basic.py 191 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3
import abc
wxchan's avatar
wxchan committed
4
import ctypes
5
import inspect
6
import json
wxchan's avatar
wxchan committed
7
import warnings
8
from collections import OrderedDict
9
from copy import deepcopy
10
from enum import Enum
11
from functools import wraps
12
from os import SEEK_END, environ
13
14
from os.path import getsize
from pathlib import Path
15
from tempfile import NamedTemporaryFile
16
from typing import TYPE_CHECKING, Any, Callable, Dict, Iterable, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
17
18
19
20

import numpy as np
import scipy.sparse

21
from .compat import (PANDAS_INSTALLED, PYARROW_INSTALLED, arrow_cffi, arrow_is_floating, arrow_is_integer, concat,
22
23
                     dt_DataTable, pa_Array, pa_ChunkedArray, pa_compute, pa_Table, pd_CategoricalDtype, pd_DataFrame,
                     pd_Series)
wxchan's avatar
wxchan committed
24
25
from .libpath import find_lib_path

26
27
28
if TYPE_CHECKING:
    from typing import Literal

29
30
31
32
33
34
35
    # typing.TypeGuard was only introduced in Python 3.10
    try:
        from typing import TypeGuard
    except ImportError:
        from typing_extensions import TypeGuard


36
37
38
39
40
41
42
43
44
__all__ = [
    'Booster',
    'Dataset',
    'LGBMDeprecationWarning',
    'LightGBMError',
    'register_logger',
    'Sequence',
]

45
_BoosterHandle = ctypes.c_void_p
46
_DatasetHandle = ctypes.c_void_p
47
48
49
50
_ctypes_int_ptr = Union[
    "ctypes._Pointer[ctypes.c_int32]",
    "ctypes._Pointer[ctypes.c_int64]"
]
51
52
53
54
_ctypes_int_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_int32]]",
    "ctypes.Array[ctypes._Pointer[ctypes.c_int64]]"
]
55
56
57
58
59
60
61
62
_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]]"
]
63
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
64
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
65
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
66
_LGBM_BoosterEvalMethodResultWithStandardDeviationType = Tuple[str, str, float, bool, float]
67
68
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], "Literal['auto']"]
_LGBM_FeatureNameConfiguration = Union[List[str], "Literal['auto']"]
69
70
71
72
73
74
_LGBM_GroupType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
75
76
77
78
_LGBM_PositionType = Union[
    np.ndarray,
    pd_Series
]
79
80
81
82
83
84
85
_LGBM_InitScoreType = Union[
    List[float],
    List[List[float]],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
]
86
87
88
89
90
91
92
93
94
_LGBM_TrainDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix,
    "Sequence",
    List["Sequence"],
95
96
    List[np.ndarray],
    pa_Table
97
]
98
_LGBM_LabelType = Union[
99
100
    List[float],
    List[int],
101
102
    np.ndarray,
    pd_Series,
103
104
105
    pd_DataFrame,
    pa_Array,
    pa_ChunkedArray,
106
]
107
108
109
110
111
112
113
114
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix
]
115
116
117
118
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
119
120
121
    pd_Series,
    pa_Array,
    pa_ChunkedArray,
122
]
123
124
125
ZERO_THRESHOLD = 1e-35


126
127
128
129
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


130
def _get_sample_count(total_nrow: int, params: str) -> int:
131
132
133
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
134
        _c_str(params),
135
136
137
138
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
139

140
141
142
143
144
145
class _MissingType(Enum):
    NONE = 'None'
    NAN = 'NaN'
    ZERO = 'Zero'


146
class _DummyLogger:
147
    def info(self, msg: str) -> None:
148
        print(msg)  # noqa: T201
149

150
    def warning(self, msg: str) -> None:
151
152
153
        warnings.warn(msg, stacklevel=3)


154
155
156
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
157
158


159
160
161
162
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


163
164
165
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
166
167
168
169
    """Register custom logger.

    Parameters
    ----------
170
    logger : Any
171
        Custom logger.
172
173
174
175
    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.
176
    """
177
178
179
180
181
182
    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
183
    _LOGGER = logger
184
185
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
186
187


188
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
189
    """Join log messages from native library which come by chunks."""
190
    msg_normalized: List[str] = []
191
192

    @wraps(func)
193
    def wrapper(msg: str) -> None:
194
195
196
197
198
199
200
201
202
203
204
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


205
def _log_info(msg: str) -> None:
206
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
207
208


209
def _log_warning(msg: str) -> None:
210
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
211
212
213


@_normalize_native_string
214
def _log_native(msg: str) -> None:
215
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
216
217


218
def _log_callback(msg: bytes) -> None:
219
    """Redirect logs from native library into Python."""
220
    _log_native(str(msg.decode('utf-8')))
221
222


223
def _load_lib() -> ctypes.CDLL:
224
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
225
226
227
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
228
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
229
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
230
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
231
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
232
233
    return lib

wxchan's avatar
wxchan committed
234

235
236
237
238
239
240
241
# 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
242

wxchan's avatar
wxchan committed
243

244
_NUMERIC_TYPES = (int, float, bool)
245
_ArrayLike = Union[List, np.ndarray, pd_Series]
246
247


248
def _safe_call(ret: int) -> None:
249
250
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
251
252
253
    Parameters
    ----------
    ret : int
254
        The return value from C API calls.
wxchan's avatar
wxchan committed
255
256
    """
    if ret != 0:
257
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
258

wxchan's avatar
wxchan committed
259

260
def _is_numeric(obj: Any) -> bool:
261
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
262
263
264
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
265
266
267
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
268
269
        return False

wxchan's avatar
wxchan committed
270

271
def _is_numpy_1d_array(data: Any) -> bool:
272
    """Check whether data is a numpy 1-D array."""
273
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
274

wxchan's avatar
wxchan committed
275

276
def _is_numpy_column_array(data: Any) -> bool:
277
278
279
280
281
282
283
    """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


284
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
285
    """Cast numpy array to given dtype."""
286
287
288
289
290
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


291
def _is_1d_list(data: Any) -> bool:
292
    """Check whether data is a 1-D list."""
293
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
294

wxchan's avatar
wxchan committed
295

296
297
298
299
300
301
302
303
304
305
306
307
308
309
def _is_list_of_numpy_arrays(data: Any) -> "TypeGuard[List[np.ndarray]]":
    return (
        isinstance(data, list)
        and all(isinstance(x, np.ndarray) for x in data)
    )


def _is_list_of_sequences(data: Any) -> "TypeGuard[List[Sequence]]":
    return (
        isinstance(data, list)
        and all(isinstance(x, Sequence) for x in data)
    )


310
311
312
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
313
        _is_numpy_1d_array(data)
314
        or _is_numpy_column_array(data)
315
        or _is_1d_list(data)
316
317
318
319
        or isinstance(data, pd_Series)
    )


320
321
def _list_to_1d_numpy(
    data: Any,
322
323
    dtype: "np.typing.DTypeLike",
    name: str
324
) -> np.ndarray:
325
    """Convert data to numpy 1-D array."""
326
    if _is_numpy_1d_array(data):
327
        return _cast_numpy_array_to_dtype(data, dtype)
328
    elif _is_numpy_column_array(data):
329
330
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
331
        return _cast_numpy_array_to_dtype(array, dtype)
332
    elif _is_1d_list(data):
wxchan's avatar
wxchan committed
333
        return np.array(data, dtype=dtype, copy=False)
334
    elif isinstance(data, pd_Series):
335
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
336
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
337
    else:
338
339
        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
340

wxchan's avatar
wxchan committed
341

342
343
344
345
346
347
348
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."""
349
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
350
351
352
353
354
355
356
357
358
359
360


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)
    )


361
362
363
364
365
def _is_pyarrow_array(data: Any) -> bool:
    """Check whether data is a PyArrow array."""
    return isinstance(data, (pa_Array, pa_ChunkedArray))


366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
def _is_pyarrow_table(data: Any) -> bool:
    """Check whether data is a PyArrow table."""
    return isinstance(data, pa_Table)


class _ArrowCArray:
    """Simple wrapper around the C representation of an Arrow type."""

    n_chunks: int
    chunks: arrow_cffi.CData
    schema: arrow_cffi.CData

    def __init__(self, n_chunks: int, chunks: arrow_cffi.CData, schema: arrow_cffi.CData):
        self.n_chunks = n_chunks
        self.chunks = chunks
        self.schema = schema

    @property
    def chunks_ptr(self) -> int:
        """Returns the address of the pointer to the list of chunks making up the array."""
        return int(arrow_cffi.cast("uintptr_t", arrow_cffi.addressof(self.chunks[0])))

    @property
    def schema_ptr(self) -> int:
        """Returns the address of the pointer to the schema of the array."""
        return int(arrow_cffi.cast("uintptr_t", self.schema))


def _export_arrow_to_c(data: pa_Table) -> _ArrowCArray:
    """Export an Arrow type to its C representation."""
    # Obtain objects to export
397
398
399
400
401
    if isinstance(data, pa_Array):
        export_objects = [data]
    elif isinstance(data, pa_ChunkedArray):
        export_objects = data.chunks
    elif isinstance(data, pa_Table):
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
        export_objects = data.to_batches()
    else:
        raise ValueError(f"data of type '{type(data)}' cannot be exported to Arrow")

    # Prepare export
    chunks = arrow_cffi.new("struct ArrowArray[]", len(export_objects))
    schema = arrow_cffi.new("struct ArrowSchema*")

    # Export all objects
    for i, obj in enumerate(export_objects):
        chunk_ptr = int(arrow_cffi.cast("uintptr_t", arrow_cffi.addressof(chunks[i])))
        if i == 0:
            schema_ptr = int(arrow_cffi.cast("uintptr_t", schema))
            obj._export_to_c(chunk_ptr, schema_ptr)
        else:
            obj._export_to_c(chunk_ptr)

    return _ArrowCArray(len(chunks), chunks, schema)



423
424
def _data_to_2d_numpy(
    data: Any,
425
426
    dtype: "np.typing.DTypeLike",
    name: str
427
) -> np.ndarray:
428
429
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
430
        return _cast_numpy_array_to_dtype(data, dtype)
431
432
433
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
434
        _check_for_bad_pandas_dtypes(data.dtypes)
435
        return _cast_numpy_array_to_dtype(data.values, dtype)
436
437
438
439
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


440
def _cfloat32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
441
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
442
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
443
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
444
    else:
445
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
446

Guolin Ke's avatar
Guolin Ke committed
447

448
def _cfloat64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
449
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
450
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
451
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
452
453
454
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
455

456
def _cint32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
457
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
458
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
459
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
460
    else:
461
462
463
        raise RuntimeError('Expected int32 pointer')


464
def _cint64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
465
466
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
467
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
468
469
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
470

wxchan's avatar
wxchan committed
471

472
def _c_str(string: str) -> ctypes.c_char_p:
473
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
474
475
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
476

477
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
478
    """Convert a Python array to C array."""
479
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
480

wxchan's avatar
wxchan committed
481

482
def _json_default_with_numpy(obj: Any) -> Any:
483
484
485
486
487
488
489
490
491
    """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


492
493
494
495
496
497
498
499
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)


500
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
501
    """Convert Python dictionary to string, which is passed to C API."""
502
    if data is None or not data:
wxchan's avatar
wxchan committed
503
504
505
        return ""
    pairs = []
    for key, val in data.items():
506
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
507
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
508
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
509
            pairs.append(f"{key}={val}")
510
        elif val is not None:
511
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
512
    return ' '.join(pairs)
513

wxchan's avatar
wxchan committed
514

515
class _TempFile:
516
517
    """Proxy class to workaround errors on Windows."""

518
519
520
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
521
            self.path = Path(self.name)
522
        return self
wxchan's avatar
wxchan committed
523

524
    def __exit__(self, exc_type, exc_val, exc_tb):
525
526
        if self.path.is_file():
            self.path.unlink()
527

wxchan's avatar
wxchan committed
528

529
class LightGBMError(Exception):
530
531
    """Error thrown by LightGBM."""

