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
_LGBM_LabelType = Union[
75
76
    List[float],
    List[int],
77
78
79
80
    np.ndarray,
    pd_Series,
    pd_DataFrame
]
81
82
83
84
85
86
87
88
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix
]
89
90
91
92
93
94
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
95
96
97
ZERO_THRESHOLD = 1e-35


98
99
100
101
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


102
def _get_sample_count(total_nrow: int, params: str) -> int:
103
104
105
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
106
        _c_str(params),
107
108
109
110
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
111

112
113
114
115
116
117
class _MissingType(Enum):
    NONE = 'None'
    NAN = 'NaN'
    ZERO = 'Zero'


118
class _DummyLogger:
119
    def info(self, msg: str) -> None:
120
121
        print(msg)

122
    def warning(self, msg: str) -> None:
123
124
125
        warnings.warn(msg, stacklevel=3)


126
127
128
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
129
130


131
132
133
134
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


135
136
137
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
138
139
140
141
    """Register custom logger.

    Parameters
    ----------
142
    logger : Any
143
        Custom logger.
144
145
146
147
    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.
148
    """
149
150
151
152
153
154
    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
155
    _LOGGER = logger
156
157
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
158
159


160
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
161
    """Join log messages from native library which come by chunks."""
162
    msg_normalized: List[str] = []
163
164

    @wraps(func)
165
    def wrapper(msg: str) -> None:
166
167
168
169
170
171
172
173
174
175
176
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


177
def _log_info(msg: str) -> None:
178
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
179
180


181
def _log_warning(msg: str) -> None:
182
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
183
184
185


@_normalize_native_string
186
def _log_native(msg: str) -> None:
187
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
188
189


190
def _log_callback(msg: bytes) -> None:
191
    """Redirect logs from native library into Python."""
192
    _log_native(str(msg.decode('utf-8')))
193
194


195
def _load_lib() -> ctypes.CDLL:
196
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
197
198
199
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
200
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
201
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
202
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
203
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
204
205
    return lib

wxchan's avatar
wxchan committed
206

207
208
209
210
211
212
213
# 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
214

wxchan's avatar
wxchan committed
215

216
_NUMERIC_TYPES = (int, float, bool)
217
_ArrayLike = Union[List, np.ndarray, pd_Series]
218
219


220
def _safe_call(ret: int) -> None:
221
222
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
223
224
225
    Parameters
    ----------
    ret : int
226
        The return value from C API calls.
wxchan's avatar
wxchan committed
227
228
    """
    if ret != 0:
229
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
230

wxchan's avatar
wxchan committed
231

232
def _is_numeric(obj: Any) -> bool:
233
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
234
235
236
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
237
238
239
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
240
241
        return False

wxchan's avatar
wxchan committed
242

243
def _is_numpy_1d_array(data: Any) -> bool:
244
    """Check whether data is a numpy 1-D array."""
245
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
246

wxchan's avatar
wxchan committed
247

248
def _is_numpy_column_array(data: Any) -> bool:
249
250
251
252
253
254
255
    """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


256
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
257
    """Cast numpy array to given dtype."""
258
259
260
261
262
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


263
def _is_1d_list(data: Any) -> bool:
264
    """Check whether data is a 1-D list."""
265
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
266

wxchan's avatar
wxchan committed
267

268
269
270
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
271
        _is_numpy_1d_array(data)
272
        or _is_numpy_column_array(data)
273
        or _is_1d_list(data)
274
275
276
277
        or isinstance(data, pd_Series)
    )


278
279
def _list_to_1d_numpy(
    data: Any,
280
    dtype: "np.typing.DTypeLike" = np.float32,
281
282
    name: str = 'list'
) -> np.ndarray:
283
    """Convert data to numpy 1-D array."""
284
    if _is_numpy_1d_array(data):
285
        return _cast_numpy_array_to_dtype(data, dtype)
286
    elif _is_numpy_column_array(data):
287
288
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
289
        return _cast_numpy_array_to_dtype(array, dtype)
290
    elif _is_1d_list(data):
wxchan's avatar
wxchan committed
291
        return np.array(data, dtype=dtype, copy=False)
292
    elif isinstance(data, pd_Series):
293
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
294
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
295
    else:
296
297
        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
298

wxchan's avatar
wxchan committed
299

300
301
302
303
304
305
306
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."""
307
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
308
309
310
311
312
313
314
315
316
317
318


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)
    )


319
320
321
322
323
def _data_to_2d_numpy(
    data: Any,
    dtype: "np.typing.DTypeLike" = np.float32,
    name: str = 'list'
) -> np.ndarray:
324
325
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
326
        return _cast_numpy_array_to_dtype(data, dtype)
327
328
329
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
330
        _check_for_bad_pandas_dtypes(data.dtypes)
331
        return _cast_numpy_array_to_dtype(data.values, dtype)
332
333
334
335
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


336
def _cfloat32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
337
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
338
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
339
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
340
    else:
341
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
342

Guolin Ke's avatar
Guolin Ke committed
343

344
def _cfloat64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
345
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
346
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
347
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
348
349
350
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
351

352
def _cint32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
353
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
354
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
355
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
356
    else:
357
358
359
        raise RuntimeError('Expected int32 pointer')


360
def _cint64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
361
362
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
363
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
364
365
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
366

wxchan's avatar
wxchan committed
367

368
def _c_str(string: str) -> ctypes.c_char_p:
369
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
370
371
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
372

373
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
374
    """Convert a Python array to C array."""
375
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
376

wxchan's avatar
wxchan committed
377

378
def _json_default_with_numpy(obj: Any) -> Any:
379
380
381
382
383
384
385
386
387
    """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


388
389
390
391
392
393
394
395
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)


396
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
397
    """Convert Python dictionary to string, which is passed to C API."""
398
    if data is None or not data:
wxchan's avatar
wxchan committed
399
400
401
        return ""
    pairs = []
    for key, val in data.items():
402
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
403
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
404
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
405
            pairs.append(f"{key}={val}")
406
        elif val is not None:
407
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
408
    return ' '.join(pairs)
409

wxchan's avatar
wxchan committed
410

411
class _TempFile:
412
413
    """Proxy class to workaround errors on Windows."""

414
415
416
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
417
            self.path = Path(self.name)
418
        return self
wxchan's avatar
wxchan committed
419

420
    def __exit__(self, exc_type, exc_val, exc_tb):
421
422
        if self.path.is_file():
            self.path.unlink()
423

wxchan's avatar
wxchan committed
424

425
class LightGBMError(Exception):
426
427
    """Error thrown by LightGBM."""

428
429
430
    pass


431
432
433
434
435
436
437
438
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
439
440
441
442
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
443
    def _get_all_param_aliases() -> Dict[str, List[str]]:
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
        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'),
463
            object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
464
465
        )
        return aliases
466
467

    @classmethod
468
469
470
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
471
472
        ret = set()
        for i in args:
473
            ret.update(cls.get_sorted(i))
474
475
        return ret

476
477
478
479
480
481
    @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])

482
    @classmethod
483
484
485
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
486
487
488
489
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
490
                    ret.update(aliases)
491
492
493
                    break
        return ret

494

495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
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)

516
517
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
518
519

    # if main_param_name was provided, keep that value and remove all aliases
520
    if main_param_name in params.keys():
521
522
523
        for param in aliases:
            params.pop(param, None)
        return params
524

525
526
527
528
529
    # 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
530

531
532
533
534
535
536
537
    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
538
539
540
541

    return params


542
_MAX_INT32 = (1 << 31) - 1
543

544
"""Macro definition of data type in C API of LightGBM"""
545
546
547
548
_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
549

550
"""Matrix is row major in Python"""
551
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
552

553
"""Macro definition of prediction type in C API of LightGBM"""
554
555
556
557
_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
558

559
"""Macro definition of sparse matrix type"""
560
561
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
562

563
"""Macro definition of feature importance type"""
564
565
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
566

567
"""Data type of data field"""
568
569
570
571
572
573
_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
574

575
"""String name to int feature importance type mapper"""
576
577
578
579
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN
}
580

wxchan's avatar
wxchan committed
581

582
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
583
    """Fix the memory of multi-dimensional sliced object."""
584
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
585
        if not data.flags.c_contiguous:
586
587
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
588
589
590
591
            return np.copy(data)
    return data


592
def _c_float_array(data):
593
    """Get pointer of float numpy array / list."""
594
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
595
        data = np.array(data, copy=False)
596
    if _is_numpy_1d_array(data):
597
        data = _convert_from_sliced_object(data)
598
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
599
600
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
601
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
602
603
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
604
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
605
        else:
606
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
607
    else:
608
        raise TypeError(f"Unknown type({type(data).__name__})")
609
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
610

wxchan's avatar
wxchan committed
611

612
def _c_int_array(data):
613
    """Get pointer of int numpy array / list."""
614
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
615
        data = np.array(data, copy=False)
616
    if _is_numpy_1d_array(data):
617
        data = _convert_from_sliced_object(data)
618
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
619
620
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
621
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
622
623
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
624
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
625
        else:
626
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
627
    else:
628
        raise TypeError(f"Unknown type({type(data).__name__})")
629
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
630

wxchan's avatar
wxchan committed
631

632
def _is_allowed_numpy_dtype(dtype: type) -> bool:
633
    float128 = getattr(np, 'float128', type(None))
634
635
636
637
    return (
        issubclass(dtype, (np.integer, np.floating, np.bool_))
        and not issubclass(dtype, (np.timedelta64, float128))
    )
638
639


