"git@developer.sourcefind.cn:tianlh/lightgbm-dcu.git" did not exist on "db4374e19bff2e655cfb795c960e7c43da760e08"
basic.py 181 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3
import abc
wxchan's avatar
wxchan committed
4
import ctypes
5
import json
wxchan's avatar
wxchan committed
6
import warnings
7
from collections import OrderedDict
8
from copy import deepcopy
9
from enum import Enum
10
from functools import wraps
11
from os import SEEK_END, environ
12
13
from os.path import getsize
from pathlib import Path
14
from tempfile import NamedTemporaryFile
15
from typing import Any, Callable, Dict, Iterable, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
16
17
18
19

import numpy as np
import scipy.sparse

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

23
24
25
26
27
28
29
30
31
__all__ = [
    'Booster',
    'Dataset',
    'LGBMDeprecationWarning',
    'LightGBMError',
    'register_logger',
    'Sequence',
]

32
_DatasetHandle = ctypes.c_void_p
33
34
35
36
37
38
39
40
41
42
43
44
_ctypes_int_ptr = Union[
    "ctypes._Pointer[ctypes.c_int32]",
    "ctypes._Pointer[ctypes.c_int64]"
]
_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]]"
]
45
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
46
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
47
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
48
49
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], str]
_LGBM_FeatureNameConfiguration = Union[List[str], str]
50
51
52
53
54
55
56
57
58
59
60
61
62
_LGBM_GroupType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
_LGBM_InitScoreType = Union[
    List[float],
    List[List[float]],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
]
63
64
65
66
67
68
69
70
71
72
73
_LGBM_TrainDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix,
    "Sequence",
    List["Sequence"],
    List[np.ndarray]
]
74
75
76
77
78
79
_LGBM_LabelType = Union[
    list,
    np.ndarray,
    pd_Series,
    pd_DataFrame
]
80
81
82
83
84
85
86
87
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix
]
88
89
90
91
92
93
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
94
95
96
ZERO_THRESHOLD = 1e-35


97
98
99
100
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


101
def _get_sample_count(total_nrow: int, params: str) -> int:
102
103
104
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
105
        _c_str(params),
106
107
108
109
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
110

111
112
113
114
115
116
class _MissingType(Enum):
    NONE = 'None'
    NAN = 'NaN'
    ZERO = 'Zero'


117
class _DummyLogger:
118
    def info(self, msg: str) -> None:
119
120
        print(msg)

121
    def warning(self, msg: str) -> None:
122
123
124
        warnings.warn(msg, stacklevel=3)


125
126
127
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
128
129


130
131
132
133
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


134
135
136
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
137
138
139
140
    """Register custom logger.

    Parameters
    ----------
141
    logger : Any
142
        Custom logger.
143
144
145
146
    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.
147
    """
148
149
150
151
152
153
    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
154
    _LOGGER = logger
155
156
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
157
158


159
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
160
    """Join log messages from native library which come by chunks."""
161
    msg_normalized: List[str] = []
162
163

    @wraps(func)
164
    def wrapper(msg: str) -> None:
165
166
167
168
169
170
171
172
173
174
175
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


176
def _log_info(msg: str) -> None:
177
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
178
179


180
def _log_warning(msg: str) -> None:
181
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
182
183
184


@_normalize_native_string
185
def _log_native(msg: str) -> None:
186
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
187
188


189
def _log_callback(msg: bytes) -> None:
190
    """Redirect logs from native library into Python."""
191
    _log_native(str(msg.decode('utf-8')))
192
193


194
def _load_lib() -> ctypes.CDLL:
195
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
196
197
198
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
199
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
200
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
201
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
202
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
203
204
    return lib

wxchan's avatar
wxchan committed
205

206
207
208
209
210
211
212
# 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
213

wxchan's avatar
wxchan committed
214

215
_NUMERIC_TYPES = (int, float, bool)
216
_ArrayLike = Union[List, np.ndarray, pd_Series]
217
218


219
def _safe_call(ret: int) -> None:
220
221
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
222
223
224
    Parameters
    ----------
    ret : int
225
        The return value from C API calls.
wxchan's avatar
wxchan committed
226
227
    """
    if ret != 0:
228
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
229

wxchan's avatar
wxchan committed
230

231
def _is_numeric(obj: Any) -> bool:
232
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
233
234
235
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
236
237
238
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
239
240
        return False

wxchan's avatar
wxchan committed
241

242
def _is_numpy_1d_array(data: Any) -> bool:
243
    """Check whether data is a numpy 1-D array."""
244
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
245

wxchan's avatar
wxchan committed
246

247
def _is_numpy_column_array(data: Any) -> bool:
248
249
250
251
252
253
254
    """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


255
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
256
    """Cast numpy array to given dtype."""
257
258
259
260
261
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


262
def _is_1d_list(data: Any) -> bool:
263
    """Check whether data is a 1-D list."""
264
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
265

wxchan's avatar
wxchan committed
266

267
268
269
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
270
        _is_numpy_1d_array(data)
271
        or _is_numpy_column_array(data)
272
        or _is_1d_list(data)
273
274
275
276
        or isinstance(data, pd_Series)
    )


277
278
def _list_to_1d_numpy(
    data: Any,
279
    dtype: "np.typing.DTypeLike" = np.float32,
280
281
    name: str = 'list'
) -> np.ndarray:
282
    """Convert data to numpy 1-D array."""
283
    if _is_numpy_1d_array(data):
284
        return _cast_numpy_array_to_dtype(data, dtype)
285
    elif _is_numpy_column_array(data):
286
287
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
288
        return _cast_numpy_array_to_dtype(array, dtype)
289
    elif _is_1d_list(data):
wxchan's avatar
wxchan committed
290
        return np.array(data, dtype=dtype, copy=False)
291
    elif isinstance(data, pd_Series):
292
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
293
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
294
    else:
295
296
        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
297

wxchan's avatar
wxchan committed
298

299
300
301
302
303
304
305
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."""
306
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
307
308
309
310
311
312
313
314
315
316
317


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)
    )


318
319
320
321
322
def _data_to_2d_numpy(
    data: Any,
    dtype: "np.typing.DTypeLike" = np.float32,
    name: str = 'list'
) -> np.ndarray:
323
324
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
325
        return _cast_numpy_array_to_dtype(data, dtype)
326
327
328
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
329
        _check_for_bad_pandas_dtypes(data.dtypes)
330
        return _cast_numpy_array_to_dtype(data.values, dtype)
331
332
333
334
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


335
def _cfloat32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
336
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
337
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
338
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
339
    else:
340
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
341

Guolin Ke's avatar
Guolin Ke committed
342

343
def _cfloat64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
344
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
345
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
346
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
347
348
349
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
350

351
def _cint32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
352
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
353
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
354
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
355
    else:
356
357
358
        raise RuntimeError('Expected int32 pointer')


359
def _cint64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
360
361
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
362
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
363
364
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
365

wxchan's avatar
wxchan committed
366

367
def _c_str(string: str) -> ctypes.c_char_p:
368
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
369
370
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
371

372
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
373
    """Convert a Python array to C array."""
374
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
375

wxchan's avatar
wxchan committed
376

377
def _json_default_with_numpy(obj: Any) -> Any:
378
379
380
381
382
383
384
385
386
    """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


387
388
389
390
391
392
393
394
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)


395
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
396
    """Convert Python dictionary to string, which is passed to C API."""
397
    if data is None or not data:
wxchan's avatar
wxchan committed
398
399
400
        return ""
    pairs = []
    for key, val in data.items():
401
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
402
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
403
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
404
            pairs.append(f"{key}={val}")
405
        elif val is not None:
406
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
407
    return ' '.join(pairs)
408

wxchan's avatar
wxchan committed
409

410
class _TempFile:
411
412
    """Proxy class to workaround errors on Windows."""

413
414
415
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
416
            self.path = Path(self.name)
417
        return self
wxchan's avatar
wxchan committed
418

419
    def __exit__(self, exc_type, exc_val, exc_tb):
420
421
        if self.path.is_file():
            self.path.unlink()
422

wxchan's avatar
wxchan committed
423

424
class LightGBMError(Exception):
425
426
    """Error thrown by LightGBM."""

427
428
429
    pass


430
431
432
433
434
435
436
437
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
438
439
440
441
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
442
    def _get_all_param_aliases() -> Dict[str, List[str]]:
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
        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))
        aliases = json.loads(
            string_buffer.value.decode('utf-8'),
462
            object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
463
464
        )
        return aliases
465
466

    @classmethod
467
468
469
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
470
471
        ret = set()
        for i in args:
472
            ret.update(cls.get_sorted(i))
473
474
        return ret

475
476
477
478
479
480
    @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])

481
    @classmethod
482
483
484
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
485
486
487
488
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
489
                    ret.update(aliases)
490
491
492
                    break
        return ret

493

494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
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)

515
516
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
517
518

    # if main_param_name was provided, keep that value and remove all aliases
519
    if main_param_name in params.keys():
520
521
522
        for param in aliases:
            params.pop(param, None)
        return params
523

524
525
526
527
528
    # 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
529

530
531
532
533
534
535
536
    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
537
538
539
540

    return params


541
_MAX_INT32 = (1 << 31) - 1
542

543
"""Macro definition of data type in C API of LightGBM"""
544
545
546
547
_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
548

549
"""Matrix is row major in Python"""
550
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
551

552
"""Macro definition of prediction type in C API of LightGBM"""
553
554
555
556
_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
557

558
"""Macro definition of sparse matrix type"""
559
560
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
561

562
"""Macro definition of feature importance type"""
563
564
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
565

566
"""Data type of data field"""
567
568
569
570
571
572
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
    "group": _C_API_DTYPE_INT32
}
wxchan's avatar
wxchan committed
573

574
"""String name to int feature importance type mapper"""
575
576
577
578
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN
}
579

wxchan's avatar
wxchan committed
580

581
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
582
    """Fix the memory of multi-dimensional sliced object."""
583
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
584
        if not data.flags.c_contiguous:
585
586
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
587
588
589
590
            return np.copy(data)
    return data


