basic.py 194 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
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
from .compat import (
    PANDAS_INSTALLED,
    PYARROW_INSTALLED,
    arrow_cffi,
    arrow_is_floating,
    arrow_is_integer,
    concat,
    dt_DataTable,
    pa_Array,
    pa_chunked_array,
    pa_ChunkedArray,
    pa_compute,
    pa_Table,
    pd_CategoricalDtype,
    pd_DataFrame,
    pd_Series,
)
wxchan's avatar
wxchan committed
38
39
from .libpath import find_lib_path

40
41
42
if TYPE_CHECKING:
    from typing import Literal

43
44
45
46
47
48
49
    # typing.TypeGuard was only introduced in Python 3.10
    try:
        from typing import TypeGuard
    except ImportError:
        from typing_extensions import TypeGuard


50
51
52
53
54
55
56
57
58
__all__ = [
    'Booster',
    'Dataset',
    'LGBMDeprecationWarning',
    'LightGBMError',
    'register_logger',
    'Sequence',
]

59
_BoosterHandle = ctypes.c_void_p
60
_DatasetHandle = ctypes.c_void_p
61
62
63
64
_ctypes_int_ptr = Union[
    "ctypes._Pointer[ctypes.c_int32]",
    "ctypes._Pointer[ctypes.c_int64]"
]
65
66
67
68
_ctypes_int_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_int32]]",
    "ctypes.Array[ctypes._Pointer[ctypes.c_int64]]"
]
69
70
71
72
73
74
75
76
_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]]"
]
77
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
78
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
79
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
80
_LGBM_BoosterEvalMethodResultWithStandardDeviationType = Tuple[str, str, float, bool, float]
81
82
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], "Literal['auto']"]
_LGBM_FeatureNameConfiguration = Union[List[str], "Literal['auto']"]
83
84
85
86
_LGBM_GroupType = Union[
    List[float],
    List[int],
    np.ndarray,
87
88
89
    pd_Series,
    pa_Array,
    pa_ChunkedArray,
90
]
91
92
93
94
_LGBM_PositionType = Union[
    np.ndarray,
    pd_Series
]
95
96
97
98
99
100
_LGBM_InitScoreType = Union[
    List[float],
    List[List[float]],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
101
102
103
    pa_Table,
    pa_Array,
    pa_ChunkedArray,
104
]
105
106
107
108
109
110
111
112
113
_LGBM_TrainDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix,
    "Sequence",
    List["Sequence"],
114
115
    List[np.ndarray],
    pa_Table
116
]
117
_LGBM_LabelType = Union[
118
119
    List[float],
    List[int],
120
121
    np.ndarray,
    pd_Series,
122
123
124
    pd_DataFrame,
    pa_Array,
    pa_ChunkedArray,
125
]
126
127
128
129
130
131
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
132
133
    scipy.sparse.spmatrix,
    pa_Table,
134
]
135
136
137
138
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
139
140
141
    pd_Series,
    pa_Array,
    pa_ChunkedArray,
142
]
143
144
145
ZERO_THRESHOLD = 1e-35


146
147
148
149
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


150
def _get_sample_count(total_nrow: int, params: str) -> int:
151
152
153
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
154
        _c_str(params),
155
156
157
158
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
159

160
161
162
163
164
165
class _MissingType(Enum):
    NONE = 'None'
    NAN = 'NaN'
    ZERO = 'Zero'


166
class _DummyLogger:
167
    def info(self, msg: str) -> None:
168
        print(msg)  # noqa: T201
169

170
    def warning(self, msg: str) -> None:
171
172
173
        warnings.warn(msg, stacklevel=3)


174
175
176
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
177
178


179
180
181
182
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


183
184
185
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
186
187
188
189
    """Register custom logger.

    Parameters
    ----------
190
    logger : Any
191
        Custom logger.
192
193
194
195
    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.
196
    """
197
198
199
200
201
202
    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
203
    _LOGGER = logger
204
205
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
206
207


208
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
209
    """Join log messages from native library which come by chunks."""
210
    msg_normalized: List[str] = []
211
212

    @wraps(func)
213
    def wrapper(msg: str) -> None:
214
215
216
217
218
219
220
221
222
223
224
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


225
def _log_info(msg: str) -> None:
226
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
227
228


229
def _log_warning(msg: str) -> None:
230
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
231
232
233


@_normalize_native_string
234
def _log_native(msg: str) -> None:
235
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
236
237


238
def _log_callback(msg: bytes) -> None:
239
    """Redirect logs from native library into Python."""
240
    _log_native(str(msg.decode('utf-8')))
241
242


243
def _load_lib() -> ctypes.CDLL:
244
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
245
246
247
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
248
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
249
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
250
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
251
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
252
253
    return lib

wxchan's avatar
wxchan committed
254

255
256
257
258
259
260
261
# 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
262

wxchan's avatar
wxchan committed
263

264
_NUMERIC_TYPES = (int, float, bool)
265
_ArrayLike = Union[List, np.ndarray, pd_Series]
266
267


268
def _safe_call(ret: int) -> None:
269
270
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
271
272
273
    Parameters
    ----------
    ret : int
274
        The return value from C API calls.
wxchan's avatar
wxchan committed
275
276
    """
    if ret != 0:
277
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
278

wxchan's avatar
wxchan committed
279

280
def _is_numeric(obj: Any) -> bool:
281
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
282
283
284
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
285
286
287
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
288
289
        return False

wxchan's avatar
wxchan committed
290

291
def _is_numpy_1d_array(data: Any) -> bool:
292
    """Check whether data is a numpy 1-D array."""
293
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
294

wxchan's avatar
wxchan committed
295

296
def _is_numpy_column_array(data: Any) -> bool:
297
298
299
300
301
302
303
    """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


304
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
305
    """Cast numpy array to given dtype."""
306
307
308
309
310
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


311
def _is_1d_list(data: Any) -> bool:
312
    """Check whether data is a 1-D list."""
313
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
314

wxchan's avatar
wxchan committed
315

316
317
318
319
320
321
322
323
324
325
326
327
328
329
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)
    )


330
331
332
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
333
        _is_numpy_1d_array(data)
334
        or _is_numpy_column_array(data)
335
        or _is_1d_list(data)
336
337
338
339
        or isinstance(data, pd_Series)
    )


340
341
def _list_to_1d_numpy(
    data: Any,
342
343
    dtype: "np.typing.DTypeLike",
    name: str
344
) -> np.ndarray:
345
    """Convert data to numpy 1-D array."""
346
    if _is_numpy_1d_array(data):
347
        return _cast_numpy_array_to_dtype(data, dtype)
348
    elif _is_numpy_column_array(data):
349
350
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
351
        return _cast_numpy_array_to_dtype(array, dtype)
352
    elif _is_1d_list(data):
wxchan's avatar
wxchan committed
353
        return np.array(data, dtype=dtype, copy=False)
354
    elif isinstance(data, pd_Series):
355
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
356
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
357
    else:
358
359
        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
360

wxchan's avatar
wxchan committed
361

362
363
364
365
366
367
368
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."""
369
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
370
371
372
373
374
375
376
377
378
379
380


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)
    )


381
def _is_pyarrow_array(data: Any) -> "TypeGuard[Union[pa_Array, pa_ChunkedArray]]":
382
383
384
385
    """Check whether data is a PyArrow array."""
    return isinstance(data, (pa_Array, pa_ChunkedArray))


386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
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
417
418
419
420
421
    if isinstance(data, pa_Array):
        export_objects = [data]
    elif isinstance(data, pa_ChunkedArray):
        export_objects = data.chunks
    elif isinstance(data, pa_Table):
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
        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)



443
444
def _data_to_2d_numpy(
    data: Any,
445
446
    dtype: "np.typing.DTypeLike",
    name: str
447
) -> np.ndarray:
448
449
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
450
        return _cast_numpy_array_to_dtype(data, dtype)
451
452
453
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
454
        _check_for_bad_pandas_dtypes(data.dtypes)
455
        return _cast_numpy_array_to_dtype(data.values, dtype)
456
457
458
459
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


460
def _cfloat32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
461
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
462
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
463
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
464
    else:
465
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
466

Guolin Ke's avatar
Guolin Ke committed
467

468
def _cfloat64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
469
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
470
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
471
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
472
473
474
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
475

476
def _cint32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
477
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
478
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
479
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
480
    else:
481
482
483
        raise RuntimeError('Expected int32 pointer')


484
def _cint64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
485
486
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
487
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
488
489
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
490

wxchan's avatar
wxchan committed
491

492
def _c_str(string: str) -> ctypes.c_char_p:
493
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
494
495
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
496

497
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
498
    """Convert a Python array to C array."""
499
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
500

wxchan's avatar
wxchan committed
501

502
def _json_default_with_numpy(obj: Any) -> Any:
503
504
505
506
507
508
509
510
511
    """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


512
513
514
515
516
517
518
519
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)


520
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
521
    """Convert Python dictionary to string, which is passed to C API."""
522
    if data is None or not data:
wxchan's avatar
wxchan committed
523
524
525
        return ""
    pairs = []
    for key, val in data.items():
526
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
527
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
528
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
529
            pairs.append(f"{key}={val}")
530
        elif val is not None:
531
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
532
    return ' '.join(pairs)
533

wxchan's avatar
wxchan committed
534

535
class _TempFile:
536
537
    """Proxy class to workaround errors on Windows."""

538
539
540
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
541
            self.path = Path(self.name)
542
        return self
wxchan's avatar
wxchan committed
543

544
    def __exit__(self, exc_type, exc_val, exc_tb):
545
546
        if self.path.is_file():
            self.path.unlink()
547

wxchan's avatar
wxchan committed
548

549
class LightGBMError(Exception):
550
551
    """Error thrown by LightGBM."""

552
553
554
    pass


555
556
557
558
559
560
561
562
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
563
564
565
566
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
567
    def _get_all_param_aliases() -> Dict[str, List[str]]:
568
569
570
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
571
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
572
573
574
575
576
577
578
579
        _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)
580
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
581
582
583
584
            _safe_call(_LIB.LGBM_DumpParamAliases(
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
585
        return json.loads(
586
            string_buffer.value.decode('utf-8'),
587
            object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
588
        )
589
590

    @classmethod
591
592
593
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
594
595
        ret = set()
        for i in args:
596
            ret.update(cls.get_sorted(i))
597
598
        return ret

599
600
601
602
603
604
    @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])

605
    @classmethod
606
607
608
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
609
610
611
612
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
613
                    ret.update(aliases)
614
615
616
                    break
        return ret

617

618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
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)

639
640
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
641
642

    # if main_param_name was provided, keep that value and remove all aliases
643
    if main_param_name in params.keys():
644
645
646
        for param in aliases:
            params.pop(param, None)
        return params
647

648
649
650
651
652
    # 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
653

654
655
656
657
658
659
660
    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
661
662
663
664

    return params


665
_MAX_INT32 = (1 << 31) - 1
666

667
"""Macro definition of data type in C API of LightGBM"""
668
669
670
671
_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
672

673
"""Matrix is row major in Python"""
674
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
675

676
"""Macro definition of prediction type in C API of LightGBM"""
677
678
679
680
_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
681

682
"""Macro definition of sparse matrix type"""
683
684
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
685

686
"""Macro definition of feature importance type"""
687
688
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
689

690
"""Data type of data field"""
691
692
693
694
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
695
696
    "group": _C_API_DTYPE_INT32,
    "position": _C_API_DTYPE_INT32
697
}
wxchan's avatar
wxchan committed
698

699
"""String name to int feature importance type mapper"""
700
701
702
703
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN
}
704

wxchan's avatar
wxchan committed
705

706
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
707
    """Fix the memory of multi-dimensional sliced object."""
708
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
709
        if not data.flags.c_contiguous:
710
711
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
712
713
714
715
            return np.copy(data)
    return data