640
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
641
642
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
643
        for column_name, pandas_dtype in pandas_dtypes_series.items()
644
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
645
646
647
648
    ]
    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)}')
649
650


651
652
653
654
655
656
def _data_from_pandas(
    data,
    feature_name: Optional[_LGBM_FeatureNameConfiguration],
    categorical_feature: Optional[_LGBM_CategoricalFeatureConfiguration],
    pandas_categorical: Optional[List[List]]
):
657
    if isinstance(data, pd_DataFrame):
658
659
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
660
        if feature_name == 'auto' or feature_name is None:
661
            data = data.rename(columns=str, copy=False)
662
        cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
663
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
664
665
666
667
668
        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.')
669
            for col, category in zip(cat_cols, pandas_categorical):
670
671
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
672
        if len(cat_cols):  # cat_cols is list
673
            data = data.copy(deep=False)  # not alter origin DataFrame
674
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
675
676
677
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
678
            if categorical_feature == 'auto':  # use cat cols from DataFrame
679
                categorical_feature = cat_cols_not_ordered
680
681
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
682
683
        if feature_name == 'auto':
            feature_name = list(data.columns)
684
        _check_for_bad_pandas_dtypes(data.dtypes)
685
686
687
        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, [])
688
689
690
691
692
693
694
695
696
697
        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)
698
699
700
701
702
703
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
704
705


706
707
708
709
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
    file_name: Optional[Union[str, Path]] = None
) -> str:
710
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
711
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
712
713
714
715
716
717
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


718
719
720
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
    model_str: Optional[str] = None
721
) -> Optional[List[List]]:
722
723
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
724
    if file_name is not None:
725
        max_offset = -getsize(file_name)
726
727
728
729
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
730
                f.seek(offset, SEEK_END)
731
732
733
734
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
735
        last_line = lines[-1].decode('utf-8').strip()
736
        if not last_line.startswith(pandas_key):
737
            last_line = lines[-2].decode('utf-8').strip()
738
    elif model_str is not None:
739
740
741
742
743
744
        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
745
746


747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
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**.

767
768
    .. versionadded:: 3.3.0

769
770
771
772
773
774
775
776
777
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
778
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
779
780
781
782
783
784
785
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
786
                return self._get_one_line(idx)
787
            elif isinstance(idx, slice):
788
789
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
790
                # Only required if using ``Dataset.subset()``.
791
                return np.array([self._get_one_line(i) for i in idx])
792
            else:
793
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
794
795
796

        Parameters
        ----------
797
        idx : int, slice[int], list[int]
798
799
800
801
            Item index.

        Returns
        -------
802
        result : numpy 1-D array or numpy 2-D array
803
            1-D array if idx is int, 2-D array if idx is slice or list.
804
805
806
807
808
809
810
811
812
        """
        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__()")


813
class _InnerPredictor:
814
815
816
817
818
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
819
820
821
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
822
    """
823

824
825
826
827
828
829
    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
    ):
830
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
831
832
833

        Parameters
        ----------
834
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
835
            Path to the model file.
836
837
838
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
839
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
840
841
842
843
844
        """
        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
845
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
846
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
847
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
848
849
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
850
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
851
852
853
854
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
855
            self.num_total_iteration = out_num_iterations.value
856
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
857
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
858
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
859
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
860
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
861
862
863
864
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
865
            self.num_total_iteration = self.current_iteration()
866
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
867
        else:
868
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
869

870
        pred_parameter = {} if pred_parameter is None else pred_parameter
871
        self.pred_parameter = _param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
872

873
    def __del__(self) -> None:
874
875
876
877
878
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
879

880
    def __getstate__(self) -> Dict[str, Any]:
881
882
883
884
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

885
886
    def predict(
        self,
887
        data: _LGBM_PredictDataType,
888
889
890
891
892
893
894
        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
895
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
896
        """Predict logic.
wxchan's avatar
wxchan committed
897
898
899

        Parameters
        ----------
900
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
901
            Data source for prediction.
902
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
903
904
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
905
906
907
908
909
910
911
912
913
914
915
        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.
916
917
918
        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
919
920
921

        Returns
        -------
922
        result : numpy array, scipy.sparse or list of scipy.sparse
923
            Prediction result.
924
            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
925
        """
wxchan's avatar
wxchan committed
926
        if isinstance(data, Dataset):
927
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
928
929
930
931
932
933
934
935
936
937
938
        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)),
                )
            )
939
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
940
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
941
        if raw_score:
942
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
943
        if pred_leaf:
944
            predict_type = _C_API_PREDICT_LEAF_INDEX
945
        if pred_contrib:
946
            predict_type = _C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
947
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
948

949
        if isinstance(data, (str, Path)):
950
            with _TempFile() as f:
wxchan's avatar
wxchan committed
951
952
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
953
                    _c_str(str(data)),
Guolin Ke's avatar
Guolin Ke committed
954
955
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
956
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
957
                    ctypes.c_int(num_iteration),
958
959
                    _c_str(self.pred_parameter),
                    _c_str(f.name)))
960
961
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
962
        elif isinstance(data, scipy.sparse.csr_matrix):
963
964
965
966
967
968
            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
969
        elif isinstance(data, scipy.sparse.csc_matrix):
970
971
972
973
974
975
            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
976
        elif isinstance(data, np.ndarray):
977
978
979
980
981
982
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
983
984
985
        elif isinstance(data, list):
            try:
                data = np.array(data)
986
            except BaseException:
987
                raise ValueError('Cannot convert data list to numpy array.')
988
989
990
991
992
993
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
994
        elif isinstance(data, dt_DataTable):
995
996
997
998
999
1000
            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
1001
1002
        else:
            try:
1003
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
1004
                csr = scipy.sparse.csr_matrix(data)
1005
            except BaseException:
1006
                raise TypeError(f'Cannot predict data for type {type(data).__name__}')
1007
1008
1009
1010
1011
1012
            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
1013
1014
        if pred_leaf:
            preds = preds.astype(np.int32)
1015
        is_sparse = isinstance(preds, scipy.sparse.spmatrix) or isinstance(preds, list)
1016
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1017
            if preds.size % nrow == 0:
1018
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1019
            else:
1020
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
1021
1022
        return preds

1023
1024
1025
1026
1027
1028
1029
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
        predict_type: int
    ) -> int:
1030
        """Get size of prediction result."""
1031
        if nrow > _MAX_INT32:
1032
            raise LightGBMError('LightGBM cannot perform prediction for data '
1033
                                f'with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n'
1034
                                'You can split your data into chunks '
1035
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
1036
1037
1038
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
1039
1040
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
1041
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1042
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
1043
1044
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
1045

1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
    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)
1059
1060
1061
1062
1063
1064
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
            predict_type=predict_type
        )
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
1093
        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]:
1094
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1095
        if len(mat.shape) != 2:
1096
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
1097

1098
        nrow = mat.shape[0]
1099
1100
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1101
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1102
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
1103
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1104
            preds = np.empty(sum(n_preds), dtype=np.float64)
1105
1106
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
1107
                # avoid memory consumption by arrays concatenation operations
1108
1109
1110
1111
1112
1113
1114
                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]
                )
1115
            return preds, nrow
wxchan's avatar
wxchan committed
1116
        else:
1117
1118
1119
1120
1121
1122
1123
            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
1124

1125
1126
1127
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1128
1129
1130
1131
1132
1133
        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,
1134
        is_csr: bool
1135
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1136
1137
1138
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1139
        if indptr_type == _C_API_DTYPE_INT32:
1140
            out_indptr = _cint32_array_to_numpy(out_ptr_indptr, indptr_len)
1141
        elif indptr_type == _C_API_DTYPE_INT64:
1142
            out_indptr = _cint64_array_to_numpy(out_ptr_indptr, indptr_len)
1143
1144
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1145
        if data_type == _C_API_DTYPE_FLOAT32:
1146
            out_data = _cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
1147
        elif data_type == _C_API_DTYPE_FLOAT64:
1148
            out_data = _cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
1149
1150
        else:
            raise TypeError("Expected float32 or float64 type for data")
1151
        out_indices = _cint32_array_to_numpy(out_ptr_indices, data_indices_len)
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
1179
        # 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

1180
1181
1182
1183
1184
1185
1186
1187
1188
    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
1189
1190
1191
1192
1193
1194
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1195
1196
1197
1198
1199
        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
1200

1201
1202
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
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
1232
        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
1233
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1234
1235
1236
1237
        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
1238
        out_ptr_indptr: _ctypes_int_ptr
1239
1240
1241
1242
1243
        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)()
1244
        out_ptr_data: _ctypes_float_ptr
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
1281
        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

