basic.py 190 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
                     dt_DataTable, pa_Array, pa_ChunkedArray, pa_Table, pd_CategoricalDtype, pd_DataFrame, pd_Series)
wxchan's avatar
wxchan committed
23
24
from .libpath import find_lib_path

25
26
27
if TYPE_CHECKING:
    from typing import Literal

28
29
30
31
32
33
34
    # typing.TypeGuard was only introduced in Python 3.10
    try:
        from typing import TypeGuard
    except ImportError:
        from typing_extensions import TypeGuard


35
36
37
38
39
40
41
42
43
__all__ = [
    'Booster',
    'Dataset',
    'LGBMDeprecationWarning',
    'LightGBMError',
    'register_logger',
    'Sequence',
]

44
_BoosterHandle = ctypes.c_void_p
45
_DatasetHandle = ctypes.c_void_p
46
47
48
49
_ctypes_int_ptr = Union[
    "ctypes._Pointer[ctypes.c_int32]",
    "ctypes._Pointer[ctypes.c_int64]"
]
50
51
52
53
_ctypes_int_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_int32]]",
    "ctypes.Array[ctypes._Pointer[ctypes.c_int64]]"
]
54
55
56
57
58
59
60
61
_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]]"
]
62
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
63
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
64
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
65
_LGBM_BoosterEvalMethodResultWithStandardDeviationType = Tuple[str, str, float, bool, float]
66
67
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], "Literal['auto']"]
_LGBM_FeatureNameConfiguration = Union[List[str], "Literal['auto']"]
68
69
70
71
72
73
_LGBM_GroupType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
74
75
76
77
_LGBM_PositionType = Union[
    np.ndarray,
    pd_Series
]
78
79
80
81
82
83
84
_LGBM_InitScoreType = Union[
    List[float],
    List[List[float]],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
]
85
86
87
88
89
90
91
92
93
_LGBM_TrainDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix,
    "Sequence",
    List["Sequence"],
94
95
    List[np.ndarray],
    pa_Table
96
]
97
_LGBM_LabelType = Union[
98
99
    List[float],
    List[int],
100
101
    np.ndarray,
    pd_Series,
102
103
104
    pd_DataFrame,
    pa_Array,
    pa_ChunkedArray,
105
]
106
107
108
109
110
111
112
113
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix
]
114
115
116
117
118
119
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
    pd_Series
]
120
121
122
ZERO_THRESHOLD = 1e-35


123
124
125
126
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


127
def _get_sample_count(total_nrow: int, params: str) -> int:
128
129
130
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
131
        _c_str(params),
132
133
134
135
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
136

137
138
139
140
141
142
class _MissingType(Enum):
    NONE = 'None'
    NAN = 'NaN'
    ZERO = 'Zero'


143
class _DummyLogger:
144
    def info(self, msg: str) -> None:
145
        print(msg)  # noqa: T201
146

147
    def warning(self, msg: str) -> None:
148
149
150
        warnings.warn(msg, stacklevel=3)


151
152
153
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
154
155


156
157
158
159
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


160
161
162
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
163
164
165
166
    """Register custom logger.

    Parameters
    ----------
167
    logger : Any
168
        Custom logger.
169
170
171
172
    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.
173
    """
174
175
176
177
178
179
    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
180
    _LOGGER = logger
181
182
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
183
184


185
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
186
    """Join log messages from native library which come by chunks."""
187
    msg_normalized: List[str] = []
188
189

    @wraps(func)
190
    def wrapper(msg: str) -> None:
191
192
193
194
195
196
197
198
199
200
201
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


202
def _log_info(msg: str) -> None:
203
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
204
205


206
def _log_warning(msg: str) -> None:
207
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
208
209
210


@_normalize_native_string
211
def _log_native(msg: str) -> None:
212
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
213
214


215
def _log_callback(msg: bytes) -> None:
216
    """Redirect logs from native library into Python."""
217
    _log_native(str(msg.decode('utf-8')))
218
219


220
def _load_lib() -> ctypes.CDLL:
221
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
222
223
224
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
225
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
226
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
227
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
228
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
229
230
    return lib

wxchan's avatar
wxchan committed
231

232
233
234
235
236
237
238
# 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
239

wxchan's avatar
wxchan committed
240

241
_NUMERIC_TYPES = (int, float, bool)
242
_ArrayLike = Union[List, np.ndarray, pd_Series]
243
244


245
def _safe_call(ret: int) -> None:
246
247
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
248
249
250
    Parameters
    ----------
    ret : int
251
        The return value from C API calls.
wxchan's avatar
wxchan committed
252
253
    """
    if ret != 0:
254
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
255

wxchan's avatar
wxchan committed
256

257
def _is_numeric(obj: Any) -> bool:
258
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
259
260
261
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
262
263
264
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
265
266
        return False

wxchan's avatar
wxchan committed
267

268
def _is_numpy_1d_array(data: Any) -> bool:
269
    """Check whether data is a numpy 1-D array."""
270
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
271

wxchan's avatar
wxchan committed
272

273
def _is_numpy_column_array(data: Any) -> bool:
274
275
276
277
278
279
280
    """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


281
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
282
    """Cast numpy array to given dtype."""
283
284
285
286
287
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


288
def _is_1d_list(data: Any) -> bool:
289
    """Check whether data is a 1-D list."""
290
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
291

wxchan's avatar
wxchan committed
292

293
294
295
296
297
298
299
300
301
302
303
304
305
306
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)
    )


307
308
309
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
310
        _is_numpy_1d_array(data)
311
        or _is_numpy_column_array(data)
312
        or _is_1d_list(data)
313
314
315
316
        or isinstance(data, pd_Series)
    )


317
318
def _list_to_1d_numpy(
    data: Any,
319
320
    dtype: "np.typing.DTypeLike",
    name: str
321
) -> np.ndarray:
322
    """Convert data to numpy 1-D array."""
323
    if _is_numpy_1d_array(data):
324
        return _cast_numpy_array_to_dtype(data, dtype)
325
    elif _is_numpy_column_array(data):
326
327
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
328
        return _cast_numpy_array_to_dtype(array, dtype)
329
    elif _is_1d_list(data):
wxchan's avatar
wxchan committed
330
        return np.array(data, dtype=dtype, copy=False)
331
    elif isinstance(data, pd_Series):
332
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
333
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
334
    else:
335
336
        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
337

wxchan's avatar
wxchan committed
338

339
340
341
342
343
344
345
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."""
346
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
347
348
349
350
351
352
353
354
355
356
357


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)
    )


358
359
360
361
362
def _is_pyarrow_array(data: Any) -> bool:
    """Check whether data is a PyArrow array."""
    return isinstance(data, (pa_Array, pa_ChunkedArray))


363
364
365
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
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
394
395
396
397
398
    if isinstance(data, pa_Array):
        export_objects = [data]
    elif isinstance(data, pa_ChunkedArray):
        export_objects = data.chunks
    elif isinstance(data, pa_Table):
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
        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)



420
421
def _data_to_2d_numpy(
    data: Any,
422
423
    dtype: "np.typing.DTypeLike",
    name: str
424
) -> np.ndarray:
425
426
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
427
        return _cast_numpy_array_to_dtype(data, dtype)
428
429
430
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
431
        _check_for_bad_pandas_dtypes(data.dtypes)
432
        return _cast_numpy_array_to_dtype(data.values, dtype)
433
434
435
436
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


437
def _cfloat32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
438
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
439
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
440
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
441
    else:
442
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
443

Guolin Ke's avatar
Guolin Ke committed
444

445
def _cfloat64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
446
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
447
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
448
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
449
450
451
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
452

453
def _cint32_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
454
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
455
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
456
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
457
    else:
458
459
460
        raise RuntimeError('Expected int32 pointer')


461
def _cint64_array_to_numpy(cptr: "ctypes._Pointer", length: int) -> np.ndarray:
462
463
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
464
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
465
466
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
467

wxchan's avatar
wxchan committed
468

469
def _c_str(string: str) -> ctypes.c_char_p:
470
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
471
472
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
473

474
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
475
    """Convert a Python array to C array."""
476
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
477

wxchan's avatar
wxchan committed
478

479
def _json_default_with_numpy(obj: Any) -> Any:
480
481
482
483
484
485
486
487
488
    """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


489
490
491
492
493
494
495
496
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)


497
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
498
    """Convert Python dictionary to string, which is passed to C API."""
499
    if data is None or not data:
wxchan's avatar
wxchan committed
500
501
502
        return ""
    pairs = []
    for key, val in data.items():
503
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
504
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
505
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
506
            pairs.append(f"{key}={val}")
507
        elif val is not None:
508
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
509
    return ' '.join(pairs)
510

wxchan's avatar
wxchan committed
511

512
class _TempFile:
513
514
    """Proxy class to workaround errors on Windows."""

515
516
517
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
518
            self.path = Path(self.name)
519
        return self
wxchan's avatar
wxchan committed
520

521
    def __exit__(self, exc_type, exc_val, exc_tb):
522
523
        if self.path.is_file():
            self.path.unlink()
524

wxchan's avatar
wxchan committed
525

526
class LightGBMError(Exception):
527
528
    """Error thrown by LightGBM."""

529
530
531
    pass


532
533
534
535
536
537
538
539
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
540
541
542
543
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
544
    def _get_all_param_aliases() -> Dict[str, List[str]]:
545
546
547
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
548
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
549
550
551
552
553
554
555
556
        _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)
557
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
558
559
560
561
            _safe_call(_LIB.LGBM_DumpParamAliases(
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
562
        return json.loads(
563
            string_buffer.value.decode('utf-8'),
564
            object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
565
        )
566
567

    @classmethod
568
569
570
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
571
572
        ret = set()
        for i in args:
573
            ret.update(cls.get_sorted(i))
574
575
        return ret

576
577
578
579
580
581
    @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])

582
    @classmethod
583
584
585
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
586
587
588
589
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
590
                    ret.update(aliases)
591
592
593
                    break
        return ret

594

595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
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)

616
617
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
618
619

    # if main_param_name was provided, keep that value and remove all aliases
620
    if main_param_name in params.keys():
621
622
623
        for param in aliases:
            params.pop(param, None)
        return params
624

625
626
627
628
629
    # 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
630

631
632
633
634
635
636
637
    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
638
639
640
641

    return params


642
_MAX_INT32 = (1 << 31) - 1
643

644
"""Macro definition of data type in C API of LightGBM"""
645
646
647
648
_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
649

650
"""Matrix is row major in Python"""
651
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
652

653
"""Macro definition of prediction type in C API of LightGBM"""
654
655
656
657
_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
658

659
"""Macro definition of sparse matrix type"""
660
661
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
662

663
"""Macro definition of feature importance type"""
664
665
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
666

667
"""Data type of data field"""
668
669
670
671
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
672
673
    "group": _C_API_DTYPE_INT32,
    "position": _C_API_DTYPE_INT32
674
}
wxchan's avatar
wxchan committed
675

676
"""String name to int feature importance type mapper"""
677
678
679
680
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN
}
681

wxchan's avatar
wxchan committed
682

683
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
684
    """Fix the memory of multi-dimensional sliced object."""
685
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
686
        if not data.flags.c_contiguous:
687
688
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
689
690
691
692
            return np.copy(data)
    return data