591
def _c_float_array(data):
592
    """Get pointer of float numpy array / list."""
593
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
594
        data = np.array(data, copy=False)
595
    if _is_numpy_1d_array(data):
596
        data = _convert_from_sliced_object(data)
597
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
598
599
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
600
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
601
602
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
603
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
604
        else:
605
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
606
    else:
607
        raise TypeError(f"Unknown type({type(data).__name__})")
608
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
609

wxchan's avatar
wxchan committed
610

611
def _c_int_array(data):
612
    """Get pointer of int numpy array / list."""
613
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
614
        data = np.array(data, copy=False)
615
    if _is_numpy_1d_array(data):
616
        data = _convert_from_sliced_object(data)
617
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
618
619
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
620
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
621
622
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
623
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
624
        else:
625
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
626
    else:
627
        raise TypeError(f"Unknown type({type(data).__name__})")
628
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
629

wxchan's avatar
wxchan committed
630

631
def _is_allowed_numpy_dtype(dtype: type) -> bool:
632
    float128 = getattr(np, 'float128', type(None))
633
634
635
636
    return (
        issubclass(dtype, (np.integer, np.floating, np.bool_))
        and not issubclass(dtype, (np.timedelta64, float128))
    )
637
638


639
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
640
641
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
642
        for column_name, pandas_dtype in pandas_dtypes_series.items()
643
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
644
645
646
647
    ]
    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)}')
648
649


650
651
652
653
654
655
def _data_from_pandas(
    data,
    feature_name: Optional[_LGBM_FeatureNameConfiguration],
    categorical_feature: Optional[_LGBM_CategoricalFeatureConfiguration],
    pandas_categorical: Optional[List[List]]
):
656
    if isinstance(data, pd_DataFrame):
657
658
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
659
        if feature_name == 'auto' or feature_name is None:
660
            data = data.rename(columns=str, copy=False)
661
        cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
662
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
663
664
665
666
667
        if pandas_categorical is None:  # train dataset
            pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
        else:
            if len(cat_cols) != len(pandas_categorical):
                raise ValueError('train and valid dataset categorical_feature do not match.')
668
            for col, category in zip(cat_cols, pandas_categorical):
669
670
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
671
        if len(cat_cols):  # cat_cols is list
672
            data = data.copy(deep=False)  # not alter origin DataFrame
673
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
674
675
676
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
677
            if categorical_feature == 'auto':  # use cat cols from DataFrame
678
                categorical_feature = cat_cols_not_ordered
679
680
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
681
682
        if feature_name == 'auto':
            feature_name = list(data.columns)
683
        _check_for_bad_pandas_dtypes(data.dtypes)
684
685
686
        df_dtypes = [dtype.type for dtype in data.dtypes]
        df_dtypes.append(np.float32)  # so that the target dtype considers floats
        target_dtype = np.find_common_type(df_dtypes, [])
687
688
689
690
691
692
693
694
695
696
        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)
697
698
699
700
701
702
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
703
704


705
706
707
708
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
    file_name: Optional[Union[str, Path]] = None
) -> str:
709
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
710
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
711
712
713
714
715
716
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


717
718
719
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
    model_str: Optional[str] = None
720
) -> Optional[List[List]]:
721
722
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
723
    if file_name is not None:
724
        max_offset = -getsize(file_name)
725
726
727
728
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
729
                f.seek(offset, SEEK_END)
730
731
732
733
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
734
        last_line = lines[-1].decode('utf-8').strip()
735
        if not last_line.startswith(pandas_key):
736
            last_line = lines[-2].decode('utf-8').strip()
737
    elif model_str is not None:
738
739
740
741
742
743
        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
744
745


746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
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**.

766
767
    .. versionadded:: 3.3.0

768
769
770
771
772
773
774
775
776
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
777
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
778
779
780
781
782
783
784
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
785
                return self._get_one_line(idx)
786
            elif isinstance(idx, slice):
787
788
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
789
                # Only required if using ``Dataset.subset()``.
790
                return np.array([self._get_one_line(i) for i in idx])
791
            else:
792
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
793
794
795

        Parameters
        ----------
796
        idx : int, slice[int], list[int]
797
798
799
800
            Item index.

        Returns
        -------
801
        result : numpy 1-D array or numpy 2-D array
802
            1-D array if idx is int, 2-D array if idx is slice or list.
803
804
805
806
807
808
809
810
811
        """
        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__()")


812
class _InnerPredictor:
813
814
815
816
817
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
818
819
820
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
821
    """
822

823
824
825
826
827
828
    def __init__(
        self,
        model_file: Optional[Union[str, Path]] = None,
        booster_handle: Optional[ctypes.c_void_p] = None,
        pred_parameter: Optional[Dict[str, Any]] = None
    ):
829
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
830
831
832

        Parameters
        ----------
833
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
834
            Path to the model file.
835
836
837
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
838
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
839
840
841
842
843
        """
        self.handle = ctypes.c_void_p()
        self.__is_manage_handle = True
        if model_file is not None:
            """Prediction task"""
Guolin Ke's avatar
Guolin Ke committed
844
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
845
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
846
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
847
848
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
849
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
850
851
852
853
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
854
            self.num_total_iteration = out_num_iterations.value
855
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
856
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
857
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
858
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
859
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
860
861
862
863
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
864
            self.num_total_iteration = self.current_iteration()
865
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
866
        else:
867
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
868

869
        pred_parameter = {} if pred_parameter is None else pred_parameter
870
        self.pred_parameter = _param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
871

872
    def __del__(self) -> None:
873
874
875
876
877
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
878

879
    def __getstate__(self) -> Dict[str, Any]:
880
881
882
883
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

884
885
    def predict(
        self,
886
        data: _LGBM_PredictDataType,
887
888
889
890
891
892
893
        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
894
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
895
        """Predict logic.
wxchan's avatar
wxchan committed
896
897
898

        Parameters
        ----------
899
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
900
            Data source for prediction.
901
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
902
903
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
904
905
906
907
908
909
910
911
912
913
914
        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.
915
916
917
        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
918
919
920

        Returns
        -------
921
        result : numpy array, scipy.sparse or list of scipy.sparse
922
            Prediction result.
923
            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
924
        """
wxchan's avatar
wxchan committed
925
        if isinstance(data, Dataset):
926
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
927
928
929
930
931
932
933
934
935
936
937
        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(
                    self.handle,
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
938
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
939
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
940
        if raw_score:
941
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
942
        if pred_leaf:
943
            predict_type = _C_API_PREDICT_LEAF_INDEX
944
        if pred_contrib:
945
            predict_type = _C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
946
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
947

948
        if isinstance(data, (str, Path)):
949
            with _TempFile() as f:
wxchan's avatar
wxchan committed
950
951
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
952
                    _c_str(str(data)),
Guolin Ke's avatar
Guolin Ke committed
953
954
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
955
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
956
                    ctypes.c_int(num_iteration),
957
958
                    _c_str(self.pred_parameter),
                    _c_str(f.name)))
959
960
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
961
        elif isinstance(data, scipy.sparse.csr_matrix):
962
963
964
965
966
967
            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
968
        elif isinstance(data, scipy.sparse.csc_matrix):
969
970
971
972
973
974
            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
975
        elif isinstance(data, np.ndarray):
976
977
978
979
980
981
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
982
983
984
        elif isinstance(data, list):
            try:
                data = np.array(data)
985
            except BaseException:
986
                raise ValueError('Cannot convert data list to numpy array.')
987
988
989
990
991
992
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
993
        elif isinstance(data, dt_DataTable):
994
995
996
997
998
999
            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
1000
1001
        else:
            try:
1002
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
1003
                csr = scipy.sparse.csr_matrix(data)
1004
            except BaseException:
1005
                raise TypeError(f'Cannot predict data for type {type(data).__name__}')
1006
1007
1008
1009
1010
1011
            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
1012
1013
        if pred_leaf:
            preds = preds.astype(np.int32)
1014
        is_sparse = isinstance(preds, scipy.sparse.spmatrix) or isinstance(preds, list)
1015
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1016
            if preds.size % nrow == 0:
1017
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1018
            else:
1019
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
1020
1021
        return preds

1022
1023
1024
1025
1026
1027
1028
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
        predict_type: int
    ) -> int:
1029
        """Get size of prediction result."""
1030
        if nrow > _MAX_INT32:
1031
            raise LightGBMError('LightGBM cannot perform prediction for data '
1032
                                f'with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n'
1033
                                'You can split your data into chunks '
1034
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
1035
1036
1037
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
1038
1039
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
1040
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1041
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
1042
1043
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
1044

1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
    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)
1058
1059
1060
1061
1062
1063
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
            predict_type=predict_type
        )
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
        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(
            self.handle,
            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]:
1093
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1094
        if len(mat.shape) != 2:
1095
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
1096

1097
        nrow = mat.shape[0]
1098
1099
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1100
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1101
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
1102
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1103
            preds = np.empty(sum(n_preds), dtype=np.float64)
1104
1105
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
1106
                # avoid memory consumption by arrays concatenation operations
1107
1108
1109
1110
1111
1112
1113
                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]
                )
1114
            return preds, nrow
wxchan's avatar
wxchan committed
1115
        else:
1116
1117
1118
1119
1120
1121
1122
            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
1123

1124
1125
1126
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1127
1128
1129
1130
1131
1132
        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,
1133
        is_csr: bool
1134
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1135
1136
1137
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1138
        if indptr_type == _C_API_DTYPE_INT32:
1139
            out_indptr = _cint32_array_to_numpy(out_ptr_indptr, indptr_len)
1140
        elif indptr_type == _C_API_DTYPE_INT64:
1141
            out_indptr = _cint64_array_to_numpy(out_ptr_indptr, indptr_len)
1142
1143
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1144
        if data_type == _C_API_DTYPE_FLOAT32:
1145
            out_data = _cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
1146
        elif data_type == _C_API_DTYPE_FLOAT64:
1147
            out_data = _cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
1148
1149
        else:
            raise TypeError("Expected float32 or float64 type for data")
1150
        out_indices = _cint32_array_to_numpy(out_ptr_indices, data_indices_len)
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
        # 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