716
717
718
def _c_float_array(
    data: np.ndarray
) -> Tuple[_ctypes_float_ptr, int, np.ndarray]:
719
    """Get pointer of float 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_float_ptr
wxchan's avatar
wxchan committed
726
727
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
728
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
729
730
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
731
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
732
        else:
733
            raise TypeError(f"Expected np.float32 or np.float64, 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
740
741
def _c_int_array(
    data: np.ndarray
) -> Tuple[_ctypes_int_ptr, int, np.ndarray]:
742
    """Get pointer of int numpy array / list."""
743
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
744
        data = np.array(data, copy=False)
745
    if _is_numpy_1d_array(data):
746
        data = _convert_from_sliced_object(data)
747
        assert data.flags.c_contiguous
748
        ptr_data: _ctypes_int_ptr
wxchan's avatar
wxchan committed
749
750
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
751
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
752
753
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
754
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
755
        else:
756
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
757
    else:
758
        raise TypeError(f"Unknown type({type(data).__name__})")
759
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
760

wxchan's avatar
wxchan committed
761

762
def _is_allowed_numpy_dtype(dtype: type) -> bool:
763
    float128 = getattr(np, 'float128', type(None))
764
765
766
767
    return (
        issubclass(dtype, (np.integer, np.floating, np.bool_))
        and not issubclass(dtype, (np.timedelta64, float128))
    )
768
769


770
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
771
772
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
773
        for column_name, pandas_dtype in pandas_dtypes_series.items()
774
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
775
776
777
778
    ]
    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)}')
779
780


781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
def _pandas_to_numpy(
    data: pd_DataFrame,
    target_dtype: "np.typing.DTypeLike"
) -> np.ndarray:
    _check_for_bad_pandas_dtypes(data.dtypes)
    try:
        # most common case (no nullable dtypes)
        return data.to_numpy(dtype=target_dtype, copy=False)
    except TypeError:
        # 1.0 <= pd version < 1.1 and nullable dtypes, least common case
        # raises error because array is casted to type(pd.NA) and there's no na_value argument
        return data.astype(target_dtype, copy=False).values
    except ValueError:
        # data has nullable dtypes, but we can specify na_value argument and copy will be made
        return data.to_numpy(dtype=target_dtype, na_value=np.nan)


798
def _data_from_pandas(
799
800
801
    data: pd_DataFrame,
    feature_name: _LGBM_FeatureNameConfiguration,
    categorical_feature: _LGBM_CategoricalFeatureConfiguration,
802
    pandas_categorical: Optional[List[List]]
803
) -> Tuple[np.ndarray, List[str], Union[List[str], List[int]], List[List]]:
804
805
806
    if len(data.shape) != 2 or data.shape[0] < 1:
        raise ValueError('Input data must be 2 dimensional and non empty.')

807
808
809
810
    # take shallow copy in case we modify categorical columns
    # whole column modifications don't change the original df
    data = data.copy(deep=False)

811
812
813
814
815
816
    # 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)]
817
    cat_cols_not_ordered: List[str] = [col for col in cat_cols if not data[col].cat.ordered]
818
819
    if pandas_categorical is None:  # train dataset
        pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
820
    else:
821
822
823
824
825
826
827
        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[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
828
829
830

    # use cat cols from DataFrame
    if categorical_feature == 'auto':
831
832
833
        categorical_feature = cat_cols_not_ordered

    df_dtypes = [dtype.type for dtype in data.dtypes]
834
835
    # so that the target dtype considers floats
    df_dtypes.append(np.float32)
836
    target_dtype = np.result_type(*df_dtypes)
837
838
839
840
841
842
843

    return (
        _pandas_to_numpy(data, target_dtype=target_dtype),
        feature_name,
        categorical_feature,
        pandas_categorical
    )
844
845


846
847
848
849
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
    file_name: Optional[Union[str, Path]] = None
) -> str:
850
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
851
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
852
853
854
855
856
857
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


858
859
860
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
    model_str: Optional[str] = None
861
) -> Optional[List[List]]:
862
863
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
864
    if file_name is not None:
865
        max_offset = -getsize(file_name)
866
867
868
869
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
870
                f.seek(offset, SEEK_END)
871
872
873
874
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
875
        last_line = lines[-1].decode('utf-8').strip()
876
        if not last_line.startswith(pandas_key):
877
            last_line = lines[-2].decode('utf-8').strip()
878
    elif model_str is not None:
879
880
881
882
883
884
        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
885
886


887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
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**.

907
908
    .. versionadded:: 3.3.0

909
910
911
912
913
914
915
916
917
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
918
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
919
920
921
922
923
924
925
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
926
                return self._get_one_line(idx)
927
            elif isinstance(idx, slice):
928
929
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
930
                # Only required if using ``Dataset.subset()``.
931
                return np.array([self._get_one_line(i) for i in idx])
932
            else:
933
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
934
935
936

        Parameters
        ----------
937
        idx : int, slice[int], list[int]
938
939
940
941
            Item index.

        Returns
        -------
942
        result : numpy 1-D array or numpy 2-D array
943
            1-D array if idx is int, 2-D array if idx is slice or list.
944
945
946
947
948
949
950
951
952
        """
        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__()")


953
class _InnerPredictor:
954
955
956
957
958
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
959
960
961
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
962
    """
963

964
965
    def __init__(
        self,
966
967
968
969
        booster_handle: _BoosterHandle,
        pandas_categorical: Optional[List[List]],
        pred_parameter: Dict[str, Any],
        manage_handle: bool
970
    ):
971
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
972
973
974

        Parameters
        ----------
975
        booster_handle : object
976
            Handle of Booster.
977
978
979
980
        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
981
            Other parameters for the prediction.
982
983
        manage_handle : bool
            If ``True``, free the corresponding Booster on the C++ side when this Python object is deleted.
wxchan's avatar
wxchan committed
984
        """
985
986
987
988
989
990
991
992
        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(
993
                self._handle,
994
995
996
997
                ctypes.byref(out_num_class)
            )
        )
        self.num_class = out_num_class.value
wxchan's avatar
wxchan committed
998

999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
    @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
1058

1059
    def __del__(self) -> None:
1060
1061
        try:
            if self.__is_manage_handle:
1062
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
1063
1064
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
1065

1066
    def __getstate__(self) -> Dict[str, Any]:
1067
1068
        this = self.__dict__.copy()
        this.pop('handle', None)
1069
        this.pop('_handle', None)
1070
1071
        return this

1072
1073
    def predict(
        self,
1074
        data: _LGBM_PredictDataType,
1075
1076
1077
1078
1079
1080
1081
        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
1082
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
1083
        """Predict logic.
wxchan's avatar
wxchan committed
1084
1085
1086

        Parameters
        ----------
1087
        data : str, pathlib.Path, numpy array, pandas DataFrame, pyarrow Table, H2O DataTable's Frame or scipy.sparse
1088
            Data source for prediction.
1089
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
1090
1091
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
        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.
1103
1104
1105
        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
1106

1107
1108
            .. versionadded:: 4.0.0

wxchan's avatar
wxchan committed
1109
1110
        Returns
        -------
1111
        result : numpy array, scipy.sparse or list of scipy.sparse
1112
            Prediction result.
1113
            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
1114
        """
wxchan's avatar
wxchan committed
1115
        if isinstance(data, Dataset):
1116
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
1117
1118
1119
1120
1121
1122
        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(
1123
                    self._handle,
1124
1125
1126
1127
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
1128
1129
1130
1131
1132
1133
1134
1135
1136

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

1137
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
1138
        if raw_score:
1139
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
1140
        if pred_leaf:
1141
            predict_type = _C_API_PREDICT_LEAF_INDEX
1142
        if pred_contrib:
1143
            predict_type = _C_API_PREDICT_CONTRIB
cbecker's avatar
cbecker committed
1144

1145
        if isinstance(data, (str, Path)):
1146
            with _TempFile() as f:
wxchan's avatar
wxchan committed
1147
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
1148
                    self._handle,
1149
                    _c_str(str(data)),
1150
                    ctypes.c_int(data_has_header),
Guolin Ke's avatar
Guolin Ke committed
1151
                    ctypes.c_int(predict_type),
1152
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1153
                    ctypes.c_int(num_iteration),
1154
1155
                    _c_str(self.pred_parameter),
                    _c_str(f.name)))
1156
1157
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
1158
        elif isinstance(data, scipy.sparse.csr_matrix):
1159
1160
1161
1162
1163
1164
            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
1165
        elif isinstance(data, scipy.sparse.csc_matrix):
1166
1167
1168
1169
1170
1171
            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
1172
        elif isinstance(data, np.ndarray):
1173
1174
1175
1176
1177
1178
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1179
1180
1181
1182
1183
1184
1185
        elif _is_pyarrow_table(data):
            preds, nrow = self.__pred_for_pyarrow_table(
                table=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1186
1187
1188
        elif isinstance(data, list):
            try:
                data = np.array(data)
1189
1190
            except BaseException as err:
                raise ValueError('Cannot convert data list to numpy array.') from err
1191
1192
1193
1194
1195
1196
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1197
        elif isinstance(data, dt_DataTable):
1198
1199
1200
1201
1202
1203
            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
1204
1205
        else:
            try:
1206
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
1207
                csr = scipy.sparse.csr_matrix(data)
1208
1209
            except BaseException as err:
                raise TypeError(f'Cannot predict data for type {type(data).__name__}') from err
1210
1211
1212
1213
1214
1215
            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
1216
1217
        if pred_leaf:
            preds = preds.astype(np.int32)
1218
        is_sparse = isinstance(preds, (list, scipy.sparse.spmatrix))
1219
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1220
            if preds.size % nrow == 0:
1221
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1222
            else:
1223
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
1224
1225
        return preds

1226
1227
1228
1229
1230
1231
1232
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
        predict_type: int
    ) -> int:
1233
        """Get size of prediction result."""
1234
        if nrow > _MAX_INT32:
1235
            raise LightGBMError('LightGBM cannot perform prediction for data '
1236
                                f'with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n'
1237
                                'You can split your data into chunks '
1238
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
1239
1240
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
1241
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1242
1243
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
1244
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1245
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
1246
1247
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
1248

1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
    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)
1262
1263
1264
1265
1266
1267
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
            predict_type=predict_type
        )
1268
1269
1270
1271
1272
1273
        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(
1274
            self._handle,
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
            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]:
1297
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1298
        if len(mat.shape) != 2:
1299
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
1300

1301
        nrow = mat.shape[0]
1302
1303
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1304
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1305
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
1306
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1307
            preds = np.empty(sum(n_preds), dtype=np.float64)
1308
1309
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
1310
                # avoid memory consumption by arrays concatenation operations
1311
1312
1313
1314
1315
1316
1317
                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]
                )
1318
            return preds, nrow
wxchan's avatar
wxchan committed
1319
        else:
1320
1321
1322
1323
1324
1325
1326
            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
1327

1328
1329
1330
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1331
1332
1333
1334
1335
1336
        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,
1337
        is_csr: bool
1338
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1339
1340
1341
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1342
        if indptr_type == _C_API_DTYPE_INT32:
1343
            out_indptr = _cint32_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1344
        elif indptr_type == _C_API_DTYPE_INT64:
1345
            out_indptr = _cint64_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1346
1347
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1348
        if data_type == _C_API_DTYPE_FLOAT32:
1349
            out_data = _cfloat32_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1350
        elif data_type == _C_API_DTYPE_FLOAT64:
1351
            out_data = _cfloat64_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1352
1353
        else:
            raise TypeError("Expected float32 or float64 type for data")
1354
        out_indices = _cint32_array_to_numpy(cptr=out_ptr_indices, length=data_indices_len)
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
        # 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

1383
1384
1385
1386
1387
1388
1389
1390
1391
    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
1392
1393
1394
1395
1396
1397
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1398
1399
1400
1401
1402
        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
1403