693
694
695
def _c_float_array(
    data: np.ndarray
) -> Tuple[_ctypes_float_ptr, int, np.ndarray]:
696
    """Get pointer of float numpy array / list."""
697
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
698
        data = np.array(data, copy=False)
699
    if _is_numpy_1d_array(data):
700
        data = _convert_from_sliced_object(data)
701
        assert data.flags.c_contiguous
702
        ptr_data: _ctypes_float_ptr
wxchan's avatar
wxchan committed
703
704
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
705
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
706
707
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
708
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
709
        else:
710
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
711
    else:
712
        raise TypeError(f"Unknown type({type(data).__name__})")
713
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
714

wxchan's avatar
wxchan committed
715

716
717
718
def _c_int_array(
    data: np.ndarray
) -> Tuple[_ctypes_int_ptr, int, np.ndarray]:
719
    """Get pointer of int numpy array / list."""
720
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
721
        data = np.array(data, copy=False)
722
    if _is_numpy_1d_array(data):
723
        data = _convert_from_sliced_object(data)
724
        assert data.flags.c_contiguous
725
        ptr_data: _ctypes_int_ptr
wxchan's avatar
wxchan committed
726
727
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
728
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
729
730
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
731
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
732
        else:
733
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
734
    else:
735
        raise TypeError(f"Unknown type({type(data).__name__})")
736
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
737

wxchan's avatar
wxchan committed
738

739
def _is_allowed_numpy_dtype(dtype: type) -> bool:
740
    float128 = getattr(np, 'float128', type(None))
741
742
743
744
    return (
        issubclass(dtype, (np.integer, np.floating, np.bool_))
        and not issubclass(dtype, (np.timedelta64, float128))
    )
745
746


747
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
748
749
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
750
        for column_name, pandas_dtype in pandas_dtypes_series.items()
751
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
752
753
754
755
    ]
    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)}')
756
757


758
def _data_from_pandas(
759
760
761
    data: pd_DataFrame,
    feature_name: _LGBM_FeatureNameConfiguration,
    categorical_feature: _LGBM_CategoricalFeatureConfiguration,
762
    pandas_categorical: Optional[List[List]]
763
764
765
766
767
768
769
770
771
772
773
774
775
) -> 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]
776
    else:
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
        if len(cat_cols) != len(pandas_categorical):
            raise ValueError('train and valid dataset categorical_feature do not match.')
        for col, category in zip(cat_cols, pandas_categorical):
            if list(data[col].cat.categories) != list(category):
                data[col] = data[col].cat.set_categories(category)
    if len(cat_cols):  # cat_cols is list
        data = data.copy(deep=False)  # not alter origin DataFrame
        data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
    if categorical_feature == 'auto':  # use cat cols from DataFrame
        categorical_feature = cat_cols_not_ordered
    else:  # use cat cols specified by user
        categorical_feature = list(categorical_feature)  # type: ignore[assignment]

    # get numpy representation of the data
    _check_for_bad_pandas_dtypes(data.dtypes)
    df_dtypes = [dtype.type for dtype in data.dtypes]
    df_dtypes.append(np.float32)  # so that the target dtype considers floats
    target_dtype = np.result_type(*df_dtypes)
    try:
        # most common case (no nullable dtypes)
        data = data.to_numpy(dtype=target_dtype, copy=False)
    except TypeError:
        # 1.0 <= pd version < 1.1 and nullable dtypes, least common case
        # raises error because array is casted to type(pd.NA) and there's no na_value argument
        data = data.astype(target_dtype, copy=False).values
    except ValueError:
        # data has nullable dtypes, but we can specify na_value argument and copy will be made
        data = data.to_numpy(dtype=target_dtype, na_value=np.nan)
805
    return data, feature_name, categorical_feature, pandas_categorical
806
807


808
809
810
811
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
    file_name: Optional[Union[str, Path]] = None
) -> str:
812
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
813
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
814
815
816
817
818
819
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


820
821
822
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
    model_str: Optional[str] = None
823
) -> Optional[List[List]]:
824
825
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
826
    if file_name is not None:
827
        max_offset = -getsize(file_name)
828
829
830
831
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
832
                f.seek(offset, SEEK_END)
833
834
835
836
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
837
        last_line = lines[-1].decode('utf-8').strip()
838
        if not last_line.startswith(pandas_key):
839
            last_line = lines[-2].decode('utf-8').strip()
840
    elif model_str is not None:
841
842
843
844
845
846
        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
847
848


849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
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**.

869
870
    .. versionadded:: 3.3.0

871
872
873
874
875
876
877
878
879
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
880
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
881
882
883
884
885
886
887
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
888
                return self._get_one_line(idx)
889
            elif isinstance(idx, slice):
890
891
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
892
                # Only required if using ``Dataset.subset()``.
893
                return np.array([self._get_one_line(i) for i in idx])
894
            else:
895
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
896
897
898

        Parameters
        ----------
899
        idx : int, slice[int], list[int]
900
901
902
903
            Item index.

        Returns
        -------
904
        result : numpy 1-D array or numpy 2-D array
905
            1-D array if idx is int, 2-D array if idx is slice or list.
906
907
908
909
910
911
912
913
914
        """
        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__()")


915
class _InnerPredictor:
916
917
918
919
920
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
921
922
923
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
924
    """
925

926
927
    def __init__(
        self,
928
929
930
931
        booster_handle: _BoosterHandle,
        pandas_categorical: Optional[List[List]],
        pred_parameter: Dict[str, Any],
        manage_handle: bool
932
    ):
933
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
934
935
936

        Parameters
        ----------
937
        booster_handle : object
938
            Handle of Booster.
939
940
941
942
        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
943
            Other parameters for the prediction.
944
945
        manage_handle : bool
            If ``True``, free the corresponding Booster on the C++ side when this Python object is deleted.
wxchan's avatar
wxchan committed
946
        """
947
948
949
950
951
952
953
954
        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(
955
                self._handle,
956
957
958
959
                ctypes.byref(out_num_class)
            )
        )
        self.num_class = out_num_class.value
wxchan's avatar
wxchan committed
960

961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
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
    @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
1020

1021
    def __del__(self) -> None:
1022
1023
        try:
            if self.__is_manage_handle:
1024
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
1025
1026
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
1027

1028
    def __getstate__(self) -> Dict[str, Any]:
1029
1030
        this = self.__dict__.copy()
        this.pop('handle', None)
1031
        this.pop('_handle', None)
1032
1033
        return this

1034
1035
    def predict(
        self,
1036
        data: _LGBM_PredictDataType,
1037
1038
1039
1040
1041
1042
1043
        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
1044
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
1045
        """Predict logic.
wxchan's avatar
wxchan committed
1046
1047
1048

        Parameters
        ----------
1049
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
1050
            Data source for prediction.
1051
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
1052
1053
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
        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.
1065
1066
1067
        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
1068

1069
1070
            .. versionadded:: 4.0.0

wxchan's avatar
wxchan committed
1071
1072
        Returns
        -------
1073
        result : numpy array, scipy.sparse or list of scipy.sparse
1074
            Prediction result.
1075
            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
1076
        """
wxchan's avatar
wxchan committed
1077
        if isinstance(data, Dataset):
1078
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
1079
1080
1081
1082
1083
1084
        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(
1085
                    self._handle,
1086
1087
1088
1089
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
1090
1091
1092
1093
1094
1095
1096
1097
1098

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

1099
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
1100
        if raw_score:
1101
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
1102
        if pred_leaf:
1103
            predict_type = _C_API_PREDICT_LEAF_INDEX
1104
        if pred_contrib:
1105
            predict_type = _C_API_PREDICT_CONTRIB
cbecker's avatar
cbecker committed
1106

1107
        if isinstance(data, (str, Path)):
1108
            with _TempFile() as f:
wxchan's avatar
wxchan committed
1109
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
1110
                    self._handle,
1111
                    _c_str(str(data)),
1112
                    ctypes.c_int(data_has_header),
Guolin Ke's avatar
Guolin Ke committed
1113
                    ctypes.c_int(predict_type),
1114
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1115
                    ctypes.c_int(num_iteration),
1116
1117
                    _c_str(self.pred_parameter),
                    _c_str(f.name)))
1118
1119
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
1120
        elif isinstance(data, scipy.sparse.csr_matrix):
1121
1122
1123
1124
1125
1126
            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
1127
        elif isinstance(data, scipy.sparse.csc_matrix):
1128
1129
1130
1131
1132
1133
            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
1134
        elif isinstance(data, np.ndarray):
1135
1136
1137
1138
1139
1140
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1141
1142
1143
        elif isinstance(data, list):
            try:
                data = np.array(data)
1144
1145
            except BaseException as err:
                raise ValueError('Cannot convert data list to numpy array.') from err
1146
1147
1148
1149
1150
1151
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1152
        elif isinstance(data, dt_DataTable):
1153
1154
1155
1156
1157
1158
            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
1159
1160
        else:
            try:
1161
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
1162
                csr = scipy.sparse.csr_matrix(data)
1163
1164
            except BaseException as err:
                raise TypeError(f'Cannot predict data for type {type(data).__name__}') from err
1165
1166
1167
1168
1169
1170
            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
1171
1172
        if pred_leaf:
            preds = preds.astype(np.int32)
1173
        is_sparse = isinstance(preds, scipy.sparse.spmatrix) or isinstance(preds, list)
1174
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1175
            if preds.size % nrow == 0:
1176
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1177
            else:
1178
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
1179
1180
        return preds

1181
1182
1183
1184
1185
1186
1187
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
        predict_type: int
    ) -> int:
1188
        """Get size of prediction result."""
1189
        if nrow > _MAX_INT32:
1190
            raise LightGBMError('LightGBM cannot perform prediction for data '
1191
                                f'with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n'
1192
                                'You can split your data into chunks '
1193
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
1194
1195
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
1196
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1197
1198
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
1199
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1200
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
1201
1202
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
1203

1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
    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)
1217
1218
1219
1220
1221
1222
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
            predict_type=predict_type
        )
1223
1224
1225
1226
1227
1228
        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(
1229
            self._handle,
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
            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]:
1252
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1253
        if len(mat.shape) != 2:
1254
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
1255

1256
        nrow = mat.shape[0]
1257
1258
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1259
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1260
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
1261
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1262
            preds = np.empty(sum(n_preds), dtype=np.float64)
1263
1264
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
1265
                # avoid memory consumption by arrays concatenation operations
1266
1267
1268
1269
1270
1271
1272
                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]
                )
1273
            return preds, nrow
wxchan's avatar
wxchan committed
1274
        else:
1275
1276
1277
1278
1279
1280
1281
            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
1282

1283
1284
1285
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1286
1287
1288
1289
1290
1291
        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,
1292
        is_csr: bool
1293
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1294
1295
1296
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1297
        if indptr_type == _C_API_DTYPE_INT32:
1298
            out_indptr = _cint32_array_to_numpy(out_ptr_indptr, indptr_len)
1299
        elif indptr_type == _C_API_DTYPE_INT64:
1300
            out_indptr = _cint64_array_to_numpy(out_ptr_indptr, indptr_len)
1301
1302
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1303
        if data_type == _C_API_DTYPE_FLOAT32:
1304
            out_data = _cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
1305
        elif data_type == _C_API_DTYPE_FLOAT64:
1306
            out_data = _cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
1307
1308
        else:
            raise TypeError("Expected float32 or float64 type for data")
1309
        out_indices = _cint32_array_to_numpy(out_ptr_indices, data_indices_len)
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
        # 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

