basic.py 199 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3

4
import abc
wxchan's avatar
wxchan committed
5
import ctypes
6
import inspect
7
import json
wxchan's avatar
wxchan committed
8
import warnings
9
from collections import OrderedDict
10
from copy import deepcopy
11
from enum import Enum
12
from functools import wraps
13
from os import SEEK_END, environ
14
15
from os.path import getsize
from pathlib import Path
16
from tempfile import NamedTemporaryFile
17
from typing import TYPE_CHECKING, Any, Callable, Dict, Iterable, Iterator, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
18
19
20
21

import numpy as np
import scipy.sparse

22
23
24
25
from .compat import (
    PANDAS_INSTALLED,
    PYARROW_INSTALLED,
    arrow_cffi,
26
    arrow_is_boolean,
27
28
29
30
31
32
33
34
35
36
37
38
39
    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
40
41
from .libpath import find_lib_path

42
43
44
if TYPE_CHECKING:
    from typing import Literal

45
46
47
48
49
50
51
    # typing.TypeGuard was only introduced in Python 3.10
    try:
        from typing import TypeGuard
    except ImportError:
        from typing_extensions import TypeGuard


52
__all__ = [
53
54
55
56
57
58
    "Booster",
    "Dataset",
    "LGBMDeprecationWarning",
    "LightGBMError",
    "register_logger",
    "Sequence",
59
60
]

61
_BoosterHandle = ctypes.c_void_p
62
_DatasetHandle = ctypes.c_void_p
63
64
_ctypes_int_ptr = Union[
    "ctypes._Pointer[ctypes.c_int32]",
65
    "ctypes._Pointer[ctypes.c_int64]",
66
]
67
68
_ctypes_int_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_int32]]",
69
    "ctypes.Array[ctypes._Pointer[ctypes.c_int64]]",
70
]
71
72
_ctypes_float_ptr = Union[
    "ctypes._Pointer[ctypes.c_float]",
73
    "ctypes._Pointer[ctypes.c_double]",
74
75
76
]
_ctypes_float_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_float]]",
77
    "ctypes.Array[ctypes._Pointer[ctypes.c_double]]",
78
]
79
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
80
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
81
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
82
_LGBM_BoosterEvalMethodResultWithStandardDeviationType = Tuple[str, str, float, bool, float]
83
84
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], "Literal['auto']"]
_LGBM_FeatureNameConfiguration = Union[List[str], "Literal['auto']"]
85
86
87
88
_LGBM_GroupType = Union[
    List[float],
    List[int],
    np.ndarray,
89
90
91
    pd_Series,
    pa_Array,
    pa_ChunkedArray,
92
]
93
94
_LGBM_PositionType = Union[
    np.ndarray,
95
    pd_Series,
96
]
97
98
99
100
101
102
_LGBM_InitScoreType = Union[
    List[float],
    List[List[float]],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
103
104
105
    pa_Table,
    pa_Array,
    pa_ChunkedArray,
106
]
107
108
109
110
111
112
113
114
115
_LGBM_TrainDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix,
    "Sequence",
    List["Sequence"],
116
    List[np.ndarray],
117
    pa_Table,
118
]
119
_LGBM_LabelType = Union[
120
121
    List[float],
    List[int],
122
123
    np.ndarray,
    pd_Series,
124
125
126
    pd_DataFrame,
    pa_Array,
    pa_ChunkedArray,
127
]
128
129
130
131
132
133
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
134
135
    scipy.sparse.spmatrix,
    pa_Table,
136
]
137
138
139
140
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
141
142
143
    pd_Series,
    pa_Array,
    pa_ChunkedArray,
144
]
145
146
147
148
149
150
151
152
153
154
155
156
157
_LGBM_SetFieldType = Union[
    List[List[float]],
    List[List[int]],
    List[float],
    List[int],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
    pa_Table,
    pa_Array,
    pa_ChunkedArray,
]

158
159
ZERO_THRESHOLD = 1e-35

160
161
_MULTICLASS_OBJECTIVES = {"multiclass", "multiclassova", "multiclass_ova", "ova", "ovr", "softmax"}

162

163
164
165
166
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


167
def _get_sample_count(total_nrow: int, params: str) -> int:
168
    sample_cnt = ctypes.c_int(0)
169
170
171
172
173
174
175
    _safe_call(
        _LIB.LGBM_GetSampleCount(
            ctypes.c_int32(total_nrow),
            _c_str(params),
            ctypes.byref(sample_cnt),
        )
    )
176
177
    return sample_cnt.value

wxchan's avatar
wxchan committed
178

179
class _MissingType(Enum):
180
181
182
    NONE = "None"
    NAN = "NaN"
    ZERO = "Zero"
183
184


185
class _DummyLogger:
186
    def info(self, msg: str) -> None:
187
        print(msg)  # noqa: T201
188

189
    def warning(self, msg: str) -> None:
190
191
192
        warnings.warn(msg, stacklevel=3)


193
194
195
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
196
197


198
199
200
201
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


202
def register_logger(
203
204
205
    logger: Any,
    info_method_name: str = "info",
    warning_method_name: str = "warning",
206
) -> None:
207
208
209
210
    """Register custom logger.