1404
1405
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1406

1407
1408
1409
1410
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

        _safe_call(_LIB.LGBM_BoosterPredictForCSR(
1411
            self._handle,
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
            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
1436
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1437
1438
1439
1440
        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
1441
        out_ptr_indptr: _ctypes_int_ptr
1442
1443
1444
1445
1446
        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)()
1447
        out_ptr_data: _ctypes_float_ptr
1448
1449
1450
1451
1452
1453
        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(
1454
            self._handle,
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
            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

1485
1486
1487
1488
1489
1490
1491
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1492
        """Predict for a CSR data."""
1493
        if predict_type == _C_API_PREDICT_CONTRIB:
1494
1495
1496
1497
1498
1499
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1500
        nrow = len(csr.indptr) - 1
1501
1502
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1503
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1504
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1505
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1506
            preds = np.empty(sum(n_preds), dtype=np.float64)
1507
1508
            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:])):
1509
                # avoid memory consumption by arrays concatenation operations
1510
1511
1512
1513
1514
1515
1516
                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]
                )
1517
1518
            return preds, nrow
        else:
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
            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,
1529
1530
1531
1532
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
1533
1534
1535
1536
1537
    ):
        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
1538
        out_ptr_indptr: _ctypes_int_ptr
1539
1540
1541
1542
1543
        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)()
1544
        out_ptr_data: _ctypes_float_ptr
1545
1546
1547
1548
1549
1550
        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(
1551
            self._handle,
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
            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
1581

1582
1583
1584
1585
1586
1587
1588
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
1589
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1590
        nrow = csc.shape[0]
1591
        if nrow > _MAX_INT32:
1592
1593
1594
1595
1596
1597
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1598
        if predict_type == _C_API_PREDICT_CONTRIB:
1599
1600
1601
1602
1603
1604
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1605
1606
1607
1608
1609
1610
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
            predict_type=predict_type
        )
1611
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1612
1613
        out_num_preds = ctypes.c_int64(0)

1614
1615
        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
1616

1617
        assert csc.shape[0] <= _MAX_INT32
1618
        csc_indices = csc.indices.astype(np.int32, copy=False)
1619

Guolin Ke's avatar
Guolin Ke committed
1620
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
1621
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
1622
            ptr_indptr,
1623
            ctypes.c_int(type_ptr_indptr),
1624
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1625
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1626
1627
1628
1629
1630
            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),
1631
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1632
            ctypes.c_int(num_iteration),
1633
            _c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1634
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1635
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1636
        if n_preds != out_num_preds.value:
1637
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1638
        return preds, nrow
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
    
    def __pred_for_pyarrow_table(
        self,
        table: pa_Table,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
        """Predict for a PyArrow table."""
        if not PYARROW_INSTALLED:
            raise LightGBMError("Cannot predict 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")

        # Prepare prediction output array
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=table.num_rows,
            predict_type=predict_type
        )
        preds = np.empty(n_preds, dtype=np.float64)
        out_num_preds = ctypes.c_int64(0)

        # Export Arrow table to C and run prediction
        c_array = _export_arrow_to_c(table)
        _safe_call(_LIB.LGBM_BoosterPredictForArrow(
            self._handle,
            ctypes.c_int64(c_array.n_chunks),
            ctypes.c_void_p(c_array.chunks_ptr),
            ctypes.c_void_p(c_array.schema_ptr),
            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, table.num_rows
wxchan's avatar
wxchan committed
1681

1682
    def current_iteration(self) -> int:
1683
1684
1685
1686
1687
1688
1689
1690
1691
        """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(
1692
            self._handle,
1693
1694
1695
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1696

1697
class Dataset:
wxchan's avatar
wxchan committed
1698
    """Dataset in LightGBM."""
1699

1700
1701
    def __init__(
        self,
1702
        data: _LGBM_TrainDataType,
1703
        label: Optional[_LGBM_LabelType] = None,
1704
        reference: Optional["Dataset"] = None,
1705
1706
1707
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1708
1709
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
1710
        params: Optional[Dict[str, Any]] = None,
1711
1712
        free_raw_data: bool = True,
        position: Optional[_LGBM_PositionType] = None,
1713
    ):
1714
        """Initialize Dataset.
1715

wxchan's avatar
wxchan committed
1716
1717
        Parameters
        ----------
1718
        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
1719
            Data source of Dataset.
1720
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1721
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1722
1723
1724
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1725
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1726
            Weight for each instance. Weights should be non-negative.
1727
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1728
1729
1730
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1731
1732
            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.
1733
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), pyarrow Array, pyarrow ChunkedArray, pyarrow Table (for multi-class task) or None, optional (default=None)
1734
            Init score for Dataset.
1735
        feature_name : list of str, or 'auto', optional (default="auto")
1736
            Feature names.
1737
            If 'auto' and data is pandas DataFrame or pyarrow Table, data columns names are used.
1738
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1739
1740
            Categorical features.
            If list of int, interpreted as indices.
1741
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1742
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1743
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1744
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1745
            All negative values in categorical features will be treated as missing values.
1746
            The output cannot be monotonically constrained with respect to a categorical feature.
1747
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1748
        params : dict or None, optional (default=None)
1749
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1750
        free_raw_data : bool, optional (default=True)
1751
            If True, raw data is freed after constructing inner Dataset.
1752
1753
        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
1754
        """
1755
        self._handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1756
1757
1758
1759
1760
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1761
        self.position = position
1762
        self.init_score = init_score
1763
1764
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1765
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1766
        self.free_raw_data = free_raw_data
1767
        self.used_indices: Optional[List[int]] = None
1768
        self._need_slice = True
1769
        self._predictor: Optional[_InnerPredictor] = None
1770
        self.pandas_categorical: Optional[List[List]] = None
1771
        self._params_back_up = None
1772
        self.version = 0
1773
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1774

1775
    def __del__(self) -> None:
1776
1777
1778
1779
        try:
            self._free_handle()
        except AttributeError:
            pass
1780

1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
    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.
        """
1798
        param_str = _param_dict_to_str(self.get_params())
1799
1800
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1801
        ptr_data, _, _ = _c_int_array(indices)
1802
1803
1804
1805
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
1806
            _c_str(param_str),
1807
1808
1809
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1810
1811
        assert sample_cnt == actual_sample_cnt.value
        return indices
1812

1813
1814
1815
1816
1817
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
        ref_dataset: _DatasetHandle
    ) -> 'Dataset':
1818
1819
1820
1821
1822
1823
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1824
1825
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1826
1827
1828
1829
1830
1831

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
1832
        self._handle = ctypes.c_void_p()
1833
1834
1835
        _safe_call(_LIB.LGBM_DatasetCreateByReference(
            ref_dataset,
            ctypes.c_int64(total_nrow),
1836
            ctypes.byref(self._handle),
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
        ))
        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
        ----------
1851
        sample_data : list of numpy array
1852
            Sample data for each column.
1853
        sample_indices : list of numpy array
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
            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.
1876
        sample_col_ptr: _ctypes_float_array = (ctypes.POINTER(ctypes.c_double) * ncol)()
1877
1878
        # c type int**
        # each int* points to start of indices for each column
1879
        indices_col_ptr: _ctypes_int_array = (ctypes.POINTER(ctypes.c_int32) * ncol)()
1880
        for i in range(ncol):
1881
1882
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1883
1884

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1885
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1886

1887
        self._handle = ctypes.c_void_p()
1888
        params_str = _param_dict_to_str(self.get_params())
1889
1890
1891
1892
1893
1894
1895
        _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),
1896
            ctypes.c_int64(total_nrow),
1897
            _c_str(params_str),
1898
            ctypes.byref(self._handle),
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
        ))
        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)
1917
        data_ptr, data_type, _ = _c_float_array(data)
1918
1919

        _safe_call(_LIB.LGBM_DatasetPushRows(
1920
            self._handle,
1921
1922
1923
1924
1925
1926
1927
1928
1929
            data_ptr,
            data_type,
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol),
            ctypes.c_int32(self._start_row),
        ))
        self._start_row += nrow
        return self

1930
    def get_params(self) -> Dict[str, Any]:
1931
1932
1933
1934
        """Get the used parameters in the Dataset.

        Returns
        -------
1935
        params : dict
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
            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",
1951
                                                "linear_tree",
1952
1953
1954
1955
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1956
                                                "precise_float_parser",
1957
1958
1959
1960
1961
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1962
1963
        else:
            return {}
1964

1965
    def _free_handle(self) -> "Dataset":
1966
1967
1968
        if self._handle is not None:
            _safe_call(_LIB.LGBM_DatasetFree(self._handle))
            self._handle = None
1969
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
1970
1971
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1972
        return self
wxchan's avatar
wxchan committed
1973

1974
1975
1976
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
1977
        data: _LGBM_TrainDataType,
1978
        used_indices: Optional[Union[List[int], np.ndarray]]
1979
    ) -> "Dataset":
Guolin Ke's avatar
Guolin Ke committed
1980
        data_has_header = False
1981
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
1982
            # check data has header or not
1983
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1984
        num_data = self.num_data()
1985
        if predictor is not None:
1986
1987
1988
1989
1990
            init_score: Union[np.ndarray, scipy.sparse.spmatrix] = predictor.predict(
                data=data,
                raw_score=True,
                data_has_header=data_has_header
            )
1991
            init_score = init_score.ravel()
1992
            if used_indices is not None:
1993
                assert not self._need_slice
1994
                if isinstance(data, (str, Path)):
1995
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1996
                    assert num_data == len(used_indices)
1997
1998
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1999
2000
2001
2002
                            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
2003
                new_init_score = np.empty(init_score.size, dtype=np.float64)
2004
2005
                for i in range(num_data):
                    for j in range(predictor.num_class):
2006
2007
2008
                        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:
2009
            init_score = np.full_like(self.init_score, fill_value=0.0, dtype=np.float64)
2010
2011
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
2012
        self.set_init_score(init_score)
2013
        return self
Guolin Ke's avatar
Guolin Ke committed
2014

2015
2016
    def _lazy_init(
        self,
2017
        data: Optional[_LGBM_TrainDataType],
2018
2019
2020
2021
2022
2023
2024
2025
        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,
2026
2027
        params: Optional[Dict[str, Any]],
        position: Optional[_LGBM_PositionType]
2028
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2029
        if data is None:
2030
            self._handle = None
Nikita Titov's avatar
Nikita Titov committed
2031
            return self
Guolin Ke's avatar
Guolin Ke committed
2032
2033
2034
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
2035
2036
2037
2038
2039
2040
2041
        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
2042

2043
        # process for args
wxchan's avatar
wxchan committed
2044
        params = {} if params is None else params
2045
        args_names = inspect.signature(self.__class__._lazy_init).parameters.keys()
2046
        for key in params.keys():
2047
            if key in args_names:
2048
2049
                _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.')
2050
        # get categorical features
2051
        if isinstance(categorical_feature, list):
2052
2053
            categorical_indices = set()
            feature_dict = {}
2054
            if isinstance(feature_name, list):
2055
2056
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
2057
                if isinstance(name, str) and name in feature_dict:
2058
                    categorical_indices.add(feature_dict[name])
2059
                elif isinstance(name, int):
2060
2061
                    categorical_indices.add(name)
                else:
2062
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
2063
            if categorical_indices:
2064
2065
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
2066
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
2067
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
2068
                            _log_warning(f'{cat_alias} in param dict is overridden.')
2069
                        params.pop(cat_alias, None)
2070
                params['categorical_column'] = sorted(categorical_indices)
2071