1282
1283
1284
1285
1286
1287
1288
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1289
        """Predict for a CSR data."""
1290
        if predict_type == _C_API_PREDICT_CONTRIB:
1291
1292
1293
1294
1295
1296
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1297
        nrow = len(csr.indptr) - 1
1298
1299
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1300
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1301
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1302
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1303
            preds = np.empty(sum(n_preds), dtype=np.float64)
1304
1305
            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:])):
1306
                # avoid memory consumption by arrays concatenation operations
1307
1308
1309
1310
1311
1312
1313
                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]
                )
1314
1315
            return preds, nrow
        else:
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
            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,
1326
1327
1328
1329
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
1330
1331
1332
1333
1334
    ):
        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
1335
        out_ptr_indptr: _ctypes_int_ptr
1336
1337
1338
1339
1340
        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)()
1341
        out_ptr_data: _ctypes_float_ptr
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
        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
1378

1379
1380
1381
1382
1383
1384
1385
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1386
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1387
        nrow = csc.shape[0]
1388
        if nrow > _MAX_INT32:
1389
1390
1391
1392
1393
1394
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1395
        if predict_type == _C_API_PREDICT_CONTRIB:
1396
1397
1398
1399
1400
1401
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1402
1403
1404
1405
1406
1407
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1408
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1409
1410
        out_num_preds = ctypes.c_int64(0)

1411
1412
        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
1413

1414
        assert csc.shape[0] <= _MAX_INT32
1415
        csc_indices = csc.indices.astype(np.int32, copy=False)
1416

Guolin Ke's avatar
Guolin Ke committed
1417
1418
1419
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
1420
            ctypes.c_int(type_ptr_indptr),
1421
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1422
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1423
1424
1425
1426
1427
            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),
1428
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1429
            ctypes.c_int(num_iteration),
1430
            _c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1431
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1432
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1433
        if n_preds != out_num_preds.value:
1434
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1435
1436
        return preds, nrow

1437
    def current_iteration(self) -> int:
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
        """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
1451

1452
class Dataset:
wxchan's avatar
wxchan committed
1453
    """Dataset in LightGBM."""
1454

1455
1456
    def __init__(
        self,
1457
        data: _LGBM_TrainDataType,
1458
        label: Optional[_LGBM_LabelType] = None,
1459
        reference: Optional["Dataset"] = None,
1460
1461
1462
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1463
1464
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
1465
1466
1467
        params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True
    ):
1468
        """Initialize Dataset.
1469

wxchan's avatar
wxchan committed
1470
1471
        Parameters
        ----------
1472
        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
1473
            Data source of Dataset.
1474
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1475
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1476
1477
1478
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1479
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1480
            Weight for each instance. Weights should be non-negative.
1481
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1482
1483
1484
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1485
1486
            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.
1487
        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)
1488
            Init score for Dataset.
1489
        feature_name : list of str, or 'auto', optional (default="auto")
1490
1491
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
1492
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1493
1494
            Categorical features.
            If list of int, interpreted as indices.
1495
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1496
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1497
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1498
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1499
            All negative values in categorical features will be treated as missing values.
1500
            The output cannot be monotonically constrained with respect to a categorical feature.
1501
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1502
        params : dict or None, optional (default=None)
1503
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1504
        free_raw_data : bool, optional (default=True)
1505
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
1506
        """
1507
        self.handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1508
1509
1510
1511
1512
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1513
        self.init_score = init_score
1514
1515
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1516
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1517
        self.free_raw_data = free_raw_data
1518
        self.used_indices: Optional[List[int]] = None
1519
        self._need_slice = True
1520
        self._predictor: Optional[_InnerPredictor] = None
1521
        self.pandas_categorical = None
1522
        self._params_back_up = None
1523
        self.version = 0
1524
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1525

1526
    def __del__(self) -> None:
1527
1528
1529
1530
        try:
            self._free_handle()
        except AttributeError:
            pass
1531

1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
    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.
        """
1549
        param_str = _param_dict_to_str(self.get_params())
1550
1551
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1552
        ptr_data, _, _ = _c_int_array(indices)
1553
1554
1555
1556
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
1557
            _c_str(param_str),
1558
1559
1560
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1561
1562
        assert sample_cnt == actual_sample_cnt.value
        return indices
1563

1564
1565
1566
1567
1568
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
        ref_dataset: _DatasetHandle
    ) -> 'Dataset':
1569
1570
1571
1572
1573
1574
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1575
1576
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601

        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
        ----------
1602
        sample_data : list of numpy array
1603
            Sample data for each column.
1604
        sample_indices : list of numpy array
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
1631
            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):
1632
1633
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1634
1635

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1636
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1637
1638

        self.handle = ctypes.c_void_p()
1639
        params_str = _param_dict_to_str(self.get_params())
1640
1641
1642
1643
1644
1645
1646
        _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),
1647
            ctypes.c_int64(total_nrow),
1648
            _c_str(params_str),
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
            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)
1668
        data_ptr, data_type, _ = _c_float_array(data)
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680

        _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

1681
    def get_params(self) -> Dict[str, Any]:
1682
1683
1684
1685
        """Get the used parameters in the Dataset.

        Returns
        -------
1686
        params : dict
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
            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",
1702
                                                "linear_tree",
1703
1704
1705
1706
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1707
                                                "precise_float_parser",
1708
1709
1710
1711
1712
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1713
1714
        else:
            return {}
1715

1716
    def _free_handle(self) -> "Dataset":
1717
        if self.handle is not None:
1718
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1719
            self.handle = None
1720
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
1721
1722
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1723
        return self
wxchan's avatar
wxchan committed
1724

1725
1726
1727
1728
1729
1730
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
        data,
        used_indices: Optional[List[int]]
    ):
Guolin Ke's avatar
Guolin Ke committed
1731
        data_has_header = False
1732
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
1733
            # check data has header or not
1734
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1735
        num_data = self.num_data()
1736
1737
1738
        if predictor is not None:
            init_score = predictor.predict(data,
                                           raw_score=True,
1739
1740
                                           data_has_header=data_has_header)
            init_score = init_score.ravel()
1741
            if used_indices is not None:
1742
                assert not self._need_slice
1743
                if isinstance(data, (str, Path)):
1744
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1745
                    assert num_data == len(used_indices)
1746
1747
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1748
1749
1750
1751
                            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
1752
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1753
1754
                for i in range(num_data):
                    for j in range(predictor.num_class):
1755
1756
1757
                        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:
1758
            init_score = np.zeros(self.init_score.shape, dtype=np.float64)
1759
1760
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1761
1762
        self.set_init_score(init_score)

1763
1764
    def _lazy_init(
        self,
1765
        data: Optional[_LGBM_TrainDataType],
1766
        label: Optional[_LGBM_LabelType] = None,
1767
        reference: Optional["Dataset"] = None,
1768
1769
1770
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1771
1772
1773
1774
1775
        predictor=None,
        feature_name='auto',
        categorical_feature='auto',
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
wxchan's avatar
wxchan committed
1776
1777
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1778
            return self
Guolin Ke's avatar
Guolin Ke committed
1779
1780
1781
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1782
1783
1784
1785
        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
1786

1787
        # process for args
wxchan's avatar
wxchan committed
1788
        params = {} if params is None else params
1789
1790
1791
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1792
        for key in params.keys():
1793
            if key in args_names:
1794
1795
                _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.')
1796
        # get categorical features
1797
1798
1799
1800
1801
1802
        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:
1803
                if isinstance(name, str) and name in feature_dict:
1804
                    categorical_indices.add(feature_dict[name])
1805
                elif isinstance(name, int):
1806
1807
                    categorical_indices.add(name)
                else:
1808
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1809
            if categorical_indices:
1810
1811
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1812
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
1813
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
1814
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1815
                        params.pop(cat_alias, None)
1816
                params['categorical_column'] = sorted(categorical_indices)
1817

1818
        params_str = _param_dict_to_str(params)
1819
        self.params = params
1820
        # process for reference dataset
wxchan's avatar
wxchan committed
1821
        ref_dataset = None
wxchan's avatar
wxchan committed
1822
        if isinstance(reference, Dataset):
1823
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1824
1825
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1826
        # start construct data
1827
        if isinstance(data, (str, Path)):
wxchan's avatar
wxchan committed
1828
1829
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1830
1831
                _c_str(str(data)),
                _c_str(params_str),
wxchan's avatar
wxchan committed
1832
1833
1834
1835
                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
1836
1837
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1838
1839
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1840
1841
1842
1843
1844
1845
1846
1847
1848
        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)
1849
        elif isinstance(data, dt_DataTable):
1850
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1851
1852
1853
1854
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1855
            except BaseException:
1856
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}')
wxchan's avatar
wxchan committed
1857
1858
1859
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1860
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1861
1862
1863
1864
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1865
1866
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1867
                _log_warning("The init_score will be overridden by the prediction of init_model.")
1868
1869
1870
1871
1872
            self._set_init_score_by_predictor(
                predictor=predictor,
                data=data,
                used_indices=None
            )
1873
1874
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1875
        elif predictor is not None:
1876
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1877
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1878
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1879

1880
1881
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
        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.
1907
        sampled = np.array([row for row in self._yield_row_from_seqlist(seqs, indices)])
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
        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

1923
1924
1925
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
1926
        ref_dataset: Optional[_DatasetHandle]
1927
    ) -> "Dataset":
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
        """
        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:
1942
            param_str = _param_dict_to_str(self.get_params())
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
            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