1179
1180
1181
1182
1183
1184
1185
1186
1187
    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
1188
1189
1190
1191
1192
1193
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1194
1195
1196
1197
1198
        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
1199

1200
1201
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1202

1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

        _safe_call(_LIB.LGBM_BoosterPredictForCSR(
            self.handle,
            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
1232
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1233
1234
1235
1236
        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
1237
        out_ptr_indptr: _ctypes_int_ptr
1238
1239
1240
1241
1242
        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)()
1243
        out_ptr_data: _ctypes_float_ptr
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
        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(
            self.handle,
            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

1281
1282
1283
1284
1285
1286
1287
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1288
        """Predict for a CSR data."""
1289
        if predict_type == _C_API_PREDICT_CONTRIB:
1290
1291
1292
1293
1294
1295
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1296
        nrow = len(csr.indptr) - 1
1297
1298
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1299
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1300
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1301
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1302
            preds = np.empty(sum(n_preds), dtype=np.float64)
1303
1304
            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:])):
1305
                # avoid memory consumption by arrays concatenation operations
1306
1307
1308
1309
1310
1311
1312
                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]
                )
1313
1314
            return preds, nrow
        else:
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
            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,
1325
1326
1327
1328
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
1329
1330
1331
1332
1333
    ):
        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
1334
        out_ptr_indptr: _ctypes_int_ptr
1335
1336
1337
1338
1339
        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)()
1340
        out_ptr_data: _ctypes_float_ptr
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
        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(
            self.handle,
            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
1377

1378
1379
1380
1381
1382
1383
1384
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1385
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1386
        nrow = csc.shape[0]
1387
        if nrow > _MAX_INT32:
1388
1389
1390
1391
1392
1393
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1394
        if predict_type == _C_API_PREDICT_CONTRIB:
1395
1396
1397
1398
1399
1400
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1401
1402
1403
1404
1405
1406
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1407
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1408
1409
        out_num_preds = ctypes.c_int64(0)

1410
1411
        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
1412

1413
        assert csc.shape[0] <= _MAX_INT32
1414
        csc_indices = csc.indices.astype(np.int32, copy=False)
1415

Guolin Ke's avatar
Guolin Ke committed
1416
1417
1418
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
1419
            ctypes.c_int(type_ptr_indptr),
1420
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1421
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1422
1423
1424
1425
1426
            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),
1427
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1428
            ctypes.c_int(num_iteration),
1429
            _c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1430
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1431
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1432
        if n_preds != out_num_preds.value:
1433
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1434
1435
        return preds, nrow

1436
    def current_iteration(self) -> int:
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
        """Get the index of the current iteration.

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

wxchan's avatar
wxchan committed
1450

1451
class Dataset:
wxchan's avatar
wxchan committed
1452
    """Dataset in LightGBM."""
1453

1454
1455
    def __init__(
        self,
1456
        data: _LGBM_TrainDataType,
1457
        label: Optional[_LGBM_LabelType] = None,
1458
        reference: Optional["Dataset"] = None,
1459
1460
1461
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1462
1463
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
1464
1465
1466
        params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True
    ):
1467
        """Initialize Dataset.
1468

wxchan's avatar
wxchan committed
1469
1470
        Parameters
        ----------
1471
        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
1472
            Data source of Dataset.
1473
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1474
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1475
1476
1477
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1478
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1479
            Weight for each instance. Weights should be non-negative.
1480
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1481
1482
1483
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1484
1485
            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.
1486
        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)
1487
            Init score for Dataset.
1488
        feature_name : list of str, or 'auto', optional (default="auto")
1489
1490
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
1491
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1492
1493
            Categorical features.
            If list of int, interpreted as indices.
1494
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1495
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1496
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1497
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1498
            All negative values in categorical features will be treated as missing values.
1499
            The output cannot be monotonically constrained with respect to a categorical feature.
1500
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1501
        params : dict or None, optional (default=None)
1502
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1503
        free_raw_data : bool, optional (default=True)
1504
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
1505
        """
1506
        self.handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1507
1508
1509
1510
1511
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1512
        self.init_score = init_score
1513
1514
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1515
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1516
        self.free_raw_data = free_raw_data
1517
        self.used_indices: Optional[List[int]] = None
1518
        self._need_slice = True
1519
        self._predictor: Optional[_InnerPredictor] = None
1520
        self.pandas_categorical = None
1521
        self._params_back_up = None
1522
        self.version = 0
1523
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1524

1525
    def __del__(self) -> None:
1526
1527
1528
1529
        try:
            self._free_handle()
        except AttributeError:
            pass
1530

1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
    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.
        """
1548
        param_str = _param_dict_to_str(self.get_params())
1549
1550
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1551
        ptr_data, _, _ = _c_int_array(indices)
1552
1553
1554
1555
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
1556
            _c_str(param_str),
1557
1558
1559
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1560
1561
        assert sample_cnt == actual_sample_cnt.value
        return indices
1562

1563
1564
1565
1566
1567
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
        ref_dataset: _DatasetHandle
    ) -> 'Dataset':
1568
1569
1570
1571
1572
1573
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1574
1575
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600

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

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

        Parameters
        ----------
1601
        sample_data : list of numpy array
1602
            Sample data for each column.
1603
        sample_indices : list of numpy array
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
            Sample data row index for each column.
        sample_cnt : int
            Number of samples.
        total_nrow : int
            Total number of rows for all input files.

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

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

        # c type: double**
        # each double* element points to start of each column of sample data.
        sample_col_ptr = (ctypes.POINTER(ctypes.c_double) * ncol)()
        # c type int**
        # each int* points to start of indices for each column
        indices_col_ptr = (ctypes.POINTER(ctypes.c_int32) * ncol)()
        for i in range(ncol):
1631
1632
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1633
1634

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1635
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1636
1637

        self.handle = ctypes.c_void_p()
1638
        params_str = _param_dict_to_str(self.get_params())
1639
1640
1641
1642
1643
1644
1645
        _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),
1646
            ctypes.c_int64(total_nrow),
1647
            _c_str(params_str),
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
            ctypes.byref(self.handle),
        ))
        return self

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

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

        Returns
        -------
        self : Dataset
            Dataset object.
        """
        nrow, ncol = data.shape
        data = data.reshape(data.size)
1667
        data_ptr, data_type, _ = _c_float_array(data)
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679

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

1680
    def get_params(self) -> Dict[str, Any]:
1681
1682
1683
1684
        """Get the used parameters in the Dataset.

        Returns
        -------
1685
        params : dict
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
            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",
1701
                                                "linear_tree",
1702
1703
1704
1705
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1706
                                                "precise_float_parser",
1707
1708
1709
1710
1711
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1712
1713
        else:
            return {}
1714

1715
    def _free_handle(self) -> "Dataset":
1716
        if self.handle is not None:
1717
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1718
            self.handle = None
1719
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
1720
1721
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1722
        return self
wxchan's avatar
wxchan committed
1723

1724
1725
1726
1727
1728
1729
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
        data,
        used_indices: Optional[List[int]]
    ):
Guolin Ke's avatar
Guolin Ke committed
1730
        data_has_header = False
1731
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
1732
            # check data has header or not
1733
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1734
        num_data = self.num_data()
1735
1736
1737
        if predictor is not None:
            init_score = predictor.predict(data,
                                           raw_score=True,
1738
1739
                                           data_has_header=data_has_header)
            init_score = init_score.ravel()
1740
            if used_indices is not None:
1741
                assert not self._need_slice
1742
                if isinstance(data, (str, Path)):
1743
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1744
                    assert num_data == len(used_indices)
1745
1746
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1747
1748
1749
1750
                            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
1751
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1752
1753
                for i in range(num_data):
                    for j in range(predictor.num_class):
1754
1755
1756
                        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:
1757
            init_score = np.zeros(self.init_score.shape, dtype=np.float64)
1758
1759
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1760
1761
        self.set_init_score(init_score)

1762
1763
    def _lazy_init(
        self,
1764
        data: Optional[_LGBM_TrainDataType],
1765
        label: Optional[_LGBM_LabelType] = None,
1766
        reference: Optional["Dataset"] = None,
1767
1768
1769
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1770
1771
1772
1773
1774
        predictor=None,
        feature_name='auto',
        categorical_feature='auto',
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
wxchan's avatar
wxchan committed
1775
1776
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1777
            return self
Guolin Ke's avatar
Guolin Ke committed
1778
1779
1780
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1781
1782
1783
1784
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
Guolin Ke's avatar
Guolin Ke committed
1785

1786
        # process for args
wxchan's avatar
wxchan committed
1787
        params = {} if params is None else params
1788
1789
1790
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1791
        for key in params.keys():
1792
            if key in args_names:
1793
1794
                _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.')
1795
        # get categorical features
1796
1797
1798
1799
1800
1801
        if categorical_feature is not None:
            categorical_indices = set()
            feature_dict = {}
            if feature_name is not None:
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
1802
                if isinstance(name, str) and name in feature_dict:
1803
                    categorical_indices.add(feature_dict[name])
1804
                elif isinstance(name, int):
1805
1806
                    categorical_indices.add(name)
                else:
1807
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1808
            if categorical_indices:
1809
1810
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1811
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
1812
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
1813
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1814
                        params.pop(cat_alias, None)
1815
                params['categorical_column'] = sorted(categorical_indices)
1816