2072
        params_str = _param_dict_to_str(params)
2073
        self.params = params
2074
        # process for reference dataset
wxchan's avatar
wxchan committed
2075
        ref_dataset = None
wxchan's avatar
wxchan committed
2076
        if isinstance(reference, Dataset):
2077
            ref_dataset = reference.construct()._handle
wxchan's avatar
wxchan committed
2078
2079
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
2080
        # start construct data
2081
        if isinstance(data, (str, Path)):
2082
            self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2083
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
2084
2085
                _c_str(str(data)),
                _c_str(params_str),
wxchan's avatar
wxchan committed
2086
                ref_dataset,
2087
                ctypes.byref(self._handle)))
wxchan's avatar
wxchan committed
2088
2089
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
2090
2091
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
2092
2093
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
2094
2095
2096
        elif _is_pyarrow_table(data):
            self.__init_from_pyarrow_table(data, params_str, ref_dataset)
            feature_name = data.column_names
2097
        elif isinstance(data, list) and len(data) > 0:
2098
            if _is_list_of_numpy_arrays(data):
2099
                self.__init_from_list_np2d(data, params_str, ref_dataset)
2100
            elif _is_list_of_sequences(data):
2101
2102
2103
2104
2105
                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)
2106
        elif isinstance(data, dt_DataTable):
2107
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
2108
2109
2110
2111
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
2112
2113
            except BaseException as err:
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}') from err
wxchan's avatar
wxchan committed
2114
2115
2116
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
2117
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
2118
2119
2120
2121
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
2122
2123
        if position is not None:
            self.set_position(position)
2124
2125
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
2126
                _log_warning("The init_score will be overridden by the prediction of init_model.")
2127
2128
2129
2130
2131
            self._set_init_score_by_predictor(
                predictor=predictor,
                data=data,
                used_indices=None
            )
2132
2133
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
2134
        elif predictor is not None:
2135
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2136
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
2137
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
2138

2139
2140
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
        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.
2166
        sampled = np.array(list(self._yield_row_from_seqlist(seqs, indices)))
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
        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

2182
2183
2184
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
2185
        ref_dataset: Optional[_DatasetHandle]
2186
    ) -> "Dataset":
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
        """
        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:
2201
            param_str = _param_dict_to_str(self.get_params())
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
            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

2215
2216
2217
2218
2219
2220
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2221
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
2222
2223
2224
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

2225
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2226
2227
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
2228
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
2229
2230
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

2231
        ptr_data, type_ptr_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2232
2233
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2234
            ctypes.c_int(type_ptr_data),
2235
2236
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
2237
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2238
            _c_str(params_str),
wxchan's avatar
wxchan committed
2239
            ref_dataset,
2240
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2241
        return self
wxchan's avatar
wxchan committed
2242

2243
2244
2245
2246
2247
2248
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2249
        """Initialize data from a list of 2-D numpy matrices."""
2250
        ncol = mats[0].shape[1]
2251
        nrow = np.empty((len(mats),), np.int32)
2252
        ptr_data: _ctypes_float_array
2253
2254
2255
2256
2257
2258
        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 = []
2259
        type_ptr_data = -1
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271

        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)
2272
            else:  # change non-float data to float data, need to copy
2273
2274
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2275
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2276
            if type_ptr_data != -1 and chunk_type_ptr_data != type_ptr_data:
2277
2278
2279
2280
2281
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

2282
        self._handle = ctypes.c_void_p()
2283
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
2284
            ctypes.c_int32(len(mats)),
2285
2286
2287
            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)),
2288
            ctypes.c_int32(ncol),
2289
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
2290
            _c_str(params_str),
2291
            ref_dataset,
2292
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2293
        return self
2294

2295
2296
2297
2298
2299
2300
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2301
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2302
        if len(csr.indices) != len(csr.data):
2303
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
2304
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2305

2306
2307
        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
2308

2309
        assert csr.shape[1] <= _MAX_INT32
2310
        csr_indices = csr.indices.astype(np.int32, copy=False)
2311

wxchan's avatar
wxchan committed
2312
2313
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2314
            ctypes.c_int(type_ptr_indptr),
2315
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
2316
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2317
2318
2319
2320
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
2321
            _c_str(params_str),
wxchan's avatar
wxchan committed
2322
            ref_dataset,
2323
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2324
        return self
wxchan's avatar
wxchan committed
2325

2326
2327
2328
2329
2330
2331
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
2332
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2333
        if len(csc.indices) != len(csc.data):
2334
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
2335
        self._handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2336

2337
2338
        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
2339

2340
        assert csc.shape[0] <= _MAX_INT32
2341
        csc_indices = csc.indices.astype(np.int32, copy=False)
2342

Guolin Ke's avatar
Guolin Ke committed
2343
2344
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
2345
            ctypes.c_int(type_ptr_indptr),
2346
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
2347
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2348
2349
2350
2351
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
2352
            _c_str(params_str),
Guolin Ke's avatar
Guolin Ke committed
2353
            ref_dataset,
2354
            ctypes.byref(self._handle)))
Nikita Titov's avatar
Nikita Titov committed
2355
        return self
Guolin Ke's avatar
Guolin Ke committed
2356

2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
    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

2383
    @staticmethod
2384
    def _compare_params_for_warning(
2385
2386
        params: Dict[str, Any],
        other_params: Dict[str, Any],
2387
2388
2389
        ignore_keys: Set[str]
    ) -> bool:
        """Compare two dictionaries with params ignoring some keys.
2390

2391
2392
2393
2394
        It is only for the warning purpose.

        Parameters
        ----------
2395
        params : dict
2396
            One dictionary with parameters to compare.
2397
        other_params : dict
2398
2399
2400
            Another dictionary with parameters to compare.
        ignore_keys : set
            Keys that should be ignored during comparing two dictionaries.
2401
2402
2403

        Returns
        -------
2404
2405
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
        """
        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

2417
    def construct(self) -> "Dataset":
2418
2419
2420
2421
2422
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2423
            Constructed Dataset object.
2424
        """
2425
        if self._handle is None:
wxchan's avatar
wxchan committed
2426
            if self.reference is not None:
2427
                reference_params = self.reference.get_params()
2428
2429
                params = self.get_params()
                if params != reference_params:
2430
2431
2432
2433
2434
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
2435
                        _log_warning('Overriding the parameters from Reference Dataset.')
2436
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2437
                if self.used_indices is None:
2438
                    # create valid
2439
                    self._lazy_init(data=self.data, label=self.label, reference=self.reference,
2440
                                    weight=self.weight, group=self.group, position=self.position,
2441
                                    init_score=self.init_score, predictor=self._predictor,
2442
                                    feature_name=self.feature_name, categorical_feature='auto', params=self.params)
wxchan's avatar
wxchan committed
2443
                else:
2444
                    # construct subset
2445
                    used_indices = _list_to_1d_numpy(self.used_indices, dtype=np.int32, name='used_indices')
2446
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2447
                    if self.reference.group is not None:
2448
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2449
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
2450
                                                  return_counts=True)
2451
                    self._handle = ctypes.c_void_p()
2452
                    params_str = _param_dict_to_str(self.params)
wxchan's avatar
wxchan committed
2453
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
2454
                        self.reference.construct()._handle,
wxchan's avatar
wxchan committed
2455
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
2456
                        ctypes.c_int32(used_indices.shape[0]),
2457
                        _c_str(params_str),
2458
                        ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
2459
2460
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2461
2462
                    if self.group is not None:
                        self.set_group(self.group)
2463
2464
                    if self.position is not None:
                        self.set_position(self.position)
wxchan's avatar
wxchan committed
2465
2466
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
2467
2468
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
2469
2470
2471
2472
2473
                        self._set_init_score_by_predictor(
                            predictor=self._predictor,
                            data=self.data,
                            used_indices=used_indices
                        )
wxchan's avatar
wxchan committed
2474
            else:
2475
                # create train
2476
                self._lazy_init(data=self.data, label=self.label, reference=None,
2477
2478
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
2479
2480
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature,
                                params=self.params, position=self.position)
wxchan's avatar
wxchan committed
2481
2482
            if self.free_raw_data:
                self.data = None
2483
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2484
        return self
wxchan's avatar
wxchan committed
2485

2486
2487
    def create_valid(
        self,
2488
        data: _LGBM_TrainDataType,
2489
        label: Optional[_LGBM_LabelType] = None,
2490
2491
2492
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2493
2494
        params: Optional[Dict[str, Any]] = None,
        position: Optional[_LGBM_PositionType] = None
2495
    ) -> "Dataset":
2496
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2497
2498
2499

        Parameters
        ----------
2500
        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
2501
            Data source of Dataset.
2502
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2503
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2504
            Label of the data.
2505
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2506
            Weight for each instance. Weights should be non-negative.
2507
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2508
2509
2510
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2511
2512
            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.
2513
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), pyarrow Array, pyarrow ChunkedArray, pyarrow Table (for multi-class task) or None, optional (default=None)
2514
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2515
        params : dict or None, optional (default=None)
2516
            Other parameters for validation Dataset.
2517
2518
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
2519
2520
2521

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2522
2523
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2524
        """
2525
        ret = Dataset(data, label=label, reference=self,
2526
                      weight=weight, group=group, position=position, init_score=init_score,
2527
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2528
        ret._predictor = self._predictor
2529
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2530
        return ret
wxchan's avatar
wxchan committed
2531

2532
2533
2534
2535
2536
    def subset(
        self,
        used_indices: List[int],
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2537
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2538
2539
2540
2541

        Parameters
        ----------
        used_indices : list of int
2542
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2543
        params : dict or None, optional (default=None)
2544
            These parameters will be passed to Dataset constructor.
2545
2546
2547
2548
2549

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2550
        """
wxchan's avatar
wxchan committed
2551
2552
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
2553
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
2554
2555
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2556
        ret._predictor = self._predictor
2557
        ret.pandas_categorical = self.pandas_categorical
2558
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2559
2560
        return ret

2561
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2562
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2563

2564
2565
2566
2567
2568
        .. 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
2569
2570
        Parameters
        ----------
2571
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2572
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2573
2574
2575
2576
2577

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2578
2579
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
2580
            self.construct()._handle,
2581
            _c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
2582
        return self
wxchan's avatar
wxchan committed
2583

2584
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2585
2586
        if not params:
            return self
2587
        params = deepcopy(params)
2588
2589
2590
2591
2592

        def update():
            if not self.params:
                self.params = params
            else:
2593
                self._params_back_up = deepcopy(self.params)
2594
2595
                self.params.update(params)

2596
        if self._handle is None:
2597
2598
2599
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2600
2601
                _c_str(_param_dict_to_str(self.params)),
                _c_str(_param_dict_to_str(params)))
2602
2603
2604
2605
2606
2607
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2608
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
2609
        return self
wxchan's avatar
wxchan committed
2610

2611
    def _reverse_update_params(self) -> "Dataset":
2612
        if self._handle is None:
2613
2614
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2615
        return self
2616

2617
2618
2619
    def set_field(
        self,
        field_name: str,
2620
        data: Optional[Union[List[List[float]], List[List[int]], List[float], List[int], np.ndarray, pd_Series, pd_DataFrame, pa_Table, pa_Array, pa_ChunkedArray]]
2621
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2622
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2623
2624
2625

        Parameters
        ----------
2626
        field_name : str
2627
            The field name of the information.
2628
        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
2629
            The data to be set.
Nikita Titov's avatar
Nikita Titov committed
2630
2631
2632
2633
2634

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2635
        """
2636
        if self._handle is None:
2637
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2638
        if data is None:
2639
            # set to None
wxchan's avatar
wxchan committed
2640
            _safe_call(_LIB.LGBM_DatasetSetField(
2641
                self._handle,
2642
                _c_str(field_name),
wxchan's avatar
wxchan committed
2643
                None,
Guolin Ke's avatar
Guolin Ke committed
2644
                ctypes.c_int(0),
2645
                ctypes.c_int(_FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2646
            return self
2647
2648

        # If the data is a arrow data, we can just pass it to C
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
        if _is_pyarrow_array(data) or _is_pyarrow_table(data):
            # If a table is being passed, we concatenate the columns. This is only valid for
            # 'init_score'.
            if _is_pyarrow_table(data):
                if field_name != "init_score":
                    raise ValueError(f"pyarrow tables are not supported for field '{field_name}'")
                data = pa_chunked_array([
                    chunk for array in data.columns for chunk in array.chunks  # type: ignore
                ])

2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
            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

2670
        dtype: "np.typing.DTypeLike"
2671
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2672
            dtype = np.float64
2673
            if _is_1d_collection(data):
2674
                data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2675
            elif _is_2d_collection(data):
2676
                data = _data_to_2d_numpy(data, dtype=dtype, name=field_name)
2677
2678
2679
2680
2681
2682
2683
                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:
2684
2685
2686
2687
            if field_name in {'group', 'position'}:
                dtype = np.int32
            else:
                dtype = np.float32
2688
            data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2689

2690
        ptr_data: Union[_ctypes_float_ptr, _ctypes_int_ptr]
2691
        if data.dtype == np.float32 or data.dtype == np.float64:
2692
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2693
        elif data.dtype == np.int32:
2694
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2695
        else:
2696
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2697
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2698
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2699
        _safe_call(_LIB.LGBM_DatasetSetField(
2700
            self._handle,
2701
            _c_str(field_name),
wxchan's avatar
wxchan committed
2702
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2703
2704
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2705
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2706
        return self
wxchan's avatar
wxchan committed
2707

2708
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2709
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2710

2711
2712
2713
2714
2715
2716
        Can only be run on a constructed Dataset.