532
533
534
    pass


535
536
537
538
539
540
541
542
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
543
544
545
546
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
547
    def _get_all_param_aliases() -> Dict[str, List[str]]:
548
549
550
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
551
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
552
553
554
555
556
557
558
559
        _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)
560
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
561
562
563
564
            _safe_call(_LIB.LGBM_DumpParamAliases(
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
565
        return json.loads(
566
            string_buffer.value.decode('utf-8'),
567
            object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
568
        )
569
570

    @classmethod
571
572
573
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
574
575
        ret = set()
        for i in args:
576
            ret.update(cls.get_sorted(i))
577
578
        return ret

579
580
581
582
583
584
    @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])

585
    @classmethod
586
587
588
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
589
590
591
592
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
593
                    ret.update(aliases)
594
595
596
                    break
        return ret

597

598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
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)

619
620
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
621
622

    # if main_param_name was provided, keep that value and remove all aliases
623
    if main_param_name in params.keys():
624
625
626
        for param in aliases:
            params.pop(param, None)
        return params
627

628
629
630
631
632
    # 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
633

634
635
636
637
638
639
640
    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
641
642
643
644

    return params


645
_MAX_INT32 = (1 << 31) - 1
646

647
"""Macro definition of data type in C API of LightGBM"""
648
649
650
651
_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
652

653
"""Matrix is row major in Python"""
654
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
655

656
"""Macro definition of prediction type in C API of LightGBM"""
657
658
659
660
_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
661

662
"""Macro definition of sparse matrix type"""
663
664
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
665

666
"""Macro definition of feature importance type"""
667
668
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
669

670
"""Data type of data field"""
671
672
673
674
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
675
676
    "group": _C_API_DTYPE_INT32,
    "position": _C_API_DTYPE_INT32
677
}
wxchan's avatar
wxchan committed
678

679
"""String name to int feature importance type mapper"""
680
681
682
683
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN
}
684

wxchan's avatar
wxchan committed
685

686
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
687
    """Fix the memory of multi-dimensional sliced object."""
688
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
689
        if not data.flags.c_contiguous:
690
691
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
692
693
694
695
            return np.copy(data)
    return data


696
697
698
def _c_float_array(
    data: np.ndarray
) -> Tuple[_ctypes_float_ptr, int, np.ndarray]:
699
    """Get pointer of float numpy array / list."""
700
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
701
        data = np.array(data, copy=False)
702
    if _is_numpy_1d_array(data):
703
        data = _convert_from_sliced_object(data)
704
        assert data.flags.c_contiguous
705
        ptr_data: _ctypes_float_ptr
wxchan's avatar
wxchan committed
706
707
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
708
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
709
710
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
711
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
712
        else:
713
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
714
    else:
715
        raise TypeError(f"Unknown type({type(data).__name__})")
716
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
717

wxchan's avatar
wxchan committed
718

719
720
721
def _c_int_array(
    data: np.ndarray
) -> Tuple[_ctypes_int_ptr, int, np.ndarray]:
722
    """Get pointer of int numpy array / list."""
723
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
724
        data = np.array(data, copy=False)
725
    if _is_numpy_1d_array(data):
726
        data = _convert_from_sliced_object(data)
727
        assert data.flags.c_contiguous
728
        ptr_data: _ctypes_int_ptr
wxchan's avatar
wxchan committed
729
730
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
731
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
732
733
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
734
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
735
        else:
736
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
737
    else:
738
        raise TypeError(f"Unknown type({type(data).__name__})")
739
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
740

wxchan's avatar
wxchan committed
741

742
def _is_allowed_numpy_dtype(dtype: type) -> bool:
743
    float128 = getattr(np, 'float128', type(None))
744
745
746
747
    return (
        issubclass(dtype, (np.integer, np.floating, np.bool_))
        and not issubclass(dtype, (np.timedelta64, float128))
    )
748
749


750
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
751
752
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
753
        for column_name, pandas_dtype in pandas_dtypes_series.items()
754
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
755
756
757
758
    ]
    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)}')
759
760


761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
def _pandas_to_numpy(
    data: pd_DataFrame,
    target_dtype: "np.typing.DTypeLike"
) -> np.ndarray:
    _check_for_bad_pandas_dtypes(data.dtypes)
    try:
        # most common case (no nullable dtypes)
        return 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
        return 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
        return data.to_numpy(dtype=target_dtype, na_value=np.nan)


778
def _data_from_pandas(
779
780
781
    data: pd_DataFrame,
    feature_name: _LGBM_FeatureNameConfiguration,
    categorical_feature: _LGBM_CategoricalFeatureConfiguration,
782
    pandas_categorical: Optional[List[List]]
783
784
785
786
787
788
789
790
791
792
793
794
795
) -> Tuple[np.ndarray, List[str], List[str], List[List]]:
    if len(data.shape) != 2 or data.shape[0] < 1:
        raise ValueError('Input data must be 2 dimensional and non empty.')

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

    # determine categorical features
    cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
    cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
    if pandas_categorical is None:  # train dataset
        pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
796
    else:
797
798
799
800
801
802
803
804
805
806
807
808
809
810
        if len(cat_cols) != len(pandas_categorical):
            raise ValueError('train and valid dataset categorical_feature do not match.')
        for col, category in zip(cat_cols, pandas_categorical):
            if list(data[col].cat.categories) != list(category):
                data[col] = data[col].cat.set_categories(category)
    if len(cat_cols):  # cat_cols is list
        data = data.copy(deep=False)  # not alter origin DataFrame
        data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
    if categorical_feature == 'auto':  # use cat cols from DataFrame
        categorical_feature = cat_cols_not_ordered
    else:  # use cat cols specified by user
        categorical_feature = list(categorical_feature)  # type: ignore[assignment]

    df_dtypes = [dtype.type for dtype in data.dtypes]
811
812
    # so that the target dtype considers floats
    df_dtypes.append(np.float32)
813
    target_dtype = np.result_type(*df_dtypes)
814
815
816
817
818
819
820

    return (
        _pandas_to_numpy(data, target_dtype=target_dtype),
        feature_name,
        categorical_feature,
        pandas_categorical
    )
821
822


823
824
825
826
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
    file_name: Optional[Union[str, Path]] = None
) -> str:
827
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
828
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
829
830
831
832
833
834
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


835
836
837
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
    model_str: Optional[str] = None
838
) -> Optional[List[List]]:
839
840
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
841
    if file_name is not None:
842
        max_offset = -getsize(file_name)
843
844
845
846
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
847
                f.seek(offset, SEEK_END)
848
849
850
851
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
852
        last_line = lines[-1].decode('utf-8').strip()
853
        if not last_line.startswith(pandas_key):
854
            last_line = lines[-2].decode('utf-8').strip()
855
    elif model_str is not None:
856
857
858
859
860
861
        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
862
863


864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
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**.

884
885
    .. versionadded:: 3.3.0

886
887
888
889
890
891
892
893
894
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
895
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
896
897
898
899
900
901
902
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
903
                return self._get_one_line(idx)
904
            elif isinstance(idx, slice):
905
906
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
907
                # Only required if using ``Dataset.subset()``.
908
                return np.array([self._get_one_line(i) for i in idx])
909
            else:
910
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
911
912
913

        Parameters
        ----------
914
        idx : int, slice[int], list[int]
915
916
917
918
            Item index.

        Returns
        -------
919
        result : numpy 1-D array or numpy 2-D array
920
            1-D array if idx is int, 2-D array if idx is slice or list.
921
922
923
924
925
926
927
928
929
        """
        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__()")


930
class _InnerPredictor:
931
932
933
934
935
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
936
937
938
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
939
    """
940

941
942
    def __init__(
        self,
943
944
945
946
        booster_handle: _BoosterHandle,
        pandas_categorical: Optional[List[List]],
        pred_parameter: Dict[str, Any],
        manage_handle: bool
947
    ):
948
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
949
950
951

        Parameters
        ----------
952
        booster_handle : object
953
            Handle of Booster.
954
955
956
957
        pandas_categorical : list of list, or None
            If provided, list of categories for ``pandas`` categorical columns.
            Where the ``i``th element of the list contains the categories for the ``i``th categorical feature.
        pred_parameter : dict
958
            Other parameters for the prediction.
959
960
        manage_handle : bool
            If ``True``, free the corresponding Booster on the C++ side when this Python object is deleted.
wxchan's avatar
wxchan committed
961
        """
962
963
964
965
966
967
968
969
        self._handle = booster_handle
        self.__is_manage_handle = manage_handle
        self.pandas_categorical = pandas_categorical
        self.pred_parameter = _param_dict_to_str(pred_parameter)

        out_num_class = ctypes.c_int(0)
        _safe_call(
            _LIB.LGBM_BoosterGetNumClasses(
970
                self._handle,
971
972
973
974
                ctypes.byref(out_num_class)
            )
        )
        self.num_class = out_num_class.value
wxchan's avatar
wxchan committed
975

976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
    @classmethod
    def from_booster(
        cls,
        booster: "Booster",
        pred_parameter: Dict[str, Any]
    ) -> "_InnerPredictor":
        """Initialize an ``_InnerPredictor`` from a ``Booster``.

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

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

        Parameters
        ----------
        model_file : str or pathlib.Path
            Path to the model file.
        pred_parameter : dict
            Other parameters for the prediction.
        """
        booster_handle = ctypes.c_void_p()
        out_num_iterations = ctypes.c_int(0)
        _safe_call(
            _LIB.LGBM_BoosterCreateFromModelfile(
                _c_str(str(model_file)),
                ctypes.byref(out_num_iterations),
                ctypes.byref(booster_handle)
            )
        )
        return cls(
            booster_handle=booster_handle,
            pandas_categorical=_load_pandas_categorical(file_name=model_file),
            pred_parameter=pred_parameter,
            manage_handle=True
        )
cbecker's avatar
cbecker committed
1035

1036
    def __del__(self) -> None:
1037
1038
        try:
            if self.__is_manage_handle:
1039
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
1040
1041
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
1042

1043
    def __getstate__(self) -> Dict[str, Any]:
1044
1045
        this = self.__dict__.copy()
        this.pop('handle', None)
1046
        this.pop('_handle', None)
1047
1048
        return this

1049
1050
    def predict(
        self,
1051
        data: _LGBM_PredictDataType,
1052
1053
1054
1055
1056
1057
1058
        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
1059
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
1060
        """Predict logic.
wxchan's avatar
wxchan committed
1061
1062
1063

        Parameters
        ----------
1064
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
1065
            Data source for prediction.
1066
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
1067
1068
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
        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.
1080
1081
1082
        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
1083

1084
1085
            .. versionadded:: 4.0.0

wxchan's avatar
wxchan committed
1086
1087
        Returns
        -------
1088
        result : numpy array, scipy.sparse or list of scipy.sparse
1089
            Prediction result.
1090
            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
1091
        """
wxchan's avatar
wxchan committed
1092
        if isinstance(data, Dataset):
1093
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
1094
1095
1096
1097
1098
1099
        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(
1100
                    self._handle,
1101
1102
1103
1104
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
1105
1106
1107
1108
1109
1110
1111
1112
1113

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

1114
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
1115
        if raw_score:
1116
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
1117
        if pred_leaf:
1118
            predict_type = _C_API_PREDICT_LEAF_INDEX
1119
        if pred_contrib:
1120
            predict_type = _C_API_PREDICT_CONTRIB
cbecker's avatar
cbecker committed
1121

1122
        if isinstance(data, (str, Path)):
1123
            with _TempFile() as f:
wxchan's avatar
wxchan committed
1124
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
1125
                    self._handle,
1126
                    _c_str(str(data)),
1127
                    ctypes.c_int(data_has_header),
Guolin Ke's avatar
Guolin Ke committed
1128
                    ctypes.c_int(predict_type),
1129
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1130
                    ctypes.c_int(num_iteration),
1131
1132
                    _c_str(self.pred_parameter),
                    _c_str(f.name)))
1133
1134
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
1135
        elif isinstance(data, scipy.sparse.csr_matrix):
1136
1137
1138
1139
1140
1141
            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
1142
        elif isinstance(data, scipy.sparse.csc_matrix):
1143
1144
1145
1146
1147
1148
            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
1149
        elif isinstance(data, np.ndarray):
1150
1151
1152
1153
1154
1155
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1156
1157
1158
        elif isinstance(data, list):
            try:
                data = np.array(data)
1159
1160
            except BaseException as err:
                raise ValueError('Cannot convert data list to numpy array.') from err
1161
1162
1163
1164
1165
1166
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1167
        elif isinstance(data, dt_DataTable):
1168
1169
1170
1171
1172
1173
            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
1174
1175
        else:
            try:
1176
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
1177
                csr = scipy.sparse.csr_matrix(data)
1178
1179
            except BaseException as err:
                raise TypeError(f'Cannot predict data for type {type(data).__name__}') from err
1180
1181
1182
1183
1184
1185
            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
1186
1187
        if pred_leaf:
            preds = preds.astype(np.int32)
1188
        is_sparse = isinstance(preds, scipy.sparse.spmatrix) or isinstance(preds, list)
1189
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1190
            if preds.size % nrow == 0:
1191
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1192
            else:
1193
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
1194
1195
        return preds