1817
        params_str = _param_dict_to_str(params)
1818
        self.params = params
1819
        # process for reference dataset
wxchan's avatar
wxchan committed
1820
        ref_dataset = None
wxchan's avatar
wxchan committed
1821
        if isinstance(reference, Dataset):
1822
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1823
1824
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1825
        # start construct data
1826
        if isinstance(data, (str, Path)):
wxchan's avatar
wxchan committed
1827
1828
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1829
1830
                _c_str(str(data)),
                _c_str(params_str),
wxchan's avatar
wxchan committed
1831
1832
1833
1834
                ref_dataset,
                ctypes.byref(self.handle)))
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
1835
1836
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1837
1838
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1839
1840
1841
1842
1843
1844
1845
1846
1847
        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)
1848
        elif isinstance(data, dt_DataTable):
1849
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1850
1851
1852
1853
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1854
            except BaseException:
1855
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}')
wxchan's avatar
wxchan committed
1856
1857
1858
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1859
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1860
1861
1862
1863
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1864
1865
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1866
                _log_warning("The init_score will be overridden by the prediction of init_model.")
1867
1868
1869
1870
1871
            self._set_init_score_by_predictor(
                predictor=predictor,
                data=data,
                used_indices=None
            )
1872
1873
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1874
        elif predictor is not None:
1875
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1876
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1877
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1878

1879
1880
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
        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.
1906
        sampled = np.array([row for row in self._yield_row_from_seqlist(seqs, indices)])
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
        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

1922
1923
1924
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
1925
        ref_dataset: Optional[_DatasetHandle]
1926
    ) -> "Dataset":
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
        """
        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:
1941
            param_str = _param_dict_to_str(self.get_params())
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
            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

1955
1956
1957
1958
1959
1960
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1961
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1962
1963
1964
1965
1966
1967
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

        self.handle = ctypes.c_void_p()
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
1968
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1969
1970
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1971
        ptr_data, type_ptr_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
1972
1973
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1974
            ctypes.c_int(type_ptr_data),
1975
1976
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
1977
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
1978
            _c_str(params_str),
wxchan's avatar
wxchan committed
1979
1980
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1981
        return self
wxchan's avatar
wxchan committed
1982

1983
1984
1985
1986
1987
1988
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1989
        """Initialize data from a list of 2-D numpy matrices."""
1990
        ncol = mats[0].shape[1]
1991
        nrow = np.empty((len(mats),), np.int32)
1992
        ptr_data: _ctypes_float_array
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
        if mats[0].dtype == np.float64:
            ptr_data = (ctypes.POINTER(ctypes.c_double) * len(mats))()
        else:
            ptr_data = (ctypes.POINTER(ctypes.c_float) * len(mats))()

        holders = []
        type_ptr_data = None

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

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

            nrow[i] = mat.shape[0]

            if mat.dtype == np.float32 or mat.dtype == np.float64:
                mats[i] = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
2012
            else:  # change non-float data to float data, need to copy
2013
2014
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2015
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2016
2017
2018
2019
2020
2021
2022
2023
            if type_ptr_data is not None and chunk_type_ptr_data != type_ptr_data:
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

        self.handle = ctypes.c_void_p()
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
2024
            ctypes.c_int32(len(mats)),
2025
2026
2027
            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)),
2028
            ctypes.c_int32(ncol),
2029
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2030
            _c_str(params_str),
2031
2032
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
2033
        return self
2034

2035
2036
2037
2038
2039
2040
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2041
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2042
        if len(csr.indices) != len(csr.data):
2043
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
wxchan's avatar
wxchan committed
2044
2045
        self.handle = ctypes.c_void_p()

2046
2047
        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
2048

2049
        assert csr.shape[1] <= _MAX_INT32
2050
        csr_indices = csr.indices.astype(np.int32, copy=False)
2051

wxchan's avatar
wxchan committed
2052
2053
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2054
            ctypes.c_int(type_ptr_indptr),
2055
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
2056
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2057
2058
2059
2060
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
2061
            _c_str(params_str),
wxchan's avatar
wxchan committed
2062
2063
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
2064
        return self
wxchan's avatar
wxchan committed
2065

2066
2067
2068
2069
2070
2071
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2072
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2073
        if len(csc.indices) != len(csc.data):
2074
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
Guolin Ke's avatar
Guolin Ke committed
2075
2076
        self.handle = ctypes.c_void_p()

2077
2078
        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
2079

2080
        assert csc.shape[0] <= _MAX_INT32
2081
        csc_indices = csc.indices.astype(np.int32, copy=False)
2082

Guolin Ke's avatar
Guolin Ke committed
2083
2084
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2085
            ctypes.c_int(type_ptr_indptr),
2086
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
2087
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2088
2089
2090
2091
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
2092
            _c_str(params_str),
Guolin Ke's avatar
Guolin Ke committed
2093
2094
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
2095
        return self
Guolin Ke's avatar
Guolin Ke committed
2096

2097
    @staticmethod
2098
2099
2100
2101
2102
2103
    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.
2104

2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
        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.
2115
2116
2117

        Returns
        -------
2118
2119
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
        """
        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

2135
    def construct(self) -> "Dataset":
2136
2137
2138
2139
2140
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2141
            Constructed Dataset object.
2142
        """
2143
        if self.handle is None:
wxchan's avatar
wxchan committed
2144
            if self.reference is not None:
2145
                reference_params = self.reference.get_params()
2146
2147
                params = self.get_params()
                if params != reference_params:
2148
2149
2150
2151
2152
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
2153
                        _log_warning('Overriding the parameters from Reference Dataset.')
2154
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2155
                if self.used_indices is None:
2156
                    # create valid
2157
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
2158
2159
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
2160
                                    feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
2161
                else:
2162
                    # construct subset
2163
                    used_indices = _list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
2164
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2165
                    if self.reference.group is not None:
2166
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2167
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
2168
                                                  return_counts=True)
2169
                    self.handle = ctypes.c_void_p()
2170
                    params_str = _param_dict_to_str(self.params)
wxchan's avatar
wxchan committed
2171
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
2172
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
2173
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
2174
                        ctypes.c_int32(used_indices.shape[0]),
2175
                        _c_str(params_str),
wxchan's avatar
wxchan committed
2176
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2177
2178
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2179
2180
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
2181
2182
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
2183
2184
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
2185
2186
2187
2188
2189
                        self._set_init_score_by_predictor(
                            predictor=self._predictor,
                            data=self.data,
                            used_indices=used_indices
                        )
wxchan's avatar
wxchan committed
2190
            else:
2191
                # create train
2192
                self._lazy_init(self.data, label=self.label,
2193
2194
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
2195
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
2196
2197
            if self.free_raw_data:
                self.data = None
2198
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2199
        return self
wxchan's avatar
wxchan committed
2200

2201
2202
    def create_valid(
        self,
2203
        data: _LGBM_TrainDataType,
2204
        label: Optional[_LGBM_LabelType] = None,
2205
2206
2207
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2208
2209
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2210
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2211
2212
2213

        Parameters
        ----------
2214
        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
2215
            Data source of Dataset.
2216
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2217
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
2218
2219
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
2220
            Weight for each instance. Weights should be non-negative.
2221
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
2222
2223
2224
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2225
2226
            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.
2227
        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)
2228
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2229
        params : dict or None, optional (default=None)
2230
            Other parameters for validation Dataset.
2231
2232
2233

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2234
2235
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2236
        """
2237
        ret = Dataset(data, label=label, reference=self,
2238
                      weight=weight, group=group, init_score=init_score,
2239
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2240
        ret._predictor = self._predictor
2241
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2242
        return ret
wxchan's avatar
wxchan committed
2243

2244
2245
2246
2247
2248
    def subset(
        self,
        used_indices: List[int],
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2249
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2250
2251
2252
2253

        Parameters
        ----------
        used_indices : list of int
2254
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2255
        params : dict or None, optional (default=None)
2256
            These parameters will be passed to Dataset constructor.
2257
2258
2259
2260
2261

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2262
        """
wxchan's avatar
wxchan committed
2263
2264
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
2265
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
2266
2267
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2268
        ret._predictor = self._predictor
2269
        ret.pandas_categorical = self.pandas_categorical
2270
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2271
2272
        return ret

2273
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2274
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2275

2276
2277
2278
2279
2280
        .. 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
2281
2282
        Parameters
        ----------
2283
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2284
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2285
2286
2287
2288
2289

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2290
2291
2292
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
2293
            _c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
2294
        return self
wxchan's avatar
wxchan committed
2295

2296
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2297
2298
        if not params:
            return self
2299
        params = deepcopy(params)
2300
2301
2302
2303
2304

        def update():
            if not self.params:
                self.params = params
            else:
2305
                self._params_back_up = deepcopy(self.params)
2306
2307
2308
2309
2310
2311
                self.params.update(params)

        if self.handle is None:
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2312
2313
                _c_str(_param_dict_to_str(self.params)),
                _c_str(_param_dict_to_str(params)))
2314
2315
2316
2317
2318
2319
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2320
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
2321
        return self
wxchan's avatar
wxchan committed
2322

2323
    def _reverse_update_params(self) -> "Dataset":
2324
        if self.handle is None:
2325
2326
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2327
        return self
2328

2329
2330
2331
2332
2333
    def set_field(
        self,
        field_name: str,
        data
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2334
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2335
2336
2337

        Parameters
        ----------
2338
        field_name : str
2339
            The field name of the information.
2340
2341
        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
2342
2343
2344
2345
2346

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2347
        """
2348
        if self.handle is None:
2349
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2350
        if data is None:
2351
            # set to None
wxchan's avatar
wxchan committed
2352
2353
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
2354
                _c_str(field_name),
wxchan's avatar
wxchan committed
2355
                None,
Guolin Ke's avatar
Guolin Ke committed
2356
                ctypes.c_int(0),
2357
                ctypes.c_int(_FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2358
            return self
2359
        dtype: "np.typing.DTypeLike"
2360
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2361
            dtype = np.float64
2362
            if _is_1d_collection(data):
2363
                data = _list_to_1d_numpy(data, dtype, name=field_name)
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
            elif _is_2d_collection(data):
                data = _data_to_2d_numpy(data, dtype, name=field_name)
                data = data.ravel(order='F')
            else:
                raise TypeError(
                    'init_score must be list, numpy 1-D array or pandas Series.\n'
                    'In multiclass classification init_score can also be a list of lists, numpy 2-D array or pandas DataFrame.'
                )
        else:
            dtype = np.int32 if field_name == 'group' else np.float32
2374
            data = _list_to_1d_numpy(data, dtype, name=field_name)
2375