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

wxchan's avatar
wxchan committed
2717
2718
        Parameters
        ----------
2719
        field_name : str
2720
            The field name of the information.
wxchan's avatar
wxchan committed
2721
2722
2723

        Returns
        -------
2724
        info : numpy array or None
2725
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2726
        """
2727
        if self._handle is None:
2728
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2729
2730
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2731
2732
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
2733
            self._handle,
2734
            _c_str(field_name),
wxchan's avatar
wxchan committed
2735
2736
2737
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
2738
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2739
2740
2741
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2742
        if out_type.value == _C_API_DTYPE_INT32:
2743
2744
2745
2746
            arr = _cint32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)),
                length=tmp_out_len.value
            )
2747
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2748
2749
2750
2751
            arr = _cfloat32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)),
                length=tmp_out_len.value
            )
2752
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2753
2754
2755
2756
            arr = _cfloat64_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)),
                length=tmp_out_len.value
            )
2757
        else:
wxchan's avatar
wxchan committed
2758
            raise TypeError("Unknown type")
2759
2760
2761
2762
2763
2764
        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
2765

2766
2767
    def set_categorical_feature(
        self,
2768
        categorical_feature: _LGBM_CategoricalFeatureConfiguration
2769
    ) -> "Dataset":
2770
        """Set categorical features.
2771
2772
2773

        Parameters
        ----------
2774
        categorical_feature : list of str or int, or 'auto'
2775
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2776
2777
2778
2779
2780

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2781
2782
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2783
            return self
2784
        if self.data is not None:
2785
2786
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2787
                return self._free_handle()
2788
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2789
                return self
2790
            else:
2791
2792
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
2793
                                 f'New categorical_feature is {list(categorical_feature)}')
2794
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2795
                return self._free_handle()
2796
        else:
2797
2798
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2799

2800
2801
2802
2803
    def _set_predictor(
        self,
        predictor: Optional[_InnerPredictor]
    ) -> "Dataset":
2804
2805
2806
2807
        """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
2808
        """
2809
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2810
            return self
2811
2812
2813
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
            if (predictor == self._predictor) and (predictor.current_iteration() == self._predictor.current_iteration()):
                return self
2814
        if self._handle is None:
Guolin Ke's avatar
Guolin Ke committed
2815
            self._predictor = predictor
2816
2817
        elif self.data is not None:
            self._predictor = predictor
2818
2819
2820
2821
2822
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
                used_indices=None
            )
2823
2824
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2825
2826
2827
2828
2829
            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
2830
        else:
2831
2832
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2833
        return self
Guolin Ke's avatar
Guolin Ke committed
2834

2835
    def set_reference(self, reference: "Dataset") -> "Dataset":
2836
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2837
2838
2839
2840

        Parameters
        ----------
        reference : Dataset
2841
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2842
2843
2844
2845
2846

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2847
        """
2848
2849
2850
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2851
        # we're done if self and reference share a common upstream reference
2852
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2853
            return self
Guolin Ke's avatar
Guolin Ke committed
2854
2855
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2856
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2857
        else:
2858
2859
            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
2860

2861
    def set_feature_name(self, feature_name: _LGBM_FeatureNameConfiguration) -> "Dataset":
2862
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2863
2864
2865

        Parameters
        ----------
2866
        feature_name : list of str
2867
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2868
2869
2870
2871
2872

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2873
        """
2874
2875
        if feature_name != 'auto':
            self.feature_name = feature_name
2876
        if self._handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2877
            if len(feature_name) != self.num_feature():
2878
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2879
            c_feature_name = [_c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2880
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
2881
                self._handle,
2882
                _c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2883
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2884
        return self
Guolin Ke's avatar
Guolin Ke committed
2885

2886
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
2887
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2888
2889
2890

        Parameters
        ----------
2891
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None
2892
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2893
2894
2895
2896
2897

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2898
2899
        """
        self.label = label
2900
        if self._handle is not None:
2901
2902
2903
            if isinstance(label, pd_DataFrame):
                if len(label.columns) > 1:
                    raise ValueError('DataFrame for label cannot have multiple columns')
2904
                label_array = np.ravel(_pandas_to_numpy(label, target_dtype=np.float32))
2905
2906
            elif _is_pyarrow_array(label):
                label_array = label
2907
            else:
2908
                label_array = _list_to_1d_numpy(label, dtype=np.float32, name='label')
2909
            self.set_field('label', label_array)
2910
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2911
        return self
Guolin Ke's avatar
Guolin Ke committed
2912

2913
2914
2915
2916
    def set_weight(
        self,
        weight: Optional[_LGBM_WeightType]
    ) -> "Dataset":
2917
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2918
2919
2920

        Parameters
        ----------
2921
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None
2922
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2923
2924
2925
2926
2927

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2928
        """
2929
2930
2931
2932
2933
2934
2935
        # Check if the weight contains values other than one
        if weight is not None:
            if _is_pyarrow_array(weight):
                if pa_compute.all(pa_compute.equal(weight, 1)).as_py():
                    weight = None
            elif np.all(weight == 1):
                weight = None
Guolin Ke's avatar
Guolin Ke committed
2936
        self.weight = weight
2937
2938

        # Set field
2939
        if self._handle is not None and weight is not None:
2940
2941
            if not _is_pyarrow_array(weight):
                weight = _list_to_1d_numpy(weight, dtype=np.float32, name='weight')
wxchan's avatar
wxchan committed
2942
            self.set_field('weight', weight)
2943
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2944
        return self
Guolin Ke's avatar
Guolin Ke committed
2945

2946
2947
2948
2949
    def set_init_score(
        self,
        init_score: Optional[_LGBM_InitScoreType]
    ) -> "Dataset":
2950
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2951
2952
2953

        Parameters
        ----------
2954
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), pyarrow Array, pyarrow ChunkedArray, pyarrow Table (for multi-class task) or None
2955
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2956
2957
2958
2959
2960

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2961
2962
        """
        self.init_score = init_score
2963
        if self._handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2964
            self.set_field('init_score', init_score)
2965
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2966
        return self
Guolin Ke's avatar
Guolin Ke committed
2967

2968
2969
2970
2971
    def set_group(
        self,
        group: Optional[_LGBM_GroupType]
    ) -> "Dataset":
2972
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2973
2974
2975

        Parameters
        ----------
2976
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None
2977
2978
2979
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2980
2981
            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
2982
2983
2984
2985
2986

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2987
2988
        """
        self.group = group
2989
        if self._handle is not None and group is not None:
2990
2991
            if not _is_pyarrow_array(group):
                group = _list_to_1d_numpy(group, dtype=np.int32, name='group')
wxchan's avatar
wxchan committed
2992
            self.set_field('group', group)
2993
2994
2995
2996
            # original values can be modified at cpp side
            constructed_group = self.get_field('group')
            if constructed_group is not None:
                self.group = np.diff(constructed_group)
Nikita Titov's avatar
Nikita Titov committed
2997
        return self
Guolin Ke's avatar
Guolin Ke committed
2998

2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
    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

3021
    def get_feature_name(self) -> List[str]:
3022
3023
3024
3025
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
3026
        feature_names : list of str
3027
3028
            The names of columns (features) in the Dataset.
        """
3029
        if self._handle is None:
3030
3031
3032
3033
3034
            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)
3035
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
3036
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
3037
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
3038
            self._handle,
3039
            ctypes.c_int(num_feature),
3040
            ctypes.byref(tmp_out_len),
3041
            ctypes.c_size_t(reserved_string_buffer_size),
3042
3043
3044
3045
            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")
3046
3047
3048
3049
        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)]
3050
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
3051
            _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
3052
                self._handle,
3053
3054
3055
3056
3057
                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))
3058
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
3059

3060
    def get_label(self) -> Optional[_LGBM_LabelType]:
3061
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3062
3063
3064

        Returns
        -------
3065
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
3066
            The label information from the Dataset.
3067
            For a constructed ``Dataset``, this will only return a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3068
        """
3069
        if self.label is None:
wxchan's avatar
wxchan committed
3070
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
3071
3072
        return self.label

3073
    def get_weight(self) -> Optional[_LGBM_WeightType]:
3074
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3075
3076
3077

        Returns
        -------
3078
        weight : list, numpy 1-D array, pandas Series or None
3079
            Weight for each data point from the Dataset. Weights should be non-negative.
3080
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3081
        """
3082
        if self.weight is None:
wxchan's avatar
wxchan committed
3083
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
3084
3085
        return self.weight

3086
    def get_init_score(self) -> Optional[_LGBM_InitScoreType]:
3087
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3088
3089
3090

        Returns
        -------
3091
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
3092
            Init score of Booster.
3093
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3094
        """
3095
        if self.init_score is None:
wxchan's avatar
wxchan committed
3096
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
3097
3098
        return self.init_score

3099
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
3100
3101
3102
3103
        """Get the raw data of the Dataset.

        Returns
        -------
3104
        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
3105
3106
            Raw data used in the Dataset construction.
        """
3107
        if self._handle is None:
3108
            raise Exception("Cannot get data before construct Dataset")
3109
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
3110
3111
            self.data = self.reference.data
            if self.data is not None:
3112
                if isinstance(self.data, (np.ndarray, scipy.sparse.spmatrix)):
Guolin Ke's avatar
Guolin Ke committed
3113
                    self.data = self.data[self.used_indices, :]
3114
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3115
                    self.data = self.data.iloc[self.used_indices].copy()
3116
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3117
                    self.data = self.data[self.used_indices, :]
3118
3119
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
3120
                elif _is_list_of_sequences(self.data) and len(self.data) > 0:
3121
                    self.data = np.array(list(self._yield_row_from_seqlist(self.data, self.used_indices)))
Guolin Ke's avatar
Guolin Ke committed
3122
                else:
3123
3124
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
3125
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
3126
3127
3128
        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.")
3129
3130
        return self.data

3131
    def get_group(self) -> Optional[_LGBM_GroupType]:
3132
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3133
3134
3135

        Returns
        -------
3136
        group : list, numpy 1-D array, pandas Series or None
3137
3138
3139
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
3140
3141
            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.
3142
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3143
        """
3144
        if self.group is None:
wxchan's avatar
wxchan committed
3145
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
3146
3147
            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
3148
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
3149
3150
        return self.group

3151
    def get_position(self) -> Optional[_LGBM_PositionType]:
3152
3153
3154
3155
        """Get the position of the Dataset.