    Parameters
    ----------
211
    logger : Any
212
        Custom logger.
213
214
215
216
    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.
217
    """
218
    if not _has_method(logger, info_method_name) or not _has_method(logger, warning_method_name):
219
        raise TypeError(f"Logger must provide '{info_method_name}' and '{warning_method_name}' method")
220
221

    global _LOGGER, _INFO_METHOD_NAME, _WARNING_METHOD_NAME
222
    _LOGGER = logger
223
224
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
225
226


227
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
228
    """Join log messages from native library which come by chunks."""
229
    msg_normalized: List[str] = []
230
231

    @wraps(func)
232
    def wrapper(msg: str) -> None:
233
        nonlocal msg_normalized
234
235
        if msg.strip() == "":
            msg = "".join(msg_normalized)
236
237
238
239
240
241
242
243
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


244
def _log_info(msg: str) -> None:
245
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
246
247


248
def _log_warning(msg: str) -> None:
249
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
250
251
252


@_normalize_native_string
253
def _log_native(msg: str) -> None:
254
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
255
256


257
def _log_callback(msg: bytes) -> None:
258
    """Redirect logs from native library into Python."""
259
    _log_native(str(msg.decode("utf-8")))
260
261


262
def _load_lib() -> ctypes.CDLL:
263
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
264
265
266
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
267
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
268
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
269
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
270
        raise LightGBMError(lib.LGBM_GetLastError().decode("utf-8"))
wxchan's avatar
wxchan committed
271
272
    return lib

wxchan's avatar
wxchan committed
273

274
275
# we don't need lib_lightgbm while building docs
_LIB: ctypes.CDLL
276
if environ.get("LIGHTGBM_BUILD_DOC", False):
277
    from unittest.mock import Mock  # isort: skip
278

279
280
281
    _LIB = Mock(ctypes.CDLL)  # type: ignore
else:
    _LIB = _load_lib()
wxchan's avatar
wxchan committed
282

wxchan's avatar
wxchan committed
283

284
_NUMERIC_TYPES = (int, float, bool)
285
286


287
def _safe_call(ret: int) -> None:
288
289
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
290
291
292
    Parameters
    ----------
    ret : int
293
        The return value from C API calls.
wxchan's avatar
wxchan committed
294
295
    """
    if ret != 0:
296
        raise LightGBMError(_LIB.LGBM_GetLastError().decode("utf-8"))
wxchan's avatar
wxchan committed
297

wxchan's avatar
wxchan committed
298

299
def _is_numeric(obj: Any) -> bool:
300
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
301
302
303
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
304
305
306
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
307
308
        return False

wxchan's avatar
wxchan committed
309

310
def _is_numpy_1d_array(data: Any) -> bool:
311
    """Check whether data is a numpy 1-D array."""
312
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
313

wxchan's avatar
wxchan committed
314

315
def _is_numpy_column_array(data: Any) -> bool:
316
317
318
319
320
321
322
    """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


323
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
324
    """Cast numpy array to given dtype."""
325
326
327
328
329
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


330
def _is_1d_list(data: Any) -> bool:
331
    """Check whether data is a 1-D list."""
332
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
333

wxchan's avatar
wxchan committed
334

335
def _is_list_of_numpy_arrays(data: Any) -> "TypeGuard[List[np.ndarray]]":
336
    return isinstance(data, list) and all(isinstance(x, np.ndarray) for x in data)
337
338
339


def _is_list_of_sequences(data: Any) -> "TypeGuard[List[Sequence]]":
340
    return isinstance(data, list) and all(isinstance(x, Sequence) for x in data)
341
342


343
344
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
345
    return _is_numpy_1d_array(data) or _is_numpy_column_array(data) or _is_1d_list(data) or isinstance(data, pd_Series)
346
347


348
349
def _list_to_1d_numpy(
    data: Any,
350
    dtype: "np.typing.DTypeLike",
351
    name: str,
352
) -> np.ndarray:
353
    """Convert data to numpy 1-D array."""
354
    if _is_numpy_1d_array(data):
355
        return _cast_numpy_array_to_dtype(data, dtype)
356
    elif _is_numpy_column_array(data):
357
        _log_warning("Converting column-vector to 1d array")
358
        array = data.ravel()
359
        return _cast_numpy_array_to_dtype(array, dtype)
360
    elif _is_1d_list(data):
361
        return np.asarray(data, dtype=dtype)
362
    elif isinstance(data, pd_Series):
363
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
364
        return np.asarray(data, dtype=dtype)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
365
    else:
366
367
368
        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
369

wxchan's avatar
wxchan committed
370

371
372
373
374
375
376
377
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."""
378
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
379
380
381
382


def _is_2d_collection(data: Any) -> bool:
    """Check whether data is a 2-D collection."""
383
    return _is_numpy_2d_array(data) or _is_2d_list(data) or isinstance(data, pd_DataFrame)
384
385