2376
        if data.dtype == np.float32 or data.dtype == np.float64:
2377
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2378
        elif data.dtype == np.int32:
2379
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2380
        else:
2381
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2382
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2383
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2384
2385
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
2386
            _c_str(field_name),
wxchan's avatar
wxchan committed
2387
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2388
2389
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2390
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2391
        return self
wxchan's avatar
wxchan committed
2392

2393
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2394
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2395
2396
2397

        Parameters
        ----------
2398
        field_name : str
2399
            The field name of the information.
wxchan's avatar
wxchan committed
2400
2401
2402

        Returns
        -------
2403
        info : numpy array or None
2404
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2405
        """
2406
        if self.handle is None:
2407
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2408
2409
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2410
2411
2412
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
            self.handle,
2413
            _c_str(field_name),
wxchan's avatar
wxchan committed
2414
2415
2416
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
2417
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2418
2419
2420
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2421
        if out_type.value == _C_API_DTYPE_INT32:
2422
            arr = _cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
2423
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2424
            arr = _cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
2425
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2426
            arr = _cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2427
        else:
wxchan's avatar
wxchan committed
2428
            raise TypeError("Unknown type")
2429
2430
2431
2432
2433
2434
        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
2435

2436
2437
    def set_categorical_feature(
        self,
2438
        categorical_feature: _LGBM_CategoricalFeatureConfiguration
2439
    ) -> "Dataset":
2440
        """Set categorical features.
2441
2442
2443

        Parameters
        ----------
2444
        categorical_feature : list of str or int, or 'auto'
2445
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2446
2447
2448
2449
2450

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2451
2452
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2453
            return self
2454
        if self.data is not None:
2455
2456
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2457
                return self._free_handle()
2458
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2459
                return self
2460
            else:
2461
2462
2463
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
                                 f'New categorical_feature is {sorted(list(categorical_feature))}')
2464
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2465
                return self._free_handle()
2466
        else:
2467
2468
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2469

2470
2471
2472
2473
    def _set_predictor(
        self,
        predictor: Optional[_InnerPredictor]
    ) -> "Dataset":
2474
2475
2476
2477
        """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
2478
        """
2479
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2480
            return self
2481
2482
2483
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
            if (predictor == self._predictor) and (predictor.current_iteration() == self._predictor.current_iteration()):
                return self
2484
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
2485
            self._predictor = predictor
2486
2487
        elif self.data is not None:
            self._predictor = predictor
2488
2489
2490
2491
2492
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
                used_indices=None
            )
2493
2494
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2495
2496
2497
2498
2499
            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
2500
        else:
2501
2502
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2503
        return self
Guolin Ke's avatar
Guolin Ke committed
2504

2505
    def set_reference(self, reference: "Dataset") -> "Dataset":
2506
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2507
2508
2509
2510

        Parameters
        ----------
        reference : Dataset
2511
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2512
2513
2514
2515
2516

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2517
        """
2518
2519
2520
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2521
        # we're done if self and reference share a common upstream reference
2522
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2523
            return self
Guolin Ke's avatar
Guolin Ke committed
2524
2525
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2526
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2527
        else:
2528
2529
            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
2530

2531
    def set_feature_name(self, feature_name: Union[List[str], str]) -> "Dataset":
2532
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2533
2534
2535

        Parameters
        ----------
2536
        feature_name : list of str
2537
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2538
2539
2540
2541
2542

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2543
        """
2544
2545
        if feature_name != 'auto':
            self.feature_name = feature_name
2546
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2547
            if len(feature_name) != self.num_feature():
2548
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2549
            c_feature_name = [_c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2550
2551
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
2552
                _c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2553
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2554
        return self
Guolin Ke's avatar
Guolin Ke committed
2555

2556
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
2557
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2558
2559
2560

        Parameters
        ----------
2561
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2562
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2563
2564
2565
2566
2567

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2568
2569
        """
        self.label = label
2570
        if self.handle is not None:
2571
2572
2573
2574
            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)
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
                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)
2586
            else:
2587
                label_array = _list_to_1d_numpy(label, name='label')
2588
            self.set_field('label', label_array)
2589
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2590
        return self
Guolin Ke's avatar
Guolin Ke committed
2591

2592
2593
2594
2595
    def set_weight(
        self,
        weight: Optional[_LGBM_WeightType]
    ) -> "Dataset":
2596
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2597
2598
2599

        Parameters
        ----------
2600
        weight : list, numpy 1-D array, pandas Series or None
2601
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2602
2603
2604
2605
2606

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2607
        """
2608
2609
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2610
        self.weight = weight
2611
        if self.handle is not None and weight is not None:
2612
            weight = _list_to_1d_numpy(weight, name='weight')
wxchan's avatar
wxchan committed
2613
            self.set_field('weight', weight)
2614
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2615
        return self
Guolin Ke's avatar
Guolin Ke committed
2616

2617
2618
2619
2620
    def set_init_score(
        self,
        init_score: Optional[_LGBM_InitScoreType]
    ) -> "Dataset":
2621
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2622
2623
2624

        Parameters
        ----------
2625
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2626
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2627
2628
2629
2630
2631

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2632
2633
        """
        self.init_score = init_score
2634
        if self.handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2635
            self.set_field('init_score', init_score)
2636
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2637
        return self
Guolin Ke's avatar
Guolin Ke committed
2638

2639
2640
2641
2642
    def set_group(
        self,
        group: Optional[_LGBM_GroupType]
    ) -> "Dataset":
2643
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2644
2645
2646

        Parameters
        ----------
2647
        group : list, numpy 1-D array, pandas Series or None
2648
2649
2650
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2651
2652
            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
2653
2654
2655
2656
2657

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2658
2659
        """
        self.group = group
2660
        if self.handle is not None and group is not None:
2661
            group = _list_to_1d_numpy(group, np.int32, name='group')
wxchan's avatar
wxchan committed
2662
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2663
        return self
Guolin Ke's avatar
Guolin Ke committed
2664

2665
    def get_feature_name(self) -> List[str]:
2666
2667
2668
2669
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2670
        feature_names : list of str
2671
2672
2673
2674
2675
2676
2677
2678
            The names of columns (features) in the Dataset.
        """
        if self.handle is None:
            raise LightGBMError("Cannot get feature_name before construct dataset")
        num_feature = self.num_feature()
        tmp_out_len = ctypes.c_int(0)
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
2679
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2680
2681
2682
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
2683
            ctypes.c_int(num_feature),
2684
            ctypes.byref(tmp_out_len),
2685
            ctypes.c_size_t(reserved_string_buffer_size),
2686
2687
2688
2689
            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")
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
                self.handle,
                ctypes.c_int(num_feature),
                ctypes.byref(tmp_out_len),
                ctypes.c_size_t(actual_string_buffer_size),
                ctypes.byref(required_string_buffer_size),
                ptr_string_buffers))
2702
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2703

2704
    def get_label(self) -> Optional[np.ndarray]:
2705
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2706
2707
2708

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2709
        label : numpy array or None
2710
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2711
        """
2712
        if self.label is None:
wxchan's avatar
wxchan committed
2713
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2714
2715
        return self.label

2716
    def get_weight(self) -> Optional[np.ndarray]:
2717
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2718
2719
2720

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2721
        weight : numpy array or None
2722
            Weight for each data point from the Dataset. Weights should be non-negative.
Guolin Ke's avatar
Guolin Ke committed
2723
        """
2724
        if self.weight is None:
wxchan's avatar
wxchan committed
2725
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2726
2727
        return self.weight

2728
    def get_init_score(self) -> Optional[np.ndarray]:
2729
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2730
2731
2732

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2733
        init_score : numpy array or None
2734
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2735
        """
2736
        if self.init_score is None:
wxchan's avatar
wxchan committed
2737
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2738
2739
        return self.init_score

2740
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
2741
2742
2743
2744
        """Get the raw data of the Dataset.

        Returns
        -------
2745
        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
2746
2747
2748
2749
            Raw data used in the Dataset construction.
        """
        if self.handle is None:
            raise Exception("Cannot get data before construct Dataset")
2750
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
2751
2752
            self.data = self.reference.data
            if self.data is not None:
2753
                if isinstance(self.data, np.ndarray) or isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2754
                    self.data = self.data[self.used_indices, :]
2755
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2756
                    self.data = self.data.iloc[self.used_indices].copy()
2757
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2758
                    self.data = self.data[self.used_indices, :]
2759
2760
2761
2762
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
                elif isinstance(self.data, list) and len(self.data) > 0 and all(isinstance(x, Sequence) for x in self.data):
                    self.data = np.array([row for row in self._yield_row_from_seqlist(self.data, self.used_indices)])
Guolin Ke's avatar
Guolin Ke committed
2763
                else:
2764
2765
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
2766
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
2767
2768
2769
        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.")
2770
2771
        return self.data

2772
    def get_group(self) -> Optional[np.ndarray]:
2773
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2774
2775
2776

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2777
        group : numpy array or None
2778
2779
2780
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2781
2782
            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
2783
        """
2784
        if self.group is None:
wxchan's avatar
wxchan committed
2785
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
2786
2787
            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
2788
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
2789
2790
        return self.group

2791
    def num_data(self) -> int:
2792
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2793
2794
2795

        Returns
        -------
2796
2797
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2798
        """
2799
        if self.handle is not None:
2800
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2801
2802
2803
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2804
        else:
2805
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2806

2807
    def num_feature(self) -> int:
2808
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2809
2810
2811

        Returns
        -------
2812
2813
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2814
        """
2815
        if self.handle is not None:
2816
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2817
2818
2819
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2820
        else:
2821
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2822

2823
    def feature_num_bin(self, feature: Union[int, str]) -> int:
2824
2825
2826
2827
        """Get the number of bins for a feature.