        Returns
        -------
3156
        position : numpy 1-D array, pandas Series or None
3157
            Position of items used in unbiased learning-to-rank task.
3158
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
3159
3160
3161
3162
3163
        """
        if self.position is None:
            self.position = self.get_field('position')
        return self.position

3164
    def num_data(self) -> int:
3165
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3166
3167
3168

        Returns
        -------
3169
3170
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3171
        """
3172
        if self._handle is not None:
3173
            ret = ctypes.c_int(0)
3174
            _safe_call(_LIB.LGBM_DatasetGetNumData(self._handle,
wxchan's avatar
wxchan committed
3175
3176
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3177
        else:
3178
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3179

3180
    def num_feature(self) -> int:
3181
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3182
3183
3184

        Returns
        -------
3185
3186
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3187
        """
3188
        if self._handle is not None:
3189
            ret = ctypes.c_int(0)
3190
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self._handle,
wxchan's avatar
wxchan committed
3191
3192
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3193
        else:
3194
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3195

3196
    def feature_num_bin(self, feature: Union[int, str]) -> int:
3197
3198
        """Get the number of bins for a feature.

3199
3200
        .. versionadded:: 4.0.0

3201
3202
        Parameters
        ----------
3203
3204
        feature : int or str
            Index or name of the feature.
3205
3206
3207
3208
3209
3210

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
3211
        if self._handle is not None:
3212
            if isinstance(feature, str):
3213
3214
3215
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
3216
            ret = ctypes.c_int(0)
3217
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self._handle,
3218
                                                         ctypes.c_int(feature_index),
3219
3220
3221
3222
3223
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

3224
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
3225
3226
3227
3228
3229
        """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.
3230
3231
3232
3233
3234

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
3235
3236
3237

        Returns
        -------
3238
3239
3240
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
3241
        head = self
3242
        ref_chain: Set[Dataset] = set()
3243
3244
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
3245
                ref_chain.add(head)
3246
3247
3248
3249
3250
3251
                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
3252
        return ref_chain
3253

3254
    def add_features_from(self, other: "Dataset") -> "Dataset":
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
        """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.
        """
3269
        if self._handle is None or other._handle is None:
3270
            raise ValueError('Both source and target Datasets must be constructed before adding features')
3271
        _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self._handle, other._handle))
Guolin Ke's avatar
Guolin Ke committed
3272
3273
3274
3275
3276
3277
3278
3279
        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))
3280
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3281
                    self.data = np.hstack((self.data, other.data.toarray()))
3282
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3283
                    self.data = np.hstack((self.data, other.data.values))
3284
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3285
3286
3287
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
3288
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3289
                sparse_format = self.data.getformat()
3290
                if isinstance(other.data, (np.ndarray, scipy.sparse.spmatrix)):
Guolin Ke's avatar
Guolin Ke committed
3291
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
3292
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3293
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
3294
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3295
3296
3297
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
3298
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3299
3300
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
3301
3302
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
3303
                if isinstance(other.data, np.ndarray):
3304
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
3305
                                       axis=1, ignore_index=True)
3306
                elif isinstance(other.data, scipy.sparse.spmatrix):
3307
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
3308
                                       axis=1, ignore_index=True)
3309
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3310
3311
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
3312
3313
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
3314
3315
3316
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
3317
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3318
                if isinstance(other.data, np.ndarray):
3319
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
3320
                elif isinstance(other.data, scipy.sparse.spmatrix):
3321
3322
3323
3324
3325
                    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
3326
3327
3328
3329
3330
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
3331
3332
            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
3333
3334
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
3335
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
3336
        self.feature_name = self.get_feature_name()
3337
3338
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
3339
3340
        self.categorical_feature = "auto"
        self.pandas_categorical = None
3341
3342
        return self

3343
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
3344
3345
3346
3347
3348
3349
        """Save Dataset to a text file.

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

        Parameters
        ----------
3350
        filename : str or pathlib.Path
3351
3352
3353
3354
3355
3356
3357
3358
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
3359
            self.construct()._handle,
3360
            _c_str(str(filename))))
3361
3362
        return self

wxchan's avatar
wxchan committed
3363

3364
3365
3366
3367
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
    Tuple[np.ndarray, np.ndarray]
]
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
        _LGBM_EvalFunctionResultType
    ],
    Callable[
        [np.ndarray, Dataset],
        List[_LGBM_EvalFunctionResultType]
    ]
]
3378
3379


3380
class Booster:
3381
    """Booster in LightGBM."""
3382

3383
3384
3385
3386
3387
3388
3389
    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
    ):
3390
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3391
3392
3393

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3394
        params : dict or None, optional (default=None)
3395
3396
3397
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3398
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3399
            Path to the model file.
3400
        model_str : str or None, optional (default=None)
3401
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3402
        """
3403
        self._handle = ctypes.c_void_p()
3404
        self._network = False
wxchan's avatar
wxchan committed
3405
        self.__need_reload_eval_info = True
3406
        self._train_data_name = "training"
3407
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3408
        self.best_iteration = -1
3409
        self.best_score: _LGBM_BoosterBestScoreType = {}
3410
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3411
        if train_set is not None:
3412
            # Training task
wxchan's avatar
wxchan committed
3413
            if not isinstance(train_set, Dataset):
3414
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
            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"]
                )
3449
            # construct booster object
3450
3451
3452
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3453
            params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3454
            _safe_call(_LIB.LGBM_BoosterCreate(
3455
                train_set._handle,
3456
                _c_str(params_str),
3457
                ctypes.byref(self._handle)))
3458
            # save reference to data
wxchan's avatar
wxchan committed
3459
            self.train_set = train_set
3460
3461
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3462
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3463
3464
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
3465
                _safe_call(_LIB.LGBM_BoosterMerge(
3466
3467
                    self._handle,
                    self.__init_predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
3468
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3469
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3470
                self._handle,
wxchan's avatar
wxchan committed
3471
3472
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3473
            # buffer for inner predict
3474
            self.__inner_predict_buffer: List[Optional[np.ndarray]] = [None]
wxchan's avatar
wxchan committed
3475
3476
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3477
            self.pandas_categorical = train_set.pandas_categorical
3478
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3479
        elif model_file is not None:
3480
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3481
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3482
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
3483
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
3484
                ctypes.byref(out_num_iterations),
3485
                ctypes.byref(self._handle)))
Guolin Ke's avatar
Guolin Ke committed
3486
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3487
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
3488
                self._handle,
wxchan's avatar
wxchan committed
3489
3490
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
3491
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3492
3493
3494
            if params:
                _log_warning('Ignoring params argument, using parameters from model file.')
            params = self._get_loaded_param()
3495
        elif model_str is not None:
3496
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3497
        else:
3498
3499
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
3500
        self.params = params
wxchan's avatar
wxchan committed
3501

3502
    def __del__(self) -> None:
3503
        try:
3504
            if self._network:
3505
3506
3507
3508
                self.free_network()
        except AttributeError:
            pass
        try:
3509
3510
            if self._handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
3511
3512
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3513

3514
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3515
3516
        return self.__deepcopy__(None)

3517
    def __deepcopy__(self, _) -> "Booster":
3518
        model_str = self.model_to_string(num_iteration=-1)
3519
        return Booster(model_str=model_str)
wxchan's avatar
wxchan committed
3520

3521
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3522
        this = self.__dict__.copy()
3523
        handle = this['_handle']
wxchan's avatar
wxchan committed
3524
3525
3526
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
3527
            this["_handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3528
3529
        return this

3530
    def __setstate__(self, state: Dict[str, Any]) -> None:
3531
        model_str = state.get('_handle', state.get('handle', None))
3532
        if model_str is not None:
wxchan's avatar
wxchan committed
3533
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3534
            out_num_iterations = ctypes.c_int(0)
3535
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3536
                _c_str(model_str),
3537
3538
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
3539
            state['_handle'] = handle
wxchan's avatar
wxchan committed
3540
3541
        self.__dict__.update(state)

3542
3543
3544
3545
    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)
3546
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3547
        _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
3548
            self._handle,
3549
3550
3551
3552
3553
3554
3555
            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)
3556
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3557
            _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
3558
                self._handle,
3559
3560
3561
3562
3563
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
        return json.loads(string_buffer.value.decode('utf-8'))

3564
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3565
3566
3567
3568
3569
3570
3571
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
3572
3573
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
3574
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3575
        return self
wxchan's avatar
wxchan committed
3576

3577
    def _free_buffer(self) -> "Booster":
3578
3579
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3580
        return self
3581

3582
3583
3584
3585
3586
3587
3588
    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":
3589
3590
3591
3592
        """Set the network configuration.

        Parameters
        ----------
3593
        machines : list, set or str
3594
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3595
        local_listen_port : int, optional (default=12400)
3596
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3597
        listen_time_out : int, optional (default=120)
3598
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3599
        num_machines : int, optional (default=1)
3600
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3601
3602
3603
3604
3605

        Returns
        -------
        self : Booster
            Booster with set network.
3606
        """
3607
3608
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
3609
        _safe_call(_LIB.LGBM_NetworkInit(_c_str(machines),
3610
3611
3612
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
3613
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3614
        return self
3615

3616
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3617
3618
3619
3620
3621
3622
3623
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3624
        _safe_call(_LIB.LGBM_NetworkFree())
3625
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3626
        return self
3627

3628
    def trees_to_dataframe(self) -> pd_DataFrame:
3629
3630
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3631
3632
3633
3634
        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.
3635
3636
3637
3638
3639
            - ``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.
3640
3641
            - ``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.
3642
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3643
3644
              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.
3645
3646
            - ``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.
3647
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3648
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3649
3650
            - ``count`` : int64, number of records in the training data that fall into this node.

3651
3652
3653
3654
3655
3656
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3657
3658
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
3659
3660
3661
3662

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

3663
        def _is_split_node(tree: Dict[str, Any]) -> bool:
3664
3665
            return 'split_index' in tree.keys()

3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
        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:
3678
                tree_num = f'{tree_index}-' if tree_index is not None else ''
3679
3680
3681
                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
3682
3683
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
3684

3685
3686
3687
3688
            def _get_split_feature(
                tree: Dict[str, Any],
                feature_names: Optional[List[str]]
            ) -> Optional[str]:
3689
3690
3691
3692
3693
3694
3695
3696
3697
                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

3698
            def _is_single_node_tree(tree: Dict[str, Any]) -> bool:
3699
                return set(tree.keys()) == {'leaf_value'}
3700
3701

            # Create the node record, and populate universal data members
3702
            node: Dict[str, Union[int, str, None]] = OrderedDict()
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
            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

3739
3740
3741
3742
3743
3744
3745
        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]]:
3746

3747
            node = create_node_record(tree=tree,
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
                                      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(
3760
                        tree=tree[child],
3761
3762
3763
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
3764
3765
                        parent_node=node['node_index']
                    )
3766
3767
3768
3769
3770
3771
3772
3773
3774
                    # 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']:
3775
            model_list.extend(tree_dict_to_node_list(tree=tree['tree_structure'],
3776
3777
3778
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

3779
        return pd_DataFrame(model_list, columns=model_list[0].keys())
3780

3781
    def set_train_data_name(self, name: str) -> "Booster":
3782
3783
3784
3785
        """Set the name to the training Dataset.