386
def _is_pyarrow_array(data: Any) -> "TypeGuard[Union[pa_Array, pa_ChunkedArray]]":
387
388
389
390
    """Check whether data is a PyArrow array."""
    return isinstance(data, (pa_Array, pa_ChunkedArray))


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
417
418
419
420
421
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
422
423
424
425
426
    if isinstance(data, pa_Array):
        export_objects = [data]
    elif isinstance(data, pa_ChunkedArray):
        export_objects = data.chunks
    elif isinstance(data, pa_Table):
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
        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)


447
448
def _data_to_2d_numpy(
    data: Any,
449
    dtype: "np.typing.DTypeLike",
450
    name: str,
451
) -> np.ndarray:
452
453
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
454
        return _cast_numpy_array_to_dtype(data, dtype)
455
456
457
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
458
        _check_for_bad_pandas_dtypes(data.dtypes)
459
        return _cast_numpy_array_to_dtype(data.values, dtype)
460
461
462
463
    raise TypeError(
        f"Wrong type({type(data).__name__}) for {name}.\n"
        "It should be list of lists, numpy 2-D array or pandas DataFrame"
    )
464
465


466
def _cfloat32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
467
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
468
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
469
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
470
    else:
471
        raise RuntimeError("Expected float pointer")
wxchan's avatar
wxchan committed
472

Guolin Ke's avatar
Guolin Ke committed
473

474
def _cfloat64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
475
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
476
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
477
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
478
    else:
479
        raise RuntimeError("Expected double pointer")
Guolin Ke's avatar
Guolin Ke committed
480

wxchan's avatar
wxchan committed
481

482
def _cint32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
483
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
484
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
485
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
486
    else:
487
        raise RuntimeError("Expected int32 pointer")
488
489


490
def _cint64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
491
492
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
493
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
494
    else:
495
        raise RuntimeError("Expected int64 pointer")
wxchan's avatar
wxchan committed
496

wxchan's avatar
wxchan committed
497

498
def _c_str(string: str) -> ctypes.c_char_p:
499
    """Convert a Python string to C string."""
500
    return ctypes.c_char_p(string.encode("utf-8"))
wxchan's avatar
wxchan committed
501

wxchan's avatar
wxchan committed
502

503
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
504
    """Convert a Python array to C array."""
505
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
506

wxchan's avatar
wxchan committed
507

508
def _json_default_with_numpy(obj: Any) -> Any:
509
510
511
512
513
514
515
516
517
    """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


518
519
520
521
522
523
524
525
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)


526
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
527
    """Convert Python dictionary to string, which is passed to C API."""
528
    if data is None or not data:
wxchan's avatar
wxchan committed
529
530
531
        return ""
    pairs = []
    for key, val in data.items():
532
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
533
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
534
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
535
            pairs.append(f"{key}={val}")
536
        elif val is not None:
537
538
            raise TypeError(f"Unknown type of parameter:{key}, got:{type(val).__name__}")
    return " ".join(pairs)
539

wxchan's avatar
wxchan committed
540

541
class _TempFile:
542
543
    """Proxy class to workaround errors on Windows."""

544
    def __enter__(self) -> "_TempFile":
545
546
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
547
            self.path = Path(self.name)
548
        return self
wxchan's avatar
wxchan committed
549

550
    def __exit__(self, exc_type: Any, exc_val: Any, exc_tb: Any) -> None:
551
552
        if self.path.is_file():
            self.path.unlink()
553

wxchan's avatar
wxchan committed
554

555
class LightGBMError(Exception):
556
557
    """Error thrown by LightGBM."""

558
559
560
    pass


561
# DeprecationWarning is not shown by default, so let's create our own with higher level
562
563
# ref: https://peps.python.org/pep-0565/#additional-use-case-for-futurewarning
class LGBMDeprecationWarning(FutureWarning):
564
565
566
567
568
569
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
570
571
572
573
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
574
    def _get_all_param_aliases() -> Dict[str, List[str]]:
575
576
577
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
578
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
579
580
581
582
583
584
585
        _safe_call(
            _LIB.LGBM_DumpParamAliases(
                ctypes.c_int64(buffer_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer,
            )
        )
586
587
588
589
        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)
590
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
591
592
593
594
595
596
597
            _safe_call(
                _LIB.LGBM_DumpParamAliases(
                    ctypes.c_int64(actual_len),
                    ctypes.byref(tmp_out_len),
                    ptr_string_buffer,
                )
            )
598
        return json.loads(
599
            string_buffer.value.decode("utf-8"), object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
600
        )
601
602

    @classmethod
603
    def get(cls, *args: str) -> Set[str]:
604
605
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
606
607
        ret = set()
        for i in args:
608
            ret.update(cls.get_sorted(i))
609
610
        return ret