        Parameters
        ----------
2828
2829
        feature : int or str
            Index or name of the feature.
2830
2831
2832
2833
2834
2835
2836

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
        if self.handle is not None:
2837
            if isinstance(feature, str):
2838
2839
2840
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
2841
2842
            ret = ctypes.c_int(0)
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self.handle,
2843
                                                         ctypes.c_int(feature_index),
2844
2845
2846
2847
2848
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

2849
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
2850
2851
2852
2853
2854
        """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.
2855
2856
2857
2858
2859

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2860
2861
2862

        Returns
        -------
2863
2864
2865
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2866
        head = self
2867
        ref_chain: Set[Dataset] = set()
2868
2869
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2870
                ref_chain.add(head)
2871
2872
2873
2874
2875
2876
                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
2877
        return ref_chain
2878

2879
    def add_features_from(self, other: "Dataset") -> "Dataset":
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
        """Add features from other Dataset to the current Dataset.

        Both Datasets must be constructed before calling this method.

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

        Returns
        -------
        self : Dataset
            Dataset with the new features added.
        """
        if self.handle is None or other.handle is None:
            raise ValueError('Both source and target Datasets must be constructed before adding features')
        _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self.handle, other.handle))
Guolin Ke's avatar
Guolin Ke committed
2897
2898
2899
2900
2901
2902
2903
2904
        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))
2905
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2906
                    self.data = np.hstack((self.data, other.data.toarray()))
2907
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2908
                    self.data = np.hstack((self.data, other.data.values))
2909
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2910
2911
2912
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
2913
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2914
                sparse_format = self.data.getformat()
2915
                if isinstance(other.data, np.ndarray) or isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2916
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
2917
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2918
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
2919
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2920
2921
2922
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
2923
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2924
2925
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
2926
2927
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
2928
                if isinstance(other.data, np.ndarray):
2929
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
2930
                                       axis=1, ignore_index=True)
2931
                elif isinstance(other.data, scipy.sparse.spmatrix):
2932
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
2933
                                       axis=1, ignore_index=True)
2934
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2935
2936
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
2937
2938
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
2939
2940
2941
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
2942
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2943
                if isinstance(other.data, np.ndarray):
2944
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
2945
                elif isinstance(other.data, scipy.sparse.spmatrix):
2946
2947
2948
2949
2950
                    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
2951
2952
2953
2954
2955
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
2956
2957
            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
2958
2959
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
2960
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
2961
        self.feature_name = self.get_feature_name()
2962
2963
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
2964
2965
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2966
2967
        return self

2968
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
2969
2970
2971
2972
2973
2974
        """Save Dataset to a text file.

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

        Parameters
        ----------
2975
        filename : str or pathlib.Path
2976
2977
2978
2979
2980
2981
2982
2983
2984
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
            self.construct().handle,
2985
            _c_str(str(filename))))
2986
2987
        return self

wxchan's avatar
wxchan committed
2988

2989
2990
2991
2992
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
    Tuple[np.ndarray, np.ndarray]
]
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
        _LGBM_EvalFunctionResultType
    ],
    Callable[
        [np.ndarray, Dataset],
        List[_LGBM_EvalFunctionResultType]
    ]
]
3003
3004


3005
class Booster:
3006
    """Booster in LightGBM."""
3007

3008
3009
3010
3011
3012
3013
3014
    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
    ):
3015
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3016
3017
3018

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3019
        params : dict or None, optional (default=None)
3020
3021
3022
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3023
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3024
            Path to the model file.
3025
        model_str : str or None, optional (default=None)
3026
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3027
        """
3028
        self.handle = None
3029
        self._network = False
wxchan's avatar
wxchan committed
3030
        self.__need_reload_eval_info = True
3031
        self._train_data_name = "training"
3032
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3033
        self.best_iteration = -1
3034
        self.best_score: _LGBM_BoosterBestScoreType = {}
3035
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3036
        if train_set is not None:
3037
            # Training task
wxchan's avatar
wxchan committed
3038
            if not isinstance(train_set, Dataset):
3039
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
            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"]
                )
3074
            # construct booster object
3075
3076
3077
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3078
            params_str = _param_dict_to_str(params)
3079
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
3080
            _safe_call(_LIB.LGBM_BoosterCreate(
3081
                train_set.handle,
3082
                _c_str(params_str),
wxchan's avatar
wxchan committed
3083
                ctypes.byref(self.handle)))
3084
            # save reference to data
wxchan's avatar
wxchan committed
3085
            self.train_set = train_set
3086
3087
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3088
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3089
3090
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
3091
3092
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
3093
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
3094
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3095
3096
3097
3098
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3099
            # buffer for inner predict
wxchan's avatar
wxchan committed
3100
3101
3102
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3103
            self.pandas_categorical = train_set.pandas_categorical
3104
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3105
        elif model_file is not None:
3106
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3107
            out_num_iterations = ctypes.c_int(0)
3108
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
3109
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
3110
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
3111
3112
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
3113
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3114
3115
3116
3117
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3118
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3119
3120
3121
            if params:
                _log_warning('Ignoring params argument, using parameters from model file.')
            params = self._get_loaded_param()
3122
        elif model_str is not None:
3123
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3124
        else:
3125
3126
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
3127
        self.params = params
wxchan's avatar
wxchan committed
3128

3129
    def __del__(self) -> None:
3130
        try:
3131
            if self._network:
3132
3133
3134
3135
3136
3137
3138
3139
                self.free_network()
        except AttributeError:
            pass
        try:
            if self.handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3140

3141
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3142
3143
        return self.__deepcopy__(None)

3144
    def __deepcopy__(self, _) -> "Booster":
3145
        model_str = self.model_to_string(num_iteration=-1)
3146
        booster = Booster(model_str=model_str)
3147
        return booster
wxchan's avatar
wxchan committed
3148

3149
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3150
3151
3152
3153
3154
        this = self.__dict__.copy()
        handle = this['handle']
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
3155
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3156
3157
        return this

3158
    def __setstate__(self, state: Dict[str, Any]) -> None:
3159
3160
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
3161
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3162
            out_num_iterations = ctypes.c_int(0)
3163
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3164
                _c_str(model_str),
3165
3166
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
3167
3168
3169
            state['handle'] = handle
        self.__dict__.update(state)

3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
    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(
            self.handle,
            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(
                self.handle,
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
        return json.loads(string_buffer.value.decode('utf-8'))

3192
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3193
3194
3195
3196
3197
3198
3199
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
3200
3201
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
3202
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3203
        return self
wxchan's avatar
wxchan committed
3204

3205
    def _free_buffer(self) -> "Booster":
3206
3207
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3208
        return self
3209

3210
3211
3212
3213
3214
3215
3216
    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":
3217
3218
3219
3220
        """Set the network configuration.

        Parameters
        ----------
3221
        machines : list, set or str
3222
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3223
        local_listen_port : int, optional (default=12400)
3224
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3225
        listen_time_out : int, optional (default=120)
3226
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3227
        num_machines : int, optional (default=1)
3228
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3229
3230
3231
3232
3233

        Returns
        -------
        self : Booster
            Booster with set network.
3234
        """
3235
3236
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
3237
        _safe_call(_LIB.LGBM_NetworkInit(_c_str(machines),
3238
3239
3240
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
3241
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3242
        return self
3243

3244
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3245
3246
3247
3248
3249
3250
3251
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3252
        _safe_call(_LIB.LGBM_NetworkFree())
3253
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3254
        return self
3255

3256
    def trees_to_dataframe(self) -> pd_DataFrame:
3257
3258
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3259
3260
3261
3262
        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.
3263
3264
3265
3266
3267
            - ``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.
3268
3269
            - ``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.
3270
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3271
3272
              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.
3273
3274
            - ``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.
3275
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3276
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3277
3278
            - ``count`` : int64, number of records in the training data that fall into this node.

3279
3280
3281
3282
3283
3284
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3285
3286
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297

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

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

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

            def _get_node_index(tree, tree_index):
3298
                tree_num = f'{tree_index}-' if tree_index is not None else ''
3299
3300
3301
                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
3302
3303
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315

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

            def _is_single_node_tree(tree):
3316
                return set(tree.keys()) == {'leaf_value'}
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389

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

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

            return node

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

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

            res = [node]

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

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

3390
        return pd_DataFrame(model_list, columns=model_list[0].keys())
3391

3392
    def set_train_data_name(self, name: str) -> "Booster":
3393
3394
3395
3396
        """Set the name to the training Dataset.

        Parameters
        ----------
3397
        name : str
Nikita Titov's avatar
Nikita Titov committed
3398
3399
3400
3401
3402
3403
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
3404
        """
3405
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
3406
        return self
wxchan's avatar
wxchan committed
3407

3408
    def add_valid(self, data: Dataset, name: str) -> "Booster":
3409
        """Add validation data.
wxchan's avatar
wxchan committed
3410
3411
3412
3413

        Parameters
        ----------
        data : Dataset
3414
            Validation data.
3415
        name : str
3416
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
3417
3418
3419
3420
3421

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
3422
        """
Guolin Ke's avatar
Guolin Ke committed
3423
        if not isinstance(data, Dataset):
3424
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3425
        if data._predictor is not self.__init_predictor:
3426
3427
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3428
3429
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
3430
            data.construct().handle))
wxchan's avatar
wxchan committed
3431
3432
3433
3434
3435
        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
3436
        return self
wxchan's avatar
wxchan committed
3437

3438
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
3439
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
3440
3441
3442
3443

        Parameters
        ----------
        params : dict
3444
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
3445
3446
3447
3448
3449

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
3450
        """
3451
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3452
3453
3454
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
                self.handle,
3455
                _c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
3456
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
3457
        return self
wxchan's avatar
wxchan committed
3458

3459
3460
3461
3462
3463
    def update(
        self,
        train_set: Optional[Dataset] = None,
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
3464
        """Update Booster for one iteration.
3465

wxchan's avatar
wxchan committed
3466
3467
        Parameters
        ----------
3468
3469
3470
3471
        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
3472
            Customized objective function.
3473
3474
3475
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3476
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3477
                    The predicted values.
3478
3479
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3480
3481
                train_data : Dataset
                    The training dataset.
3482
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3483
3484
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3485
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3486
3487
                    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
3488