1196
1197
1198
1199
1200
1201
1202
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
        predict_type: int
    ) -> int:
1203
        """Get size of prediction result."""
1204
        if nrow > _MAX_INT32:
1205
            raise LightGBMError('LightGBM cannot perform prediction for data '
1206
                                f'with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n'
1207
                                'You can split your data into chunks '
1208
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
1209
1210
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
1211
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1212
1213
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
1214
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1215
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
1216
1217
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
1218

1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
    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)
1232
1233
1234
1235
1236
1237
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
            predict_type=predict_type
        )
1238
1239
1240
1241
1242
1243
        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(
1244
            self._handle,
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
            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]:
1267
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1268
        if len(mat.shape) != 2:
1269
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
1270

1271
        nrow = mat.shape[0]
1272
1273
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1274
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1275
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
1276
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1277
            preds = np.empty(sum(n_preds), dtype=np.float64)
1278
1279
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
1280
                # avoid memory consumption by arrays concatenation operations
1281
1282
1283
1284
1285
1286
1287
                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]
                )
1288
            return preds, nrow
wxchan's avatar
wxchan committed
1289
        else:
1290
1291
1292
1293
1294
1295
1296
            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
1297

1298
1299
1300
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1301
1302
1303
1304
1305
1306
        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,
1307
        is_csr: bool
1308
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1309
1310
1311
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1312
        if indptr_type == _C_API_DTYPE_INT32:
1313
            out_indptr = _cint32_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1314
        elif indptr_type == _C_API_DTYPE_INT64:
1315
            out_indptr = _cint64_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1316
1317
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1318
        if data_type == _C_API_DTYPE_FLOAT32:
1319
            out_data = _cfloat32_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1320
        elif data_type == _C_API_DTYPE_FLOAT64:
1321
            out_data = _cfloat64_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1322
1323
        else:
            raise TypeError("Expected float32 or float64 type for data")
1324
        out_indices = _cint32_array_to_numpy(cptr=out_ptr_indices, length=data_indices_len)
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
        # 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

1353
1354
1355
1356
1357
1358
1359
1360
1361
    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
1362
1363
1364
1365
1366
1367
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1368
1369
1370
1371
1372
        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
1373

1374
1375
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1376

1377
1378
1379
1380
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

        _safe_call(_LIB.LGBM_BoosterPredictForCSR(
1381
            self._handle,
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
            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
1406
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1407
1408
1409
1410
        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
1411
        out_ptr_indptr: _ctypes_int_ptr
1412
1413
1414
1415
1416
        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)()
1417
        out_ptr_data: _ctypes_float_ptr
1418
1419
1420
1421
1422
1423
        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(
1424
            self._handle,
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
            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

1455
1456
1457
1458
1459
1460
1461
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1462
        """Predict for a CSR data."""
1463
        if predict_type == _C_API_PREDICT_CONTRIB:
1464
1465
1466
1467
1468
1469
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1470
        nrow = len(csr.indptr) - 1
1471
1472
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1473
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1474
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1475
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1476
            preds = np.empty(sum(n_preds), dtype=np.float64)
1477
1478
            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:])):
1479
                # avoid memory consumption by arrays concatenation operations
1480
1481
1482
1483
1484
1485
1486
                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]
                )
1487
1488
            return preds, nrow
        else:
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
            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,
1499
1500
1501
1502
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
1503
1504
1505
1506
1507
    ):
        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
1508
        out_ptr_indptr: _ctypes_int_ptr
1509
1510
1511
1512
1513
        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)()
1514
        out_ptr_data: _ctypes_float_ptr
1515
1516
1517
1518
1519
1520
        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(
1521
            self._handle,
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
            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
1551

1552
1553
1554
1555
1556
1557
1558
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1559
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1560
        nrow = csc.shape[0]
1561
        if nrow > _MAX_INT32:
1562
1563
1564
1565
1566
1567
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1568
        if predict_type == _C_API_PREDICT_CONTRIB:
1569
1570
1571
1572
1573
1574
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1575
1576
1577
1578
1579
1580
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1581
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1582
1583
        out_num_preds = ctypes.c_int64(0)

1584
1585
        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
1586

1587
        assert csc.shape[0] <= _MAX_INT32
1588
        csc_indices = csc.indices.astype(np.int32, copy=False)
1589

Guolin Ke's avatar
Guolin Ke committed
1590
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
1591
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1592
            ptr_indptr,
1593
            ctypes.c_int(type_ptr_indptr),
1594
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1595
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1596
1597
1598
1599
1600
            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),
1601
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1602
            ctypes.c_int(num_iteration),
1603
            _c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1604
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1605
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1606
        if n_preds != out_num_preds.value:
1607
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1608
1609
        return preds, nrow

1610
    def current_iteration(self) -> int:
1611
1612
1613
1614
1615
1616
1617
1618
1619
        """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(
1620
            self._handle,
1621
1622
1623
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1624

1625
class Dataset:
wxchan's avatar
wxchan committed
1626
    """Dataset in LightGBM."""
1627

1628
1629
    def __init__(
        self,
1630
        data: _LGBM_TrainDataType,
1631
        label: Optional[_LGBM_LabelType] = None,
1632
        reference: Optional["Dataset"] = None,
1633
1634
1635
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1636
1637
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
1638
        params: Optional[Dict[str, Any]] = None,
1639
1640
        free_raw_data: bool = True,
        position: Optional[_LGBM_PositionType] = None,
1641
    ):
1642
        """Initialize Dataset.
1643

wxchan's avatar
wxchan committed
1644
1645
        Parameters
        ----------
1646
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence, list of numpy array or pyarrow Table
wxchan's avatar
wxchan committed
1647
            Data source of Dataset.
1648
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1649
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1650
1651
1652
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1653
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1654
            Weight for each instance. Weights should be non-negative.
1655
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1656
1657
1658
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1659
1660
            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.
1661
        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)
1662
            Init score for Dataset.
1663
        feature_name : list of str, or 'auto', optional (default="auto")
1664
            Feature names.
1665
            If 'auto' and data is pandas DataFrame or pyarrow Table, data columns names are used.
1666
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1667
1668
            Categorical features.
            If list of int, interpreted as indices.
1669
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1670
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1671
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1672
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1673
            All negative values in categorical features will be treated as missing values.
1674
            The output cannot be monotonically constrained with respect to a categorical feature.
1675
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1676
        params : dict or None, optional (default=None)
1677
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1678
        free_raw_data : bool, optional (default=True)
1679
            If True, raw data is freed after constructing inner Dataset.
1680
1681
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
wxchan's avatar
wxchan committed
1682
        """
1683
        self._handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1684
1685
1686
1687
1688
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1689
        self.position = position
1690
        self.init_score = init_score
1691
1692
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1693
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1694
        self.free_raw_data = free_raw_data
1695
        self.used_indices: Optional[List[int]] = None
1696
        self._need_slice = True
1697
        self._predictor: Optional[_InnerPredictor] = None
1698
        self.pandas_categorical: Optional[List[List]] = None
1699
        self._params_back_up = None
1700
        self.version = 0
1701
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1702

1703
    def __del__(self) -> None:
1704
1705
1706
1707
        try:
            self._free_handle()
        except AttributeError:
            pass
1708

1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
    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.
        """
1726
        param_str = _param_dict_to_str(self.get_params())
1727
1728
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1729
        ptr_data, _, _ = _c_int_array(indices)
1730
1731
1732
1733
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
1734
            _c_str(param_str),
1735
1736
1737
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1738
1739
        assert sample_cnt == actual_sample_cnt.value
        return indices
1740

1741
1742
1743
1744
1745
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
        ref_dataset: _DatasetHandle
    ) -> 'Dataset':
1746
1747
1748
1749
1750
1751
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1752
1753
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1754
1755
1756
1757
1758
1759

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
1760
        self._handle = ctypes.c_void_p()
1761
1762
1763
        _safe_call(_LIB.LGBM_DatasetCreateByReference(
            ref_dataset,
            ctypes.c_int64(total_nrow),
1764
            ctypes.byref(self._handle),
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
        ))
        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
        ----------
1779
        sample_data : list of numpy array
1780
            Sample data for each column.
1781
        sample_indices : list of numpy array
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
            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.
1804
        sample_col_ptr: _ctypes_float_array = (ctypes.POINTER(ctypes.c_double) * ncol)()
1805
1806
        # c type int**
        # each int* points to start of indices for each column
1807
        indices_col_ptr: _ctypes_int_array = (ctypes.POINTER(ctypes.c_int32) * ncol)()
1808
        for i in range(ncol):
1809
1810
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1811
1812

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1813
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1814

1815
        self._handle = ctypes.c_void_p()
1816
        params_str = _param_dict_to_str(self.get_params())
1817
1818
1819
1820
1821
1822
1823
        _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),
1824
            ctypes.c_int64(total_nrow),
1825
            _c_str(params_str),
1826
            ctypes.byref(self._handle),
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
        ))
        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)
1845
        data_ptr, data_type, _ = _c_float_array(data)
1846
1847

        _safe_call(_LIB.LGBM_DatasetPushRows(
1848
            self._handle,
1849
1850
1851
1852
1853
1854
1855
1856
1857
            data_ptr,
            data_type,
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol),
            ctypes.c_int32(self._start_row),
        ))
        self._start_row += nrow
        return self

1858
    def get_params(self) -> Dict[str, Any]:
1859
1860
1861
1862
        """Get the used parameters in the Dataset.

        Returns
        -------
1863
        params : dict
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
            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",
1879
                                                "linear_tree",
1880
1881
1882
1883
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1884
                                                "precise_float_parser",
1885
1886
1887
1888
1889
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1890
1891
        else:
            return {}
1892

1893
    def _free_handle(self) -> "Dataset":
1894
1895
1896
        if self._handle is not None:
            _safe_call(_LIB.LGBM_DatasetFree(self._handle))
            self._handle = None
1897
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
1898
1899
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1900
        return self
wxchan's avatar
wxchan committed
1901

1902
1903
1904
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
1905
        data: _LGBM_TrainDataType,
1906
        used_indices: Optional[Union[List[int], np.ndarray]]
1907
    ) -> "Dataset":
Guolin Ke's avatar
Guolin Ke committed
1908
        data_has_header = False
1909
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
1910
            # check data has header or not
1911
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1912
        num_data = self.num_data()
1913
        if predictor is not None:
1914
1915
1916
1917
1918
            init_score: Union[np.ndarray, scipy.sparse.spmatrix] = predictor.predict(
                data=data,
                raw_score=True,
                data_has_header=data_has_header
            )
1919
            init_score = init_score.ravel()
1920
            if used_indices is not None:
1921
                assert not self._need_slice
1922
                if isinstance(data, (str, Path)):
1923
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1924
                    assert num_data == len(used_indices)
1925
1926
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1927
1928
1929
1930
                            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
1931
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1932
1933
                for i in range(num_data):
                    for j in range(predictor.num_class):
1934
1935
1936
                        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:
1937
            init_score = np.full_like(self.init_score, fill_value=0.0, dtype=np.float64)
1938
1939
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1940
        self.set_init_score(init_score)
1941
        return self
Guolin Ke's avatar
Guolin Ke committed
1942