611
612
613
614
615
616
    @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])

617
    @classmethod
618
    def get_by_alias(cls, *args: str) -> Set[str]:
619
620
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
621
622
623
624
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
625
                    ret.update(aliases)
626
627
628
                    break
        return ret

629

630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
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)

651
652
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
653
654

    # if main_param_name was provided, keep that value and remove all aliases
655
    if main_param_name in params.keys():
656
657
658
        for param in aliases:
            params.pop(param, None)
        return params
659

660
661
662
663
664
    # 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
665

666
667
668
669
670
671
672
    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
673
674
675
676

    return params


677
_MAX_INT32 = (1 << 31) - 1
678

679
"""Macro definition of data type in C API of LightGBM"""
680
681
682
683
_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
684

685
"""Matrix is row major in Python"""
686
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
687

688
"""Macro definition of prediction type in C API of LightGBM"""
689
690
691
692
_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
693

694
"""Macro definition of sparse matrix type"""
695
696
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
697

698
"""Macro definition of feature importance type"""
699
700
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
701

702
"""Data type of data field"""
703
704
705
706
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
707
    "group": _C_API_DTYPE_INT32,
708
    "position": _C_API_DTYPE_INT32,
709
}
wxchan's avatar
wxchan committed
710

711
"""String name to int feature importance type mapper"""
712
713
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
714
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN,
715
}
716

wxchan's avatar
wxchan committed
717

718
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
719
    """Fix the memory of multi-dimensional sliced object."""
720
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
721
        if not data.flags.c_contiguous:
722
723
724
725
            _log_warning(
                "Usage of np.ndarray subset (sliced data) is not recommended "
                "due to it will double the peak memory cost in LightGBM."
            )
726
727
728
729
            return np.copy(data)
    return data


730
def _c_float_array(data: np.ndarray) -> Tuple[_ctypes_float_ptr, int, np.ndarray]:
731
    """Get pointer of float numpy array / list."""
732
    if _is_1d_list(data):
733
        data = np.asarray(data)
734
    if _is_numpy_1d_array(data):
735
        data = _convert_from_sliced_object(data)
736
        assert data.flags.c_contiguous
737
        ptr_data: _ctypes_float_ptr
wxchan's avatar
wxchan committed
738
739
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
740
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
741
742
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
743
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
744
        else:
745
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
746
    else:
747
        raise TypeError(f"Unknown type({type(data).__name__})")
748
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
749

wxchan's avatar
wxchan committed
750

751
def _c_int_array(data: np.ndarray) -> Tuple[_ctypes_int_ptr, int, np.ndarray]:
752
    """Get pointer of int numpy array / list."""
753
    if _is_1d_list(data):
754
        data = np.asarray(data)
755
    if _is_numpy_1d_array(data):
756
        data = _convert_from_sliced_object(data)
757
        assert data.flags.c_contiguous
758
        ptr_data: _ctypes_int_ptr
wxchan's avatar
wxchan committed
759
760
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
761
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
762
763
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
764
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
765
        else:
766
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
767
    else:
768
        raise TypeError(f"Unknown type({type(data).__name__})")
769
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
770

wxchan's avatar
wxchan committed
771

772
def _is_allowed_numpy_dtype(dtype: type) -> bool:
773
774
    float128 = getattr(np, "float128", type(None))
    return issubclass(dtype, (np.integer, np.floating, np.bool_)) and not issubclass(dtype, (np.timedelta64, float128))
775
776


777
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
778
    bad_pandas_dtypes = [
779
        f"{column_name}: {pandas_dtype}"
780
        for column_name, pandas_dtype in pandas_dtypes_series.items()
781
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
782
783
    ]
    if bad_pandas_dtypes:
784
785
786
787
        raise ValueError(
            'pandas dtypes must be int, float or bool.\n'
            f'Fields with bad pandas dtypes: {", ".join(bad_pandas_dtypes)}'
        )
788
789


790
791
def _pandas_to_numpy(
    data: pd_DataFrame,
792
    target_dtype: "np.typing.DTypeLike",
793
794
795
796
797
798
799
800
801
802
803
804
805
806
) -> 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)


807
def _data_from_pandas(
808
809
810
    data: pd_DataFrame,
    feature_name: _LGBM_FeatureNameConfiguration,
    categorical_feature: _LGBM_CategoricalFeatureConfiguration,
811
    pandas_categorical: Optional[List[List]],
812
) -> Tuple[np.ndarray, List[str], Union[List[str], List[int]], List[List]]:
813
    if len(data.shape) != 2 or data.shape[0] < 1:
814
        raise ValueError("Input data must be 2 dimensional and non empty.")
815

816
817
818
819
    # take shallow copy in case we modify categorical columns
    # whole column modifications don't change the original df
    data = data.copy(deep=False)