3489
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3490
            and grad and hess should be returned in the same format.
3491

wxchan's avatar
wxchan committed
3492
3493
        Returns
        -------
3494
3495
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3496
        """
3497
        # need reset training data
3498
3499
3500
3501
3502
3503
        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
3504
            if not isinstance(train_set, Dataset):
3505
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3506
            if train_set._predictor is not self.__init_predictor:
3507
3508
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3509
3510
3511
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
3512
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
3513
            self.__inner_predict_buffer[0] = None
3514
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3515
3516
        is_finished = ctypes.c_int(0)
        if fobj is None:
3517
            if self.__set_objective_to_none:
3518
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3519
3520
3521
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
3522
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3523
3524
            return is_finished.value == 1
        else:
3525
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3526
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3527
3528
3529
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

3530
3531
3532
3533
3534
    def __boost(
        self,
        grad: np.ndarray,
        hess: np.ndarray
    ) -> bool:
3535
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3536

Nikita Titov's avatar
Nikita Titov committed
3537
3538
        .. note::

3539
3540
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3541
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3542
            and grad and hess should be returned in the same format.
3543

wxchan's avatar
wxchan committed
3544
3545
        Parameters
        ----------
3546
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3547
3548
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3549
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3550
3551
            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
3552
3553
3554

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3555
3556
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3557
        """
3558
3559
3560
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3561
3562
        grad = _list_to_1d_numpy(grad, name='gradient')
        hess = _list_to_1d_numpy(hess, name='hessian')
3563
3564
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3565
        if len(grad) != len(hess):
3566
3567
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3568
        if len(grad) != num_train_data * self.__num_class:
3569
3570
3571
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3572
                f"number of models per one iteration ({self.__num_class})"
3573
            )
wxchan's avatar
wxchan committed
3574
3575
3576
3577
3578
3579
        is_finished = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterUpdateOneIterCustom(
            self.handle,
            grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            ctypes.byref(is_finished)))
3580
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3581
3582
        return is_finished.value == 1

3583
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3584
3585
3586
3587
3588
3589
3590
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3591
3592
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
3593
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3594
        return self
wxchan's avatar
wxchan committed
3595

3596
    def current_iteration(self) -> int:
3597
3598
3599
3600
3601
3602
3603
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3604
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3605
3606
3607
3608
3609
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3610
    def num_model_per_iteration(self) -> int:
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
        """Get number of models per iteration.

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

3624
    def num_trees(self) -> int:
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
        """Get number of weak sub-models.

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

3638
    def upper_bound(self) -> float:
3639
3640
3641
3642
        """Get upper bound value of a model.

        Returns
        -------
3643
        upper_bound : float
3644
3645
3646
3647
3648
3649
3650
3651
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

3652
    def lower_bound(self) -> float:
3653
3654
3655
3656
        """Get lower bound value of a model.

        Returns
        -------
3657
        lower_bound : float
3658
3659
3660
3661
3662
3663
3664
3665
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

3666
3667
3668
3669
3670
3671
    def eval(
        self,
        data: Dataset,
        name: str,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3672
        """Evaluate for data.
wxchan's avatar
wxchan committed
3673
3674
3675

        Parameters
        ----------
3676
3677
        data : Dataset
            Data for the evaluating.
3678
        name : str
3679
            Name of the data.
3680
        feval : callable, list of callable, or None, optional (default=None)
3681
            Customized evaluation function.
3682
            Each evaluation function should accept two parameters: preds, eval_data,
3683
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3684

3685
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3686
                    The predicted values.
3687
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3688
                    If custom objective function is used, predicted values are returned before any transformation,
3689
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3690
                eval_data : Dataset
3691
                    A ``Dataset`` to evaluate.
3692
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3693
                    The name of evaluation function (without whitespace).
3694
3695
3696
3697
3698
                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
3699
3700
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3701
        result : list
3702
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3703
        """
Guolin Ke's avatar
Guolin Ke committed
3704
3705
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3706
3707
3708
3709
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3710
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3711
3712
3713
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3714
        # need to push new valid data
wxchan's avatar
wxchan committed
3715
3716
3717
3718
3719
3720
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

3721
3722
3723
3724
    def eval_train(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3725
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3726
3727
3728

        Parameters
        ----------
3729
        feval : callable, list of callable, or None, optional (default=None)
3730
            Customized evaluation function.
3731
            Each evaluation function should accept two parameters: preds, eval_data,
3732
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3733

3734
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3735
                    The predicted values.
3736
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3737
                    If custom objective function is used, predicted values are returned before any transformation,
3738
                    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
3739
                eval_data : Dataset
3740
                    The training dataset.
3741
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3742
                    The name of evaluation function (without whitespace).
3743
3744
3745
3746
3747
                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
3748
3749
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3750
        result : list
3751
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3752
        """
3753
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
3754

3755
3756
3757
3758
    def eval_valid(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3759
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
3760
3761
3762

        Parameters
        ----------
3763
        feval : callable, list of callable, or None, optional (default=None)
3764
            Customized evaluation function.
3765
            Each evaluation function should accept two parameters: preds, eval_data,
3766
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3767

3768
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3769
                    The predicted values.
3770
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3771
                    If custom objective function is used, predicted values are returned before any transformation,
3772
                    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
3773
                eval_data : Dataset
3774
                    The validation dataset.
3775
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3776
                    The name of evaluation function (without whitespace).
3777
3778
3779
3780
3781
                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
3782
3783
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3784
        result : list
3785
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3786
        """
3787
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
3788
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
3789

3790
3791
3792
3793
3794
3795
3796
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> "Booster":
3797
        """Save Booster to file.
wxchan's avatar
wxchan committed
3798
3799
3800

        Parameters
        ----------
3801
        filename : str or pathlib.Path
3802
            Filename to save Booster.
3803
3804
3805
3806
        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
3807
        start_iteration : int, optional (default=0)
3808
            Start index of the iteration that should be saved.
3809
        importance_type : str, optional (default="split")
3810
3811
3812
            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
3813
3814
3815
3816
3817

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
3818
        """
3819
        if num_iteration is None:
3820
            num_iteration = self.best_iteration
3821
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3822
3823
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
3824
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3825
            ctypes.c_int(num_iteration),
3826
            ctypes.c_int(importance_type_int),
3827
            _c_str(str(filename))))
3828
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
3829
        return self
wxchan's avatar
wxchan committed
3830

3831
3832
3833
3834
3835
    def shuffle_models(
        self,
        start_iteration: int = 0,
        end_iteration: int = -1
    ) -> "Booster":
3836
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
3837

3838
3839
3840
        Parameters
        ----------
        start_iteration : int, optional (default=0)
3841
            The first iteration that will be shuffled.
3842
3843
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
3844
            If <= 0, means the last available iteration.
3845

Nikita Titov's avatar
Nikita Titov committed
3846
3847
3848
3849
        Returns
        -------
        self : Booster
            Booster with shuffled models.
3850
        """
3851
3852
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
3853
3854
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
3855
        return self
3856

3857
    def model_from_string(self, model_str: str) -> "Booster":
3858
3859
3860
3861
        """Load Booster from a string.

        Parameters
        ----------
3862
        model_str : str
3863
3864
3865
3866
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3867
        self : Booster
3868
3869
            Loaded Booster object.
        """
3870
3871
3872
3873
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
3874
3875
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3876
            _c_str(model_str),
3877
3878
3879
3880
3881
3882
3883
            ctypes.byref(out_num_iterations),
            ctypes.byref(self.handle)))
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
            self.handle,
            ctypes.byref(out_num_class)))
        self.__num_class = out_num_class.value
3884
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
3885
3886
        return self

3887
3888
3889
3890
3891
3892
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> str:
3893
        """Save Booster to string.
3894

3895
3896
3897
3898
3899
3900
        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
3901
        start_iteration : int, optional (default=0)
3902
            Start index of the iteration that should be saved.
3903
        importance_type : str, optional (default="split")
3904
3905
3906
            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.
3907
3908
3909

        Returns
        -------
3910
        str_repr : str
3911
3912
            String representation of Booster.
        """
3913
        if num_iteration is None:
3914
            num_iteration = self.best_iteration
3915
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3916
        buffer_len = 1 << 20
3917
        tmp_out_len = ctypes.c_int64(0)
3918
3919
3920
3921
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
            self.handle,
3922
            ctypes.c_int(start_iteration),
3923
            ctypes.c_int(num_iteration),
3924
            ctypes.c_int(importance_type_int),
3925
            ctypes.c_int64(buffer_len),
3926
3927
3928
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
3929
        # if buffer length is not long enough, re-allocate a buffer
3930
3931
3932
3933
3934
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _safe_call(_LIB.LGBM_BoosterSaveModelToString(
                self.handle,
3935
                ctypes.c_int(start_iteration),
3936
                ctypes.c_int(num_iteration),
3937
                ctypes.c_int(importance_type_int),
3938
                ctypes.c_int64(actual_len),
3939
3940
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
3941
        ret = string_buffer.value.decode('utf-8')
3942
3943
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
3944

3945
3946
3947
3948
3949
3950
3951
    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
3952
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
3953