1956
1957
1958
1959
1960
1961
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1962
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1963
1964
1965
1966
1967
1968
        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)
1969
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1970
1971
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1972
        ptr_data, type_ptr_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
1973
1974
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1975
            ctypes.c_int(type_ptr_data),
1976
1977
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
1978
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
1979
            _c_str(params_str),
wxchan's avatar
wxchan committed
1980
1981
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1982
        return self
wxchan's avatar
wxchan committed
1983

1984
1985
1986
1987
1988
1989
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1990
        """Initialize data from a list of 2-D numpy matrices."""
1991
        ncol = mats[0].shape[1]
1992
        nrow = np.empty((len(mats),), np.int32)
1993
        ptr_data: _ctypes_float_array
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
        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)
2013
            else:  # change non-float data to float data, need to copy
2014
2015
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2016
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2017
2018
2019
2020
2021
2022
2023
2024
            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(
2025
            ctypes.c_int32(len(mats)),
2026
2027
2028
            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)),
2029
            ctypes.c_int32(ncol),
2030
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2031
            _c_str(params_str),
2032
2033
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
2034
        return self
2035

2036
2037
2038
2039
2040
2041
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2042
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2043
        if len(csr.indices) != len(csr.data):
2044
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
wxchan's avatar
wxchan committed
2045
2046
        self.handle = ctypes.c_void_p()

2047
2048
        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
2049

2050
        assert csr.shape[1] <= _MAX_INT32
2051
        csr_indices = csr.indices.astype(np.int32, copy=False)
2052

wxchan's avatar
wxchan committed
2053
2054
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2055
            ctypes.c_int(type_ptr_indptr),
2056
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
2057
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2058
2059
2060
2061
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
2062
            _c_str(params_str),
wxchan's avatar
wxchan committed
2063
2064
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
2065
        return self
wxchan's avatar
wxchan committed
2066

2067
2068
2069
2070
2071
2072
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2073
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2074
        if len(csc.indices) != len(csc.data):
2075
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
Guolin Ke's avatar
Guolin Ke committed
2076
2077
        self.handle = ctypes.c_void_p()

2078
2079
        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
2080

2081
        assert csc.shape[0] <= _MAX_INT32
2082
        csc_indices = csc.indices.astype(np.int32, copy=False)
2083

Guolin Ke's avatar
Guolin Ke committed
2084
2085
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2086
            ctypes.c_int(type_ptr_indptr),
2087
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
2088
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2089
2090
2091
2092
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
2093
            _c_str(params_str),
Guolin Ke's avatar
Guolin Ke committed
2094
2095
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
2096
        return self
Guolin Ke's avatar
Guolin Ke committed
2097

2098
    @staticmethod
2099
2100
2101
2102
2103
2104
    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.
2105

2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
        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.
2116
2117
2118

        Returns
        -------
2119
2120
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
        """
        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

2136
    def construct(self) -> "Dataset":
2137
2138
2139
2140
2141
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2142
            Constructed Dataset object.
2143
        """
2144
        if self.handle is None:
wxchan's avatar
wxchan committed
2145
            if self.reference is not None:
2146
                reference_params = self.reference.get_params()
2147
2148
                params = self.get_params()
                if params != reference_params:
2149
2150
2151
2152
2153
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
2154
                        _log_warning('Overriding the parameters from Reference Dataset.')
2155
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2156
                if self.used_indices is None:
2157
                    # create valid
2158
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
2159
2160
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
2161
                                    feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
2162
                else:
2163
                    # construct subset
2164
                    used_indices = _list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
2165
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2166
                    if self.reference.group is not None:
2167
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2168
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
2169
                                                  return_counts=True)
2170
                    self.handle = ctypes.c_void_p()
2171
                    params_str = _param_dict_to_str(self.params)
wxchan's avatar
wxchan committed
2172
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
2173
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
2174
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
2175
                        ctypes.c_int32(used_indices.shape[0]),
2176
                        _c_str(params_str),
wxchan's avatar
wxchan committed
2177
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2178
2179
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2180
2181
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
2182
2183
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
2184
2185
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
2186
2187
2188
2189
2190
                        self._set_init_score_by_predictor(
                            predictor=self._predictor,
                            data=self.data,
                            used_indices=used_indices
                        )
wxchan's avatar
wxchan committed
2191
            else:
2192
                # create train
2193
                self._lazy_init(self.data, label=self.label,
2194
2195
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
2196
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
2197
2198
            if self.free_raw_data:
                self.data = None
2199
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2200
        return self
wxchan's avatar
wxchan committed
2201

2202
2203
    def create_valid(
        self,
2204
        data: _LGBM_TrainDataType,
2205
        label: Optional[_LGBM_LabelType] = None,
2206
2207
2208
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2209
2210
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2211
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2212
2213
2214

        Parameters
        ----------
2215
        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
2216
            Data source of Dataset.
2217
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2218
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
2219
2220
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
2221
            Weight for each instance. Weights should be non-negative.
2222
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
2223
2224
2225
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2226
2227
            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.
2228
        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)
2229
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2230
        params : dict or None, optional (default=None)
2231
            Other parameters for validation Dataset.
2232
2233
2234

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2235
2236
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2237
        """
2238
        ret = Dataset(data, label=label, reference=self,
2239
                      weight=weight, group=group, init_score=init_score,
2240
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2241
        ret._predictor = self._predictor
2242
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2243
        return ret
wxchan's avatar
wxchan committed
2244

2245
2246
2247
2248
2249
    def subset(
        self,
        used_indices: List[int],
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2250
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2251
2252
2253
2254

        Parameters
        ----------
        used_indices : list of int
2255
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2256
        params : dict or None, optional (default=None)
2257
            These parameters will be passed to Dataset constructor.
2258
2259
2260
2261
2262

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2263
        """
wxchan's avatar
wxchan committed
2264
2265
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
2266
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
2267
2268
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2269
        ret._predictor = self._predictor
2270
        ret.pandas_categorical = self.pandas_categorical
2271
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2272
2273
        return ret

2274
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2275
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2276

2277
2278
2279
2280
2281
        .. 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
2282
2283
        Parameters
        ----------
2284
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2285
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2286
2287
2288
2289
2290

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2291
2292
2293
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
2294
            _c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
2295
        return self
wxchan's avatar
wxchan committed
2296

2297
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2298
2299
        if not params:
            return self
2300
        params = deepcopy(params)
2301
2302
2303
2304
2305

        def update():
            if not self.params:
                self.params = params
            else:
2306
                self._params_back_up = deepcopy(self.params)
2307
2308
2309
2310
2311
2312
                self.params.update(params)

        if self.handle is None:
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2313
2314
                _c_str(_param_dict_to_str(self.params)),
                _c_str(_param_dict_to_str(params)))
2315
2316
2317
2318
2319
2320
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2321
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
2322
        return self
wxchan's avatar
wxchan committed
2323

2324
    def _reverse_update_params(self) -> "Dataset":
2325
        if self.handle is None:
2326
2327
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2328
        return self
2329

2330
2331
2332
2333
2334
    def set_field(
        self,
        field_name: str,
        data
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2335
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2336
2337
2338

        Parameters
        ----------
2339
        field_name : str
2340
            The field name of the information.
2341
2342
        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
2343
2344
2345
2346
2347

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2348
        """
2349
        if self.handle is None:
2350
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2351
        if data is None:
2352
            # set to None
wxchan's avatar
wxchan committed
2353
2354
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
2355
                _c_str(field_name),
wxchan's avatar
wxchan committed
2356
                None,
Guolin Ke's avatar
Guolin Ke committed
2357
                ctypes.c_int(0),
2358
                ctypes.c_int(_FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2359
            return self
2360
        dtype: "np.typing.DTypeLike"
2361
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2362
            dtype = np.float64
2363
            if _is_1d_collection(data):
2364
                data = _list_to_1d_numpy(data, dtype, name=field_name)
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
            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
2375
            data = _list_to_1d_numpy(data, dtype, name=field_name)
2376

2377
        if data.dtype == np.float32 or data.dtype == np.float64:
2378
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2379
        elif data.dtype == np.int32:
2380
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2381
        else:
2382
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2383
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2384
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2385
2386
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
2387
            _c_str(field_name),
wxchan's avatar
wxchan committed
2388
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2389
2390
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2391
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2392
        return self
wxchan's avatar
wxchan committed
2393

2394
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2395
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2396
2397
2398

        Parameters
        ----------
2399
        field_name : str
2400
            The field name of the information.
wxchan's avatar
wxchan committed
2401
2402
2403

        Returns
        -------
2404
        info : numpy array or None
2405
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2406
        """
2407
        if self.handle is None:
2408
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2409
2410
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2411
2412
2413
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
            self.handle,
2414
            _c_str(field_name),
wxchan's avatar
wxchan committed
2415
2416
2417
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
2418
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2419
2420
2421
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2422
        if out_type.value == _C_API_DTYPE_INT32:
2423
            arr = _cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
2424
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2425
            arr = _cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
2426
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2427
            arr = _cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2428
        else:
wxchan's avatar
wxchan committed
2429
            raise TypeError("Unknown type")
2430
2431
2432
2433
2434
2435
        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
2436