820
    # determine feature names
821
    if feature_name == "auto":
822
823
824
825
        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)]
826
    cat_cols_not_ordered: List[str] = [col for col in cat_cols if not data[col].cat.ordered]
827
828
    if pandas_categorical is None:  # train dataset
        pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
829
    else:
830
        if len(cat_cols) != len(pandas_categorical):
831
            raise ValueError("train and valid dataset categorical_feature do not match.")
832
833
834
835
836
        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})
837
838

    # use cat cols from DataFrame
839
    if categorical_feature == "auto":
840
841
842
        categorical_feature = cat_cols_not_ordered

    df_dtypes = [dtype.type for dtype in data.dtypes]
843
844
    # so that the target dtype considers floats
    df_dtypes.append(np.float32)
845
    target_dtype = np.result_type(*df_dtypes)
846
847
848
849
850

    return (
        _pandas_to_numpy(data, target_dtype=target_dtype),
        feature_name,
        categorical_feature,
851
        pandas_categorical,
852
    )
853
854


855
856
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
857
    file_name: Optional[Union[str, Path]] = None,
858
) -> str:
859
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
860
    pandas_str = f"\npandas_categorical:{categorical_json}\n"
861
    if file_name is not None:
862
        with open(file_name, "a") as f:
863
864
865
866
            f.write(pandas_str)
    return pandas_str


867
868
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
869
    model_str: Optional[str] = None,
870
) -> Optional[List[List]]:
871
    pandas_key = "pandas_categorical:"
872
    offset = -len(pandas_key)
873
    if file_name is not None:
874
        max_offset = -getsize(file_name)
875
        with open(file_name, "rb") as f:
876
877
878
            while True:
                if offset < max_offset:
                    offset = max_offset
879
                f.seek(offset, SEEK_END)
880
881
882
883
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
884
        last_line = lines[-1].decode("utf-8").strip()
885
        if not last_line.startswith(pandas_key):
886
            last_line = lines[-2].decode("utf-8").strip()
887
    elif model_str is not None:
888
        idx = model_str.rfind("\n", 0, offset)
889
890
        last_line = model_str[idx:].strip()
    if last_line.startswith(pandas_key):
891
        return json.loads(last_line[len(pandas_key) :])
892
893
    else:
        return None
894
895


896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
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**.

916
917
    .. versionadded:: 3.3.0

918
919
920
921
922
923
924
925
926
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
927
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
928
929
930
931
932
933
934
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
935
                return self._get_one_line(idx)
936
            elif isinstance(idx, slice):
937
938
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
939
                # Only required if using ``Dataset.subset()``.
940
                return np.array([self._get_one_line(i) for i in idx])
941
            else:
942
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
943
944
945

        Parameters
        ----------
946
        idx : int, slice[int], list[int]
947
948
949
950
            Item index.

        Returns
        -------
951
        result : numpy 1-D array or numpy 2-D array
952
            1-D array if idx is int, 2-D array if idx is slice or list.
953
954
955
956
957
958
959
960
961
        """
        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__()")


962
class _InnerPredictor:
963
964
965
966
967
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
968
969
970
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
971
    """
972

973
974
    def __init__(
        self,
975
976
977
        booster_handle: _BoosterHandle,
        pandas_categorical: Optional[List[List]],
        pred_parameter: Dict[str, Any],
978
        manage_handle: bool,
979
    ):
980
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
981
982
983

        Parameters
        ----------
984
        booster_handle : object
985
            Handle of Booster.
986
987
988
989
        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
990
            Other parameters for the prediction.
991
992
        manage_handle : bool
            If ``True``, free the corresponding Booster on the C++ side when this Python object is deleted.
wxchan's avatar
wxchan committed
993
        """
994
995
996
997
998
999
1000
1001
        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(
1002
                self._handle,
1003
                ctypes.byref(out_num_class),
1004
1005
1006
            )
        )
        self.num_class = out_num_class.value
wxchan's avatar
wxchan committed
1007

1008
1009
1010
1011
    @classmethod
    def from_booster(
        cls,
        booster: "Booster",
1012
        pred_parameter: Dict[str, Any],
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
    ) -> "_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,
1027
                ctypes.byref(out_cur_iter),
1028
1029
1030
1031
1032
1033
            )
        )
        return cls(
            booster_handle=booster._handle,
            pandas_categorical=booster.pandas_categorical,
            pred_parameter=pred_parameter,
1034
            manage_handle=False,
1035
1036
1037
1038
1039
1040
        )

    @classmethod
    def from_model_file(
        cls,
        model_file: Union[str, Path],
1041
        pred_parameter: Dict[str, Any],
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
    ) -> "_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),
1058
                ctypes.byref(booster_handle),
1059
1060
1061
1062
1063
1064
            )
        )
        return cls(
            booster_handle=booster_handle,
            pandas_categorical=_load_pandas_categorical(file_name=model_file),
            pred_parameter=pred_parameter,
1065
            manage_handle=True,
1066
        )
cbecker's avatar
cbecker committed
1067

1068
    def __del__(self) -> None:
1069
1070
        try:
            if self.__is_manage_handle:
1071
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
1072
1073
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
1074

