type : str, optional (default="split")
4103
4104
4105
            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.
4106
4107
4108
4109
4110
4111
4112
4113
4114
        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.
4115

wxchan's avatar
wxchan committed
4116
4117
        Returns
        -------
4118
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
4119
            JSON format of Booster.
wxchan's avatar
wxchan committed
4120
        """
4121
        if num_iteration is None:
4122
            num_iteration = self.best_iteration
4123
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4124
        buffer_len = 1 << 20
4125
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
4126
4127
4128
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterDumpModel(
4129
            self._handle,
4130
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4131
            ctypes.c_int(num_iteration),
4132
            ctypes.c_int(importance_type_int),
4133
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
4134
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4135
            ptr_string_buffer))
wxchan's avatar
wxchan committed
4136
        actual_len = tmp_out_len.value
4137
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
4138
4139
4140
4141
        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(
4142
                self._handle,
4143
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4144
                ctypes.c_int(num_iteration),
4145
                ctypes.c_int(importance_type_int),
4146
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
4147
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4148
                ptr_string_buffer))
4149
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
4150
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
4151
                                                          default=_json_default_with_numpy))
4152
        return ret
wxchan's avatar
wxchan committed
4153

4154
4155
    def predict(
        self,
4156
        data: _LGBM_PredictDataType,
4157
4158
4159
4160
4161
4162
4163
4164
        start_iteration: int = 0,
        num_iteration: Optional[int] = None,
        raw_score: bool = False,
        pred_leaf: bool = False,
        pred_contrib: bool = False,
        data_has_header: bool = False,
        validate_features: bool = False,
        **kwargs: Any
4165
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4166
        """Make a prediction.
wxchan's avatar
wxchan committed
4167
4168
4169

        Parameters
        ----------
4170
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
4171
            Data source for prediction.
4172
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4173
        start_iteration : int, optional (default=0)
4174
            Start index of the iteration to predict.
4175
            If <= 0, starts from the first iteration.
4176
        num_iteration : int or None, optional (default=None)
4177
4178
4179
4180
            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).
4181
4182
4183
4184
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4185
4186
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4187

Nikita Titov's avatar
Nikita Titov committed
4188
4189
4190
4191
4192
4193
4194
            .. 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.
4195

4196
4197
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4198
            Used only if data is str.
4199
4200
4201
        validate_features : bool, optional (default=False)
            If True, ensure that the features used to predict match the ones used to train.
            Used only if data is pandas DataFrame.
4202
4203
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4204
4205
4206

        Returns
        -------
4207
        result : numpy array, scipy.sparse or list of scipy.sparse
4208
            Prediction result.
4209
            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
4210
        """
4211
4212
4213
4214
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs),
        )
4215
        if num_iteration is None:
4216
            if start_iteration <= 0:
4217
4218
4219
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
        return predictor.predict(
            data=data,
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            raw_score=raw_score,
            pred_leaf=pred_leaf,
            pred_contrib=pred_contrib,
            data_has_header=data_has_header,
            validate_features=validate_features
        )
wxchan's avatar
wxchan committed
4230

4231
4232
    def refit(
        self,
4233
        data: _LGBM_TrainDataType,
4234
        label: _LGBM_LabelType,
4235
4236
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4237
4238
4239
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4240
4241
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
4242
4243
4244
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4245
        **kwargs
4246
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4247
4248
4249
4250
        """Refit the existing Booster by new data.

        Parameters
        ----------
4251
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence or list of numpy array
Guolin Ke's avatar
Guolin Ke committed
4252
            Data source for refit.
4253
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4254
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
4255
4256
            Label for refit.
        decay_rate : float, optional (default=0.9)
4257
4258
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4259
4260
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
4261
4262
4263

            .. versionadded:: 4.0.0

4264
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
4265
            Weight for each ``data`` instance. Weights should be non-negative.
4266
4267
4268

            .. versionadded:: 4.0.0

4269
4270
4271
4272
4273
4274
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
            Group/query size for ``data``.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
            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.
4275
4276
4277

            .. versionadded:: 4.0.0

4278
4279
        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)
            Init score for ``data``.
4280
4281
4282

            .. versionadded:: 4.0.0

4283
4284
4285
        feature_name : list of str, or 'auto', optional (default="auto")
            Feature names for ``data``.
            If 'auto' and data is pandas DataFrame, data columns names are used.
4286
4287
4288