2437
2438
    def set_categorical_feature(
        self,
2439
        categorical_feature: _LGBM_CategoricalFeatureConfiguration
2440
    ) -> "Dataset":
2441
        """Set categorical features.
2442
2443
2444

        Parameters
        ----------
2445
        categorical_feature : list of str or int, or 'auto'
2446
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2447
2448
2449
2450
2451

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2452
2453
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2454
            return self
2455
        if self.data is not None:
2456
2457
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2458
                return self._free_handle()
2459
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2460
                return self
2461
            else:
2462
2463
2464
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
                                 f'New categorical_feature is {sorted(list(categorical_feature))}')
2465
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2466
                return self._free_handle()
2467
        else:
2468
2469
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2470

2471
2472
2473
2474
    def _set_predictor(
        self,
        predictor: Optional[_InnerPredictor]
    ) -> "Dataset":
2475
2476
2477
2478
        """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
2479
        """
2480
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2481
            return self
2482
2483
2484
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
            if (predictor == self._predictor) and (predictor.current_iteration() == self._predictor.current_iteration()):
                return self
2485
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
2486
            self._predictor = predictor
2487
2488
        elif self.data is not None:
            self._predictor = predictor
2489
2490
2491
2492
2493
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
                used_indices=None
            )
2494
2495
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2496
2497
2498
2499
2500
            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
2501
        else:
2502
2503
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2504
        return self
Guolin Ke's avatar
Guolin Ke committed
2505

2506
    def set_reference(self, reference: "Dataset") -> "Dataset":
2507
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2508
2509
2510
2511

        Parameters
        ----------
        reference : Dataset
2512
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2513
2514
2515
2516
2517

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2518
        """
2519
2520
2521
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2522
        # we're done if self and reference share a common upstream reference
2523
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2524
            return self
Guolin Ke's avatar
Guolin Ke committed
2525
2526
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2527
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2528
        else:
2529
2530
            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
2531

2532
    def set_feature_name(self, feature_name: Union[List[str], str]) -> "Dataset":
2533
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2534
2535
2536

        Parameters
        ----------
2537
        feature_name : list of str
2538
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2539
2540
2541
2542
2543

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2544
        """
2545
2546
        if feature_name != 'auto':
            self.feature_name = feature_name
2547
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2548
            if len(feature_name) != self.num_feature():
2549
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2550
            c_feature_name = [_c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2551
2552
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
2553
                _c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2554
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2555
        return self
Guolin Ke's avatar
Guolin Ke committed
2556

2557
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
2558
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2559
2560
2561

        Parameters
        ----------
2562
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2563
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2564
2565
2566
2567
2568

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2569
2570
        """
        self.label = label
2571
        if self.handle is not None:
2572
2573
2574
2575
            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)
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
                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)
2587
            else:
2588
                label_array = _list_to_1d_numpy(label, name='label')
2589
            self.set_field('label', label_array)
2590
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2591
        return self
Guolin Ke's avatar
Guolin Ke committed
2592

2593
2594
2595
2596
    def set_weight(
        self,
        weight: Optional[_LGBM_WeightType]
    ) -> "Dataset":
2597
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2598
2599
2600

        Parameters
        ----------
2601
        weight : list, numpy 1-D array, pandas Series or None
2602
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2603
2604
2605
2606
2607

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2608
        """
2609
2610
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2611
        self.weight = weight
2612
        if self.handle is not None and weight is not None:
2613
            weight = _list_to_1d_numpy(weight, name='weight')
wxchan's avatar
wxchan committed
2614
            self.set_field('weight', weight)
2615
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2616
        return self
Guolin Ke's avatar
Guolin Ke committed
2617

2618
2619
2620
2621
    def set_init_score(
        self,
        init_score: Optional[_LGBM_InitScoreType]
    ) -> "Dataset":
2622
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2623
2624
2625

        Parameters
        ----------
2626
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2627
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2628
2629
2630
2631
2632

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2633
2634
        """
        self.init_score = init_score
2635
        if self.handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2636
            self.set_field('init_score', init_score)
2637
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2638
        return self
Guolin Ke's avatar
Guolin Ke committed
2639

2640
2641
2642
2643
    def set_group(
        self,
        group: Optional[_LGBM_GroupType]
    ) -> "Dataset":
2644
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2645
2646
2647

        Parameters
        ----------
2648
        group : list, numpy 1-D array, pandas Series or None
2649
2650
2651
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2652
2653
            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
2654
2655
2656
2657
2658

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2659
2660
        """
        self.group = group
2661
        if self.handle is not None and group is not None:
2662
            group = _list_to_1d_numpy(group, np.int32, name='group')
wxchan's avatar
wxchan committed
2663
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2664
        return self
Guolin Ke's avatar
Guolin Ke committed
2665

2666
    def get_feature_name(self) -> List[str]:
2667
2668
2669
2670
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2671
        feature_names : list of str
2672
2673
2674
2675
2676
2677
2678
2679
            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)
2680
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2681
2682
2683
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
2684
            ctypes.c_int(num_feature),
2685
            ctypes.byref(tmp_out_len),
2686
            ctypes.c_size_t(reserved_string_buffer_size),
2687
2688
2689
2690
            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")
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
        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))
2703
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2704

2705
    def get_label(self) -> Optional[np.ndarray]:
2706
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2707
2708
2709

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2710
        label : numpy array or None
2711
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2712
        """
2713
        if self.label is None:
wxchan's avatar
wxchan committed
2714
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2715
2716
        return self.label

2717
    def get_weight(self) -> Optional[np.ndarray]:
2718
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2719
2720
2721

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2722
        weight : numpy array or None
2723
            Weight for each data point from the Dataset. Weights should be non-negative.
Guolin Ke's avatar
Guolin Ke committed
2724
        """
2725
        if self.weight is None:
wxchan's avatar
wxchan committed
2726
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2727
2728
        return self.weight

2729
    def get_init_score(self) -> Optional[np.ndarray]:
2730
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2731
2732
2733

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2734
        init_score : numpy array or None
2735
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2736
        """
2737
        if self.init_score is None:
wxchan's avatar
wxchan committed
2738
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2739
2740
        return self.init_score

2741
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
2742
2743
2744
2745
        """Get the raw data of the Dataset.

        Returns
        -------
2746
        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
2747
2748
2749
2750
            Raw data used in the Dataset construction.
        """
        if self.handle is None:
            raise Exception("Cannot get data before construct Dataset")
2751
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
2752
2753
            self.data = self.reference.data
            if self.data is not None:
2754
                if isinstance(self.data, np.ndarray) or isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2755
                    self.data = self.data[self.used_indices, :]
2756
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2757
                    self.data = self.data.iloc[self.used_indices].copy()
2758
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2759
                    self.data = self.data[self.used_indices, :]
2760
2761
2762
2763
                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
2764
                else:
2765
2766
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
2767
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
2768
2769
2770
        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.")
2771
2772
        return self.data

2773
    def get_group(self) -> Optional[np.ndarray]:
2774
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2775
2776
2777

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2778
        group : numpy array or None
2779
2780
2781
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2782
2783
            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
2784
        """
2785
        if self.group is None:
wxchan's avatar
wxchan committed
2786
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
2787
2788
            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
2789
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
2790
2791
        return self.group

2792
    def num_data(self) -> int:
2793
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2794
2795
2796

        Returns
        -------
2797
2798
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2799
        """
2800
        if self.handle is not None:
2801
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2802
2803
2804
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2805
        else:
2806
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2807

2808
    def num_feature(self) -> int:
2809
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2810
2811
2812

        Returns
        -------
2813
2814
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2815
        """
2816
        if self.handle is not None:
2817
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2818
2819
2820
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2821
        else:
2822
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2823

2824
    def feature_num_bin(self, feature: Union[int, str]) -> int:
2825
2826
2827
2828
        """Get the number of bins for a feature.

        Parameters
        ----------
2829
2830
        feature : int or str
            Index or name of the feature.
2831
2832
2833
2834
2835
2836
2837

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
        if self.handle is not None:
2838
            if isinstance(feature, str):
2839
2840
2841
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
2842
2843
            ret = ctypes.c_int(0)
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self.handle,
2844
                                                         ctypes.c_int(feature_index),
2845
2846
2847
2848
2849
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

2850
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
2851
2852
2853
2854
2855
        """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.
2856
2857
2858
2859
2860

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2861
2862
2863

        Returns
        -------
2864
2865
2866
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2867
        head = self
2868
        ref_chain: Set[Dataset] = set()
2869
2870
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2871
                ref_chain.add(head)
2872
2873
2874
2875
2876
2877
                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
2878
        return ref_chain
2879

2880
    def add_features_from(self, other: "Dataset") -> "Dataset":
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
        """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
2898
2899
2900
2901
2902
2903
2904
2905
        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))
2906
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2907
                    self.data = np.hstack((self.data, other.data.toarray()))
2908
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2909
                    self.data = np.hstack((self.data, other.data.values))
2910
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2911
2912
2913
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
2914
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2915
                sparse_format = self.data.getformat()
2916
                if isinstance(other.data, np.ndarray) or isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2917
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
2918
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2919
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
2920
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2921
2922
2923
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
2924
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2925
2926
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
2927
2928
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
2929
                if isinstance(other.data, np.ndarray):
2930
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
2931
                                       axis=1, ignore_index=True)
2932
                elif isinstance(other.data, scipy.sparse.spmatrix):
2933
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
2934
                                       axis=1, ignore_index=True)
2935
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2936
2937
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
2938
2939
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
2940
2941
2942
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
2943
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2944
                if isinstance(other.data, np.ndarray):
2945
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
2946
                elif isinstance(other.data, scipy.sparse.spmatrix):
2947
2948
2949
2950
2951
                    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
2952
2953
2954
2955
2956
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
2957
2958
            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
2959
2960
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
2961
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
2962
        self.feature_name = self.get_feature_name()
2963
2964
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
2965
2966
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2967
2968
        return self

2969
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
2970
2971
2972
2973
2974
2975
        """Save Dataset to a text file.