1943
1944
    def _lazy_init(
        self,
1945
        data: Optional[_LGBM_TrainDataType],
1946
1947
1948
1949
1950
1951
1952
1953
        label: Optional[_LGBM_LabelType],
        reference: Optional["Dataset"],
        weight: Optional[_LGBM_WeightType],
        group: Optional[_LGBM_GroupType],
        init_score: Optional[_LGBM_InitScoreType],
        predictor: Optional[_InnerPredictor],
        feature_name: _LGBM_FeatureNameConfiguration,
        categorical_feature: _LGBM_CategoricalFeatureConfiguration,
1954
1955
        params: Optional[Dict[str, Any]],
        position: Optional[_LGBM_PositionType]
1956
    ) -> "Dataset":
wxchan's avatar
wxchan committed
1957
        if data is None:
1958
            self._handle = None
Nikita Titov's avatar
Nikita Titov committed
1959
            return self
Guolin Ke's avatar
Guolin Ke committed
1960
1961
1962
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1963
1964
1965
1966
1967
1968
1969
        if isinstance(data, pd_DataFrame):
            data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(
                data=data,
                feature_name=feature_name,
                categorical_feature=categorical_feature,
                pandas_categorical=self.pandas_categorical
            )
Guolin Ke's avatar
Guolin Ke committed
1970

1971
        # process for args
wxchan's avatar
wxchan committed
1972
        params = {} if params is None else params
1973
        args_names = inspect.signature(self.__class__._lazy_init).parameters.keys()
1974
        for key in params.keys():
1975
            if key in args_names:
1976
1977
                _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.')
1978
        # get categorical features
1979
        if isinstance(categorical_feature, list):
1980
1981
            categorical_indices = set()
            feature_dict = {}
1982
            if isinstance(feature_name, list):
1983
1984
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
1985
                if isinstance(name, str) and name in feature_dict:
1986
                    categorical_indices.add(feature_dict[name])
1987
                elif isinstance(name, int):
1988
1989
                    categorical_indices.add(name)
                else:
1990
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1991
            if categorical_indices:
1992
1993
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1994
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
1995
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
1996
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1997
                        params.pop(cat_alias, None)
1998
                params['categorical_column'] = sorted(categorical_indices)
1999

2000
        params_str = _param_dict_to_str(params)
2001
        self.params = params
2002
        # process for reference dataset
wxchan's avatar
wxchan committed
2003
        ref_dataset = None
wxchan's avatar
wxchan committed
2004
        if isinstance(reference, Dataset):
2005
            ref_dataset = reference.construct()._handle
wxchan's avatar
wxchan committed
2006
2007
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
2008
        # start construct data
2009
        if isinstance(data, (str, Path)):
2010
            self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2011
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
2012
2013
                _c_str(str(data)),
                _c_str(params_str),
wxchan's avatar
wxchan committed
2014
                ref_dataset,
2015
                ctypes.byref(self._handle)))
wxchan's avatar
wxchan committed
2016
2017
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
2018
2019
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
2020
2021
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
2022
2023
2024
        elif _is_pyarrow_table(data):
            self.__init_from_pyarrow_table(data, params_str, ref_dataset)
            feature_name = data.column_names
2025
        elif isinstance(data, list) and len(data) > 0:
2026
            if _is_list_of_numpy_arrays(data):
2027
                self.__init_from_list_np2d(data, params_str, ref_dataset)
2028
            elif _is_list_of_sequences(data):
2029
2030
2031
2032
2033
                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)
2034
        elif isinstance(data, dt_DataTable):
2035
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
2036
2037
2038
2039
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
2040
2041
            except BaseException as err:
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}') from err
wxchan's avatar
wxchan committed
2042
2043
2044
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
2045
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
2046
2047
2048
2049
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
2050
2051
        if position is not None:
            self.set_position(position)
2052
2053
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
2054
                _log_warning("The init_score will be overridden by the prediction of init_model.")
2055
2056
2057
2058
2059
            self._set_init_score_by_predictor(
                predictor=predictor,
                data=data,
                used_indices=None
            )
2060
2061
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
2062
        elif predictor is not None:
2063
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2064
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
2065
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
2066

2067
2068
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
        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.
2094
        sampled = np.array(list(self._yield_row_from_seqlist(seqs, indices)))
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
        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

2110
2111
2112
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
2113
        ref_dataset: Optional[_DatasetHandle]
2114
    ) -> "Dataset":
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
        """
        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:
2129
            param_str = _param_dict_to_str(self.get_params())
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
            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

2143
2144
2145
2146
2147
2148
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2149
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
2150
2151
2152
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

2153
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2154
2155
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
2156
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
2157
2158
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

2159
        ptr_data, type_ptr_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2160
2161
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2162
            ctypes.c_int(type_ptr_data),
2163
2164
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
2165
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2166
            _c_str(params_str),
wxchan's avatar
wxchan committed
2167
            ref_dataset,
2168
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2169
        return self
wxchan's avatar
wxchan committed
2170

2171
2172
2173
2174
2175
2176
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2177
        """Initialize data from a list of 2-D numpy matrices."""
2178
        ncol = mats[0].shape[1]
2179
        nrow = np.empty((len(mats),), np.int32)
2180
        ptr_data: _ctypes_float_array
2181
2182
2183
2184
2185
2186
        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 = []
2187
        type_ptr_data = -1
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199

        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)
2200
            else:  # change non-float data to float data, need to copy
2201
2202
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2203
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2204
            if type_ptr_data != -1 and chunk_type_ptr_data != type_ptr_data:
2205
2206
2207
2208
2209
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

2210
        self._handle = ctypes.c_void_p()
2211
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
2212
            ctypes.c_int32(len(mats)),
2213
2214
2215
            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)),
2216
            ctypes.c_int32(ncol),
2217
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2218
            _c_str(params_str),
2219
            ref_dataset,
2220
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2221
        return self
2222

2223
2224
2225
2226
2227
2228
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2229
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2230
        if len(csr.indices) != len(csr.data):
2231
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
2232
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2233

2234
2235
        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
2236

2237
        assert csr.shape[1] <= _MAX_INT32
2238
        csr_indices = csr.indices.astype(np.int32, copy=False)
2239

wxchan's avatar
wxchan committed
2240
2241
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2242
            ctypes.c_int(type_ptr_indptr),
2243
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
2244
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2245
2246
2247
2248
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
2249
            _c_str(params_str),
wxchan's avatar
wxchan committed
2250
            ref_dataset,
2251
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2252
        return self
wxchan's avatar
wxchan committed
2253

2254
2255
2256
2257
2258
2259
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2260
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2261
        if len(csc.indices) != len(csc.data):
2262
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
2263
        self._handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2264

2265
2266
        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
2267

2268
        assert csc.shape[0] <= _MAX_INT32
2269
        csc_indices = csc.indices.astype(np.int32, copy=False)
2270

Guolin Ke's avatar
Guolin Ke committed
2271
2272
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2273
            ctypes.c_int(type_ptr_indptr),
2274
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
2275
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2276
2277
2278
2279
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
2280
            _c_str(params_str),
Guolin Ke's avatar
Guolin Ke committed
2281
            ref_dataset,
2282
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2283
        return self
Guolin Ke's avatar
Guolin Ke committed
2284

2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
    def __init_from_pyarrow_table(
        self,
        table: pa_Table,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
        """Initialize data from a PyArrow table."""
        if not PYARROW_INSTALLED:
            raise LightGBMError("Cannot init dataframe from Arrow without `pyarrow` installed.")

        # Check that the input is valid: we only handle numbers (for now)
        if not all(arrow_is_integer(t) or arrow_is_floating(t) for t in table.schema.types):
            raise ValueError("Arrow table may only have integer or floating point datatypes")

        # Export Arrow table to C
        c_array = _export_arrow_to_c(table)
        self._handle = ctypes.c_void_p()
        _safe_call(_LIB.LGBM_DatasetCreateFromArrow(
            ctypes.c_int64(c_array.n_chunks),
            ctypes.c_void_p(c_array.chunks_ptr),
            ctypes.c_void_p(c_array.schema_ptr),
            _c_str(params_str),
            ref_dataset,
            ctypes.byref(self._handle)))
        return self

2311
    @staticmethod
2312
    def _compare_params_for_warning(
2313
2314
        params: Dict[str, Any],
        other_params: Dict[str, Any],
2315
2316
2317
        ignore_keys: Set[str]
    ) -> bool:
        """Compare two dictionaries with params ignoring some keys.
2318

2319
2320
2321
2322
        It is only for the warning purpose.

        Parameters
        ----------
2323
        params : dict
2324
            One dictionary with parameters to compare.
2325
        other_params : dict
2326
2327
2328
            Another dictionary with parameters to compare.
        ignore_keys : set
            Keys that should be ignored during comparing two dictionaries.
2329
2330
2331

        Returns
        -------
2332
2333
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
        """
        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

2345
    def construct(self) -> "Dataset":
2346
2347
2348
2349
2350
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2351
            Constructed Dataset object.
2352
        """
2353
        if self._handle is None:
wxchan's avatar
wxchan committed
2354
            if self.reference is not None:
2355
                reference_params = self.reference.get_params()
2356
2357
                params = self.get_params()
                if params != reference_params:
2358
2359
2360
2361
2362
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
2363
                        _log_warning('Overriding the parameters from Reference Dataset.')
2364
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2365
                if self.used_indices is None:
2366
                    # create valid
2367
                    self._lazy_init(data=self.data, label=self.label, reference=self.reference,
2368
                                    weight=self.weight, group=self.group, position=self.position,
2369
                                    init_score=self.init_score, predictor=self._predictor,
2370
                                    feature_name=self.feature_name, categorical_feature='auto', params=self.params)
wxchan's avatar
wxchan committed
2371
                else:
2372
                    # construct subset
2373
                    used_indices = _list_to_1d_numpy(self.used_indices, dtype=np.int32, name='used_indices')
2374
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2375
                    if self.reference.group is not None:
2376
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2377
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
2378
                                                  return_counts=True)
2379
                    self._handle = ctypes.c_void_p()
2380
                    params_str = _param_dict_to_str(self.params)
wxchan's avatar
wxchan committed
2381
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
2382
                        self.reference.construct()._handle,
wxchan's avatar
wxchan committed
2383
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
2384
                        ctypes.c_int32(used_indices.shape[0]),
2385
                        _c_str(params_str),
2386
                        ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
2387
2388
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2389
2390
                    if self.group is not None:
                        self.set_group(self.group)
2391
2392
                    if self.position is not None:
                        self.set_position(self.position)
wxchan's avatar
wxchan committed
2393
2394
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
2395
2396
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
2397
2398
2399
2400
2401
                        self._set_init_score_by_predictor(
                            predictor=self._predictor,
                            data=self.data,
                            used_indices=used_indices
                        )
wxchan's avatar
wxchan committed
2402
            else:
2403
                # create train
2404
                self._lazy_init(data=self.data, label=self.label, reference=None,
2405
2406
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
2407
2408
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature,
                                params=self.params, position=self.position)
wxchan's avatar
wxchan committed
2409
2410
            if self.free_raw_data:
                self.data = None
2411
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2412
        return self
wxchan's avatar
wxchan committed
2413

2414
2415
    def create_valid(
        self,
2416
        data: _LGBM_TrainDataType,
2417
        label: Optional[_LGBM_LabelType] = None,
2418
2419
2420
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2421
2422
        params: Optional[Dict[str, Any]] = None,
        position: Optional[_LGBM_PositionType] = None
2423
    ) -> "Dataset":
2424
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2425
2426
2427

        Parameters
        ----------
2428
        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
2429
            Data source of Dataset.
2430
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2431
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2432
            Label of the data.
2433
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2434
            Weight for each instance. Weights should be non-negative.
2435
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
2436
2437
2438
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2439
2440
            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.
2441
        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)
2442
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2443
        params : dict or None, optional (default=None)
2444
            Other parameters for validation Dataset.
2445
2446
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
2447
2448
2449

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2450
2451
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2452
        """
2453
        ret = Dataset(data, label=label, reference=self,
2454
                      weight=weight, group=group, position=position, init_score=init_score,
2455
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2456
        ret._predictor = self._predictor
2457
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2458
        return ret
wxchan's avatar
wxchan committed
2459

2460
2461
2462
2463
2464
    def subset(
        self,
        used_indices: List[int],
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2465
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2466
2467
2468
2469

        Parameters
        ----------
        used_indices : list of int
2470
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2471
        params : dict or None, optional (default=None)
2472
            These parameters will be passed to Dataset constructor.
2473
2474
2475
2476
2477

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2478
        """
wxchan's avatar
wxchan committed
2479
2480
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
2481
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
2482
2483
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2484
        ret._predictor = self._predictor
2485
        ret.pandas_categorical = self.pandas_categorical
2486
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2487
2488
        return ret