1075
    def __getstate__(self) -> Dict[str, Any]:
1076
        this = self.__dict__.copy()
1077
1078
        this.pop("handle", None)
        this.pop("_handle", None)
1079
1080
        return this

1081
1082
    def predict(
        self,
1083
        data: _LGBM_PredictDataType,
1084
1085
1086
1087
1088
1089
        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,
1090
        validate_features: bool = False,
1091
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
1092
        """Predict logic.
wxchan's avatar
wxchan committed
1093
1094
1095

        Parameters
        ----------
1096
        data : str, pathlib.Path, numpy array, pandas DataFrame, pyarrow Table, H2O DataTable's Frame or scipy.sparse
1097
            Data source for prediction.
1098
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
1099
1100
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
        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.
1112
1113
1114
        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
1115

1116
1117
            .. versionadded:: 4.0.0

wxchan's avatar
wxchan committed
1118
1119
        Returns
        -------
1120
        result : numpy array, scipy.sparse or list of scipy.sparse
1121
            Prediction result.
1122
            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
1123
        """
wxchan's avatar
wxchan committed
1124
        if isinstance(data, Dataset):
1125
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
1126
1127
1128
        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))()
1129
            ptr_names[:] = [x.encode("utf-8") for x in data_names]
1130
1131
            _safe_call(
                _LIB.LGBM_BoosterValidateFeatureNames(
1132
                    self._handle,
1133
1134
1135
1136
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
1137
1138
1139
1140
1141
1142

        if isinstance(data, pd_DataFrame):
            data = _data_from_pandas(
                data=data,
                feature_name="auto",
                categorical_feature="auto",
1143
                pandas_categorical=self.pandas_categorical,
1144
1145
            )[0]

1146
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
1147
        if raw_score:
1148
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
1149
        if pred_leaf:
1150
            predict_type = _C_API_PREDICT_LEAF_INDEX
1151
        if pred_contrib:
1152
            predict_type = _C_API_PREDICT_CONTRIB
cbecker's avatar
cbecker committed
1153

1154
        if isinstance(data, (str, Path)):
1155
            with _TempFile() as f:
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
                _safe_call(
                    _LIB.LGBM_BoosterPredictForFile(
                        self._handle,
                        _c_str(str(data)),
                        ctypes.c_int(data_has_header),
                        ctypes.c_int(predict_type),
                        ctypes.c_int(start_iteration),
                        ctypes.c_int(num_iteration),
                        _c_str(self.pred_parameter),
                        _c_str(f.name),
                    )
                )
1168
1169
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
1170
        elif isinstance(data, scipy.sparse.csr_matrix):
1171
1172
1173
1174
            preds, nrow = self.__pred_for_csr(
                csr=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1175
                predict_type=predict_type,
1176
            )
Guolin Ke's avatar
Guolin Ke committed
1177
        elif isinstance(data, scipy.sparse.csc_matrix):
1178
1179
1180
1181
            preds, nrow = self.__pred_for_csc(
                csc=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1182
                predict_type=predict_type,
1183
            )
wxchan's avatar
wxchan committed
1184
        elif isinstance(data, np.ndarray):
1185
1186
1187
1188
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1189
                predict_type=predict_type,
1190
            )
1191
1192
1193
1194
1195
        elif _is_pyarrow_table(data):
            preds, nrow = self.__pred_for_pyarrow_table(
                table=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1196
                predict_type=predict_type,
1197
            )
1198
1199
1200
        elif isinstance(data, list):
            try:
                data = np.array(data)
1201
            except BaseException as err:
1202
                raise ValueError("Cannot convert data list to numpy array.") from err
1203
1204
1205
1206
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1207
                predict_type=predict_type,
1208
            )
1209
        elif isinstance(data, dt_DataTable):
1210
1211
1212
1213
            preds, nrow = self.__pred_for_np2d(
                mat=data.to_numpy(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1214
                predict_type=predict_type,
1215
            )
wxchan's avatar
wxchan committed
1216
1217
        else:
            try:
1218
                _log_warning("Converting data to scipy sparse matrix.")
wxchan's avatar
wxchan committed
1219
                csr = scipy.sparse.csr_matrix(data)
1220
            except BaseException as err:
1221
                raise TypeError(f"Cannot predict data for type {type(data).__name__}") from err
1222
1223
1224
1225
            preds, nrow = self.__pred_for_csr(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1226
                predict_type=predict_type,
1227
            )
wxchan's avatar
wxchan committed
1228
1229
        if pred_leaf:
            preds = preds.astype(np.int32)
1230
        is_sparse = isinstance(preds, (list, scipy.sparse.spmatrix))
1231
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1232
            if preds.size % nrow == 0:
1233
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1234
            else:
1235
                raise ValueError(f"Length of predict result ({preds.size}) cannot be divide nrow ({nrow})")
wxchan's avatar
wxchan committed
1236
1237
        return preds