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

        Parameters
        ----------
2976
        filename : str or pathlib.Path
2977
2978
2979
2980
2981
2982
2983
2984
2985
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
            self.construct().handle,
2986
            _c_str(str(filename))))
2987
2988
        return self

wxchan's avatar
wxchan committed
2989

2990
2991
2992
2993
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
    Tuple[np.ndarray, np.ndarray]
]
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
        _LGBM_EvalFunctionResultType
    ],
    Callable[
        [np.ndarray, Dataset],
        List[_LGBM_EvalFunctionResultType]
    ]
]
3004
3005


3006
class Booster:
3007
    """Booster in LightGBM."""
3008

3009
3010
3011
3012
3013
3014
3015
    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
    ):
3016
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3017
3018
3019

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3020
        params : dict or None, optional (default=None)
3021
3022
3023
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3024
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3025
            Path to the model file.
3026
        model_str : str or None, optional (default=None)
3027
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3028
        """
3029
        self.handle = None
3030
        self._network = False
wxchan's avatar
wxchan committed
3031
        self.__need_reload_eval_info = True
3032
        self._train_data_name = "training"
3033
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3034
        self.best_iteration = -1
3035
        self.best_score: _LGBM_BoosterBestScoreType = {}
3036
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3037
        if train_set is not None:
3038
            # Training task
wxchan's avatar
wxchan committed
3039
            if not isinstance(train_set, Dataset):
3040
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
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
3074
            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"]
                )
3075
            # construct booster object
3076
3077
3078
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3079
            params_str = _param_dict_to_str(params)
3080
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
3081
            _safe_call(_LIB.LGBM_BoosterCreate(
3082
                train_set.handle,
3083
                _c_str(params_str),
wxchan's avatar
wxchan committed
3084
                ctypes.byref(self.handle)))
3085
            # save reference to data
wxchan's avatar
wxchan committed
3086
            self.train_set = train_set
3087
3088
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3089
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3090
3091
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
3092
3093
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
3094
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
3095
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3096
3097
3098
3099
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3100
            # buffer for inner predict
wxchan's avatar
wxchan committed
3101
3102
3103
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3104
            self.pandas_categorical = train_set.pandas_categorical
3105
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3106
        elif model_file is not None:
3107
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3108
            out_num_iterations = ctypes.c_int(0)
3109
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
3110
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
3111
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
3112
3113
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
3114
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3115
3116
3117
3118
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3119
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3120
3121
3122
            if params:
                _log_warning('Ignoring params argument, using parameters from model file.')
            params = self._get_loaded_param()
3123
        elif model_str is not None:
3124
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3125
        else:
3126
3127
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
3128
        self.params = params
wxchan's avatar
wxchan committed
3129

3130
    def __del__(self) -> None:
3131
        try:
3132
            if self._network:
3133
3134
3135
3136
3137
3138
3139
3140
                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
3141

3142
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3143
3144
        return self.__deepcopy__(None)

3145
    def __deepcopy__(self, _) -> "Booster":
3146
        model_str = self.model_to_string(num_iteration=-1)
3147
        booster = Booster(model_str=model_str)
3148
        return booster
wxchan's avatar
wxchan committed
3149

3150
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3151
3152
3153
3154
3155
        this = self.__dict__.copy()
        handle = this['handle']
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
3156
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3157
3158
        return this

3159
    def __setstate__(self, state: Dict[str, Any]) -> None:
3160
3161
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
3162
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3163
            out_num_iterations = ctypes.c_int(0)
3164
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3165
                _c_str(model_str),
3166
3167
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
3168
3169
3170
            state['handle'] = handle
        self.__dict__.update(state)

3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
    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'))

3193
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3194
3195
3196
3197
3198
3199
3200
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
3201
3202
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
3203
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3204
        return self
wxchan's avatar
wxchan committed
3205

3206
    def _free_buffer(self) -> "Booster":
3207
3208
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3209
        return self
3210

3211
3212
3213
3214
3215
3216
3217
    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":
3218
3219
3220
3221
        """Set the network configuration.

        Parameters
        ----------
3222
        machines : list, set or str
3223
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3224
        local_listen_port : int, optional (default=12400)
3225
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3226
        listen_time_out : int, optional (default=120)
3227
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3228
        num_machines : int, optional (default=1)
3229
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3230
3231
3232
3233
3234

        Returns
        -------
        self : Booster
            Booster with set network.
3235
        """
3236
3237
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
3238
        _safe_call(_LIB.LGBM_NetworkInit(_c_str(machines),
3239
3240
3241
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
3242
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3243
        return self
3244

3245
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3246
3247
3248
3249
3250
3251
3252
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3253
        _safe_call(_LIB.LGBM_NetworkFree())
3254
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3255
        return self
3256

3257
    def trees_to_dataframe(self) -> pd_DataFrame:
3258
3259
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3260
3261
3262
3263
        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.
3264
3265
3266
3267
3268
            - ``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.
3269
3270
            - ``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.
3271
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3272
3273
              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.
3274
3275
            - ``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.
3276
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3277
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3278
3279
            - ``count`` : int64, number of records in the training data that fall into this node.

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

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

3292
        def _is_split_node(tree: Dict[str, Any]) -> bool:
3293
3294
            return 'split_index' in tree.keys()

3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
        def create_node_record(
            tree: Dict[str, Any],
            node_depth: int = 1,
            tree_index: Optional[int] = None,
            feature_names: Optional[List[str]] = None,
            parent_node: Optional[str] = None
        ) -> Dict[str, Any]:

            def _get_node_index(
                tree: Dict[str, Any],
                tree_index: Optional[int]
            ) -> str:
3307
                tree_num = f'{tree_index}-' if tree_index is not None else ''
3308
3309
3310
                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
3311
3312
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
3313

3314
3315
3316
3317
            def _get_split_feature(
                tree: Dict[str, Any],
                feature_names: Optional[List[str]]
            ) -> Optional[str]:
3318
3319
3320
3321
3322
3323
3324
3325
3326
                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

3327
            def _is_single_node_tree(tree: Dict[str, Any]) -> bool:
3328
                return set(tree.keys()) == {'leaf_value'}
3329
3330

            # Create the node record, and populate universal data members
3331
            node: Dict[str, Union[int, str, None]] = OrderedDict()
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
            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

3368
3369
3370
3371
3372
3373
3374
        def tree_dict_to_node_list(
            tree: Dict[str, Any],
            node_depth: int = 1,
            tree_index: Optional[int] = None,
            feature_names: Optional[List[str]] = None,
            parent_node: Optional[str] = None
        ) -> List[Dict[str, Any]]:
3375

3376
            node = create_node_record(tree=tree,
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
                                      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(
3389
                        tree=tree[child],
3390
3391
3392
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
3393
3394
                        parent_node=node['node_index']
                    )
3395
3396
3397
3398
3399
3400
3401
3402
3403
                    # 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']:
3404
            model_list.extend(tree_dict_to_node_list(tree=tree['tree_structure'],
3405
3406
3407
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

3408
        return pd_DataFrame(model_list, columns=model_list[0].keys())
3409

3410
    def set_train_data_name(self, name: str) -> "Booster":
3411
3412
3413
3414
        """Set the name to the training Dataset.