2489
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2490
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2491

2492
2493
2494
2495
2496
        .. 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
2497
2498
        Parameters
        ----------
2499
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2500
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2501
2502
2503
2504
2505

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2506
2507
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
2508
            self.construct()._handle,
2509
            _c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
2510
        return self
wxchan's avatar
wxchan committed
2511

2512
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2513
2514
        if not params:
            return self
2515
        params = deepcopy(params)
2516
2517
2518
2519
2520

        def update():
            if not self.params:
                self.params = params
            else:
2521
                self._params_back_up = deepcopy(self.params)
2522
2523
                self.params.update(params)

2524
        if self._handle is None:
2525
2526
2527
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2528
2529
                _c_str(_param_dict_to_str(self.params)),
                _c_str(_param_dict_to_str(params)))
2530
2531
2532
2533
2534
2535
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2536
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
2537
        return self
wxchan's avatar
wxchan committed
2538

2539
    def _reverse_update_params(self) -> "Dataset":
2540
        if self._handle is None:
2541
2542
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2543
        return self
2544

2545
2546
2547
    def set_field(
        self,
        field_name: str,
2548
        data: Optional[Union[List[List[float]], List[List[int]], List[float], List[int], np.ndarray, pd_Series, pd_DataFrame, pa_Array, pa_ChunkedArray]]
2549
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2550
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2551
2552
2553

        Parameters
        ----------
2554
        field_name : str
2555
            The field name of the information.
2556
        data : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), pyarrow Array, pyarrow ChunkedArray or None
2557
            The data to be set.
Nikita Titov's avatar
Nikita Titov committed
2558
2559
2560
2561
2562

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2563
        """
2564
        if self._handle is None:
2565
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2566
        if data is None:
2567
            # set to None
wxchan's avatar
wxchan committed
2568
            _safe_call(_LIB.LGBM_DatasetSetField(
2569
                self._handle,
2570
                _c_str(field_name),
wxchan's avatar
wxchan committed
2571
                None,
Guolin Ke's avatar
Guolin Ke committed
2572
                ctypes.c_int(0),
2573
                ctypes.c_int(_FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2574
            return self
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588

        # If the data is a arrow data, we can just pass it to C
        if _is_pyarrow_array(data):
            c_array = _export_arrow_to_c(data)
            _safe_call(_LIB.LGBM_DatasetSetFieldFromArrow(
                self._handle,
                _c_str(field_name),
                ctypes.c_int64(c_array.n_chunks),
                ctypes.c_void_p(c_array.chunks_ptr),
                ctypes.c_void_p(c_array.schema_ptr),
            ))
            self.version += 1
            return self

2589
        dtype: "np.typing.DTypeLike"
2590
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2591
            dtype = np.float64
2592
            if _is_1d_collection(data):
2593
                data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2594
            elif _is_2d_collection(data):
2595
                data = _data_to_2d_numpy(data, dtype=dtype, name=field_name)
2596
2597
2598
2599
2600
2601
2602
                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:
2603
            dtype = np.int32 if (field_name == 'group' or field_name == 'position') else np.float32
2604
            data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2605

2606
        ptr_data: Union[_ctypes_float_ptr, _ctypes_int_ptr]
2607
        if data.dtype == np.float32 or data.dtype == np.float64:
2608
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2609
        elif data.dtype == np.int32:
2610
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2611
        else:
2612
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2613
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2614
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2615
        _safe_call(_LIB.LGBM_DatasetSetField(
2616
            self._handle,
2617
            _c_str(field_name),
wxchan's avatar
wxchan committed
2618
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2619
2620
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2621
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2622
        return self
wxchan's avatar
wxchan committed
2623

2624
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2625
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2626

2627
2628
2629
2630
2631
2632
        Can only be run on a constructed Dataset.

        Unlike ``get_group()``, ``get_init_score()``, ``get_label()``, ``get_position()``, and ``get_weight()``,
        this method ignores any raw data passed into ``lgb.Dataset()`` on the Python side, and will only read
        data from the constructed C++ ``Dataset`` object.

wxchan's avatar
wxchan committed
2633
2634
        Parameters
        ----------
2635
        field_name : str
2636
            The field name of the information.
wxchan's avatar
wxchan committed
2637
2638
2639

        Returns
        -------
2640
        info : numpy array or None
2641
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2642
        """
2643
        if self._handle is None:
2644
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2645
2646
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2647
2648
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
2649
            self._handle,
2650
            _c_str(field_name),
wxchan's avatar
wxchan committed
2651
2652
2653
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
2654
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2655
2656
2657
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2658
        if out_type.value == _C_API_DTYPE_INT32:
2659
2660
2661
2662
            arr = _cint32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)),
                length=tmp_out_len.value
            )
2663
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2664
2665
2666
2667
            arr = _cfloat32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)),
                length=tmp_out_len.value
            )
2668
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2669
2670
2671
2672
            arr = _cfloat64_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)),
                length=tmp_out_len.value
            )
2673
        else:
wxchan's avatar
wxchan committed
2674
            raise TypeError("Unknown type")
2675
2676
2677
2678
2679
2680
        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
2681

2682
2683
    def set_categorical_feature(
        self,
2684
        categorical_feature: _LGBM_CategoricalFeatureConfiguration
2685
    ) -> "Dataset":
2686
        """Set categorical features.
2687
2688
2689

        Parameters
        ----------
2690
        categorical_feature : list of str or int, or 'auto'
2691
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2692
2693
2694
2695
2696

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2697
2698
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2699
            return self
2700
        if self.data is not None:
2701
2702
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2703
                return self._free_handle()
2704
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2705
                return self
2706
            else:
2707
2708
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
2709
                                 f'New categorical_feature is {list(categorical_feature)}')
2710
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2711
                return self._free_handle()
2712
        else:
2713
2714
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2715

2716
2717
2718
2719
    def _set_predictor(
        self,
        predictor: Optional[_InnerPredictor]
    ) -> "Dataset":
2720
2721
2722
2723
        """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
2724
        """
2725
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2726
            return self
2727
2728
2729
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
            if (predictor == self._predictor) and (predictor.current_iteration() == self._predictor.current_iteration()):
                return self
2730
        if self._handle is None:
Guolin Ke's avatar
Guolin Ke committed
2731
            self._predictor = predictor
2732
2733
        elif self.data is not None:
            self._predictor = predictor
2734
2735
2736
2737
2738
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
                used_indices=None
            )
2739
2740
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2741
2742
2743
2744
2745
            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
2746
        else:
2747
2748
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2749
        return self
Guolin Ke's avatar
Guolin Ke committed
2750

2751
    def set_reference(self, reference: "Dataset") -> "Dataset":
2752
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2753
2754
2755
2756

        Parameters
        ----------
        reference : Dataset
2757
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2758
2759
2760
2761
2762

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2763
        """
2764
2765
2766
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2767
        # we're done if self and reference share a common upstream reference
2768
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2769
            return self
Guolin Ke's avatar
Guolin Ke committed
2770
2771
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2772
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2773
        else:
2774
2775
            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
2776

2777
    def set_feature_name(self, feature_name: _LGBM_FeatureNameConfiguration) -> "Dataset":
2778
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2779
2780
2781

        Parameters
        ----------
2782
        feature_name : list of str
2783
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2784
2785
2786
2787
2788

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2789
        """
2790
2791
        if feature_name != 'auto':
            self.feature_name = feature_name
2792
        if self._handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2793
            if len(feature_name) != self.num_feature():
2794
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2795
            c_feature_name = [_c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2796
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
2797
                self._handle,
2798
                _c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2799
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2800
        return self
Guolin Ke's avatar
Guolin Ke committed
2801

2802
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
2803
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2804
2805
2806

        Parameters
        ----------
2807
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None
2808
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2809
2810
2811
2812
2813

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2814
2815
        """
        self.label = label
2816
        if self._handle is not None:
2817
2818
2819
            if isinstance(label, pd_DataFrame):
                if len(label.columns) > 1:
                    raise ValueError('DataFrame for label cannot have multiple columns')
2820
                label_array = np.ravel(_pandas_to_numpy(label, target_dtype=np.float32))
2821
2822
            elif _is_pyarrow_array(label):
                label_array = label
2823
            else:
2824
                label_array = _list_to_1d_numpy(label, dtype=np.float32, name='label')
2825
            self.set_field('label', label_array)
2826
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2827
        return self
Guolin Ke's avatar
Guolin Ke committed
2828

2829
2830
2831
2832
    def set_weight(
        self,
        weight: Optional[_LGBM_WeightType]
    ) -> "Dataset":
2833
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2834
2835
2836

        Parameters
        ----------
2837
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None
2838
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2839
2840
2841
2842
2843

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2844
        """
2845
2846
2847
2848
2849
2850
2851
        # Check if the weight contains values other than one
        if weight is not None:
            if _is_pyarrow_array(weight):
                if pa_compute.all(pa_compute.equal(weight, 1)).as_py():
                    weight = None
            elif np.all(weight == 1):
                weight = None
Guolin Ke's avatar
Guolin Ke committed
2852
        self.weight = weight
2853
2854

        # Set field
2855
        if self._handle is not None and weight is not None:
2856
2857
            if not _is_pyarrow_array(weight):
                weight = _list_to_1d_numpy(weight, dtype=np.float32, name='weight')
wxchan's avatar
wxchan committed
2858
            self.set_field('weight', weight)
2859
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2860
        return self
Guolin Ke's avatar
Guolin Ke committed
2861

2862
2863
2864
2865
    def set_init_score(
        self,
        init_score: Optional[_LGBM_InitScoreType]
    ) -> "Dataset":
2866
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2867
2868
2869

        Parameters
        ----------
2870
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2871
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2872
2873
2874
2875
2876

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2877
2878
        """
        self.init_score = init_score
2879
        if self._handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2880
            self.set_field('init_score', init_score)
2881
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2882
        return self
Guolin Ke's avatar
Guolin Ke committed
2883

2884
2885
2886
2887
    def set_group(
        self,
        group: Optional[_LGBM_GroupType]
    ) -> "Dataset":
2888
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2889
2890
2891

        Parameters
        ----------
2892
        group : list, numpy 1-D array, pandas Series or None
2893
2894
2895
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2896
2897
            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
2898
2899
2900
2901
2902

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2903
2904
        """
        self.group = group
2905
        if self._handle is not None and group is not None:
2906
            group = _list_to_1d_numpy(group, dtype=np.int32, name='group')
wxchan's avatar
wxchan committed
2907
            self.set_field('group', group)
2908
2909
2910
2911
            # original values can be modified at cpp side
            constructed_group = self.get_field('group')
            if constructed_group is not None:
                self.group = np.diff(constructed_group)
Nikita Titov's avatar
Nikita Titov committed
2912
        return self
Guolin Ke's avatar
Guolin Ke committed
2913

2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
    def set_position(
        self,
        position: Optional[_LGBM_PositionType]
    ) -> "Dataset":
        """Set position of Dataset (used for ranking).

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

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

2936
    def get_feature_name(self) -> List[str]:
2937
2938
2939
2940
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2941
        feature_names : list of str
2942
2943
            The names of columns (features) in the Dataset.
        """
2944
        if self._handle is None:
2945
2946
2947
2948
2949
            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)
2950
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2951
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
2952
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
2953
            self._handle,
2954
            ctypes.c_int(num_feature),
2955
            ctypes.byref(tmp_out_len),
2956
            ctypes.c_size_t(reserved_string_buffer_size),
2957
2958
2959
2960
            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")
2961
2962
2963
2964
        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)]
2965
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
2966
            _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
2967
                self._handle,
2968
2969
2970
2971
2972
                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))
2973
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2974

2975
    def get_label(self) -> Optional[_LGBM_LabelType]:
2976
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2977
2978
2979

        Returns
        -------
2980
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2981
            The label information from the Dataset.
2982
            For a constructed ``Dataset``, this will only return a numpy array.
Guolin Ke's avatar
Guolin Ke committed
2983
        """
2984
        if self.label is None:
wxchan's avatar
wxchan committed
2985
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2986
2987
        return self.label

2988
    def get_weight(self) -> Optional[_LGBM_WeightType]:
2989
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2990
2991
2992

        Returns
        -------
2993
        weight : list, numpy 1-D array, pandas Series or None
2994
            Weight for each data point from the Dataset. Weights should be non-negative.
2995
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
2996
        """
2997
        if self.weight is None:
wxchan's avatar
wxchan committed
2998
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2999
3000
        return self.weight

3001
    def get_init_score(self) -> Optional[_LGBM_InitScoreType]:
3002
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3003
3004
3005

        Returns
        -------
3006
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
3007
            Init score of Booster.
3008
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3009
        """
3010
        if self.init_score is None:
wxchan's avatar
wxchan committed
3011
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
3012
3013
        return self.init_score

3014
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
3015
3016
3017
3018
        """Get the raw data of the Dataset.