            .. versionadded:: 4.0.0

4289
4290
4291
4292
4293
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
            Categorical features for ``data``.
            If list of int, interpreted as indices.
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
4294
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4295
4296
4297
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
            All negative values in categorical features will be treated as missing values.
            The output cannot be monotonically constrained with respect to a categorical feature.
4298
            Floating point numbers in categorical features will be rounded towards 0.
4299
4300
4301

            .. versionadded:: 4.0.0

4302
4303
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
4304
4305
4306

            .. versionadded:: 4.0.0

4307
4308
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
4309
4310
4311

            .. versionadded:: 4.0.0

4312
4313
4314
        validate_features : bool, optional (default=False)
            If True, ensure that the features used to refit the model match the original ones.
            Used only if data is pandas DataFrame.
4315
4316
4317

            .. versionadded:: 4.0.0

4318
4319
        **kwargs
            Other parameters for refit.
4320
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4321
4322
4323
4324
4325
4326

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4327
4328
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
4329
4330
        if dataset_params is None:
            dataset_params = {}
4331
4332
4333
4334
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs)
        )
4335
        leaf_preds: np.ndarray = predictor.predict(  # type: ignore[assignment]
4336
4337
4338
4339
4340
            data=data,
            start_iteration=-1,
            pred_leaf=True,
            validate_features=validate_features
        )
4341
        nrow, ncol = leaf_preds.shape
4342
        out_is_linear = ctypes.c_int(0)
4343
        _safe_call(_LIB.LGBM_BoosterGetLinear(
4344
            self._handle,
4345
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
4346
4347
4348
4349
4350
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
4351
        new_params["linear_tree"] = bool(out_is_linear.value)
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
        new_params.update(dataset_params)
        train_set = Dataset(
            data=data,
            label=label,
            reference=reference,
            weight=weight,
            group=group,
            init_score=init_score,
            feature_name=feature_name,
            categorical_feature=categorical_feature,
            params=new_params,
            free_raw_data=free_raw_data,
        )
4365
        new_params['refit_decay_rate'] = decay_rate
4366
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4367
4368
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
4369
4370
            new_booster._handle,
            predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
4371
        leaf_preds = leaf_preds.reshape(-1)
4372
        ptr_data, _, _ = _c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
4373
        _safe_call(_LIB.LGBM_BoosterRefit(
4374
            new_booster._handle,
Guolin Ke's avatar
Guolin Ke committed
4375
            ptr_data,
4376
4377
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
4378
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4379
4380
        return new_booster

4381
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
        """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.
        """
4396
4397
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
4398
            self._handle,
4399
4400
4401
4402
4403
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

4404
4405
4406
4407
4408
4409
4410
4411
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
    ) -> 'Booster':
        """Set the output of a leaf.

4412
4413
        .. versionadded:: 4.0.0

4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
        Parameters
        ----------
        tree_id : int
            The index of the tree.
        leaf_id : int
            The index of the leaf in the tree.
        value : float
            Value to set as the output of the leaf.

        Returns
        -------
        self : Booster
            Booster with the leaf output set.
        """
        _safe_call(
            _LIB.LGBM_BoosterSetLeafValue(
4430
                self._handle,
4431
4432
4433
4434
4435
4436
4437
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.c_double(value)
            )
        )
        return self

4438
    def num_feature(self) -> int:
4439
4440
4441
4442
4443
4444
4445
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4446
4447
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
4448
            self._handle,
4449
4450
4451
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

4452
    def feature_name(self) -> List[str]:
4453
        """Get names of features.
wxchan's avatar
wxchan committed
4454
4455
4456

        Returns
        -------
4457
        result : list of str
4458
            List with names of features.
wxchan's avatar
wxchan committed
4459
        """
4460
        num_feature = self.num_feature()
4461
        # Get name of features
wxchan's avatar
wxchan committed
4462
        tmp_out_len = ctypes.c_int(0)
4463
4464
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
4465
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
4466
4467
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
4468
            self._handle,
4469
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
4470
            ctypes.byref(tmp_out_len),
4471
            ctypes.c_size_t(reserved_string_buffer_size),
4472
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4473
4474
4475
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
4476
4477
4478
4479
4480
4481
        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(
4482
                self._handle,
4483
4484
4485
4486
4487
                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))
4488
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
4489

4490
4491
4492
4493
4494
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
4495
        """Get feature importances.
4496

4497
4498
        Parameters
        ----------
4499
        importance_type : str, optional (default="split")
4500
4501
4502
            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.
4503
4504
4505
4506
        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).
4507