1238
1239
1240
1241
1242
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
1243
        predict_type: int,
1244
    ) -> int:
1245
        """Get size of prediction result."""
1246
        if nrow > _MAX_INT32:
1247
1248
1249
1250
1251
1252
            raise LightGBMError(
                "LightGBM cannot perform prediction for data "
                f"with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n"
                "You can split your data into chunks "
                "and then concatenate predictions for them"
            )
Guolin Ke's avatar
Guolin Ke committed
1253
        n_preds = ctypes.c_int64(0)
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
        _safe_call(
            _LIB.LGBM_BoosterCalcNumPredict(
                self._handle,
                ctypes.c_int(nrow),
                ctypes.c_int(predict_type),
                ctypes.c_int(start_iteration),
                ctypes.c_int(num_iteration),
                ctypes.byref(n_preds),
            )
        )
Guolin Ke's avatar
Guolin Ke committed
1264
        return n_preds.value
wxchan's avatar
wxchan committed
1265

1266
1267
1268
1269
1270
1271
    def __inner_predict_np2d(
        self,
        mat: np.ndarray,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
1272
        preds: Optional[np.ndarray],
1273
1274
    ) -> Tuple[np.ndarray, int]:
        if mat.dtype == np.float32 or mat.dtype == np.float64:
1275
            data = np.asarray(mat.reshape(mat.size), dtype=mat.dtype)
1276
1277
1278
        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)
1279
1280
1281
1282
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
1283
            predict_type=predict_type,
1284
        )
1285
1286
1287
1288
1289
        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)
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
        _safe_call(
            _LIB.LGBM_BoosterPredictForMat(
                self._handle,
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int32(mat.shape[0]),
                ctypes.c_int32(mat.shape[1]),
                ctypes.c_int(_C_API_IS_ROW_MAJOR),
                ctypes.c_int(predict_type),
                ctypes.c_int(start_iteration),
                ctypes.c_int(num_iteration),
                _c_str(self.pred_parameter),
                ctypes.byref(out_num_preds),
                preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double)),
            )
        )
1306
1307
1308
1309
1310
1311
1312
1313
1314
        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,
1315
        predict_type: int,
1316
    ) -> Tuple[np.ndarray, int]:
1317
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1318
        if len(mat.shape) != 2:
1319
            raise ValueError("Input numpy.ndarray or list must be 2 dimensional")
wxchan's avatar
wxchan committed
1320

1321
        nrow = mat.shape[0]
1322
1323
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1324
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1325
1326
1327
1328
            n_preds = [
                self.__get_num_preds(start_iteration, num_iteration, i, predict_type)
                for i in np.diff([0] + list(sections) + [nrow])
            ]
1329
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1330
            preds = np.empty(sum(n_preds), dtype=np.float64)
1331
1332
1333
            for chunk, (start_idx_pred, end_idx_pred) in zip(
                np.array_split(mat, sections), zip(n_preds_sections, n_preds_sections[1:])
            ):
1334
                # avoid memory consumption by arrays concatenation operations
1335
1336
1337
1338
1339
                self.__inner_predict_np2d(
                    mat=chunk,
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
1340
                    preds=preds[start_idx_pred:end_idx_pred],
1341
                )
1342
            return preds, nrow
wxchan's avatar
wxchan committed
1343
        else:
1344
1345
1346
1347
1348
            return self.__inner_predict_np2d(
                mat=mat,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
1349
                preds=None,
1350
            )
wxchan's avatar
wxchan committed
1351

1352
1353
1354
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1355
1356
1357
1358
1359
1360
        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,
1361
        is_csr: bool,
1362
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1363
1364
1365
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1366
        if indptr_type == _C_API_DTYPE_INT32:
1367
            out_indptr = _cint32_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1368
        elif indptr_type == _C_API_DTYPE_INT64:
1369
            out_indptr = _cint64_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1370
1371
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1372
        if data_type == _C_API_DTYPE_FLOAT32:
1373
            out_data = _cfloat32_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1374
        elif data_type == _C_API_DTYPE_FLOAT64:
1375
            out_data = _cfloat64_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1376
1377
        else:
            raise TypeError("Expected float32 or float64 type for data")
1378
        out_indices = _cint32_array_to_numpy(cptr=out_ptr_indices, length=data_indices_len)
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
        # 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]
1390
1391
            cs_indices = out_indices[offset + cs_indptr[0] : offset + matrix_indptr_len]
            cs_data = out_data[offset + cs_indptr[0] : offset + matrix_indptr_len]
1392
1393
1394
1395
1396
1397
1398
1399
1400
            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
1401
1402
1403
1404
1405
1406
1407
1408
1409
        _safe_call(
            _LIB.LGBM_BoosterFreePredictSparse(
                out_ptr_indptr,
                out_ptr_indices,
                out_ptr_data,
                ctypes.c_int(indptr_type),
                ctypes.c_int(data_type),
            )
        )
1410
1411
1412
1413
        if len(cs_output_matrices) == 1:
            return cs_output_matrices[0]
        return cs_output_matrices