        Returns
        -------
3019
        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
3020
3021
            Raw data used in the Dataset construction.
        """
3022
        if self._handle is None:
3023
            raise Exception("Cannot get data before construct Dataset")
3024
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
3025
3026
            self.data = self.reference.data
            if self.data is not None:
3027
                if isinstance(self.data, np.ndarray) or isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3028
                    self.data = self.data[self.used_indices, :]
3029
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3030
                    self.data = self.data.iloc[self.used_indices].copy()
3031
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3032
                    self.data = self.data[self.used_indices, :]
3033
3034
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
3035
                elif _is_list_of_sequences(self.data) and len(self.data) > 0:
3036
                    self.data = np.array(list(self._yield_row_from_seqlist(self.data, self.used_indices)))
Guolin Ke's avatar
Guolin Ke committed
3037
                else:
3038
3039
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
3040
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
3041
3042
3043
        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.")
3044
3045
        return self.data

3046
    def get_group(self) -> Optional[_LGBM_GroupType]:
3047
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3048
3049
3050

        Returns
        -------
3051
        group : list, numpy 1-D array, pandas Series or None
3052
3053
3054
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
3055
3056
            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.
3057
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3058
        """
3059
        if self.group is None:
wxchan's avatar
wxchan committed
3060
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
3061
3062
            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
3063
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
3064
3065
        return self.group

3066
    def get_position(self) -> Optional[_LGBM_PositionType]:
3067
3068
3069
3070
        """Get the position of the Dataset.

        Returns
        -------
3071
        position : numpy 1-D array, pandas Series or None
3072
            Position of items used in unbiased learning-to-rank task.
3073
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
3074
3075
3076
3077
3078
        """
        if self.position is None:
            self.position = self.get_field('position')
        return self.position

3079
    def num_data(self) -> int:
3080
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3081
3082
3083

        Returns
        -------
3084
3085
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3086
        """
3087
        if self._handle is not None:
3088
            ret = ctypes.c_int(0)
3089
            _safe_call(_LIB.LGBM_DatasetGetNumData(self._handle,
wxchan's avatar
wxchan committed
3090
3091
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3092
        else:
3093
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3094

3095
    def num_feature(self) -> int:
3096
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3097
3098
3099

        Returns
        -------
3100
3101
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3102
        """
3103
        if self._handle is not None:
3104
            ret = ctypes.c_int(0)
3105
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self._handle,
wxchan's avatar
wxchan committed
3106
3107
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3108
        else:
3109
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3110

3111
    def feature_num_bin(self, feature: Union[int, str]) -> int:
3112
3113
        """Get the number of bins for a feature.

3114
3115
        .. versionadded:: 4.0.0

3116
3117
        Parameters
        ----------
3118
3119
        feature : int or str
            Index or name of the feature.
3120
3121
3122
3123
3124
3125

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
3126
        if self._handle is not None:
3127
            if isinstance(feature, str):
3128
3129
3130
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
3131
            ret = ctypes.c_int(0)
3132
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self._handle,
3133
                                                         ctypes.c_int(feature_index),
3134
3135
3136
3137
3138
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

3139
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
3140
3141
3142
3143
3144
        """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.
3145
3146
3147
3148
3149

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
3150
3151
3152

        Returns
        -------
3153
3154
3155
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
3156
        head = self
3157
        ref_chain: Set[Dataset] = set()
3158
3159
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
3160
                ref_chain.add(head)
3161
3162
3163
3164
3165
3166
                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
3167
        return ref_chain
3168

3169
    def add_features_from(self, other: "Dataset") -> "Dataset":
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
        """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.
        """
3184
        if self._handle is None or other._handle is None:
3185
            raise ValueError('Both source and target Datasets must be constructed before adding features')
3186
        _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self._handle, other._handle))
Guolin Ke's avatar
Guolin Ke committed
3187
3188
3189
3190
3191
3192
3193
3194
        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))
3195
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3196
                    self.data = np.hstack((self.data, other.data.toarray()))
3197
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3198
                    self.data = np.hstack((self.data, other.data.values))
3199
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3200
3201
3202
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
3203
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3204
                sparse_format = self.data.getformat()
3205
                if isinstance(other.data, np.ndarray) or isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3206
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
3207
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3208
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
3209
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3210
3211
3212
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
3213
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3214
3215
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
3216
3217
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
3218
                if isinstance(other.data, np.ndarray):
3219
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
3220
                                       axis=1, ignore_index=True)
3221
                elif isinstance(other.data, scipy.sparse.spmatrix):
3222
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
3223
                                       axis=1, ignore_index=True)
3224
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3225
3226
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
3227
3228
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
3229
3230
3231
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
3232
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3233
                if isinstance(other.data, np.ndarray):
3234
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
3235
                elif isinstance(other.data, scipy.sparse.spmatrix):
3236
3237
3238
3239
3240
                    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
3241
3242
3243
3244
3245
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
3246
3247
            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
3248
3249
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
3250
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
3251
        self.feature_name = self.get_feature_name()
3252
3253
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
3254
3255
        self.categorical_feature = "auto"
        self.pandas_categorical = None
3256
3257
        return self

3258
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
3259
3260
3261
3262
3263
3264
        """Save Dataset to a text file.

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

        Parameters
        ----------
3265
        filename : str or pathlib.Path
3266
3267
3268
3269
3270
3271
3272
3273
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
3274
            self.construct()._handle,
3275
            _c_str(str(filename))))
3276
3277
        return self

wxchan's avatar
wxchan committed
3278

3279
3280
3281
3282
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
    Tuple[np.ndarray, np.ndarray]
]
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
        _LGBM_EvalFunctionResultType
    ],
    Callable[
        [np.ndarray, Dataset],
        List[_LGBM_EvalFunctionResultType]
    ]
]
3293
3294


3295
class Booster:
3296
    """Booster in LightGBM."""
3297

3298
3299
3300
3301
3302
3303
3304
    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
    ):
3305
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3306
3307
3308

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3309
        params : dict or None, optional (default=None)
3310
3311
3312
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3313
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3314
            Path to the model file.
3315
        model_str : str or None, optional (default=None)
3316
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3317
        """
3318
        self._handle = ctypes.c_void_p()
3319
        self._network = False
wxchan's avatar
wxchan committed
3320
        self.__need_reload_eval_info = True
3321
        self._train_data_name = "training"
3322
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3323
        self.best_iteration = -1
3324
        self.best_score: _LGBM_BoosterBestScoreType = {}
3325
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3326
        if train_set is not None:
3327
            # Training task
wxchan's avatar
wxchan committed
3328
            if not isinstance(train_set, Dataset):
3329
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
            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"]
                )
3364
            # construct booster object
3365
3366
3367
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3368
            params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3369
            _safe_call(_LIB.LGBM_BoosterCreate(
3370
                train_set._handle,
3371
                _c_str(params_str),
3372
                ctypes.byref(self._handle)))
3373
            # save reference to data
wxchan's avatar
wxchan committed
3374
            self.train_set = train_set
3375
3376
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3377
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3378
3379
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
3380
                _safe_call(_LIB.LGBM_BoosterMerge(
3381
3382
                    self._handle,
                    self.__init_predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
3383
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3384
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3385
                self._handle,
wxchan's avatar
wxchan committed
3386
3387
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3388
            # buffer for inner predict
3389
            self.__inner_predict_buffer: List[Optional[np.ndarray]] = [None]
wxchan's avatar
wxchan committed
3390
3391
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3392
            self.pandas_categorical = train_set.pandas_categorical
3393
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3394
        elif model_file is not None:
3395
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3396
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3397
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
3398
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
3399
                ctypes.byref(out_num_iterations),
3400
                ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
3401
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3402
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3403
                self._handle,
wxchan's avatar
wxchan committed
3404
3405
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3406
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3407
3408
3409
            if params:
                _log_warning('Ignoring params argument, using parameters from model file.')
            params = self._get_loaded_param()
3410
        elif model_str is not None:
3411
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3412
        else:
3413
3414
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
3415
        self.params = params
wxchan's avatar
wxchan committed
3416

3417
    def __del__(self) -> None:
3418
        try:
3419
            if self._network:
3420
3421
3422
3423
                self.free_network()
        except AttributeError:
            pass
        try:
3424
3425
            if self._handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
3426
3427
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3428

3429
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3430
3431
        return self.__deepcopy__(None)

3432
    def __deepcopy__(self, _) -> "Booster":
3433
        model_str = self.model_to_string(num_iteration=-1)
3434
        return Booster(model_str=model_str)
wxchan's avatar
wxchan committed
3435

3436
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3437
        this = self.__dict__.copy()
3438
        handle = this['_handle']
wxchan's avatar
wxchan committed
3439
3440
3441
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
3442
            this["_handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3443
3444
        return this

3445
    def __setstate__(self, state: Dict[str, Any]) -> None:
3446
        model_str = state.get('_handle', state.get('handle', None))
3447
        if model_str is not None:
wxchan's avatar
wxchan committed
3448
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3449
            out_num_iterations = ctypes.c_int(0)
3450
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3451
                _c_str(model_str),
3452
3453
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
3454
            state['_handle'] = handle
wxchan's avatar
wxchan committed
3455
3456
        self.__dict__.update(state)

3457
3458
3459
3460
    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)
3461
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3462
        _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
3463
            self._handle,
3464
3465
3466
3467
3468
3469
3470
            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)
3471
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3472
            _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
3473
                self._handle,
3474
3475
3476
3477
3478
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
        return json.loads(string_buffer.value.decode('utf-8'))

3479
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3480
3481
3482
3483
3484
3485
3486
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
3487
3488
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
3489
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3490
        return self
wxchan's avatar
wxchan committed
3491

3492
    def _free_buffer(self) -> "Booster":
3493
3494
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3495
        return self
3496

3497
3498
3499
3500
3501
3502
3503
    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":
3504
3505
3506
3507
        """Set the network configuration.

        Parameters
        ----------
3508
        machines : list, set or str
3509
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3510
        local_listen_port : int, optional (default=12400)
3511
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3512
        listen_time_out : int, optional (default=120)
3513
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3514
        num_machines : int, optional (default=1)
3515
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3516
3517
3518
3519
3520

        Returns
        -------
        self : Booster
            Booster with set network.
3521
        """
3522
3523
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
3524
        _safe_call(_LIB.LGBM_NetworkInit(_c_str(machines),
3525
3526
3527
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
3528
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3529
        return self
3530

3531
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3532
3533
3534
3535
3536
3537
3538
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3539
        _safe_call(_LIB.LGBM_NetworkFree())
3540
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3541
        return self
3542

3543
    def trees_to_dataframe(self) -> pd_DataFrame:
3544
3545
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3546
3547
3548
3549
        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.
3550
3551
3552
3553
3554
            - ``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.
3555
3556
            - ``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.
3557
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3558
3559
              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.
3560
3561
            - ``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.
3562
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3563
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3564
3565
            - ``count`` : int64, number of records in the training data that fall into this node.