3954
3955
        Parameters
        ----------
3956
3957
3958
3959
        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
3960
        start_iteration : int, optional (default=0)
3961
            Start index of the iteration that should be dumped.
3962
        importance_type : str, optional (default="split")
3963
3964
3965
            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.
3966
3967
3968
3969
3970
3971
3972
3973
3974
        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.
3975

wxchan's avatar
wxchan committed
3976
3977
        Returns
        -------
3978
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
3979
            JSON format of Booster.
wxchan's avatar
wxchan committed
3980
        """
3981
        if num_iteration is None:
3982
            num_iteration = self.best_iteration
3983
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3984
        buffer_len = 1 << 20
3985
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
3986
3987
3988
3989
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterDumpModel(
            self.handle,
3990
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3991
            ctypes.c_int(num_iteration),
3992
            ctypes.c_int(importance_type_int),
3993
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
3994
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3995
            ptr_string_buffer))
wxchan's avatar
wxchan committed
3996
        actual_len = tmp_out_len.value
3997
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
3998
3999
4000
4001
4002
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _safe_call(_LIB.LGBM_BoosterDumpModel(
                self.handle,
4003
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4004
                ctypes.c_int(num_iteration),
4005
                ctypes.c_int(importance_type_int),
4006
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
4007
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4008
                ptr_string_buffer))
4009
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
4010
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
4011
                                                          default=_json_default_with_numpy))
4012
        return ret
wxchan's avatar
wxchan committed
4013

4014
4015
    def predict(
        self,
4016
        data: _LGBM_PredictDataType,
4017
4018
4019
4020
4021
4022
4023
4024
        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
4025
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4026
        """Make a prediction.
wxchan's avatar
wxchan committed
4027
4028
4029

        Parameters
        ----------
4030
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
4031
            Data source for prediction.
4032
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4033
        start_iteration : int, optional (default=0)
4034
            Start index of the iteration to predict.
4035
            If <= 0, starts from the first iteration.
4036
        num_iteration : int or None, optional (default=None)
4037
4038
4039
4040
            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).
4041
4042
4043
4044
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4045
4046
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4047

Nikita Titov's avatar
Nikita Titov committed
4048
4049
4050
4051
4052
4053
4054
            .. 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.
4055

4056
4057
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4058
            Used only if data is str.
4059
4060
4061
        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.
4062
4063
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4064
4065
4066

        Returns
        -------
4067
        result : numpy array, scipy.sparse or list of scipy.sparse
4068
            Prediction result.
4069
            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
4070
        """
4071
        predictor = self._to_predictor(deepcopy(kwargs))
4072
        if num_iteration is None:
4073
            if start_iteration <= 0:
4074
4075
4076
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
        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
4087

4088
4089
    def refit(
        self,
4090
        data: _LGBM_TrainDataType,
4091
        label: _LGBM_LabelType,
4092
4093
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4094
4095
4096
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4097
4098
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
4099
4100
4101
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4102
        **kwargs
4103
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4104
4105
4106
4107
        """Refit the existing Booster by new data.

        Parameters
        ----------
4108
        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
4109
            Data source for refit.
4110
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4111
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
4112
4113
            Label for refit.
        decay_rate : float, optional (default=0.9)
4114
4115
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4116
4117
4118
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
4119
            Weight for each ``data`` instance. Weights should be non-negative.
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
        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.
        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``.
        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.
        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.
4136
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4137
4138
4139
            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.
4140
            Floating point numbers in categorical features will be rounded towards 0.
4141
4142
4143
4144
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
4145
4146
4147
        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.
4148
4149
        **kwargs
            Other parameters for refit.
4150
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4151
4152
4153
4154
4155
4156

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4157
4158
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
4159
4160
        if dataset_params is None:
            dataset_params = {}
4161
        predictor = self._to_predictor(deepcopy(kwargs))
4162
4163
4164
4165
4166
4167
        leaf_preds = predictor.predict(
            data=data,
            start_iteration=-1,
            pred_leaf=True,
            validate_features=validate_features
        )
4168
        nrow, ncol = leaf_preds.shape
4169
        out_is_linear = ctypes.c_int(0)
4170
4171
4172
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
4173
4174
4175
4176
4177
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
4178
        new_params["linear_tree"] = bool(out_is_linear.value)
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
        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,
        )
4192
        new_params['refit_decay_rate'] = decay_rate
4193
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4194
4195
4196
4197
4198
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
4199
        ptr_data, _, _ = _c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
4200
4201
4202
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
4203
4204
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
4205
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4206
4207
        return new_booster

4208
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
        """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.
        """
4223
4224
4225
4226
4227
4228
4229
4230
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
            self.handle,
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
    ) -> 'Booster':
        """Set 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.
        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(
                self.handle,
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.c_double(value)
            )
        )
        return self

4263
4264
4265
4266
    def _to_predictor(
        self,
        pred_parameter: Optional[Dict[str, Any]] = None
    ) -> _InnerPredictor:
4267
        """Convert to predictor."""
4268
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
4269
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
4270
4271
        return predictor

4272
    def num_feature(self) -> int:
4273
4274
4275
4276
4277
4278
4279
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4280
4281
4282
4283
4284
4285
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
            self.handle,
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

4286
    def feature_name(self) -> List[str]:
4287
        """Get names of features.
wxchan's avatar
wxchan committed
4288
4289
4290

        Returns
        -------
4291
        result : list of str
4292
            List with names of features.
wxchan's avatar
wxchan committed
4293
        """
4294
        num_feature = self.num_feature()
4295
        # Get name of features
wxchan's avatar
wxchan committed
4296
        tmp_out_len = ctypes.c_int(0)
4297
4298
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
4299
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
4300
4301
4302
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
4303
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
4304
            ctypes.byref(tmp_out_len),
4305
            ctypes.c_size_t(reserved_string_buffer_size),
4306
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4307
4308
4309
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
                self.handle,
                ctypes.c_int(num_feature),
                ctypes.byref(tmp_out_len),
                ctypes.c_size_t(actual_string_buffer_size),
                ctypes.byref(required_string_buffer_size),
                ptr_string_buffers))
4322
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
4323

4324
4325
4326
4327
4328
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
4329
        """Get feature importances.
4330

4331
4332
        Parameters
        ----------
4333
        importance_type : str, optional (default="split")
4334
4335
4336
            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.
4337
4338
4339
4340
        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).
4341

4342
4343
        Returns
        -------
4344
4345
        result : numpy array
            Array with feature importances.
4346
        """
4347
4348
        if iteration is None:
            iteration = self.best_iteration
4349
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4350
        result = np.empty(self.num_feature(), dtype=np.float64)
4351
4352
4353
4354
4355
        _safe_call(_LIB.LGBM_BoosterFeatureImportance(
            self.handle,
            ctypes.c_int(iteration),
            ctypes.c_int(importance_type_int),
            result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
4356
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
4357
            return result.astype(np.int32)
4358
4359
        else:
            return result
4360

4361
4362
4363
4364
4365
4366
    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]:
4367
4368
4369
4370
        """Get split value histogram for the specified feature.

        Parameters
        ----------
4371
        feature : int or str
4372
4373
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
4374
            If str, interpreted as name.
4375

Nikita Titov's avatar
Nikita Titov committed
4376
4377
4378
            .. warning::

                Categorical features are not supported.
4379

4380
        bins : int, str or None, optional (default=None)
4381
4382
4383
            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.
4384
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
        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.
        """
4399
        def add(root: Dict[str, Any]) -> None:
4400
4401
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
4402
                if feature_names is not None and isinstance(feature, str):
4403
4404
4405
4406
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
4407
                    if isinstance(root['threshold'], str):
4408
4409
4410
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
                add(root['left_child'])
                add(root['right_child'])

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

4421
        if bins is None or isinstance(bins, int) and xgboost_style:
4422
4423
4424
4425
4426
4427
4428
            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:
4429
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
4430
4431
4432
4433
4434
            else:
                return ret
        else:
            return hist, bin_edges

4435
4436
4437
4438
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
4439
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]]
4440
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4441
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
4442
        if data_idx >= self.__num_dataset:
4443
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4444
4445
4446
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
4447
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
4448
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
4449
4450
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
4451
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4452
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4453
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
4454
            if tmp_out_len.value != self.__num_inner_eval:
4455
                raise ValueError("Wrong length of eval results")
4456
            for i in range(self.__num_inner_eval):
4457
4458
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
4459
4460
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
4461
4462
4463
4464
4465
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
4466
4467
4468
4469
4470
4471
4472
4473
4474
            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
4475
4476
4477
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

4478
    def __inner_predict(self, data_idx: int) -> np.ndarray:
4479
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
4480
        if data_idx >= self.__num_dataset:
4481
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4482
4483
4484
4485
4486
        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
4487
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
4488
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
4489
4490
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
4491
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
4492
4493
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
4494
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4495
4496
4497
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
4498
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
4499
            self.__is_predicted_cur_iter[data_idx] = True
4500
4501
4502
4503
4504
        result = self.__inner_predict_buffer[data_idx]
        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
4505

4506
    def __get_eval_info(self) -> None:
4507
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
4508
4509
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
4510
            out_num_eval = ctypes.c_int(0)
4511
            # Get num of inner evals
wxchan's avatar
wxchan committed
4512
4513
4514
4515
4516
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
                self.handle,
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
4517
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
4518
                tmp_out_len = ctypes.c_int(0)
4519
4520
4521
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
4522
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
4523
                ]
wxchan's avatar
wxchan committed
4524
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
4525
4526
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
4527
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
4528
                    ctypes.byref(tmp_out_len),
4529
                    ctypes.c_size_t(reserved_string_buffer_size),
4530
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4531
4532
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
4533
                    raise ValueError("Length of eval names doesn't equal with num_evals")
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
                actual_string_buffer_size = required_string_buffer_size.value
                # if buffer length is not long enough, reallocate buffers
                if reserved_string_buffer_size < actual_string_buffer_size:
                    string_buffers = [
                        ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(self.__num_inner_eval)
                    ]
                    ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
                    _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                        self.handle,
                        ctypes.c_int(self.__num_inner_eval),
                        ctypes.byref(tmp_out_len),
                        ctypes.c_size_t(actual_string_buffer_size),
                        ctypes.byref(required_string_buffer_size),
                        ptr_string_buffers))
                self.__name_inner_eval = [
                    string_buffers[i].value.decode('utf-8') for i in range(self.__num_inner_eval)
                ]
                self.__higher_better_inner_eval = [
                    name.startswith(('auc', 'ndcg@', 'map@', 'average_precision')) for name in self.__name_inner_eval
                ]