4508
4509
        Returns
        -------
4510
4511
        result : numpy array
            Array with feature importances.
4512
        """
4513
4514
        if iteration is None:
            iteration = self.best_iteration
4515
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4516
        result = np.empty(self.num_feature(), dtype=np.float64)
4517
        _safe_call(_LIB.LGBM_BoosterFeatureImportance(
4518
            self._handle,
4519
4520
4521
            ctypes.c_int(iteration),
            ctypes.c_int(importance_type_int),
            result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
4522
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
4523
            return result.astype(np.int32)
4524
4525
        else:
            return result
4526

4527
4528
4529
4530
4531
4532
    def get_split_value_histogram(
        self,
        feature: Union[int, str],
        bins: Optional[Union[int, str]] = None,
        xgboost_style: bool = False
    ) -> Union[Tuple[np.ndarray, np.ndarray], np.ndarray, pd_DataFrame]:
4533
4534
4535
4536
        """Get split value histogram for the specified feature.

        Parameters
        ----------
4537
        feature : int or str
4538
4539
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
4540
            If str, interpreted as name.
4541

Nikita Titov's avatar
Nikita Titov committed
4542
4543
4544
            .. warning::

                Categorical features are not supported.
4545

4546
        bins : int, str or None, optional (default=None)
4547
4548
4549
            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.
4550
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
        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.
        """
4565
        def add(root: Dict[str, Any]) -> None:
4566
4567
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
4568
                if feature_names is not None and isinstance(feature, str):
4569
4570
4571
4572
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
4573
                    if isinstance(root['threshold'], str):
4574
4575
4576
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
4577
4578
4579
4580
4581
4582
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
4583
        values: List[float] = []
4584
4585
4586
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

4587
        if bins is None or isinstance(bins, int) and xgboost_style:
4588
4589
4590
4591
4592
4593
4594
            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:
4595
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
4596
4597
4598
4599
4600
            else:
                return ret
        else:
            return hist, bin_edges

4601
4602
4603
4604
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
4605
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]]
4606
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4607
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
4608
        if data_idx >= self.__num_dataset:
4609
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4610
4611
4612
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
4613
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
4614
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
4615
            _safe_call(_LIB.LGBM_BoosterGetEval(
4616
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4617
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4618
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4619
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
4620
            if tmp_out_len.value != self.__num_inner_eval:
4621
                raise ValueError("Wrong length of eval results")
4622
            for i in range(self.__num_inner_eval):
4623
4624
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
4625
4626
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
4627
4628
4629
4630
4631
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
4632
4633
4634
4635
4636
4637
4638
4639
4640
            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
4641
4642
4643
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

4644
    def __inner_predict(self, data_idx: int) -> np.ndarray:
4645
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
4646
        if data_idx >= self.__num_dataset:
4647
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4648
4649
4650
4651
4652
        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
4653
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
4654
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
4655
4656
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
4657
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))  # type: ignore[union-attr]
wxchan's avatar
wxchan committed
4658
            _safe_call(_LIB.LGBM_BoosterGetPredict(
4659
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4660
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4661
4662
                ctypes.byref(tmp_out_len),
                data_ptr))
4663
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):  # type: ignore[arg-type]
4664
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
4665
            self.__is_predicted_cur_iter[data_idx] = True
4666
        result: np.ndarray = self.__inner_predict_buffer[data_idx]  # type: ignore[assignment]
4667
4668
4669
4670
        if self.__num_class > 1:
            num_data = result.size // self.__num_class
            result = result.reshape(num_data, self.__num_class, order='F')
        return result
wxchan's avatar
wxchan committed
4671

4672
    def __get_eval_info(self) -> None:
4673
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
4674
4675
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
4676
            out_num_eval = ctypes.c_int(0)
4677
            # Get num of inner evals
wxchan's avatar
wxchan committed
4678
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
4679
                self._handle,
wxchan's avatar
wxchan committed
4680
4681
4682
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
4683
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
4684
                tmp_out_len = ctypes.c_int(0)
4685
4686
4687
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
4688
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
4689
                ]
wxchan's avatar
wxchan committed
4690
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
4691
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
4692
                    self._handle,
4693
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
4694
                    ctypes.byref(tmp_out_len),
4695
                    ctypes.c_size_t(reserved_string_buffer_size),
4696
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4697
4698
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
4699
                    raise ValueError("Length of eval names doesn't equal with num_evals")
4700
4701
4702
4703
4704
4705
4706
4707
                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(
4708
                        self._handle,
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
                        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
                ]