        Parameters
        ----------
3786
        name : str
Nikita Titov's avatar
Nikita Titov committed
3787
3788
3789
3790
3791
3792
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
3793
        """
3794
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
3795
        return self
wxchan's avatar
wxchan committed
3796

3797
    def add_valid(self, data: Dataset, name: str) -> "Booster":
3798
        """Add validation data.
wxchan's avatar
wxchan committed
3799
3800
3801
3802

        Parameters
        ----------
        data : Dataset
3803
            Validation data.
3804
        name : str
3805
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
3806
3807
3808
3809
3810

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
3811
        """
Guolin Ke's avatar
Guolin Ke committed
3812
        if not isinstance(data, Dataset):
3813
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3814
        if data._predictor is not self.__init_predictor:
3815
3816
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3817
        _safe_call(_LIB.LGBM_BoosterAddValidData(
3818
3819
            self._handle,
            data.construct()._handle))
wxchan's avatar
wxchan committed
3820
3821
3822
3823
3824
        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
3825
        return self
wxchan's avatar
wxchan committed
3826

3827
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
3828
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
3829
3830
3831
3832

        Parameters
        ----------
        params : dict
3833
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
3834
3835
3836
3837
3838

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
3839
        """
3840
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3841
3842
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
3843
                self._handle,
3844
                _c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
3845
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
3846
        return self
wxchan's avatar
wxchan committed
3847

3848
3849
3850
3851
3852
    def update(
        self,
        train_set: Optional[Dataset] = None,
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
3853
        """Update Booster for one iteration.
3854

wxchan's avatar
wxchan committed
3855
3856
        Parameters
        ----------
3857
3858
3859
3860
        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
3861
            Customized objective function.
3862
3863
3864
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3865
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3866
                    The predicted values.
3867
3868
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3869
3870
                train_data : Dataset
                    The training dataset.
3871
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3872
3873
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3874
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3875
3876
                    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
3877

3878
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3879
            and grad and hess should be returned in the same format.
3880

wxchan's avatar
wxchan committed
3881
3882
        Returns
        -------
3883
3884
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3885
        """
3886
        # need reset training data
3887
3888
3889
3890
3891
3892
        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
3893
            if not isinstance(train_set, Dataset):
3894
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3895
            if train_set._predictor is not self.__init_predictor:
3896
3897
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3898
3899
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
3900
3901
                self._handle,
                self.train_set.construct()._handle))
wxchan's avatar
wxchan committed
3902
            self.__inner_predict_buffer[0] = None
3903
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3904
3905
        is_finished = ctypes.c_int(0)
        if fobj is None:
3906
            if self.__set_objective_to_none:
3907
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3908
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
3909
                self._handle,
wxchan's avatar
wxchan committed
3910
                ctypes.byref(is_finished)))
3911
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3912
3913
            return is_finished.value == 1
        else:
3914
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3915
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3916
3917
3918
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

3919
3920
3921
3922
3923
    def __boost(
        self,
        grad: np.ndarray,
        hess: np.ndarray
    ) -> bool:
3924
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3925

Nikita Titov's avatar
Nikita Titov committed
3926
3927
        .. note::

3928
3929
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3930
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3931
            and grad and hess should be returned in the same format.
3932

wxchan's avatar
wxchan committed
3933
3934
        Parameters
        ----------
3935
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3936
3937
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3938
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3939
3940
            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
3941
3942
3943

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3944
3945
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3946
        """
3947
3948
3949
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3950
3951
        grad = _list_to_1d_numpy(grad, dtype=np.float32, name='gradient')
        hess = _list_to_1d_numpy(hess, dtype=np.float32, name='hessian')
3952
3953
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3954
        if len(grad) != len(hess):
3955
3956
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3957
        if len(grad) != num_train_data * self.__num_class:
3958
3959
3960
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3961
                f"number of models per one iteration ({self.__num_class})"
3962
            )
wxchan's avatar
wxchan committed
3963
3964
        is_finished = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterUpdateOneIterCustom(
3965
            self._handle,
wxchan's avatar
wxchan committed
3966
3967
3968
            grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            ctypes.byref(is_finished)))
3969
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3970
3971
        return is_finished.value == 1

3972
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3973
3974
3975
3976
3977
3978
3979
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3980
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
3981
            self._handle))
3982
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3983
        return self
wxchan's avatar
wxchan committed
3984

3985
    def current_iteration(self) -> int:
3986
3987
3988
3989
3990
3991
3992
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3993
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3994
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
3995
            self._handle,
wxchan's avatar
wxchan committed
3996
3997
3998
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3999
    def num_model_per_iteration(self) -> int:
4000
4001
4002
4003
4004
4005
4006
4007
4008
        """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(
4009
            self._handle,
4010
4011
4012
            ctypes.byref(model_per_iter)))
        return model_per_iter.value

4013
    def num_trees(self) -> int:
4014
4015
4016
4017
4018
4019
4020
4021
4022
        """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(
4023
            self._handle,
4024
4025
4026
            ctypes.byref(num_trees)))
        return num_trees.value

4027
    def upper_bound(self) -> float:
4028
4029
4030
4031
        """Get upper bound value of a model.

        Returns
        -------
4032
        upper_bound : float
4033
4034
4035
4036
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
4037
            self._handle,
4038
4039
4040
            ctypes.byref(ret)))
        return ret.value

4041
    def lower_bound(self) -> float:
4042
4043
4044
4045
        """Get lower bound value of a model.

        Returns
        -------
4046
        lower_bound : float
4047
4048
4049
4050
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
4051
            self._handle,
4052
4053
4054
            ctypes.byref(ret)))
        return ret.value

4055
4056
4057
4058
4059
4060
    def eval(
        self,
        data: Dataset,
        name: str,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4061
        """Evaluate for data.
wxchan's avatar
wxchan committed
4062
4063
4064

        Parameters
        ----------
4065
4066
        data : Dataset
            Data for the evaluating.
4067
        name : str
4068
            Name of the data.
4069
        feval : callable, list of callable, or None, optional (default=None)
4070
            Customized evaluation function.
4071
            Each evaluation function should accept two parameters: preds, eval_data,
4072
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4073

4074
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4075
                    The predicted values.
4076
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4077
                    If custom objective function is used, predicted values are returned before any transformation,
4078
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
4079
                eval_data : Dataset
4080
                    A ``Dataset`` to evaluate.
4081
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4082
                    The name of evaluation function (without whitespace).
4083
4084
4085
4086
4087
                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
4088
4089
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4090
        result : list
4091
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4092
        """
Guolin Ke's avatar
Guolin Ke committed
4093
4094
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
4095
4096
4097
4098
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
4099
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
4100
4101
4102
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
4103
        # need to push new valid data
wxchan's avatar
wxchan committed
4104
4105
4106
4107
4108
4109
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

4110
4111
4112
4113
    def eval_train(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4114
        """Evaluate for training data.
wxchan's avatar
wxchan committed
4115
4116
4117

        Parameters
        ----------
4118
        feval : callable, list of callable, or None, optional (default=None)
4119
            Customized evaluation function.
4120
            Each evaluation function should accept two parameters: preds, eval_data,
4121
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4122

4123
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4124
                    The predicted values.
4125
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4126
                    If custom objective function is used, predicted values are returned before any transformation,
4127
                    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
4128
                eval_data : Dataset
4129
                    The training dataset.
4130
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4131
                    The name of evaluation function (without whitespace).
4132
4133
4134
4135
4136
                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
4137
4138
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4139
        result : list
4140
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4141
        """
4142
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
4143

4144
4145
4146
4147
    def eval_valid(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4148
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
4149
4150
4151

        Parameters
        ----------
4152
        feval : callable, list of callable, or None, optional (default=None)
4153
            Customized evaluation function.
4154
            Each evaluation function should accept two parameters: preds, eval_data,
4155
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4156

4157
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4158
                    The predicted values.
4159
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4160
                    If custom objective function is used, predicted values are returned before any transformation,
4161
                    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
4162
                eval_data : Dataset
4163
                    The validation dataset.
4164
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4165
                    The name of evaluation function (without whitespace).
4166
4167
4168
4169
4170
                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
4171
4172
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4173
        result : list
4174
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4175
        """
4176
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
4177
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
4178

4179
4180
4181
4182
4183
4184
4185
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> "Booster":
4186
        """Save Booster to file.
wxchan's avatar
wxchan committed
4187
4188
4189

        Parameters
        ----------
4190
        filename : str or pathlib.Path
4191
            Filename to save Booster.
4192
4193
4194
4195
        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
4196
        start_iteration : int, optional (default=0)
4197
            Start index of the iteration that should be saved.
4198
        importance_type : str, optional (default="split")
4199
4200
4201
            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
4202
4203
4204
4205
4206

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
4207
        """
4208
        if num_iteration is None:
4209
            num_iteration = self.best_iteration
4210
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4211
        _safe_call(_LIB.LGBM_BoosterSaveModel(
4212
            self._handle,
4213
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4214
            ctypes.c_int(num_iteration),
4215
            ctypes.c_int(importance_type_int),
4216
            _c_str(str(filename))))
4217
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
4218
        return self
wxchan's avatar
wxchan committed
4219

4220
4221
4222
4223
4224
    def shuffle_models(
        self,
        start_iteration: int = 0,
        end_iteration: int = -1
    ) -> "Booster":
4225
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
4226

4227
4228
4229
        Parameters
        ----------
        start_iteration : int, optional (default=0)
4230
            The first iteration that will be shuffled.
4231
4232
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
4233
            If <= 0, means the last available iteration.
4234

Nikita Titov's avatar
Nikita Titov committed
4235
4236
4237
4238
        Returns
        -------
        self : Booster
            Booster with shuffled models.
4239
        """
4240
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
4241
            self._handle,
Guolin Ke's avatar
Guolin Ke committed
4242
4243
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
4244
        return self
4245

4246
    def model_from_string(self, model_str: str) -> "Booster":
4247
4248
4249
4250
        """Load Booster from a string.

        Parameters
        ----------
4251
        model_str : str
4252
4253
4254
4255
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4256
        self : Booster
4257
4258
            Loaded Booster object.
        """
4259
4260
4261
        # ensure that existing Booster is freed before replacing it
        # with a new one createdfrom file
        _safe_call(_LIB.LGBM_BoosterFree(self._handle))
4262
        self._free_buffer()
4263
        self._handle = ctypes.c_void_p()
4264
4265
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
4266
            _c_str(model_str),
4267
            ctypes.byref(out_num_iterations),
4268
            ctypes.byref(self._handle)))
4269
4270
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
4271
            self._handle,
4272
4273
            ctypes.byref(out_num_class)))
        self.__num_class = out_num_class.value
4274
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
4275
4276
        return self

4277
4278
4279
4280
4281
4282
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> str:
4283
        """Save Booster to string.
4284

4285
4286
4287
4288
4289
4290
        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
4291
        start_iteration : int, optional (default=0)
4292
            Start index of the iteration that should be saved.
4293
        importance_type : str, optional (default="split")
4294
4295
4296
            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.
4297
4298
4299

        Returns
        -------
4300
        str_repr : str
4301
4302
            String representation of Booster.
        """
4303
        if num_iteration is None:
4304
            num_iteration = self.best_iteration
4305
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4306
        buffer_len = 1 << 20
4307
        tmp_out_len = ctypes.c_int64(0)
4308
        string_buffer = ctypes.create_string_buffer(buffer_len)
4309
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4310
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
4311
            self._handle,
4312
            ctypes.c_int(start_iteration),
4313
            ctypes.c_int(num_iteration),
4314
            ctypes.c_int(importance_type_int),
4315
            ctypes.c_int64(buffer_len),
4316
4317
4318
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
4319
        # if buffer length is not long enough, re-allocate a buffer
4320
4321
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
4322
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4323
            _safe_call(_LIB.LGBM_BoosterSaveModelToString(
4324
                self._handle,
4325
                ctypes.c_int(start_iteration),
4326
                ctypes.c_int(num_iteration),
4327
                ctypes.c_int(importance_type_int),
4328
                ctypes.c_int64(actual_len),
4329
4330
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
4331
        ret = string_buffer.value.decode('utf-8')
4332
4333
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
4334

4335
4336
4337
4338
4339
4340
4341
    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
4342
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
4343