        Parameters
        ----------
3415
        name : str
Nikita Titov's avatar
Nikita Titov committed
3416
3417
3418
3419
3420
3421
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
3422
        """
3423
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
3424
        return self
wxchan's avatar
wxchan committed
3425

3426
    def add_valid(self, data: Dataset, name: str) -> "Booster":
3427
        """Add validation data.
wxchan's avatar
wxchan committed
3428
3429
3430
3431

        Parameters
        ----------
        data : Dataset
3432
            Validation data.
3433
        name : str
3434
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
3435
3436
3437
3438
3439

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
3440
        """
Guolin Ke's avatar
Guolin Ke committed
3441
        if not isinstance(data, Dataset):
3442
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3443
        if data._predictor is not self.__init_predictor:
3444
3445
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3446
3447
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
3448
            data.construct().handle))
wxchan's avatar
wxchan committed
3449
3450
3451
3452
3453
        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
3454
        return self
wxchan's avatar
wxchan committed
3455

3456
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
3457
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
3458
3459
3460
3461

        Parameters
        ----------
        params : dict
3462
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
3463
3464
3465
3466
3467

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
3468
        """
3469
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3470
3471
3472
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
                self.handle,
3473
                _c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
3474
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
3475
        return self
wxchan's avatar
wxchan committed
3476

3477
3478
3479
3480
3481
    def update(
        self,
        train_set: Optional[Dataset] = None,
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
3482
        """Update Booster for one iteration.
3483

wxchan's avatar
wxchan committed
3484
3485
        Parameters
        ----------
3486
3487
3488
3489
        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
3490
            Customized objective function.
3491
3492
3493
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3494
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3495
                    The predicted values.
3496
3497
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3498
3499
                train_data : Dataset
                    The training dataset.
3500
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3501
3502
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3503
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3504
3505
                    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
3506

3507
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3508
            and grad and hess should be returned in the same format.
3509

wxchan's avatar
wxchan committed
3510
3511
        Returns
        -------
3512
3513
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3514
        """
3515
        # need reset training data
3516
3517
3518
3519
3520
3521
        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
3522
            if not isinstance(train_set, Dataset):
3523
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3524
            if train_set._predictor is not self.__init_predictor:
3525
3526
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3527
3528
3529
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
3530
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
3531
            self.__inner_predict_buffer[0] = None
3532
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3533
3534
        is_finished = ctypes.c_int(0)
        if fobj is None:
3535
            if self.__set_objective_to_none:
3536
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3537
3538
3539
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
3540
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3541
3542
            return is_finished.value == 1
        else:
3543
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3544
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3545
3546
3547
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

3548
3549
3550
3551
3552
    def __boost(
        self,
        grad: np.ndarray,
        hess: np.ndarray
    ) -> bool:
3553
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3554

Nikita Titov's avatar
Nikita Titov committed
3555
3556
        .. note::

3557
3558
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3559
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3560
            and grad and hess should be returned in the same format.
3561

wxchan's avatar
wxchan committed
3562
3563
        Parameters
        ----------
3564
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3565
3566
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3567
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3568
3569
            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
3570
3571
3572

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3573
3574
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3575
        """
3576
3577
3578
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3579
3580
        grad = _list_to_1d_numpy(grad, name='gradient')
        hess = _list_to_1d_numpy(hess, name='hessian')
3581
3582
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3583
        if len(grad) != len(hess):
3584
3585
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3586
        if len(grad) != num_train_data * self.__num_class:
3587
3588
3589
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3590
                f"number of models per one iteration ({self.__num_class})"
3591
            )
wxchan's avatar
wxchan committed
3592
3593
3594
3595
3596
3597
        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)))
3598
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3599
3600
        return is_finished.value == 1

3601
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3602
3603
3604
3605
3606
3607
3608
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3609
3610
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
3611
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3612
        return self
wxchan's avatar
wxchan committed
3613

3614
    def current_iteration(self) -> int:
3615
3616
3617
3618
3619
3620
3621
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3622
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3623
3624
3625
3626
3627
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3628
    def num_model_per_iteration(self) -> int:
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
        """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

3642
    def num_trees(self) -> int:
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
        """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

3656
    def upper_bound(self) -> float:
3657
3658
3659
3660
        """Get upper bound value of a model.

        Returns
        -------
3661
        upper_bound : float
3662
3663
3664
3665
3666
3667
3668
3669
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

3670
    def lower_bound(self) -> float:
3671
3672
3673
3674
        """Get lower bound value of a model.

        Returns
        -------
3675
        lower_bound : float
3676
3677
3678
3679
3680
3681
3682
3683
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

3684
3685
3686
3687
3688
3689
    def eval(
        self,
        data: Dataset,
        name: str,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3690
        """Evaluate for data.
wxchan's avatar
wxchan committed
3691
3692
3693

        Parameters
        ----------
3694
3695
        data : Dataset
            Data for the evaluating.
3696
        name : str
3697
            Name of the data.
3698
        feval : callable, list of callable, or None, optional (default=None)
3699
            Customized evaluation function.
3700
            Each evaluation function should accept two parameters: preds, eval_data,
3701
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3702

3703
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3704
                    The predicted values.
3705
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3706
                    If custom objective function is used, predicted values are returned before any transformation,
3707
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3708
                eval_data : Dataset
3709
                    A ``Dataset`` to evaluate.
3710
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3711
                    The name of evaluation function (without whitespace).
3712
3713
3714
3715
3716
                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
3717
3718
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3719
        result : list
3720
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3721
        """
Guolin Ke's avatar
Guolin Ke committed
3722
3723
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3724
3725
3726
3727
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3728
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3729
3730
3731
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3732
        # need to push new valid data
wxchan's avatar
wxchan committed
3733
3734
3735
3736
3737
3738
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

3739
3740
3741
3742
    def eval_train(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3743
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3744
3745
3746

        Parameters
        ----------
3747
        feval : callable, list of callable, or None, optional (default=None)
3748
            Customized evaluation function.
3749
            Each evaluation function should accept two parameters: preds, eval_data,
3750
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3751

3752
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3753
                    The predicted values.
3754
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3755
                    If custom objective function is used, predicted values are returned before any transformation,
3756
                    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
3757
                eval_data : Dataset
3758
                    The training dataset.
3759
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3760
                    The name of evaluation function (without whitespace).
3761
3762
3763
3764
3765
                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
3766
3767
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3768
        result : list
3769
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3770
        """
3771
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
3772

3773
3774
3775
3776
    def eval_valid(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3777
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
3778
3779
3780

        Parameters
        ----------
3781
        feval : callable, list of callable, or None, optional (default=None)
3782
            Customized evaluation function.
3783
            Each evaluation function should accept two parameters: preds, eval_data,
3784
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3785

3786
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3787
                    The predicted values.
3788
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3789
                    If custom objective function is used, predicted values are returned before any transformation,
3790
                    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
3791
                eval_data : Dataset
3792
                    The validation dataset.
3793
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3794
                    The name of evaluation function (without whitespace).
3795
3796
3797
3798
3799
                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
3800
3801
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3802
        result : list
3803
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3804
        """
3805
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
3806
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
3807

3808
3809
3810
3811
3812
3813
3814
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> "Booster":
3815
        """Save Booster to file.
wxchan's avatar
wxchan committed
3816
3817
3818

        Parameters
        ----------
3819
        filename : str or pathlib.Path
3820
            Filename to save Booster.
3821
3822
3823
3824
        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
3825
        start_iteration : int, optional (default=0)
3826
            Start index of the iteration that should be saved.
3827
        importance_type : str, optional (default="split")
3828
3829
3830
            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
3831
3832
3833
3834
3835

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
3836
        """
3837
        if num_iteration is None:
3838
            num_iteration = self.best_iteration
3839
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3840
3841
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
3842
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3843
            ctypes.c_int(num_iteration),
3844
            ctypes.c_int(importance_type_int),
3845
            _c_str(str(filename))))
3846
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
3847
        return self
wxchan's avatar
wxchan committed
3848

3849
3850
3851
3852
3853
    def shuffle_models(
        self,
        start_iteration: int = 0,
        end_iteration: int = -1
    ) -> "Booster":
3854
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
3855

3856
3857
3858
        Parameters
        ----------
        start_iteration : int, optional (default=0)
3859
            The first iteration that will be shuffled.
3860
3861
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
3862
            If <= 0, means the last available iteration.
3863

Nikita Titov's avatar
Nikita Titov committed
3864
3865
3866
3867
        Returns
        -------
        self : Booster
            Booster with shuffled models.
3868
        """
3869
3870
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
3871
3872
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
3873
        return self
3874

3875
    def model_from_string(self, model_str: str) -> "Booster":
3876
3877
3878
3879
        """Load Booster from a string.

        Parameters
        ----------
3880
        model_str : str
3881
3882
3883
3884
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3885
        self : Booster
3886
3887
            Loaded Booster object.
        """
3888
3889
3890
3891
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
3892
3893
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3894
            _c_str(model_str),
3895
3896
3897
3898
3899
3900
3901
            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
3902
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
3903
3904
        return self

3905
3906
3907
3908
3909
3910
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> str:
3911
        """Save Booster to string.
3912

3913
3914
3915
3916
3917
3918
        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
3919
        start_iteration : int, optional (default=0)
3920
            Start index of the iteration that should be saved.
3921
        importance_type : str, optional (default="split")
3922
3923
3924
            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.
3925
3926
3927

        Returns
        -------
3928
        str_repr : str
3929
3930
            String representation of Booster.
        """
3931
        if num_iteration is None:
3932
            num_iteration = self.best_iteration
3933
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3934
        buffer_len = 1 << 20
3935
        tmp_out_len = ctypes.c_int64(0)
3936
3937
3938
3939
        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,
3940
            ctypes.c_int(start_iteration),
3941
            ctypes.c_int(num_iteration),
3942
            ctypes.c_int(importance_type_int),
3943
            ctypes.c_int64(buffer_len),
3944
3945
3946
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
3947
        # if buffer length is not long enough, re-allocate a buffer
3948
3949
3950
3951
3952
        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,
3953
                ctypes.c_int(start_iteration),
3954
                ctypes.c_int(num_iteration),
3955
                ctypes.c_int(importance_type_int),
3956
                ctypes.c_int64(actual_len),
3957
3958
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
3959
        ret = string_buffer.value.decode('utf-8')
3960
3961
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
3962