3566
3567
3568
3569
3570
3571
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3572
3573
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
3574
3575
3576
3577

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

3578
        def _is_split_node(tree: Dict[str, Any]) -> bool:
3579
3580
            return 'split_index' in tree.keys()

3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
        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:
3593
                tree_num = f'{tree_index}-' if tree_index is not None else ''
3594
3595
3596
                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
3597
3598
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
3599

3600
3601
3602
3603
            def _get_split_feature(
                tree: Dict[str, Any],
                feature_names: Optional[List[str]]
            ) -> Optional[str]:
3604
3605
3606
3607
3608
3609
3610
3611
3612
                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

3613
            def _is_single_node_tree(tree: Dict[str, Any]) -> bool:
3614
                return set(tree.keys()) == {'leaf_value'}
3615
3616

            # Create the node record, and populate universal data members
3617
            node: Dict[str, Union[int, str, None]] = OrderedDict()
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
            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

3654
3655
3656
3657
3658
3659
3660
        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]]:
3661

3662
            node = create_node_record(tree=tree,
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
                                      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(
3675
                        tree=tree[child],
3676
3677
3678
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
3679
3680
                        parent_node=node['node_index']
                    )
3681
3682
3683
3684
3685
3686
3687
3688
3689
                    # 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']:
3690
            model_list.extend(tree_dict_to_node_list(tree=tree['tree_structure'],
3691
3692
3693
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

3694
        return pd_DataFrame(model_list, columns=model_list[0].keys())
3695

3696
    def set_train_data_name(self, name: str) -> "Booster":
3697
3698
3699
3700
        """Set the name to the training Dataset.

        Parameters
        ----------
3701
        name : str
Nikita Titov's avatar
Nikita Titov committed
3702
3703
3704
3705
3706
3707
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
3708
        """
3709
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
3710
        return self
wxchan's avatar
wxchan committed
3711

3712
    def add_valid(self, data: Dataset, name: str) -> "Booster":
3713
        """Add validation data.
wxchan's avatar
wxchan committed
3714
3715
3716
3717

        Parameters
        ----------
        data : Dataset
3718
            Validation data.
3719
        name : str
3720
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
3721
3722
3723
3724
3725

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
3726
        """
Guolin Ke's avatar
Guolin Ke committed
3727
        if not isinstance(data, Dataset):
3728
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3729
        if data._predictor is not self.__init_predictor:
3730
3731
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3732
        _safe_call(_LIB.LGBM_BoosterAddValidData(
3733
3734
            self._handle,
            data.construct()._handle))
wxchan's avatar
wxchan committed
3735
3736
3737
3738
3739
        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
3740
        return self
wxchan's avatar
wxchan committed
3741

3742
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
3743
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
3744
3745
3746
3747

        Parameters
        ----------
        params : dict
3748
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
3749
3750
3751
3752
3753

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
3754
        """
3755
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3756
3757
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
3758
                self._handle,
3759
                _c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
3760
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
3761
        return self
wxchan's avatar
wxchan committed
3762

3763
3764
3765
3766
3767
    def update(
        self,
        train_set: Optional[Dataset] = None,
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
3768
        """Update Booster for one iteration.
3769

wxchan's avatar
wxchan committed
3770
3771
        Parameters
        ----------
3772
3773
3774
3775
        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
3776
            Customized objective function.
3777
3778
3779
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3780
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3781
                    The predicted values.
3782
3783
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3784
3785
                train_data : Dataset
                    The training dataset.
3786
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3787
3788
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3789
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3790
3791
                    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
3792

3793
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3794
            and grad and hess should be returned in the same format.
3795

wxchan's avatar
wxchan committed
3796
3797
        Returns
        -------
3798
3799
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3800
        """
3801
        # need reset training data
3802
3803
3804
3805
3806
3807
        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
3808
            if not isinstance(train_set, Dataset):
3809
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3810
            if train_set._predictor is not self.__init_predictor:
3811
3812
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3813
3814
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
3815
3816
                self._handle,
                self.train_set.construct()._handle))
wxchan's avatar
wxchan committed
3817
            self.__inner_predict_buffer[0] = None
3818
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3819
3820
        is_finished = ctypes.c_int(0)
        if fobj is None:
3821
            if self.__set_objective_to_none:
3822
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3823
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
3824
                self._handle,
wxchan's avatar
wxchan committed
3825
                ctypes.byref(is_finished)))
3826
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3827
3828
            return is_finished.value == 1
        else:
3829
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3830
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3831
3832
3833
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

3834
3835
3836
3837
3838
    def __boost(
        self,
        grad: np.ndarray,
        hess: np.ndarray
    ) -> bool:
3839
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3840

Nikita Titov's avatar
Nikita Titov committed
3841
3842
        .. note::

3843
3844
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3845
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3846
            and grad and hess should be returned in the same format.
3847

wxchan's avatar
wxchan committed
3848
3849
        Parameters
        ----------
3850
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3851
3852
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3853
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3854
3855
            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
3856
3857
3858

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3859
3860
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3861
        """
3862
3863
3864
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3865
3866
        grad = _list_to_1d_numpy(grad, dtype=np.float32, name='gradient')
        hess = _list_to_1d_numpy(hess, dtype=np.float32, name='hessian')
3867
3868
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3869
        if len(grad) != len(hess):
3870
3871
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3872
        if len(grad) != num_train_data * self.__num_class:
3873
3874
3875
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3876
                f"number of models per one iteration ({self.__num_class})"
3877
            )
wxchan's avatar
wxchan committed
3878
3879
        is_finished = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterUpdateOneIterCustom(
3880
            self._handle,
wxchan's avatar
wxchan committed
3881
3882
3883
            grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            ctypes.byref(is_finished)))
3884
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3885
3886
        return is_finished.value == 1

3887
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3888
3889
3890
3891
3892
3893
3894
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3895
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
3896
            self._handle))
3897
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3898
        return self
wxchan's avatar
wxchan committed
3899

3900
    def current_iteration(self) -> int:
3901
3902
3903
3904
3905
3906
3907
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3908
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3909
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
3910
            self._handle,
wxchan's avatar
wxchan committed
3911
3912
3913
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3914
    def num_model_per_iteration(self) -> int:
3915
3916
3917
3918
3919
3920
3921
3922
3923
        """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(
3924
            self._handle,
3925
3926
3927
            ctypes.byref(model_per_iter)))
        return model_per_iter.value

3928
    def num_trees(self) -> int:
3929
3930
3931
3932
3933
3934
3935
3936
3937
        """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(
3938
            self._handle,
3939
3940
3941
            ctypes.byref(num_trees)))
        return num_trees.value

3942
    def upper_bound(self) -> float:
3943
3944
3945
3946
        """Get upper bound value of a model.

        Returns
        -------
3947
        upper_bound : float
3948
3949
3950
3951
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
3952
            self._handle,
3953
3954
3955
            ctypes.byref(ret)))
        return ret.value

3956
    def lower_bound(self) -> float:
3957
3958
3959
3960
        """Get lower bound value of a model.

        Returns
        -------
3961
        lower_bound : float
3962
3963
3964
3965
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
3966
            self._handle,
3967
3968
3969
            ctypes.byref(ret)))
        return ret.value

3970
3971
3972
3973
3974
3975
    def eval(
        self,
        data: Dataset,
        name: str,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3976
        """Evaluate for data.
wxchan's avatar
wxchan committed
3977
3978
3979

        Parameters
        ----------
3980
3981
        data : Dataset
            Data for the evaluating.
3982
        name : str
3983
            Name of the data.
3984
        feval : callable, list of callable, or None, optional (default=None)
3985
            Customized evaluation function.
3986
            Each evaluation function should accept two parameters: preds, eval_data,
3987
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3988

3989
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3990
                    The predicted values.
3991
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3992
                    If custom objective function is used, predicted values are returned before any transformation,
3993
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3994
                eval_data : Dataset
3995
                    A ``Dataset`` to evaluate.
3996
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3997
                    The name of evaluation function (without whitespace).
3998
3999
4000
4001
4002
                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
4003
4004
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4005
        result : list
4006
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4007
        """
Guolin Ke's avatar
Guolin Ke committed
4008
4009
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
4010
4011
4012
4013
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
4014
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
4015
4016
4017
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
4018
        # need to push new valid data
wxchan's avatar
wxchan committed
4019
4020
4021
4022
4023
4024
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

4025
4026
4027
4028
    def eval_train(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4029
        """Evaluate for training data.
wxchan's avatar
wxchan committed
4030
4031
4032

        Parameters
        ----------
4033
        feval : callable, list of callable, or None, optional (default=None)
4034
            Customized evaluation function.
4035
            Each evaluation function should accept two parameters: preds, eval_data,
4036
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4037

4038
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4039
                    The predicted values.
4040
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4041
                    If custom objective function is used, predicted values are returned before any transformation,
4042
                    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
4043
                eval_data : Dataset
4044
                    The training dataset.
4045
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4046
                    The name of evaluation function (without whitespace).
4047
4048
4049
4050
4051
                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
4052
4053
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4054
        result : list
4055
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4056
        """
4057
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
4058

4059
4060
4061
4062
    def eval_valid(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4063
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
4064
4065
4066

        Parameters
        ----------
4067
        feval : callable, list of callable, or None, optional (default=None)
4068
            Customized evaluation function.
4069
            Each evaluation function should accept two parameters: preds, eval_data,
4070
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4071

4072
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4073
                    The predicted values.
4074
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4075
                    If custom objective function is used, predicted values are returned before any transformation,
4076
                    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
4077
                eval_data : Dataset
4078
                    The validation dataset.
4079
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4080
                    The name of evaluation function (without whitespace).
4081
4082
4083
4084
4085
                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
4086
4087
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4088
        result : list
4089
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4090
        """
4091
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
4092
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
4093

4094
4095
4096
4097
4098
4099
4100
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> "Booster":
4101
        """Save Booster to file.
wxchan's avatar
wxchan committed
4102
4103
4104

        Parameters
        ----------
4105
        filename : str or pathlib.Path
4106
            Filename to save Booster.
4107
4108
4109
4110
        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
4111
        start_iteration : int, optional (default=0)
4112
            Start index of the iteration that should be saved.
4113
        importance_type : str, optional (default="split")
4114
4115
4116
            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
4117
4118
4119
4120
4121

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
4122
        """
4123
        if num_iteration is None:
4124
            num_iteration = self.best_iteration
4125
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4126
        _safe_call(_LIB.LGBM_BoosterSaveModel(
4127
            self._handle,
4128
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4129
            ctypes.c_int(num_iteration),
4130
            ctypes.c_int(importance_type_int),
4131
            _c_str(str(filename))))
4132
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
4133
        return self
wxchan's avatar
wxchan committed
4134

4135
4136
4137
4138
4139
    def shuffle_models(
        self,
        start_iteration: int = 0,
        end_iteration: int = -1
    ) -> "Booster":
4140
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
4141

4142
4143
4144
        Parameters
        ----------
        start_iteration : int, optional (default=0)
4145
            The first iteration that will be shuffled.
4146
4147
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
4148
            If <= 0, means the last available iteration.
4149

Nikita Titov's avatar
Nikita Titov committed
4150
4151
4152
4153
        Returns
        -------
        self : Booster
            Booster with shuffled models.
4154
        """
4155
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
4156
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
4157
4158
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
4159
        return self
4160

4161
    def model_from_string(self, model_str: str) -> "Booster":
4162
4163
4164
4165
        """Load Booster from a string.