4344
4345
        Parameters
        ----------
4346
4347
4348
4349
        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
4350
        start_iteration : int, optional (default=0)
4351
            Start index of the iteration that should be dumped.
4352
        importance_type : str, optional (default="split")
4353
4354
4355
            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.
4356
4357
4358
4359
4360
4361
4362
4363
4364
        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.
4365

wxchan's avatar
wxchan committed
4366
4367
        Returns
        -------
4368
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
4369
            JSON format of Booster.
wxchan's avatar
wxchan committed
4370
        """
4371
        if num_iteration is None:
4372
            num_iteration = self.best_iteration
4373
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4374
        buffer_len = 1 << 20
4375
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
4376
        string_buffer = ctypes.create_string_buffer(buffer_len)
4377
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
wxchan's avatar
wxchan committed
4378
        _safe_call(_LIB.LGBM_BoosterDumpModel(
4379
            self._handle,
4380
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4381
            ctypes.c_int(num_iteration),
4382
            ctypes.c_int(importance_type_int),
4383
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
4384
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4385
            ptr_string_buffer))
wxchan's avatar
wxchan committed
4386
        actual_len = tmp_out_len.value
4387
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
4388
4389
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
4390
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
wxchan's avatar
wxchan committed
4391
            _safe_call(_LIB.LGBM_BoosterDumpModel(
4392
                self._handle,
4393
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4394
                ctypes.c_int(num_iteration),
4395
                ctypes.c_int(importance_type_int),
4396
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
4397
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4398
                ptr_string_buffer))
4399
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
4400
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
4401
                                                          default=_json_default_with_numpy))
4402
        return ret
wxchan's avatar
wxchan committed
4403

4404
4405
    def predict(
        self,
4406
        data: _LGBM_PredictDataType,
4407
4408
4409
4410
4411
4412
4413
4414
        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
4415
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4416
        """Make a prediction.
wxchan's avatar
wxchan committed
4417
4418
4419

        Parameters
        ----------
4420
        data : str, pathlib.Path, numpy array, pandas DataFrame, pyarrow Table, H2O DataTable's Frame or scipy.sparse
4421
            Data source for prediction.
4422
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4423
        start_iteration : int, optional (default=0)
4424
            Start index of the iteration to predict.
4425
            If <= 0, starts from the first iteration.
4426
        num_iteration : int or None, optional (default=None)
4427
4428
4429
4430
            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).
4431
4432
4433
4434
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4435
4436
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4437

Nikita Titov's avatar
Nikita Titov committed
4438
4439
4440
4441
4442
4443
4444
            .. 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.
4445

4446
4447
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4448
            Used only if data is str.
4449
4450
4451
        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.
4452
4453
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4454
4455
4456

        Returns
        -------
4457
        result : numpy array, scipy.sparse or list of scipy.sparse
4458
            Prediction result.
4459
            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
4460
        """
4461
4462
4463
4464
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs),
        )
4465
        if num_iteration is None:
4466
            if start_iteration <= 0:
4467
4468
4469
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
        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
4480

4481
4482
    def refit(
        self,
4483
        data: _LGBM_TrainDataType,
4484
        label: _LGBM_LabelType,
4485
4486
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4487
4488
4489
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4490
4491
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
4492
4493
4494
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4495
        **kwargs
4496
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4497
4498
4499
4500
        """Refit the existing Booster by new data.

        Parameters
        ----------
4501
        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
4502
            Data source for refit.
4503
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4504
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array or pyarrow ChunkedArray
Guolin Ke's avatar
Guolin Ke committed
4505
4506
            Label for refit.
        decay_rate : float, optional (default=0.9)
4507
4508
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4509
4510
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
4511
4512
4513

            .. versionadded:: 4.0.0

4514
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
4515
            Weight for each ``data`` instance. Weights should be non-negative.
4516
4517
4518

            .. versionadded:: 4.0.0

4519
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
4520
4521
4522
4523
4524
            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.
4525
4526
4527

            .. versionadded:: 4.0.0

4528
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), pyarrow Array, pyarrow ChunkedArray, pyarrow Table (for multi-class task) or None, optional (default=None)
4529
            Init score for ``data``.
4530
4531
4532

            .. versionadded:: 4.0.0

4533
4534
4535
        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.
4536
4537
4538

            .. versionadded:: 4.0.0

4539
4540
4541
4542
4543
        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.
4544
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4545
4546
4547
            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.
4548
            Floating point numbers in categorical features will be rounded towards 0.
4549
4550
4551

            .. versionadded:: 4.0.0

4552
4553
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
4554
4555
4556

            .. versionadded:: 4.0.0

4557
4558
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
4559
4560
4561

            .. versionadded:: 4.0.0

4562
4563
4564
        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.
4565
4566
4567

            .. versionadded:: 4.0.0

4568
4569
        **kwargs
            Other parameters for refit.
4570
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4571
4572
4573
4574
4575
4576

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4577
4578
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
4579
4580
        if dataset_params is None:
            dataset_params = {}
4581
4582
4583
4584
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs)
        )
4585
        leaf_preds: np.ndarray = predictor.predict(  # type: ignore[assignment]
4586
4587
4588
4589
4590
            data=data,
            start_iteration=-1,
            pred_leaf=True,
            validate_features=validate_features
        )
4591
        nrow, ncol = leaf_preds.shape
4592
        out_is_linear = ctypes.c_int(0)
4593
        _safe_call(_LIB.LGBM_BoosterGetLinear(
4594
            self._handle,
4595
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
4596
4597
4598
4599
4600
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
4601
        new_params["linear_tree"] = bool(out_is_linear.value)
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
        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,
        )
4615
        new_params['refit_decay_rate'] = decay_rate
4616
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4617
4618
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
4619
4620
            new_booster._handle,
            predictor._handle))
Guolin Ke's avatar
Guolin Ke committed
4621
        leaf_preds = leaf_preds.reshape(-1)
4622
        ptr_data, _, _ = _c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
4623
        _safe_call(_LIB.LGBM_BoosterRefit(
4624
            new_booster._handle,
Guolin Ke's avatar
Guolin Ke committed
4625
            ptr_data,
4626
4627
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
4628
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4629
4630
        return new_booster

4631
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
        """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.
        """
4646
4647
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
4648
            self._handle,
4649
4650
4651
4652
4653
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

4654
4655
4656
4657
4658
4659
4660
4661
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
    ) -> 'Booster':
        """Set the output of a leaf.

4662
4663
        .. versionadded:: 4.0.0

4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
        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(
4680
                self._handle,
4681
4682
4683
4684
4685
4686
4687
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.c_double(value)
            )
        )
        return self

4688
    def num_feature(self) -> int:
4689
4690
4691
4692
4693
4694
4695
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4696
4697
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
4698
            self._handle,
4699
4700
4701
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

4702
    def feature_name(self) -> List[str]:
4703
        """Get names of features.
wxchan's avatar
wxchan committed
4704
4705
4706

        Returns
        -------
4707
        result : list of str
4708
            List with names of features.
wxchan's avatar
wxchan committed
4709
        """
4710
        num_feature = self.num_feature()
4711
        # Get name of features
wxchan's avatar
wxchan committed
4712
        tmp_out_len = ctypes.c_int(0)
4713
4714
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
4715
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
4716
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
wxchan's avatar
wxchan committed
4717
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
4718
            self._handle,
4719
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
4720
            ctypes.byref(tmp_out_len),
4721
            ctypes.c_size_t(reserved_string_buffer_size),
4722
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4723
4724
4725
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
4726
4727
4728
4729
        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)]
4730
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
4731
            _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
4732
                self._handle,
4733
4734
4735
4736
4737
                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))
4738
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
4739

4740
4741
4742
4743
4744
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
4745
        """Get feature importances.
4746

4747
4748
        Parameters
        ----------
4749
        importance_type : str, optional (default="split")
4750
4751
4752
            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.
4753
4754
4755
4756
        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).
4757

4758
4759
        Returns
        -------
4760
4761
        result : numpy array
            Array with feature importances.
4762
        """
4763
4764
        if iteration is None:
            iteration = self.best_iteration
4765
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4766
        result = np.empty(self.num_feature(), dtype=np.float64)
4767
        _safe_call(_LIB.LGBM_BoosterFeatureImportance(
4768
            self._handle,
4769
4770
4771
            ctypes.c_int(iteration),
            ctypes.c_int(importance_type_int),
            result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
4772
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
4773
            return result.astype(np.int32)
4774
4775
        else:
            return result
4776

4777
4778
4779
4780
4781
4782
    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]:
4783
4784
4785
4786
        """Get split value histogram for the specified feature.

        Parameters
        ----------
4787
        feature : int or str
4788
4789
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
4790
            If str, interpreted as name.
4791

Nikita Titov's avatar
Nikita Titov committed
4792
4793
4794
            .. warning::

                Categorical features are not supported.
4795

4796
        bins : int, str or None, optional (default=None)
4797
4798
4799
            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.
4800
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
        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.
        """
4815
        def add(root: Dict[str, Any]) -> None:
4816
4817
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
4818
                if feature_names is not None and isinstance(feature, str):
4819
4820
4821
4822
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
4823
                    if isinstance(root['threshold'], str):
4824
4825
4826
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
4827
4828
4829
4830
4831
4832
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
4833
        values: List[float] = []
4834
4835
4836
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

4837
        if bins is None or isinstance(bins, int) and xgboost_style:
4838
4839
4840
4841
4842
4843
4844
            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:
4845
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
4846
4847
4848
4849
4850
            else:
                return ret
        else:
            return hist, bin_edges

4851
4852
4853
4854
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
4855
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]]
4856
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4857
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
4858
        if data_idx >= self.__num_dataset:
4859
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4860
4861
4862
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
4863
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
4864
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
4865
            _safe_call(_LIB.LGBM_BoosterGetEval(
4866
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4867
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4868
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4869
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
4870
            if tmp_out_len.value != self.__num_inner_eval:
4871
                raise ValueError("Wrong length of eval results")
4872
            for i in range(self.__num_inner_eval):
4873
4874
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
4875
4876
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
4877
4878
4879
4880
4881
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
4882
4883
4884
4885
4886
4887
4888
4889
4890
            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
4891
4892
4893
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

4894
    def __inner_predict(self, data_idx: int) -> np.ndarray:
4895
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
4896
        if data_idx >= self.__num_dataset:
4897
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4898
4899
4900
4901
4902
        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
4903
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
4904
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
4905
4906
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
4907
            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
4908
            _safe_call(_LIB.LGBM_BoosterGetPredict(
4909
                self._handle,
Guolin Ke's avatar
Guolin Ke committed
4910
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4911
4912
                ctypes.byref(tmp_out_len),
                data_ptr))
4913
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):  # type: ignore[arg-type]
4914
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
4915
            self.__is_predicted_cur_iter[data_idx] = True
4916
        result: np.ndarray = self.__inner_predict_buffer[data_idx]  # type: ignore[assignment]
4917
4918
4919
4920
        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
4921

4922
    def __get_eval_info(self) -> None:
4923
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
4924
4925
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
4926
            out_num_eval = ctypes.c_int(0)
4927
            # Get num of inner evals
wxchan's avatar
wxchan committed
4928
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
4929
                self._handle,
wxchan's avatar
wxchan committed
4930
4931
4932
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
4933
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
4934
                tmp_out_len = ctypes.c_int(0)
4935
4936
4937
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
4938
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
4939
                ]
4940
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
wxchan's avatar
wxchan committed
4941
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
4942
                    self._handle,
4943
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
4944
                    ctypes.byref(tmp_out_len),
4945
                    ctypes.c_size_t(reserved_string_buffer_size),
4946
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4947
4948
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
4949
                    raise ValueError("Length of eval names doesn't equal with num_evals")
4950
4951
4952
4953
4954
4955
                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)
                    ]
4956
                    ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
4957
                    _safe_call(_LIB.LGBM_BoosterGetEvalNames(
4958
                        self._handle,
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
                        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
                ]