3963
3964
3965
3966
3967
3968
3969
    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
3970
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
3971

3972
3973
        Parameters
        ----------
3974
3975
3976
3977
        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
3978
        start_iteration : int, optional (default=0)
3979
            Start index of the iteration that should be dumped.
3980
        importance_type : str, optional (default="split")
3981
3982
3983
            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.
3984
3985
3986
3987
3988
3989
3990
3991
3992
        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.
3993

wxchan's avatar
wxchan committed
3994
3995
        Returns
        -------
3996
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
3997
            JSON format of Booster.
wxchan's avatar
wxchan committed
3998
        """
3999
        if num_iteration is None:
4000
            num_iteration = self.best_iteration
4001
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4002
        buffer_len = 1 << 20
4003
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
4004
4005
4006
4007
        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,
4008
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4009
            ctypes.c_int(num_iteration),
4010
            ctypes.c_int(importance_type_int),
4011
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
4012
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4013
            ptr_string_buffer))
wxchan's avatar
wxchan committed
4014
        actual_len = tmp_out_len.value
4015
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
4016
4017
4018
4019
4020
        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,
4021
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4022
                ctypes.c_int(num_iteration),
4023
                ctypes.c_int(importance_type_int),
4024
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
4025
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4026
                ptr_string_buffer))
4027
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
4028
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
4029
                                                          default=_json_default_with_numpy))
4030
        return ret
wxchan's avatar
wxchan committed
4031

4032
4033
    def predict(
        self,
4034
        data: _LGBM_PredictDataType,
4035
4036
4037
4038
4039
4040
4041
4042
        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
4043
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4044
        """Make a prediction.
wxchan's avatar
wxchan committed
4045
4046
4047

        Parameters
        ----------
4048
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
4049
            Data source for prediction.
4050
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4051
        start_iteration : int, optional (default=0)
4052
            Start index of the iteration to predict.
4053
            If <= 0, starts from the first iteration.
4054
        num_iteration : int or None, optional (default=None)
4055
4056
4057
4058
            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).
4059
4060
4061
4062
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4063
4064
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4065

Nikita Titov's avatar
Nikita Titov committed
4066
4067
4068
4069
4070
4071
4072
            .. 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.
4073

4074
4075
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4076
            Used only if data is str.
4077
4078
4079
        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.
4080
4081
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4082
4083
4084

        Returns
        -------
4085
        result : numpy array, scipy.sparse or list of scipy.sparse
4086
            Prediction result.
4087
            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
4088
        """
4089
        predictor = self._to_predictor(deepcopy(kwargs))
4090
        if num_iteration is None:
4091
            if start_iteration <= 0:
4092
4093
4094
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
        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
4105

4106
4107
    def refit(
        self,
4108
        data: _LGBM_TrainDataType,
4109
        label: _LGBM_LabelType,
4110
4111
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4112
4113
4114
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4115
4116
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
4117
4118
4119
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4120
        **kwargs
4121
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4122
4123
4124
4125
        """Refit the existing Booster by new data.

        Parameters
        ----------
4126
        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
4127
            Data source for refit.
4128
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4129
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
4130
4131
            Label for refit.
        decay_rate : float, optional (default=0.9)
4132
4133
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4134
4135
4136
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
4137
            Weight for each ``data`` instance. Weights should be non-negative.
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
        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.
4154
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4155
4156
4157
            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.
4158
            Floating point numbers in categorical features will be rounded towards 0.
4159
4160
4161
4162
        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``.
4163
4164
4165
        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.
4166
4167
        **kwargs
            Other parameters for refit.
4168
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4169
4170
4171
4172
4173
4174

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4175
4176
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
4177
4178
        if dataset_params is None:
            dataset_params = {}
4179
        predictor = self._to_predictor(deepcopy(kwargs))
4180
4181
4182
4183
4184
4185
        leaf_preds = predictor.predict(
            data=data,
            start_iteration=-1,
            pred_leaf=True,
            validate_features=validate_features
        )
4186
        nrow, ncol = leaf_preds.shape
4187
        out_is_linear = ctypes.c_int(0)
4188
4189
4190
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
4191
4192
4193
4194
4195
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
4196
        new_params["linear_tree"] = bool(out_is_linear.value)
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
        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,
        )
4210
        new_params['refit_decay_rate'] = decay_rate
4211
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4212
4213
4214
4215
4216
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
4217
        ptr_data, _, _ = _c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
4218
4219
4220
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
4221
4222
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
4223
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4224
4225
        return new_booster

4226
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
        """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.
        """
4241
4242
4243
4244
4245
4246
4247
4248
        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

4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
    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

4281
4282
4283
4284
    def _to_predictor(
        self,
        pred_parameter: Optional[Dict[str, Any]] = None
    ) -> _InnerPredictor:
4285
        """Convert to predictor."""
4286
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
4287
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
4288
4289
        return predictor

4290
    def num_feature(self) -> int:
4291
4292
4293
4294
4295
4296
4297
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4298
4299
4300
4301
4302
4303
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
            self.handle,
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

4304
    def feature_name(self) -> List[str]:
4305
        """Get names of features.
wxchan's avatar
wxchan committed
4306
4307
4308

        Returns
        -------
4309
        result : list of str
4310
            List with names of features.
wxchan's avatar
wxchan committed
4311
        """
4312
        num_feature = self.num_feature()
4313
        # Get name of features
wxchan's avatar
wxchan committed
4314
        tmp_out_len = ctypes.c_int(0)
4315
4316
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
4317
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
4318
4319
4320
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
4321
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
4322
            ctypes.byref(tmp_out_len),
4323
            ctypes.c_size_t(reserved_string_buffer_size),
4324
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4325
4326
4327
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
        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))
4340
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
4341

4342
4343
4344
4345
4346
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
4347
        """Get feature importances.
4348

4349
4350
        Parameters
        ----------
4351
        importance_type : str, optional (default="split")
4352
4353
4354
            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.
4355
4356
4357
4358
        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).
4359

4360
4361
        Returns
        -------
4362
4363
        result : numpy array
            Array with feature importances.
4364
        """
4365
4366
        if iteration is None:
            iteration = self.best_iteration
4367
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4368
        result = np.empty(self.num_feature(), dtype=np.float64)
4369
4370
4371
4372
4373
        _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))))
4374
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
4375
            return result.astype(np.int32)
4376
4377
        else:
            return result
4378

4379
4380
4381
4382
4383
4384
    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]:
4385
4386
4387
4388
        """Get split value histogram for the specified feature.

        Parameters
        ----------
4389
        feature : int or str
4390
4391
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
4392
            If str, interpreted as name.
4393

Nikita Titov's avatar
Nikita Titov committed
4394
4395
4396
            .. warning::

                Categorical features are not supported.
4397

4398
        bins : int, str or None, optional (default=None)
4399
4400
4401
            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.
4402
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
        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.
        """
4417
        def add(root: Dict[str, Any]) -> None:
4418
4419
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
4420
                if feature_names is not None and isinstance(feature, str):
4421
4422
4423
4424
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
4425
                    if isinstance(root['threshold'], str):
4426
4427
4428
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
4429
4430
4431
4432
4433
4434
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
4435
        values: List[float] = []
4436
4437
4438
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

4439
        if bins is None or isinstance(bins, int) and xgboost_style:
4440
4441
4442
4443
4444
4445
4446
            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:
4447
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
4448
4449
4450
4451
4452
            else:
                return ret
        else:
            return hist, bin_edges

4453
4454
4455
4456
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
4457
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]]
4458
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4459
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
4460
        if data_idx >= self.__num_dataset:
4461
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4462
4463
4464
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
4465
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
4466
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
4467
4468
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
4469
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4470
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4471
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
4472
            if tmp_out_len.value != self.__num_inner_eval:
4473
                raise ValueError("Wrong length of eval results")
4474
            for i in range(self.__num_inner_eval):
4475
4476
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
4477
4478
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
4479
4480
4481
4482
4483
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
4484
4485
4486
4487
4488
4489
4490
4491
4492
            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
4493
4494
4495
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

4496
    def __inner_predict(self, data_idx: int) -> np.ndarray:
4497
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
4498
        if data_idx >= self.__num_dataset:
4499
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4500
4501
4502
4503
4504
        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
4505
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
4506
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
4507
4508
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
4509
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
4510
4511
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
4512
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4513
4514
4515
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
4516
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
4517
            self.__is_predicted_cur_iter[data_idx] = True
4518
4519
4520
4521
4522
        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
4523

4524
    def __get_eval_info(self) -> None:
4525
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
4526
4527
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
4528
            out_num_eval = ctypes.c_int(0)
4529
            # Get num of inner evals
wxchan's avatar
wxchan committed
4530
4531
4532
4533
4534
            _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:
4535
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
4536
                tmp_out_len = ctypes.c_int(0)
4537
4538
4539
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
4540
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
4541
                ]
wxchan's avatar
wxchan committed
4542
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
4543
4544
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
4545
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
4546
                    ctypes.byref(tmp_out_len),
4547
                    ctypes.c_size_t(reserved_string_buffer_size),
4548
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4549
4550
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
4551
                    raise ValueError("Length of eval names doesn't equal with num_evals")
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
                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
                ]