1338
1339
1340
1341
1342
1343
1344
1345
1346
    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
1347
1348
1349
1350
1351
1352
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1353
1354
1355
1356
1357
        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
1358

1359
1360
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1361

1362
1363
1364
1365
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

        _safe_call(_LIB.LGBM_BoosterPredictForCSR(
1366
            self._handle,
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
            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
1391
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1392
1393
1394
1395
        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
1396
        out_ptr_indptr: _ctypes_int_ptr
1397
1398
1399
1400
1401
        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)()
1402
        out_ptr_data: _ctypes_float_ptr
1403
1404
1405
1406
1407
1408
        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(
1409
            self._handle,
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
            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

1440
1441
1442
1443
1444
1445
1446
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1447
        """Predict for a CSR data."""
1448
        if predict_type == _C_API_PREDICT_CONTRIB:
1449
1450
1451
1452
1453
1454
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1455
        nrow = len(csr.indptr) - 1
1456
1457
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1458
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1459
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1460
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1461
            preds = np.empty(sum(n_preds), dtype=np.float64)
1462
1463
            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:])):
1464
                # avoid memory consumption by arrays concatenation operations
1465
1466
1467
1468
1469
1470
1471
                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]
                )
1472
1473
            return preds, nrow
        else:
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
            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,
1484
1485
1486
1487
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
1488
1489
1490
1491
1492
    ):
        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
1493
        out_ptr_indptr: _ctypes_int_ptr
1494
1495
1496
1497
1498
        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)()
1499
        out_ptr_data: _ctypes_float_ptr
1500
1501
1502
1503
1504
1505
        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(
1506
            self._handle,
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
            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
1536

1537
1538
1539
1540
1541
1542
1543
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1544
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1545
        nrow = csc.shape[0]
1546
        if nrow > _MAX_INT32:
1547
1548
1549
1550
1551
1552
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1553
        if predict_type == _C_API_PREDICT_CONTRIB:
1554
1555
1556
1557
1558
1559
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1560
1561
1562
1563
1564
1565
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1566
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1567
1568
        out_num_preds = ctypes.c_int64(0)

1569
1570
        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
1571

1572
        assert csc.shape[0] <= _MAX_INT32
1573
        csc_indices = csc.indices.astype(np.int32, copy=False)
1574

Guolin Ke's avatar
Guolin Ke committed
1575
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
1576
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1577
            ptr_indptr,
1578
            ctypes.c_int(type_ptr_indptr),
1579
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1580
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1581
1582
1583
1584
1585
            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),
1586
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1587
            ctypes.c_int(num_iteration),
1588
            _c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1589
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1590
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1591
        if n_preds != out_num_preds.value:
1592
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1593
1594
        return preds, nrow

1595
    def current_iteration(self) -> int:
1596
1597
1598
1599
1600
1601
1602
1603
1604
        """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(
1605
            self._handle,
1606
1607
1608
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1609

1610
class Dataset:
wxchan's avatar
wxchan committed
1611
    """Dataset in LightGBM."""
1612

1613
1614
    def __init__(
        self,
1615
        data: _LGBM_TrainDataType,
1616
        label: Optional[_LGBM_LabelType] = None,
1617
        reference: Optional["Dataset"] = None,
1618
1619
1620
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1621
1622
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
1623
        params: Optional[Dict[str, Any]] = None,
1624
1625
        free_raw_data: bool = True,
        position: Optional[_LGBM_PositionType] = None,
1626
    ):
1627
        """Initialize Dataset.
1628

wxchan's avatar
wxchan committed
1629
1630
        Parameters
        ----------
1631
        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
1632
            Data source of Dataset.
1633
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1634
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1635
1636
1637
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1638
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1639
            Weight for each instance. Weights should be non-negative.
1640
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1641
1642
1643
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1644
1645
            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.
1646
        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)
1647
            Init score for Dataset.
1648
        feature_name : list of str, or 'auto', optional (default="auto")
1649
            Feature names.
1650
            If 'auto' and data is pandas DataFrame or pyarrow Table, data columns names are used.
1651
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1652
1653
            Categorical features.
            If list of int, interpreted as indices.
1654
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1655
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1656
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1657
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1658
            All negative values in categorical features will be treated as missing values.
1659
            The output cannot be monotonically constrained with respect to a categorical feature.
1660
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1661
        params : dict or None, optional (default=None)
1662
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1663
        free_raw_data : bool, optional (default=True)
1664
            If True, raw data is freed after constructing inner Dataset.
1665
1666
        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
1667
        """
1668
        self._handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1669
1670
1671
1672
1673
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1674
        self.position = position
1675
        self.init_score = init_score
1676
1677
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1678
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1679
        self.free_raw_data = free_raw_data
1680
        self.used_indices: Optional[List[int]] = None
1681
        self._need_slice = True
1682
        self._predictor: Optional[_InnerPredictor] = None
1683
        self.pandas_categorical: Optional[List[List]] = None
1684
        self._params_back_up = None
1685
        self.version = 0
1686
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1687

1688
    def __del__(self) -> None:
1689
1690
1691
1692
        try:
            self._free_handle()
        except AttributeError:
            pass
1693

1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
    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.
        """
1711
        param_str = _param_dict_to_str(self.get_params())
1712
1713
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1714
        ptr_data, _, _ = _c_int_array(indices)
1715
1716
1717
1718
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
1719
            _c_str(param_str),
1720
1721
1722
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1723
1724
        assert sample_cnt == actual_sample_cnt.value
        return indices
1725

1726
1727
1728
1729
1730
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
        ref_dataset: _DatasetHandle
    ) -> 'Dataset':
1731
1732
1733
1734
1735
1736
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1737
1738
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1739
1740
1741
1742
1743
1744

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
1745
        self._handle = ctypes.c_void_p()
1746
1747
1748
        _safe_call(_LIB.LGBM_DatasetCreateByReference(
            ref_dataset,
            ctypes.c_int64(total_nrow),
1749
            ctypes.byref(self._handle),
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
        ))
        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
        ----------
1764
        sample_data : list of numpy array
1765
            Sample data for each column.
1766
        sample_indices : list of numpy array
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
            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.
1789
        sample_col_ptr: _ctypes_float_array = (ctypes.POINTER(ctypes.c_double) * ncol)()
1790
1791
        # c type int**
        # each int* points to start of indices for each column
1792
        indices_col_ptr: _ctypes_int_array = (ctypes.POINTER(ctypes.c_int32) * ncol)()
1793
        for i in range(ncol):
1794
1795
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1796
1797

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1798
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1799

1800
        self._handle = ctypes.c_void_p()
1801
        params_str = _param_dict_to_str(self.get_params())
1802
1803
1804
1805
1806
1807
1808
        _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),
1809
            ctypes.c_int64(total_nrow),
1810
            _c_str(params_str),
1811
            ctypes.byref(self._handle),
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
        ))
        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)
1830
        data_ptr, data_type, _ = _c_float_array(data)
1831
1832

        _safe_call(_LIB.LGBM_DatasetPushRows(
1833
            self._handle,
1834
1835
1836
1837
1838
1839
1840
1841
1842
            data_ptr,
            data_type,
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol),
            ctypes.c_int32(self._start_row),
        ))
        self._start_row += nrow
        return self

1843
    def get_params(self) -> Dict[str, Any]:
1844
1845
1846
1847
        """Get the used parameters in the Dataset.

        Returns
        -------
1848
        params : dict
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
            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",
1864
                                                "linear_tree",
1865
1866
1867
1868
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1869
                                                "precise_float_parser",
1870
1871
1872
1873
1874
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1875
1876
        else:
            return {}
1877

1878
    def _free_handle(self) -> "Dataset":
1879
1880
1881
        if self._handle is not None:
            _safe_call(_LIB.LGBM_DatasetFree(self._handle))
            self._handle = None
1882
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
1883
1884
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1885
        return self
wxchan's avatar
wxchan committed
1886

1887
1888
1889
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
1890
        data: _LGBM_TrainDataType,
1891
        used_indices: Optional[Union[List[int], np.ndarray]]
1892
    ) -> "Dataset":
Guolin Ke's avatar
Guolin Ke committed
1893
        data_has_header = False
1894
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
1895
            # check data has header or not
1896
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1897
        num_data = self.num_data()
1898
        if predictor is not None:
1899
1900
1901
1902
1903
            init_score: Union[np.ndarray, scipy.sparse.spmatrix] = predictor.predict(
                data=data,
                raw_score=True,
                data_has_header=data_has_header
            )
1904
            init_score = init_score.ravel()
1905
            if used_indices is not None:
1906
                assert not self._need_slice
1907
                if isinstance(data, (str, Path)):
1908
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1909
                    assert num_data == len(used_indices)
1910
1911
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1912
1913
1914
1915
                            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
1916
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1917
1918
                for i in range(num_data):
                    for j in range(predictor.num_class):
1919
1920
1921
                        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:
1922
            init_score = np.full_like(self.init_score, fill_value=0.0, dtype=np.float64)
1923
1924
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1925
        self.set_init_score(init_score)
1926
        return self
Guolin Ke's avatar
Guolin Ke committed
1927

1928
1929
    def _lazy_init(
        self,
1930
        data: Optional[_LGBM_TrainDataType],
1931
1932
1933
1934
1935
1936
1937
1938
        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,
1939
1940
        params: Optional[Dict[str, Any]],
        position: Optional[_LGBM_PositionType]
1941
    ) -> "Dataset":
wxchan's avatar
wxchan committed
1942
        if data is None:
1943
            self._handle = None
Nikita Titov's avatar
Nikita Titov committed
1944
            return self
Guolin Ke's avatar
Guolin Ke committed
1945
1946
1947
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1948
1949
1950
1951
1952
1953
1954
        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
1955

1956
        # process for args
wxchan's avatar
wxchan committed
1957
        params = {} if params is None else params
1958
        args_names = inspect.signature(self.__class__._lazy_init).parameters.keys()
1959
        for key in params.keys():
1960
            if key in args_names:
1961
1962
                _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.')
1963
        # get categorical features
1964
        if isinstance(categorical_feature, list):
1965
1966
            categorical_indices = set()
            feature_dict = {}
1967
            if isinstance(feature_name, list):
1968
1969
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
1970
                if isinstance(name, str) and name in feature_dict:
1971
                    categorical_indices.add(feature_dict[name])
1972
                elif isinstance(name, int):
1973
1974
                    categorical_indices.add(name)
                else:
1975
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1976
            if categorical_indices:
1977
1978
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1979
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
1980
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
1981
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1982
                        params.pop(cat_alias, None)
1983
                params['categorical_column'] = sorted(categorical_indices)
1984