        Parameters
        ----------
4166
        model_str : str
4167
4168
4169
4170
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4171
        self : Booster
4172
4173
            Loaded Booster object.
        """
4174
4175
4176
        # ensure that existing Booster is freed before replacing it
        # with a new one createdfrom file
        _safe_call(_LIB.LGBM_BoosterFree(self._handle))
4177
        self._free_buffer()
4178
        self._handle = ctypes.c_void_p()
4179
4180
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
4181
            _c_str(model_str),
4182
            ctypes.byref(out_num_iterations),
4183
            ctypes.byref(self._handle)))
4184
4185
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
4186
            self._handle,
4187
4188
            ctypes.byref(out_num_class)))
        self.__num_class = out_num_class.value
4189
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
4190
4191
        return self

4192
4193
4194
4195
4196
4197
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> str:
4198
        """Save Booster to string.
4199

4200
4201
4202
4203
4204
4205
        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
4206
        start_iteration : int, optional (default=0)
4207
            Start index of the iteration that should be saved.
4208
        importance_type : str, optional (default="split")
4209
4210
4211
            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.
4212
4213
4214

        Returns
        -------
4215
        str_repr : str
4216
4217
            String representation of Booster.
        """
4218
        if num_iteration is None:
4219
            num_iteration = self.best_iteration
4220
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4221
        buffer_len = 1 << 20
4222
        tmp_out_len = ctypes.c_int64(0)
4223
        string_buffer = ctypes.create_string_buffer(buffer_len)
4224
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4225
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
4226
            self._handle,
4227
            ctypes.c_int(start_iteration),
4228
            ctypes.c_int(num_iteration),
4229
            ctypes.c_int(importance_type_int),
4230
            ctypes.c_int64(buffer_len),
4231
4232
4233
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
4234
        # if buffer length is not long enough, re-allocate a buffer
4235
4236
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
4237
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4238
            _safe_call(_LIB.LGBM_BoosterSaveModelToString(
4239
                self._handle,
4240
                ctypes.c_int(start_iteration),
4241
                ctypes.c_int(num_iteration),
4242
                ctypes.c_int(importance_type_int),
4243
                ctypes.c_int64(actual_len),
4244
4245
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
4246
        ret = string_buffer.value.decode('utf-8')
4247
4248
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
4249

4250
4251
4252
4253
4254
4255
4256
    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
4257
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
4258

4259
4260
        Parameters
        ----------
4261
4262
4263
4264
        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
4265
        start_iteration : int, optional (default=0)
4266
            Start index of the iteration that should be dumped.
4267
        importance_type : str, optional (default="split")
4268
4269
4270
            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.
4271
4272
4273
4274
4275
4276
4277
4278
4279
        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.
4280

wxchan's avatar
wxchan committed
4281
4282
        Returns
        -------
4283
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
4284
            JSON format of Booster.
wxchan's avatar
wxchan committed
4285
        """
4286
        if num_iteration is None:
4287
            num_iteration = self.best_iteration
4288
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4289
        buffer_len = 1 << 20
4290
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
4291
        string_buffer = ctypes.create_string_buffer(buffer_len)
4292
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
wxchan's avatar
wxchan committed
4293
        _safe_call(_LIB.LGBM_BoosterDumpModel(
4294
            self._handle,
4295
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4296
            ctypes.c_int(num_iteration),
4297
            ctypes.c_int(importance_type_int),
4298
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
4299
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4300
            ptr_string_buffer))
wxchan's avatar
wxchan committed
4301
        actual_len = tmp_out_len.value
4302
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
4303
4304
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
4305
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
wxchan's avatar
wxchan committed
4306
            _safe_call(_LIB.LGBM_BoosterDumpModel(
4307
                self._handle,
4308
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4309
                ctypes.c_int(num_iteration),
4310
                ctypes.c_int(importance_type_int),
4311
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
4312
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4313
                ptr_string_buffer))
4314
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
4315
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
4316
                                                          default=_json_default_with_numpy))
4317
        return ret
wxchan's avatar
wxchan committed
4318

4319
4320
    def predict(
        self,
4321
        data: _LGBM_PredictDataType,
4322
4323
4324
4325
4326
4327
4328
4329
        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
4330
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4331
        """Make a prediction.
wxchan's avatar
wxchan committed
4332
4333
4334

        Parameters
        ----------
4335
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
4336
            Data source for prediction.
4337
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4338
        start_iteration : int, optional (default=0)
4339
            Start index of the iteration to predict.
4340
            If <= 0, starts from the first iteration.
4341
        num_iteration : int or None, optional (default=None)
4342
4343
4344
4345
            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).
4346
4347
4348
4349
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4350
4351
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4352

Nikita Titov's avatar
Nikita Titov committed
4353
4354
4355
4356
4357
4358
4359
            .. 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.
4360

4361
4362
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4363
            Used only if data is str.
4364
4365
4366
        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.
4367
4368
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4369
4370
4371

        Returns
        -------
4372
        result : numpy array, scipy.sparse or list of scipy.sparse
4373
            Prediction result.
4374
            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
4375
        """
4376
4377
4378
4379
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs),
        )
4380
        if num_iteration is None:
4381
            if start_iteration <= 0:
4382
4383
4384
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
        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
4395

4396
4397
    def refit(
        self,
4398
        data: _LGBM_TrainDataType,
4399
        label: _LGBM_LabelType,
4400
4401
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4402
4403
4404
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4405
4406
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
4407
4408
4409
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4410
        **kwargs
4411
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4412
4413
4414
4415
        """Refit the existing Booster by new data.

        Parameters
        ----------
4416
        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
4417
            Data source for refit.
4418
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4419
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array or pyarrow ChunkedArray
Guolin Ke's avatar
Guolin Ke committed
4420
4421
            Label for refit.
        decay_rate : float, optional (default=0.9)
4422
4423
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4424
4425
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
4426
4427
4428

            .. versionadded:: 4.0.0

4429
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
4430
            Weight for each ``data`` instance. Weights should be non-negative.
4431
4432
4433

            .. versionadded:: 4.0.0

4434
4435
4436
4437
4438
4439
        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.
4440
4441
4442

            .. versionadded:: 4.0.0

4443
4444
        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``.
4445
4446
4447

            .. versionadded:: 4.0.0

4448
4449
4450
        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.
4451
4452
4453

            .. versionadded:: 4.0.0

4454
4455
4456
4457
4458
        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.
4459
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4460
4461
4462
            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.
4463
            Floating point numbers in categorical features will be rounded towards 0.
4464
4465
4466

            .. versionadded:: 4.0.0

4467
4468
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
4469
4470
4471

            .. versionadded:: 4.0.0

4472
4473
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
4474
4475
4476

            .. versionadded:: 4.0.0

4477
4478
4479
        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.
4480
4481
4482

            .. versionadded:: 4.0.0

4483
4484
        **kwargs
            Other parameters for refit.
4485
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4486
4487
4488
4489
4490
4491

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4492
4493
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
4494
4495
        if dataset_params is None:
            dataset_params = {}
4496
4497
4498
4499
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs)
        )
4500
        leaf_preds: np.ndarray = predictor.predict(  # type: ignore[assignment]
4501
4502
4503
4504
4505
            data=data,
            start_iteration=-1,
            pred_leaf=True,
            validate_features=validate_features
        )
4506
        nrow, ncol = leaf_preds.shape
4507
        out_is_linear = ctypes.c_int(0)
4508
        _safe_call(_LIB.LGBM_BoosterGetLinear(
4509
            self._handle,
4510
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
4511
4512
4513
4514
4515
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
4516
        new_params["linear_tree"] = bool(out_is_linear.value)
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
        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,
        )
4530
        new_params['refit_decay_rate'] = decay_rate
4531
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4532
4533
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
4534
4535
            new_booster._handle,
            predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
4536
        leaf_preds = leaf_preds.reshape(-1)
4537
        ptr_data, _, _ = _c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
4538
        _safe_call(_LIB.LGBM_BoosterRefit(
4539
            new_booster._handle,
Guolin Ke's avatar
Guolin Ke committed
4540
            ptr_data,
4541
4542
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
4543
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4544
4545
        return new_booster

4546
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
        """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.
        """
4561
4562
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
4563
            self._handle,
4564
4565
4566
4567
4568
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

4569
4570
4571
4572
4573
4574
4575
4576
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
    ) -> 'Booster':
        """Set the output of a leaf.

4577
4578
        .. versionadded:: 4.0.0

4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
        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(
4595
                self._handle,
4596
4597
4598
4599
4600
4601
4602
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.c_double(value)
            )
        )
        return self

4603
    def num_feature(self) -> int:
4604
4605
4606
4607
4608
4609
4610
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4611
4612
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
4613
            self._handle,
4614
4615
4616
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

4617
    def feature_name(self) -> List[str]:
4618
        """Get names of features.
wxchan's avatar
wxchan committed
4619
4620
4621

        Returns
        -------
4622
        result : list of str
4623
            List with names of features.
wxchan's avatar
wxchan committed
4624
        """
4625
        num_feature = self.num_feature()
4626
        # Get name of features
wxchan's avatar
wxchan committed
4627
        tmp_out_len = ctypes.c_int(0)
4628
4629
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
4630
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
4631
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
wxchan's avatar
wxchan committed
4632
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
4633
            self._handle,
4634
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
4635
            ctypes.byref(tmp_out_len),
4636
            ctypes.c_size_t(reserved_string_buffer_size),
4637
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4638
4639
4640
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
4641
4642
4643
4644
        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)]
4645
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
4646
            _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
4647
                self._handle,
4648
4649
4650
4651
4652
                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))
4653
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
4654

4655
4656
4657
4658
4659
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
4660
        """Get feature importances.
4661

4662
4663
        Parameters
        ----------
4664
        importance_type : str, optional (default="split")
4665
4666
4667
            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.
4668
4669
4670
4671
        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).
4672

4673
4674
        Returns
        -------
4675
4676
        result : numpy array
            Array with feature importances.
4677
        """
4678
4679
        if iteration is None:
            iteration = self.best_iteration
4680
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4681
        result = np.empty(self.num_feature(), dtype=np.float64)
4682
        _safe_call(_LIB.LGBM_BoosterFeatureImportance(
4683
            self._handle,
4684
4685
4686
            ctypes.c_int(iteration),
            ctypes.c_int(importance_type_int),
            result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
4687
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
4688
            return result.astype(np.int32)
4689
4690
        else:
            return result
4691

4692
4693
4694
4695
4696
4697
    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]:
4698
4699
4700
4701
        """Get split value histogram for the specified feature.

        Parameters
        ----------
4702
        feature : int or str
4703
4704
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
4705
            If str, interpreted as name.
4706

Nikita Titov's avatar
Nikita Titov committed
4707
4708
4709
            .. warning::

                Categorical features are not supported.
4710

4711
        bins : int, str or None, optional (default=None)
4712
4713
4714
            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.
4715
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
        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.
        """
4730
        def add(root: Dict[str, Any]) -> None:
4731
4732
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
4733
                if feature_names is not None and isinstance(feature, str):
4734
4735
4736
4737
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
4738
                    if isinstance(root['threshold'], str):
4739
4740
4741
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
4742
4743
4744
4745
4746
4747
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
4748
        values: List[float] = []
4749
4750
4751
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

4752
        if bins is None or isinstance(bins, int) and xgboost_style:
4753
4754
4755
4756
4757
4758
4759
            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:
4760
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
4761
4762
4763
4764
4765
            else:
                return ret
        else:
            return hist, bin_edges

4766
4767
4768
4769
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
4770
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]]
4771
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4772
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
4773
        if data_idx >= self.__num_dataset:
4774
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4775
4776
4777
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
4778
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
4779
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
4780
            _safe_call(_LIB.LGBM_BoosterGetEval(
4781
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4782
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4783
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4784
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
4785
            if tmp_out_len.value != self.__num_inner_eval:
4786
                raise ValueError("Wrong length of eval results")
4787
            for i in range(self.__num_inner_eval):
4788
4789
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
4790
4791
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
4792
4793
4794
4795
4796
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
4797
4798
4799
4800
4801
4802
4803
4804
4805
            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
4806
4807
4808
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

4809
    def __inner_predict(self, data_idx: int) -> np.ndarray:
4810
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
4811
        if data_idx >= self.__num_dataset:
4812
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4813
4814
4815
4816
4817
        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
4818
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
4819
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
4820
4821
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
4822
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))  # type: ignore[union-attr]
wxchan's avatar
wxchan committed
4823
            _safe_call(_LIB.LGBM_BoosterGetPredict(
4824
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4825
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4826
4827
                ctypes.byref(tmp_out_len),
                data_ptr))
4828
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):  # type: ignore[arg-type]
4829
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
4830
            self.__is_predicted_cur_iter[data_idx] = True
4831
        result: np.ndarray = self.__inner_predict_buffer[data_idx]  # type: ignore[assignment]
4832
4833
4834
4835
        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
4836

4837
    def __get_eval_info(self) -> None:
4838
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
4839
4840
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
4841
            out_num_eval = ctypes.c_int(0)
4842
            # Get num of inner evals
wxchan's avatar
wxchan committed
4843
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
4844
                self._handle,
wxchan's avatar
wxchan committed
4845
4846
4847
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
4848
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
4849
                tmp_out_len = ctypes.c_int(0)
4850
4851
4852
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
4853
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
4854
                ]
4855
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
wxchan's avatar
wxchan committed
4856
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
4857
                    self._handle,
4858
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
4859
                    ctypes.byref(tmp_out_len),
4860
                    ctypes.c_size_t(reserved_string_buffer_size),
4861
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4862
4863
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
4864
                    raise ValueError("Length of eval names doesn't equal with num_evals")
4865
4866
4867
4868
4869
4870
                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)
                    ]
4871
                    ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
4872
                    _safe_call(_LIB.LGBM_BoosterGetEvalNames(
4873
                        self._handle,
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
                        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
                ]