1985
        params_str = _param_dict_to_str(params)
1986
        self.params = params
1987
        # process for reference dataset
wxchan's avatar
wxchan committed
1988
        ref_dataset = None
wxchan's avatar
wxchan committed
1989
        if isinstance(reference, Dataset):
1990
            ref_dataset = reference.construct()._handle
wxchan's avatar
wxchan committed
1991
1992
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1993
        # start construct data
1994
        if isinstance(data, (str, Path)):
1995
            self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1996
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1997
1998
                _c_str(str(data)),
                _c_str(params_str),
wxchan's avatar
wxchan committed
1999
                ref_dataset,
2000
                ctypes.byref(self._handle)))
wxchan's avatar
wxchan committed
2001
2002
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
2003
2004
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
2005
2006
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
2007
2008
2009
        elif _is_pyarrow_table(data):
            self.__init_from_pyarrow_table(data, params_str, ref_dataset)
            feature_name = data.column_names
2010
        elif isinstance(data, list) and len(data) > 0:
2011
            if _is_list_of_numpy_arrays(data):
2012
                self.__init_from_list_np2d(data, params_str, ref_dataset)
2013
            elif _is_list_of_sequences(data):
2014
2015
2016
2017
2018
                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)
2019
        elif isinstance(data, dt_DataTable):
2020
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
2021
2022
2023
2024
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
2025
2026
            except BaseException as err:
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}') from err
wxchan's avatar
wxchan committed
2027
2028
2029
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
2030
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
2031
2032
2033
2034
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
2035
2036
        if position is not None:
            self.set_position(position)
2037
2038
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
2039
                _log_warning("The init_score will be overridden by the prediction of init_model.")
2040
2041
2042
2043
2044
            self._set_init_score_by_predictor(
                predictor=predictor,
                data=data,
                used_indices=None
            )
2045
2046
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
2047
        elif predictor is not None:
2048
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2049
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
2050
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
2051

2052
2053
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
        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.
2079
        sampled = np.array(list(self._yield_row_from_seqlist(seqs, indices)))
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
        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

2095
2096
2097
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
2098
        ref_dataset: Optional[_DatasetHandle]
2099
    ) -> "Dataset":
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
        """
        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:
2114
            param_str = _param_dict_to_str(self.get_params())
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
            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

2128
2129
2130
2131
2132
2133
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2134
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
2135
2136
2137
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

2138
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2139
2140
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
2141
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
2142
2143
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

2144
        ptr_data, type_ptr_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2145
2146
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2147
            ctypes.c_int(type_ptr_data),
2148
2149
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
2150
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2151
            _c_str(params_str),
wxchan's avatar
wxchan committed
2152
            ref_dataset,
2153
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2154
        return self
wxchan's avatar
wxchan committed
2155

2156
2157
2158
2159
2160
2161
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2162
        """Initialize data from a list of 2-D numpy matrices."""
2163
        ncol = mats[0].shape[1]
2164
        nrow = np.empty((len(mats),), np.int32)
2165
        ptr_data: _ctypes_float_array
2166
2167
2168
2169
2170
2171
        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 = []
2172
        type_ptr_data = -1
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184

        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)
2185
            else:  # change non-float data to float data, need to copy
2186
2187
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2188
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2189
            if type_ptr_data != -1 and chunk_type_ptr_data != type_ptr_data:
2190
2191
2192
2193
2194
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

2195
        self._handle = ctypes.c_void_p()
2196
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
2197
            ctypes.c_int32(len(mats)),
2198
2199
2200
            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)),
2201
            ctypes.c_int32(ncol),
2202
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2203
            _c_str(params_str),
2204
            ref_dataset,
2205
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2206
        return self
2207

2208
2209
2210
2211
2212
2213
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2214
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2215
        if len(csr.indices) != len(csr.data):
2216
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
2217
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2218

2219
2220
        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
2221

2222
        assert csr.shape[1] <= _MAX_INT32
2223
        csr_indices = csr.indices.astype(np.int32, copy=False)
2224

wxchan's avatar
wxchan committed
2225
2226
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2227
            ctypes.c_int(type_ptr_indptr),
2228
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
2229
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2230
2231
2232
2233
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
2234
            _c_str(params_str),
wxchan's avatar
wxchan committed
2235
            ref_dataset,
2236
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2237
        return self
wxchan's avatar
wxchan committed
2238

2239
2240
2241
2242
2243
2244
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2245
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2246
        if len(csc.indices) != len(csc.data):
2247
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
2248
        self._handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2249

2250
2251
        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
2252

2253
        assert csc.shape[0] <= _MAX_INT32
2254
        csc_indices = csc.indices.astype(np.int32, copy=False)
2255

Guolin Ke's avatar
Guolin Ke committed
2256
2257
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2258
            ctypes.c_int(type_ptr_indptr),
2259
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
2260
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2261
2262
2263
2264
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
2265
            _c_str(params_str),
Guolin Ke's avatar
Guolin Ke committed
2266
            ref_dataset,
2267
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2268
        return self
Guolin Ke's avatar
Guolin Ke committed
2269

2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
    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

2296
    @staticmethod
2297
    def _compare_params_for_warning(
2298
2299
        params: Dict[str, Any],
        other_params: Dict[str, Any],
2300
2301
2302
        ignore_keys: Set[str]
    ) -> bool:
        """Compare two dictionaries with params ignoring some keys.
2303

2304
2305
2306
2307
        It is only for the warning purpose.

        Parameters
        ----------
2308
        params : dict
2309
            One dictionary with parameters to compare.
2310
        other_params : dict
2311
2312
2313
            Another dictionary with parameters to compare.
        ignore_keys : set
            Keys that should be ignored during comparing two dictionaries.
2314
2315
2316

        Returns
        -------
2317
2318
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
        """
        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

2330
    def construct(self) -> "Dataset":
2331
2332
2333
2334
2335
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2336
            Constructed Dataset object.
2337
        """
2338
        if self._handle is None:
wxchan's avatar
wxchan committed
2339
            if self.reference is not None:
2340
                reference_params = self.reference.get_params()
2341
2342
                params = self.get_params()
                if params != reference_params:
2343
2344
2345
2346
2347
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
2348
                        _log_warning('Overriding the parameters from Reference Dataset.')
2349
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2350
                if self.used_indices is None:
2351
                    # create valid
2352
                    self._lazy_init(data=self.data, label=self.label, reference=self.reference,
2353
                                    weight=self.weight, group=self.group, position=self.position,
2354
                                    init_score=self.init_score, predictor=self._predictor,
2355
                                    feature_name=self.feature_name, categorical_feature='auto', params=self.params)
wxchan's avatar
wxchan committed
2356
                else:
2357
                    # construct subset
2358
                    used_indices = _list_to_1d_numpy(self.used_indices, dtype=np.int32, name='used_indices')
2359
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2360
                    if self.reference.group is not None:
2361
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2362
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
2363
                                                  return_counts=True)
2364
                    self._handle = ctypes.c_void_p()
2365
                    params_str = _param_dict_to_str(self.params)
wxchan's avatar
wxchan committed
2366
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
2367
                        self.reference.construct()._handle,
wxchan's avatar
wxchan committed
2368
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
2369
                        ctypes.c_int32(used_indices.shape[0]),
2370
                        _c_str(params_str),
2371
                        ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
2372
2373
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2374
2375
                    if self.group is not None:
                        self.set_group(self.group)
2376
2377
                    if self.position is not None:
                        self.set_position(self.position)
wxchan's avatar
wxchan committed
2378
2379
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
2380
2381
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
2382
2383
2384
2385
2386
                        self._set_init_score_by_predictor(
                            predictor=self._predictor,
                            data=self.data,
                            used_indices=used_indices
                        )
wxchan's avatar
wxchan committed
2387
            else:
2388
                # create train
2389
                self._lazy_init(data=self.data, label=self.label, reference=None,
2390
2391
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
2392
2393
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature,
                                params=self.params, position=self.position)
wxchan's avatar
wxchan committed
2394
2395
            if self.free_raw_data:
                self.data = None
2396
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2397
        return self
wxchan's avatar
wxchan committed
2398

2399
2400
    def create_valid(
        self,
2401
        data: _LGBM_TrainDataType,
2402
        label: Optional[_LGBM_LabelType] = None,
2403
2404
2405
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2406
2407
        params: Optional[Dict[str, Any]] = None,
        position: Optional[_LGBM_PositionType] = None
2408
    ) -> "Dataset":
2409
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2410
2411
2412

        Parameters
        ----------
2413
        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
2414
            Data source of Dataset.
2415
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2416
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2417
2418
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
2419
            Weight for each instance. Weights should be non-negative.
2420
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
2421
2422
2423
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2424
2425
            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.
2426
        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)
2427
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2428
        params : dict or None, optional (default=None)
2429
            Other parameters for validation Dataset.
2430
2431
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
2432
2433
2434

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2435
2436
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2437
        """
2438
        ret = Dataset(data, label=label, reference=self,
2439
                      weight=weight, group=group, position=position, init_score=init_score,
2440
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2441
        ret._predictor = self._predictor
2442
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2443
        return ret
wxchan's avatar
wxchan committed
2444

2445
2446
2447
2448
2449
    def subset(
        self,
        used_indices: List[int],
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2450
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2451
2452
2453
2454

        Parameters
        ----------
        used_indices : list of int
2455
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2456
        params : dict or None, optional (default=None)
2457
            These parameters will be passed to Dataset constructor.
2458
2459
2460
2461
2462

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2463
        """
wxchan's avatar
wxchan committed
2464
2465
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
2466
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
2467
2468
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2469
        ret._predictor = self._predictor
2470
        ret.pandas_categorical = self.pandas_categorical
2471
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2472
2473
        return ret

2474
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2475
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2476

2477
2478
2479
2480
2481
        .. 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
2482
2483
        Parameters
        ----------
2484
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2485
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2486
2487
2488
2489
2490

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2491
2492
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
2493
            self.construct()._handle,
2494
            _c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
2495
        return self
wxchan's avatar
wxchan committed
2496

2497
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2498
2499
        if not params:
            return self
2500
        params = deepcopy(params)
2501
2502
2503
2504
2505

        def update():
            if not self.params:
                self.params = params
            else:
2506
                self._params_back_up = deepcopy(self.params)
2507
2508
                self.params.update(params)

2509
        if self._handle is None:
2510
2511
2512
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2513
2514
                _c_str(_param_dict_to_str(self.params)),
                _c_str(_param_dict_to_str(params)))
2515
2516
2517
2518
2519
2520
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2521
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
2522
        return self
wxchan's avatar
wxchan committed
2523

2524
    def _reverse_update_params(self) -> "Dataset":
2525
        if self._handle is None:
2526
2527
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2528
        return self
2529

2530
2531
2532
    def set_field(
        self,
        field_name: str,
2533
        data: Optional[Union[List[List[float]], List[List[int]], List[float], List[int], np.ndarray, pd_Series, pd_DataFrame, pa_Array, pa_ChunkedArray]]
2534
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2535
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2536
2537
2538

        Parameters
        ----------
2539
        field_name : str
2540
            The field name of the information.
2541
        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
2542
            The data to be set.
Nikita Titov's avatar
Nikita Titov committed
2543
2544
2545
2546
2547

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2548
        """
2549
        if self._handle is None:
2550
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2551
        if data is None:
2552
            # set to None
wxchan's avatar
wxchan committed
2553
            _safe_call(_LIB.LGBM_DatasetSetField(
2554
                self._handle,
2555
                _c_str(field_name),
wxchan's avatar
wxchan committed
2556
                None,
Guolin Ke's avatar
Guolin Ke committed
2557
                ctypes.c_int(0),
2558
                ctypes.c_int(_FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2559
            return self
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573

        # 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

2574
        dtype: "np.typing.DTypeLike"
2575
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2576
            dtype = np.float64
2577
            if _is_1d_collection(data):
2578
                data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2579
            elif _is_2d_collection(data):
2580
                data = _data_to_2d_numpy(data, dtype=dtype, name=field_name)
2581
2582
2583
2584
2585
2586
2587
                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:
2588
            dtype = np.int32 if (field_name == 'group' or field_name == 'position') else np.float32
2589
            data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2590

2591
        ptr_data: Union[_ctypes_float_ptr, _ctypes_int_ptr]
2592
        if data.dtype == np.float32 or data.dtype == np.float64:
2593
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2594
        elif data.dtype == np.int32:
2595
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2596
        else:
2597
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2598
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2599
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2600
        _safe_call(_LIB.LGBM_DatasetSetField(
2601
            self._handle,
2602
            _c_str(field_name),
wxchan's avatar
wxchan committed
2603
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2604
2605
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2606
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2607
        return self
wxchan's avatar
wxchan committed
2608

2609
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2610
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2611
2612
2613

        Parameters
        ----------
2614
        field_name : str
2615
            The field name of the information.
wxchan's avatar
wxchan committed
2616
2617
2618

        Returns
        -------
2619
        info : numpy array or None
2620
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2621
        """
2622
        if self._handle is None:
2623
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2624
2625
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2626
2627
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
2628
            self._handle,
2629
            _c_str(field_name),
wxchan's avatar
wxchan committed
2630
2631
2632
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
2633
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2634
2635
2636
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2637
        if out_type.value == _C_API_DTYPE_INT32:
2638
            arr = _cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
2639
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2640
            arr = _cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
2641
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2642
            arr = _cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2643
        else:
wxchan's avatar
wxchan committed
2644
            raise TypeError("Unknown type")
2645
2646
2647
2648
2649
2650
        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
2651

2652
2653
    def set_categorical_feature(
        self,
2654
        categorical_feature: _LGBM_CategoricalFeatureConfiguration
2655
    ) -> "Dataset":
2656
        """Set categorical features.
2657
2658
2659

        Parameters
        ----------
2660
        categorical_feature : list of str or int, or 'auto'
2661
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2662
2663
2664
2665
2666

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2667
2668
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2669
            return self
2670
        if self.data is not None:
2671
2672
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2673
                return self._free_handle()
2674
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2675
                return self
2676
            else:
2677
2678
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
2679
                                 f'New categorical_feature is {list(categorical_feature)}')
2680
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2681
                return self._free_handle()
2682
        else:
2683
2684
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2685

2686
2687
2688
2689
    def _set_predictor(
        self,
        predictor: Optional[_InnerPredictor]
    ) -> "Dataset":
2690
2691
2692
2693
        """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
2694
        """
2695
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2696
            return self
2697
2698
2699
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
            if (predictor == self._predictor) and (predictor.current_iteration() == self._predictor.current_iteration()):
                return self
2700
        if self._handle is None:
Guolin Ke's avatar
Guolin Ke committed
2701
            self._predictor = predictor
2702
2703
        elif self.data is not None:
            self._predictor = predictor
2704
2705
2706
2707
2708
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
                used_indices=None
            )
2709
2710
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2711
2712
2713
2714
2715
            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
2716
        else:
2717
2718
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2719
        return self
Guolin Ke's avatar
Guolin Ke committed
2720

2721
    def set_reference(self, reference: "Dataset") -> "Dataset":
2722
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2723
2724
2725
2726

        Parameters
        ----------
        reference : Dataset
2727
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2728
2729
2730
2731
2732

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2733
        """
2734
2735
2736
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2737
        # we're done if self and reference share a common upstream reference
2738
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2739
            return self
Guolin Ke's avatar
Guolin Ke committed
2740
2741
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2742
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2743
        else:
2744
2745
            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
2746

2747
    def set_feature_name(self, feature_name: _LGBM_FeatureNameConfiguration) -> "Dataset":
2748
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2749
2750
2751

        Parameters
        ----------
2752
        feature_name : list of str
2753
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2754
2755
2756
2757
2758

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2759
        """
2760
2761
        if feature_name != 'auto':
            self.feature_name = feature_name
2762
        if self._handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2763
            if len(feature_name) != self.num_feature():
2764
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2765
            c_feature_name = [_c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2766
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
2767
                self._handle,
2768
                _c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2769
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2770
        return self
Guolin Ke's avatar
Guolin Ke committed
2771

2772
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
2773
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2774
2775
2776

        Parameters
        ----------
2777
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None
2778
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2779
2780
2781
2782
2783

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2784
2785
        """
        self.label = label
2786
        if self._handle is not None:
2787
2788
2789
2790
            if isinstance(label, pd_DataFrame):
                if len(label.columns) > 1:
                    raise ValueError('DataFrame for label cannot have multiple columns')
                _check_for_bad_pandas_dtypes(label.dtypes)
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
                try:
                    # most common case (no nullable dtypes)
                    label = label.to_numpy(dtype=np.float32, copy=False)
                except TypeError:
                    # 1.0 <= pd version < 1.1 and nullable dtypes, least common case
                    # raises error because array is casted to type(pd.NA) and there's no na_value argument
                    label = label.astype(np.float32, copy=False).values
                except ValueError:
                    # data has nullable dtypes, but we can specify na_value argument and copy will be made
                    label = label.to_numpy(dtype=np.float32, na_value=np.nan)
                label_array = np.ravel(label)
2802
2803
            elif _is_pyarrow_array(label):
                label_array = label
2804
            else:
2805
                label_array = _list_to_1d_numpy(label, dtype=np.float32, name='label')
2806
            self.set_field('label', label_array)
2807
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2808
        return self
Guolin Ke's avatar
Guolin Ke committed
2809

2810
2811
2812
2813
    def set_weight(
        self,
        weight: Optional[_LGBM_WeightType]
    ) -> "Dataset":
2814
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2815
2816
2817

        Parameters
        ----------
2818
        weight : list, numpy 1-D array, pandas Series or None
2819
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2820
2821
2822
2823
2824

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2825
        """
2826
2827
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2828
        self.weight = weight
2829
        if self._handle is not None and weight is not None:
2830
            weight = _list_to_1d_numpy(weight, dtype=np.float32, name='weight')
wxchan's avatar
wxchan committed
2831
            self.set_field('weight', weight)
2832
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2833
        return self
Guolin Ke's avatar
Guolin Ke committed
2834

2835
2836
2837
2838
    def set_init_score(
        self,
        init_score: Optional[_LGBM_InitScoreType]
    ) -> "Dataset":
2839
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2840
2841
2842

        Parameters
        ----------
2843
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2844
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2845
2846
2847
2848
2849

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2850
2851
        """
        self.init_score = init_score
2852
        if self._handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2853
            self.set_field('init_score', init_score)
2854
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2855
        return self
Guolin Ke's avatar
Guolin Ke committed
2856

2857
2858
2859
2860
    def set_group(
        self,
        group: Optional[_LGBM_GroupType]
    ) -> "Dataset":
2861
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2862
2863
2864

        Parameters
        ----------
2865
        group : list, numpy 1-D array, pandas Series or None
2866
2867
2868
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2869
2870
            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
2871
2872
2873
2874
2875

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2876
2877
        """
        self.group = group
2878
        if self._handle is not None and group is not None:
2879
            group = _list_to_1d_numpy(group, dtype=np.int32, name='group')
wxchan's avatar
wxchan committed
2880
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2881
        return self
Guolin Ke's avatar
Guolin Ke committed
2882

2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
    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

2905
    def get_feature_name(self) -> List[str]:
2906
2907
2908
2909
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2910
        feature_names : list of str
2911
2912
            The names of columns (features) in the Dataset.
        """
2913
        if self._handle is None:
2914
2915
2916
2917
2918
            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)
2919
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2920
2921
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
2922
            self._handle,
2923
            ctypes.c_int(num_feature),
2924
            ctypes.byref(tmp_out_len),
2925
            ctypes.c_size_t(reserved_string_buffer_size),
2926
2927
2928
2929
            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")
2930
2931
2932
2933
2934
2935
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
2936
                self._handle,
2937
2938
2939
2940
2941
                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))
2942
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2943

2944
    def get_label(self) -> Optional[np.ndarray]:
2945
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2946
2947
2948

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2949
        label : numpy array or None
2950
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2951
        """
2952
        if self.label is None:
wxchan's avatar
wxchan committed
2953
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2954
2955
        return self.label

2956
    def get_weight(self) -> Optional[np.ndarray]:
2957
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2958
2959
2960

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2961
        weight : numpy array or None
2962
            Weight for each data point from the Dataset. Weights should be non-negative.
Guolin Ke's avatar
Guolin Ke committed
2963
        """
2964
        if self.weight is None:
wxchan's avatar
wxchan committed
2965
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2966
2967
        return self.weight

2968
    def get_init_score(self) -> Optional[np.ndarray]:
2969
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2970
2971
2972

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2973
        init_score : numpy array or None
2974
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2975
        """
2976
        if self.init_score is None:
wxchan's avatar
wxchan committed
2977
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2978
2979
        return self.init_score

2980
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
2981
2982
2983
2984
        """Get the raw data of the Dataset.

        Returns
        -------
2985
        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
2986
2987
            Raw data used in the Dataset construction.
        """
2988
        if self._handle is None:
2989
            raise Exception("Cannot get data before construct Dataset")
2990
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
2991
2992
            self.data = self.reference.data
            if self.data is not None:
2993
                if isinstance(self.data, np.ndarray) or isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
2994
                    self.data = self.data[self.used_indices, :]
2995
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2996
                    self.data = self.data.iloc[self.used_indices].copy()
2997
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2998
                    self.data = self.data[self.used_indices, :]
2999
3000
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
3001
                elif _is_list_of_sequences(self.data) and len(self.data) > 0:
3002
                    self.data = np.array(list(self._yield_row_from_seqlist(self.data, self.used_indices)))
Guolin Ke's avatar
Guolin Ke committed
3003
                else:
3004
3005
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
3006
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
3007
3008
3009
        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.")
3010
3011
        return self.data

3012
    def get_group(self) -> Optional[np.ndarray]:
3013
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3014
3015
3016

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3017
        group : numpy array or None
3018
3019
3020
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
3021
3022
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
Guolin Ke's avatar
Guolin Ke committed
3023
        """
3024
        if self.group is None:
wxchan's avatar
wxchan committed
3025
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
3026
3027
            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
3028
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
3029
3030
        return self.group

3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
    def get_position(self) -> Optional[np.ndarray]:
        """Get the position of the Dataset.

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

3043
    def num_data(self) -> int:
3044
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3045
3046
3047

        Returns
        -------
3048
3049
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3050
        """
3051
        if self._handle is not None:
3052
            ret = ctypes.c_int(0)
3053
            _safe_call(_LIB.LGBM_DatasetGetNumData(self._handle,
wxchan's avatar
wxchan committed
3054
3055
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3056
        else:
3057
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3058

3059
    def num_feature(self) -> int:
3060
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3061
3062
3063

        Returns
        -------
3064
3065
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3066
        """
3067
        if self._handle is not None:
3068
            ret = ctypes.c_int(0)
3069
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self._handle,
wxchan's avatar
wxchan committed
3070
3071
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3072
        else:
3073
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3074

3075
    def feature_num_bin(self, feature: Union[int, str]) -> int:
3076
3077
        """Get the number of bins for a feature.

3078
3079
        .. versionadded:: 4.0.0

3080
3081
        Parameters
        ----------
3082
3083
        feature : int or str
            Index or name of the feature.
3084
3085
3086
3087
3088
3089

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
3090
        if self._handle is not None:
3091
            if isinstance(feature, str):
3092
3093
3094
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
3095
            ret = ctypes.c_int(0)
3096
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self._handle,
3097
                                                         ctypes.c_int(feature_index),
3098
3099
3100
3101
3102
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

3103
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
3104
3105
3106
3107
3108
        """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.
3109
3110
3111
3112
3113

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
3114
3115
3116

        Returns
        -------
3117
3118
3119
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
3120
        head = self
3121
        ref_chain: Set[Dataset] = set()
3122
3123
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
3124
                ref_chain.add(head)
3125
3126
3127
3128
3129
3130
                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
3131
        return ref_chain
3132

3133
    def add_features_from(self, other: "Dataset") -> "Dataset":
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
        """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.
        """
3148
        if self._handle is None or other._handle is None:
3149
            raise ValueError('Both source and target Datasets must be constructed before adding features')
3150
        _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self._handle, other._handle))
Guolin Ke's avatar
Guolin Ke committed
3151
3152
3153
3154
3155
3156
3157
3158
        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))
3159
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3160
                    self.data = np.hstack((self.data, other.data.toarray()))
3161
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3162
                    self.data = np.hstack((self.data, other.data.values))
3163
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3164
3165
3166
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
3167
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3168
                sparse_format = self.data.getformat()
3169
                if isinstance(other.data, np.ndarray) or isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3170
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
3171
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3172
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
3173
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3174
3175
3176
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
3177
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3178
3179
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
3180
3181
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
3182
                if isinstance(other.data, np.ndarray):
3183
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
3184
                                       axis=1, ignore_index=True)
3185
                elif isinstance(other.data, scipy.sparse.spmatrix):
3186
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
3187
                                       axis=1, ignore_index=True)
3188
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3189
3190
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
3191
3192
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
3193
3194
3195
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
3196
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3197
                if isinstance(other.data, np.ndarray):
3198
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
3199
                elif isinstance(other.data, scipy.sparse.spmatrix):
3200
3201
3202
3203
3204
                    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
3205
3206
3207
3208
3209
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
3210
3211
            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
3212
3213
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
3214
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
3215
        self.feature_name = self.get_feature_name()
3216
3217
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
3218
3219
        self.categorical_feature = "auto"
        self.pandas_categorical = None
3220
3221
        return self

3222
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
3223
3224
3225
3226
3227
3228
        """Save Dataset to a text file.

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

        Parameters
        ----------
3229
        filename : str or pathlib.Path
3230
3231
3232
3233
3234
3235
3236
3237
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
3238
            self.construct()._handle,
3239
            _c_str(str(filename))))
3240
3241
        return self

wxchan's avatar
wxchan committed
3242

3243
3244
3245
3246
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
    Tuple[np.ndarray, np.ndarray]
]
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
        _LGBM_EvalFunctionResultType
    ],
    Callable[
        [np.ndarray, Dataset],
        List[_LGBM_EvalFunctionResultType]
    ]
]
3257
3258


3259
class Booster:
3260
    """Booster in LightGBM."""
3261

3262
3263
3264
3265
3266
3267
3268
    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
    ):
3269
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3270
3271
3272

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3273
        params : dict or None, optional (default=None)
3274
3275
3276
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3277
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3278
            Path to the model file.
3279
        model_str : str or None, optional (default=None)
3280
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3281
        """
3282
        self._handle = ctypes.c_void_p()
3283
        self._network = False
wxchan's avatar
wxchan committed
3284
        self.__need_reload_eval_info = True
3285
        self._train_data_name = "training"
3286
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3287
        self.best_iteration = -1
3288
        self.best_score: _LGBM_BoosterBestScoreType = {}
3289
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3290
        if train_set is not None:
3291
            # Training task
wxchan's avatar
wxchan committed
3292
            if not isinstance(train_set, Dataset):
3293
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
            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"]
                )
3328
            # construct booster object
3329
3330
3331
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3332
            params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3333
            _safe_call(_LIB.LGBM_BoosterCreate(
3334
                train_set._handle,
3335
                _c_str(params_str),
3336
                ctypes.byref(self._handle)))
3337
            # save reference to data
wxchan's avatar
wxchan committed
3338
            self.train_set = train_set
3339
3340
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3341
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3342
3343
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
3344
                _safe_call(_LIB.LGBM_BoosterMerge(
3345
3346
                    self._handle,
                    self.__init_predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
3347
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3348
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3349
                self._handle,
wxchan's avatar
wxchan committed
3350
3351
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3352
            # buffer for inner predict
3353
            self.__inner_predict_buffer: List[Optional[np.ndarray]] = [None]
wxchan's avatar
wxchan committed
3354
3355
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3356
            self.pandas_categorical = train_set.pandas_categorical
3357
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3358
        elif model_file is not None:
3359
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3360
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3361
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
3362
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
3363
                ctypes.byref(out_num_iterations),
3364
                ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
3365
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3366
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3367
                self._handle,
wxchan's avatar
wxchan committed
3368
3369
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3370
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3371
3372
3373
            if params:
                _log_warning('Ignoring params argument, using parameters from model file.')
            params = self._get_loaded_param()
3374
        elif model_str is not None:
3375
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3376
        else:
3377
3378
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
3379
        self.params = params
wxchan's avatar
wxchan committed
3380

3381
    def __del__(self) -> None:
3382
        try:
3383
            if self._network:
3384
3385
3386
3387
                self.free_network()
        except AttributeError:
            pass
        try:
3388
3389
            if self._handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
3390
3391
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3392

3393
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3394
3395
        return self.__deepcopy__(None)

3396
    def __deepcopy__(self, _) -> "Booster":
3397
        model_str = self.model_to_string(num_iteration=-1)
3398
        return Booster(model_str=model_str)
wxchan's avatar
wxchan committed
3399

3400
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3401
        this = self.__dict__.copy()
3402
        handle = this['_handle']
wxchan's avatar
wxchan committed
3403
3404
3405
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
3406
            this["_handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3407
3408
        return this

3409
    def __setstate__(self, state: Dict[str, Any]) -> None:
3410
        model_str = state.get('_handle', state.get('handle', None))
3411
        if model_str is not None:
wxchan's avatar
wxchan committed
3412
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3413
            out_num_iterations = ctypes.c_int(0)
3414
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3415
                _c_str(model_str),
3416
3417
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
3418
            state['_handle'] = handle
wxchan's avatar
wxchan committed
3419
3420
        self.__dict__.update(state)

3421
3422
3423
3424
    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)
3425
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3426
        _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
3427
            self._handle,
3428
3429
3430
3431
3432
3433
3434
            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)
3435
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3436
            _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
3437
                self._handle,
3438
3439
3440
3441
3442
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
        return json.loads(string_buffer.value.decode('utf-8'))

3443
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3444
3445
3446
3447
3448
3449
3450
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
3451
3452
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
3453
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3454
        return self
wxchan's avatar
wxchan committed
3455

3456
    def _free_buffer(self) -> "Booster":
3457
3458
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3459
        return self
3460

3461
3462
3463
3464
3465
3466
3467
    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":
3468
3469
3470
3471
        """Set the network configuration.

        Parameters
        ----------
3472
        machines : list, set or str
3473
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3474
        local_listen_port : int, optional (default=12400)
3475
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3476
        listen_time_out : int, optional (default=120)
3477
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3478
        num_machines : int, optional (default=1)
3479
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3480
3481
3482
3483
3484

        Returns
        -------
        self : Booster
            Booster with set network.
3485
        """
3486
3487
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
3488
        _safe_call(_LIB.LGBM_NetworkInit(_c_str(machines),
3489
3490
3491
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
3492
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3493
        return self
3494

3495
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3496
3497
3498
3499
3500
3501
3502
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3503
        _safe_call(_LIB.LGBM_NetworkFree())
3504
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3505
        return self
3506

3507
    def trees_to_dataframe(self) -> pd_DataFrame:
3508
3509
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3510
3511
3512
3513
        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.
3514
3515
3516
3517
3518
            - ``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.
3519
3520
            - ``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.
3521
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3522
3523
              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.
3524
3525
            - ``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.
3526
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3527
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3528
3529
            - ``count`` : int64, number of records in the training data that fall into this node.

3530
3531
3532
3533
3534
3535
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3536
3537
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
3538
3539
3540
3541

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

3542
        def _is_split_node(tree: Dict[str, Any]) -> bool:
3543
3544
            return 'split_index' in tree.keys()

3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
        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:
3557
                tree_num = f'{tree_index}-' if tree_index is not None else ''
3558
3559
3560
                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
3561
3562
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
3563

3564
3565
3566
3567
            def _get_split_feature(
                tree: Dict[str, Any],
                feature_names: Optional[List[str]]
            ) -> Optional[str]:
3568
3569
3570
3571
3572
3573
3574
3575
3576
                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

3577
            def _is_single_node_tree(tree: Dict[str, Any]) -> bool:
3578
                return set(tree.keys()) == {'leaf_value'}
3579
3580

            # Create the node record, and populate universal data members
3581
            node: Dict[str, Union[int, str, None]] = OrderedDict()
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
            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

3618
3619
3620
3621
3622
3623
3624
        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]]:
3625

3626
            node = create_node_record(tree=tree,
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
                                      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(
3639
                        tree=tree[child],
3640
3641
3642
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
3643
3644
                        parent_node=node['node_index']
                    )
3645
3646
3647
3648
3649
3650
3651
3652
3653
                    # 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']:
3654
            model_list.extend(tree_dict_to_node_list(tree=tree['tree_structure'],
3655
3656
3657
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

3658
        return pd_DataFrame(model_list, columns=model_list[0].keys())
3659

3660
    def set_train_data_name(self, name: str) -> "Booster":
3661
3662
3663
3664
        """Set the name to the training Dataset.

        Parameters
        ----------
3665
        name : str
Nikita Titov's avatar
Nikita Titov committed
3666
3667
3668
3669
3670
3671
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
3672
        """
3673
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
3674
        return self
wxchan's avatar
wxchan committed
3675

3676
    def add_valid(self, data: Dataset, name: str) -> "Booster":
3677
        """Add validation data.
wxchan's avatar
wxchan committed
3678
3679
3680
3681

        Parameters
        ----------
        data : Dataset
3682
            Validation data.
3683
        name : str
3684
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
3685
3686
3687
3688
3689

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
3690
        """
Guolin Ke's avatar
Guolin Ke committed
3691
        if not isinstance(data, Dataset):
3692
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3693
        if data._predictor is not self.__init_predictor:
3694
3695
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3696
        _safe_call(_LIB.LGBM_BoosterAddValidData(
3697
3698
            self._handle,
            data.construct()._handle))
wxchan's avatar
wxchan committed
3699
3700
3701
3702
3703
        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
3704
        return self
wxchan's avatar
wxchan committed
3705

3706
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
3707
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
3708
3709
3710
3711

        Parameters
        ----------
        params : dict
3712
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
3713
3714
3715
3716
3717

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
3718
        """
3719
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3720
3721
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
3722
                self._handle,
3723
                _c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
3724
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
3725
        return self
wxchan's avatar
wxchan committed
3726

3727
3728
3729
3730
3731
    def update(
        self,
        train_set: Optional[Dataset] = None,
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
3732
        """Update Booster for one iteration.
3733

wxchan's avatar
wxchan committed
3734
3735
        Parameters
        ----------
3736
3737
3738
3739
        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
3740
            Customized objective function.
3741
3742
3743
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3744
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3745
                    The predicted values.
3746
3747
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3748
3749
                train_data : Dataset
                    The training dataset.
3750
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3751
3752
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3753
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3754
3755
                    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
3756

3757
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3758
            and grad and hess should be returned in the same format.
3759

wxchan's avatar
wxchan committed
3760
3761
        Returns
        -------
3762
3763
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3764
        """
3765
        # need reset training data
3766
3767
3768
3769
3770
3771
        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
3772
            if not isinstance(train_set, Dataset):
3773
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3774
            if train_set._predictor is not self.__init_predictor:
3775
3776
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3777
3778
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
3779
3780
                self._handle,
                self.train_set.construct()._handle))
wxchan's avatar
wxchan committed
3781
            self.__inner_predict_buffer[0] = None
3782
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3783
3784
        is_finished = ctypes.c_int(0)
        if fobj is None:
3785
            if self.__set_objective_to_none:
3786
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3787
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
3788
                self._handle,
wxchan's avatar
wxchan committed
3789
                ctypes.byref(is_finished)))
3790
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3791
3792
            return is_finished.value == 1
        else:
3793
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3794
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3795
3796
3797
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

3798
3799
3800
3801
3802
    def __boost(
        self,
        grad: np.ndarray,
        hess: np.ndarray
    ) -> bool:
3803
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3804

Nikita Titov's avatar
Nikita Titov committed
3805
3806
        .. note::

3807
3808
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3809
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3810
            and grad and hess should be returned in the same format.
3811

wxchan's avatar
wxchan committed
3812
3813
        Parameters
        ----------
3814
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3815
3816
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3817
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3818
3819
            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
3820
3821
3822

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3823
3824
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3825
        """
3826
3827
3828
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3829
3830
        grad = _list_to_1d_numpy(grad, dtype=np.float32, name='gradient')
        hess = _list_to_1d_numpy(hess, dtype=np.float32, name='hessian')
3831
3832
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3833
        if len(grad) != len(hess):
3834
3835
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3836
        if len(grad) != num_train_data * self.__num_class:
3837
3838
3839
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3840
                f"number of models per one iteration ({self.__num_class})"
3841
            )
wxchan's avatar
wxchan committed
3842
3843
        is_finished = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterUpdateOneIterCustom(
3844
            self._handle,
wxchan's avatar
wxchan committed
3845
3846
3847
            grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            ctypes.byref(is_finished)))
3848
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3849
3850
        return is_finished.value == 1

3851
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3852
3853
3854
3855
3856
3857
3858
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3859
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
3860
            self._handle))
3861
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3862
        return self
wxchan's avatar
wxchan committed
3863

3864
    def current_iteration(self) -> int:
3865
3866
3867
3868
3869
3870
3871
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3872
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3873
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
3874
            self._handle,
wxchan's avatar
wxchan committed
3875
3876
3877
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3878
    def num_model_per_iteration(self) -> int:
3879
3880
3881
3882
3883
3884
3885
3886
3887
        """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(
3888
            self._handle,
3889
3890
3891
            ctypes.byref(model_per_iter)))
        return model_per_iter.value

3892
    def num_trees(self) -> int:
3893
3894
3895
3896
3897
3898
3899
3900
3901
        """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(
3902
            self._handle,
3903
3904
3905
            ctypes.byref(num_trees)))
        return num_trees.value

3906
    def upper_bound(self) -> float:
3907
3908
3909
3910
        """Get upper bound value of a model.

        Returns
        -------
3911
        upper_bound : float
3912
3913
3914
3915
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
3916
            self._handle,
3917
3918
3919
            ctypes.byref(ret)))
        return ret.value

3920
    def lower_bound(self) -> float:
3921
3922
3923
3924
        """Get lower bound value of a model.

        Returns
        -------
3925
        lower_bound : float
3926
3927
3928
3929
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
3930
            self._handle,
3931
3932
3933
            ctypes.byref(ret)))
        return ret.value

3934
3935
3936
3937
3938
3939
    def eval(
        self,
        data: Dataset,
        name: str,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3940
        """Evaluate for data.
wxchan's avatar
wxchan committed
3941
3942
3943

        Parameters
        ----------
3944
3945
        data : Dataset
            Data for the evaluating.
3946
        name : str
3947
            Name of the data.
3948
        feval : callable, list of callable, or None, optional (default=None)
3949
            Customized evaluation function.
3950
            Each evaluation function should accept two parameters: preds, eval_data,
3951
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3952

3953
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3954
                    The predicted values.
3955
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3956
                    If custom objective function is used, predicted values are returned before any transformation,
3957
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3958
                eval_data : Dataset
3959
                    A ``Dataset`` to evaluate.
3960
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3961
                    The name of evaluation function (without whitespace).
3962
3963
3964
3965
3966
                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
3967
3968
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3969
        result : list
3970
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3971
        """
Guolin Ke's avatar
Guolin Ke committed
3972
3973
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3974
3975
3976
3977
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3978
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3979
3980
3981
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3982
        # need to push new valid data
wxchan's avatar
wxchan committed
3983
3984
3985
3986
3987
3988
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

3989
3990
3991
3992
    def eval_train(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3993
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3994
3995
3996

        Parameters
        ----------
3997
        feval : callable, list of callable, or None, optional (default=None)
3998
            Customized evaluation function.
3999
            Each evaluation function should accept two parameters: preds, eval_data,
4000
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4001

4002
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4003
                    The predicted values.
4004
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4005
                    If custom objective function is used, predicted values are returned before any transformation,
4006
                    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
4007
                eval_data : Dataset
4008
                    The training dataset.
4009
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4010
                    The name of evaluation function (without whitespace).
4011
4012
4013
4014
4015
                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
4016
4017
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4018
        result : list
4019
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4020
        """
4021
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
4022

4023
4024
4025
4026
    def eval_valid(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4027
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
4028
4029
4030

        Parameters
        ----------
4031
        feval : callable, list of callable, or None, optional (default=None)
4032
            Customized evaluation function.
4033
            Each evaluation function should accept two parameters: preds, eval_data,
4034
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4035

4036
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4037
                    The predicted values.
4038
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4039
                    If custom objective function is used, predicted values are returned before any transformation,
4040
                    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
4041
                eval_data : Dataset
4042
                    The validation dataset.
4043
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4044
                    The name of evaluation function (without whitespace).
4045
4046
4047
4048
4049
                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
4050
4051
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4052
        result : list
4053
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4054
        """
4055
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
4056
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
4057

4058
4059
4060
4061
4062
4063
4064
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> "Booster":
4065
        """Save Booster to file.
wxchan's avatar
wxchan committed
4066
4067
4068

        Parameters
        ----------
4069
        filename : str or pathlib.Path
4070
            Filename to save Booster.
4071
4072
4073
4074
        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
4075
        start_iteration : int, optional (default=0)
4076
            Start index of the iteration that should be saved.
4077
        importance_type : str, optional (default="split")
4078
4079
4080
            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
4081
4082
4083
4084
4085

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
4086
        """
4087
        if num_iteration is None:
4088
            num_iteration = self.best_iteration
4089
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4090
        _safe_call(_LIB.LGBM_BoosterSaveModel(
4091
            self._handle,
4092
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4093
            ctypes.c_int(num_iteration),
4094
            ctypes.c_int(importance_type_int),
4095
            _c_str(str(filename))))
4096
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
4097
        return self
wxchan's avatar
wxchan committed
4098

4099
4100
4101
4102
4103
    def shuffle_models(
        self,
        start_iteration: int = 0,
        end_iteration: int = -1
    ) -> "Booster":
4104
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
4105

4106
4107
4108
        Parameters
        ----------
        start_iteration : int, optional (default=0)
4109
            The first iteration that will be shuffled.
4110
4111
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
4112
            If <= 0, means the last available iteration.
4113

Nikita Titov's avatar
Nikita Titov committed
4114
4115
4116
4117
        Returns
        -------
        self : Booster
            Booster with shuffled models.
4118
        """
4119
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
4120
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
4121
4122
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
4123
        return self
4124

4125
    def model_from_string(self, model_str: str) -> "Booster":
4126
4127
4128
4129
        """Load Booster from a string.

        Parameters
        ----------
4130
        model_str : str
4131
4132
4133
4134
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4135
        self : Booster
4136
4137
            Loaded Booster object.
        """
4138
4139
4140
        # ensure that existing Booster is freed before replacing it
        # with a new one createdfrom file
        _safe_call(_LIB.LGBM_BoosterFree(self._handle))
4141
        self._free_buffer()
4142
        self._handle = ctypes.c_void_p()
4143
4144
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
4145
            _c_str(model_str),
4146
            ctypes.byref(out_num_iterations),
4147
            ctypes.byref(self._handle)))
4148
4149
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
4150
            self._handle,
4151
4152
            ctypes.byref(out_num_class)))
        self.__num_class = out_num_class.value
4153
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
4154
4155
        return self

4156
4157
4158
4159
4160
4161
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> str:
4162
        """Save Booster to string.
4163

4164
4165
4166
4167
4168
4169
        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
4170
        start_iteration : int, optional (default=0)
4171
            Start index of the iteration that should be saved.
4172
        importance_type : str, optional (default="split")
4173
4174
4175
            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.
4176
4177
4178

        Returns
        -------
4179
        str_repr : str
4180
4181
            String representation of Booster.
        """
4182
        if num_iteration is None:
4183
            num_iteration = self.best_iteration
4184
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4185
        buffer_len = 1 << 20
4186
        tmp_out_len = ctypes.c_int64(0)
4187
        string_buffer = ctypes.create_string_buffer(buffer_len)
4188
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4189
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
4190
            self._handle,
4191
            ctypes.c_int(start_iteration),
4192
            ctypes.c_int(num_iteration),
4193
            ctypes.c_int(importance_type_int),
4194
            ctypes.c_int64(buffer_len),
4195
4196
4197
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
4198
        # if buffer length is not long enough, re-allocate a buffer
4199
4200
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
4201
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4202
            _safe_call(_LIB.LGBM_BoosterSaveModelToString(
4203
                self._handle,
4204
                ctypes.c_int(start_iteration),
4205
                ctypes.c_int(num_iteration),
4206
                ctypes.c_int(importance_type_int),
4207
                ctypes.c_int64(actual_len),
4208
4209
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
4210
        ret = string_buffer.value.decode('utf-8')
4211
4212
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
4213

4214
4215
4216
4217
4218
4219
4220
    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
4221
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
4222

4223
4224
        Parameters
        ----------
4225
4226
4227
4228
        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
4229
        start_iteration : int, optional (default=0)
4230
            Start index of the iteration that should be dumped.
4231
        importance_type : str, optional (default="split")
4232
4233
4234
            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.
4235
4236
4237
4238
4239
4240
4241
4242
4243
        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.
4244

wxchan's avatar
wxchan committed
4245
4246
        Returns
        -------
4247
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
4248
            JSON format of Booster.
wxchan's avatar
wxchan committed
4249
        """
4250
        if num_iteration is None:
4251
            num_iteration = self.best_iteration
4252
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4253
        buffer_len = 1 << 20
4254
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
4255
        string_buffer = ctypes.create_string_buffer(buffer_len)
4256
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
wxchan's avatar
wxchan committed
4257
        _safe_call(_LIB.LGBM_BoosterDumpModel(
4258
            self._handle,
4259
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4260
            ctypes.c_int(num_iteration),
4261
            ctypes.c_int(importance_type_int),
4262
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
4263
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4264
            ptr_string_buffer))
wxchan's avatar
wxchan committed
4265
        actual_len = tmp_out_len.value
4266
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
4267
4268
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
4269
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
wxchan's avatar
wxchan committed
4270
            _safe_call(_LIB.LGBM_BoosterDumpModel(
4271
                self._handle,
4272
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4273
                ctypes.c_int(num_iteration),
4274
                ctypes.c_int(importance_type_int),
4275
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
4276
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4277
                ptr_string_buffer))
4278
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
4279
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
4280
                                                          default=_json_default_with_numpy))
4281
        return ret
wxchan's avatar
wxchan committed
4282

4283
4284
    def predict(
        self,
4285
        data: _LGBM_PredictDataType,
4286
4287
4288
4289
4290
4291
4292
4293
        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
4294
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4295
        """Make a prediction.
wxchan's avatar
wxchan committed
4296
4297
4298

        Parameters
        ----------
4299
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
4300
            Data source for prediction.
4301
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4302
        start_iteration : int, optional (default=0)
4303
            Start index of the iteration to predict.
4304
            If <= 0, starts from the first iteration.
4305
        num_iteration : int or None, optional (default=None)
4306
4307
4308
4309
            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).
4310
4311
4312
4313
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4314
4315
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4316

Nikita Titov's avatar
Nikita Titov committed
4317
4318
4319
4320
4321
4322
4323
            .. 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.
4324

4325
4326
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4327
            Used only if data is str.
4328
4329
4330
        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.
4331
4332
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4333
4334
4335

        Returns
        -------
4336
        result : numpy array, scipy.sparse or list of scipy.sparse
4337
            Prediction result.
4338
            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
4339
        """
4340
4341
4342
4343
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs),
        )
4344
        if num_iteration is None:
4345
            if start_iteration <= 0:
4346
4347
4348
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
        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
4359

4360
4361
    def refit(
        self,
4362
        data: _LGBM_TrainDataType,
4363
        label: _LGBM_LabelType,
4364
4365
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4366
4367
4368
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4369
4370
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
4371
4372
4373
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4374
        **kwargs
4375
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4376
4377
4378
4379
        """Refit the existing Booster by new data.

        Parameters
        ----------
4380
        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
4381
            Data source for refit.
4382
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4383
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array or pyarrow ChunkedArray
Guolin Ke's avatar
Guolin Ke committed
4384
4385
            Label for refit.
        decay_rate : float, optional (default=0.9)
4386
4387
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4388
4389
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
4390
4391
4392

            .. versionadded:: 4.0.0

4393
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
4394
            Weight for each ``data`` instance. Weights should be non-negative.
4395
4396
4397

            .. versionadded:: 4.0.0

4398
4399
4400
4401
4402
4403
        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.
4404
4405
4406

            .. versionadded:: 4.0.0

4407
4408
        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``.
4409
4410
4411

            .. versionadded:: 4.0.0

4412
4413
4414
        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.
4415
4416
4417

            .. versionadded:: 4.0.0

4418
4419
4420
4421
4422
        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.
4423
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4424
4425
4426
            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.
4427
            Floating point numbers in categorical features will be rounded towards 0.
4428
4429
4430

            .. versionadded:: 4.0.0

4431
4432
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
4433
4434
4435

            .. versionadded:: 4.0.0

4436
4437
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
4438
4439
4440

            .. versionadded:: 4.0.0

4441
4442
4443
        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.
4444
4445
4446

            .. versionadded:: 4.0.0

4447
4448
        **kwargs
            Other parameters for refit.
4449
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4450
4451
4452
4453
4454
4455

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4456
4457
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
4458
4459
        if dataset_params is None:
            dataset_params = {}
4460
4461
4462
4463
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs)
        )
4464
        leaf_preds: np.ndarray = predictor.predict(  # type: ignore[assignment]
4465
4466
4467
4468
4469
            data=data,
            start_iteration=-1,
            pred_leaf=True,
            validate_features=validate_features
        )
4470
        nrow, ncol = leaf_preds.shape
4471
        out_is_linear = ctypes.c_int(0)
4472
        _safe_call(_LIB.LGBM_BoosterGetLinear(
4473
            self._handle,
4474
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
4475
4476
4477
4478
4479
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
4480
        new_params["linear_tree"] = bool(out_is_linear.value)
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
        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,
        )
4494
        new_params['refit_decay_rate'] = decay_rate
4495
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4496
4497
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
4498
4499
            new_booster._handle,
            predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
4500
        leaf_preds = leaf_preds.reshape(-1)
4501
        ptr_data, _, _ = _c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
4502
        _safe_call(_LIB.LGBM_BoosterRefit(
4503
            new_booster._handle,
Guolin Ke's avatar
Guolin Ke committed
4504
            ptr_data,
4505
4506
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
4507
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4508
4509
        return new_booster

4510
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
        """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.
        """
4525
4526
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
4527
            self._handle,
4528
4529
4530
4531
4532
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

4533
4534
4535
4536
4537
4538
4539
4540
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
    ) -> 'Booster':
        """Set the output of a leaf.

4541
4542
        .. versionadded:: 4.0.0

4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
        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(
4559
                self._handle,
4560
4561
4562
4563
4564
4565
4566
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.c_double(value)
            )
        )
        return self

4567
    def num_feature(self) -> int:
4568
4569
4570
4571
4572
4573
4574
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4575
4576
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
4577
            self._handle,
4578
4579
4580
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

4581
    def feature_name(self) -> List[str]:
4582
        """Get names of features.
wxchan's avatar
wxchan committed
4583
4584
4585

        Returns
        -------
4586
        result : list of str
4587
            List with names of features.
wxchan's avatar
wxchan committed
4588
        """
4589
        num_feature = self.num_feature()
4590
        # Get name of features
wxchan's avatar
wxchan committed
4591
        tmp_out_len = ctypes.c_int(0)
4592
4593
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
4594
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
4595
4596
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
4597
            self._handle,
4598
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
4599
            ctypes.byref(tmp_out_len),
4600
            ctypes.c_size_t(reserved_string_buffer_size),
4601
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4602
4603
4604
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
4605
4606
4607
4608
4609
4610
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
4611
                self._handle,
4612
4613
4614
4615
4616
                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))
4617
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
4618

4619
4620
4621
4622
4623
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
4624
        """Get feature importances.
4625

4626
4627
        Parameters
        ----------
4628
        importance_type : str, optional (default="split")
4629
4630
4631
            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.
4632
4633
4634
4635
        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).
4636

4637
4638
        Returns
        -------
4639
4640
        result : numpy array
            Array with feature importances.
4641
        """
4642
4643
        if iteration is None:
            iteration = self.best_iteration
4644
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4645
        result = np.empty(self.num_feature(), dtype=np.float64)
4646
        _safe_call(_LIB.LGBM_BoosterFeatureImportance(
4647
            self._handle,
4648
4649
4650
            ctypes.c_int(iteration),
            ctypes.c_int(importance_type_int),
            result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
4651
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
4652
            return result.astype(np.int32)
4653
4654
        else:
            return result
4655

4656
4657
4658
4659
4660
4661
    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]:
4662
4663
4664
4665
        """Get split value histogram for the specified feature.

        Parameters
        ----------
4666
        feature : int or str
4667
4668
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
4669
            If str, interpreted as name.
4670

Nikita Titov's avatar
Nikita Titov committed
4671
4672
4673
            .. warning::

                Categorical features are not supported.
4674

4675
        bins : int, str or None, optional (default=None)
4676
4677
4678
            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.
4679
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
        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.
        """
4694
        def add(root: Dict[str, Any]) -> None:
4695
4696
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
4697
                if feature_names is not None and isinstance(feature, str):
4698
4699
4700
4701
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
4702
                    if isinstance(root['threshold'], str):
4703
4704
4705
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
4706
4707
4708
4709
4710
4711
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
4712
        values: List[float] = []
4713
4714
4715
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

4716
        if bins is None or isinstance(bins, int) and xgboost_style:
4717
4718
4719
4720
4721
4722
4723
            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:
4724
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
4725
4726
4727
4728
4729
            else:
                return ret
        else:
            return hist, bin_edges

4730
4731
4732
4733
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
4734
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]]
4735
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4736
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
4737
        if data_idx >= self.__num_dataset:
4738
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4739
4740
4741
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
4742
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
4743
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
4744
            _safe_call(_LIB.LGBM_BoosterGetEval(
4745
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4746
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4747
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4748
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
4749
            if tmp_out_len.value != self.__num_inner_eval:
4750
                raise ValueError("Wrong length of eval results")
4751
            for i in range(self.__num_inner_eval):
4752
4753
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
4754
4755
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
4756
4757
4758
4759
4760
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
4761
4762
4763
4764
4765
4766
4767
4768
4769
            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
4770
4771
4772
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

4773
    def __inner_predict(self, data_idx: int) -> np.ndarray:
4774
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
4775
        if data_idx >= self.__num_dataset:
4776
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4777
4778
4779
4780
4781
        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
4782
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
4783
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
4784
4785
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
4786
            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
4787
            _safe_call(_LIB.LGBM_BoosterGetPredict(
4788
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4789
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4790
4791
                ctypes.byref(tmp_out_len),
                data_ptr))
4792
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):  # type: ignore[arg-type]
4793
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
4794
            self.__is_predicted_cur_iter[data_idx] = True
4795
        result: np.ndarray = self.__inner_predict_buffer[data_idx]  # type: ignore[assignment]
4796
4797
4798
4799
        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
4800

4801
    def __get_eval_info(self) -> None:
4802
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
4803
4804
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
4805
            out_num_eval = ctypes.c_int(0)
4806
            # Get num of inner evals
wxchan's avatar
wxchan committed
4807
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
4808
                self._handle,
wxchan's avatar
wxchan committed
4809
4810
4811
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
4812
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
4813
                tmp_out_len = ctypes.c_int(0)
4814
4815
4816
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
4817
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
4818
                ]
wxchan's avatar
wxchan committed
4819
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
4820
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
4821
                    self._handle,
4822
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
4823
                    ctypes.byref(tmp_out_len),
4824
                    ctypes.c_size_t(reserved_string_buffer_size),
4825
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4826
4827
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
4828
                    raise ValueError("Length of eval names doesn't equal with num_evals")
4829
4830
4831
4832
4833
4834
4835
4836
                actual_string_buffer_size = required_string_buffer_size.value
                # if buffer length is not long enough, reallocate buffers
                if reserved_string_buffer_size < actual_string_buffer_size:
                    string_buffers = [
                        ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(self.__num_inner_eval)
                    ]
                    ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
                    _safe_call(_LIB.LGBM_BoosterGetEvalNames(
4837
                        self._handle,
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
                        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
                ]