1414
1415
1416
1417
1418
1419
    def __inner_predict_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
1420
        preds: Optional[np.ndarray],
1421
1422
    ) -> Tuple[np.ndarray, int]:
        nrow = len(csr.indptr) - 1
1423
1424
1425
1426
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
1427
            predict_type=predict_type,
1428
        )
1429
1430
1431
1432
1433
        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
1434

1435
1436
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1437

1438
1439
1440
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
        _safe_call(
            _LIB.LGBM_BoosterPredictForCSR(
                self._handle,
                ptr_indptr,
                ctypes.c_int(type_ptr_indptr),
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csr.indptr)),
                ctypes.c_int64(len(csr.data)),
                ctypes.c_int64(csr.shape[1]),
                ctypes.c_int(predict_type),
                ctypes.c_int(start_iteration),
                ctypes.c_int(num_iteration),
                _c_str(self.pred_parameter),
                ctypes.byref(out_num_preds),
                preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double)),
            )
        )
1460
1461
1462
1463
1464
1465
1466
1467
1468
        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,
1469
        predict_type: int,
1470
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1471
1472
1473
1474
        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
1475
        out_ptr_indptr: _ctypes_int_ptr
1476
1477
1478
1479
1480
        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)()
1481
        out_ptr_data: _ctypes_float_ptr
1482
1483
1484
1485
1486
        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)
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
        _safe_call(
            _LIB.LGBM_BoosterPredictSparseOutput(
                self._handle,
                ptr_indptr,
                ctypes.c_int(type_ptr_indptr),
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csr.indptr)),
                ctypes.c_int64(len(csr.data)),
                ctypes.c_int64(csr.shape[1]),
                ctypes.c_int(predict_type),
                ctypes.c_int(start_iteration),
                ctypes.c_int(num_iteration),
                _c_str(self.pred_parameter),
                ctypes.c_int(matrix_type),
                out_shape.ctypes.data_as(ctypes.POINTER(ctypes.c_int64)),
                ctypes.byref(out_ptr_indptr),
                ctypes.byref(out_ptr_indices),
                ctypes.byref(out_ptr_data),
            )
        )
1509
1510
1511
1512
1513
1514
1515
1516
        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,
1517
            is_csr=True,
1518
1519
1520
1521
        )
        nrow = len(csr.indptr) - 1
        return matrices, nrow

1522
1523
1524
1525
1526
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
1527
        predict_type: int,
1528
    ) -> Tuple[np.ndarray, int]:
1529
        """Predict for a CSR data."""
1530
        if predict_type == _C_API_PREDICT_CONTRIB:
1531
1532
1533
1534
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1535
                predict_type=predict_type,
1536
            )
1537
        nrow = len(csr.indptr) - 1
1538
1539
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1540
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1541
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1542
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1543
            preds = np.empty(sum(n_preds), dtype=np.float64)
1544
1545
1546
            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:])
            ):
1547
                # avoid memory consumption by arrays concatenation operations
1548
1549
1550
1551
1552
                self.__inner_predict_csr(
                    csr=csr[start_idx:end_idx],
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
1553
                    preds=preds[start_idx_pred:end_idx_pred],
1554
                )
1555
1556
            return preds, nrow
        else:
1557
1558
1559
1560
1561
            return self.__inner_predict_csr(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
1562
                preds=None,
1563
1564
1565
1566
            )

    def __inner_predict_sparse_csc(
        self,
1567
1568
1569
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
1570
        predict_type: int,
1571
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1572
1573
1574
1575
        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
1576
        out_ptr_indptr: _ctypes_int_ptr
1577
1578
1579
1580
1581
        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)()
1582
        out_ptr_data: _ctypes_float_ptr
1583
1584
1585
1586
1587
        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)
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
        _safe_call(
            _LIB.LGBM_BoosterPredictSparseOutput(
                self._handle,
                ptr_indptr,
                ctypes.c_int(type_ptr_indptr),
                csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csc.indptr)),
                ctypes.c_int64(len(csc.data)),
                ctypes.c_int64(csc.shape[0]),
                ctypes.c_int(predict_type),
                ctypes.c_int(start_iteration),
                ctypes.c_int(num_iteration),
                _c_str(self.pred_parameter),
                ctypes.c_int(matrix_type),
                out_shape.ctypes.data_as(ctypes.POINTER(ctypes.c_int64)),
                ctypes.byref(out_ptr_indptr),
                ctypes.byref(out_ptr_indices),
                ctypes.byref(out_ptr_data),
            )
        )
1610
1611
1612
1613
1614
1615
1616
1617
        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,
1618
            is_csr=False,
1619
1620
1621
        )
        nrow = csc.shape[0]
        return matrices, nrow
Guolin Ke's avatar
Guolin Ke committed
1622

1623
1624
1625
1626
1627
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
1628
        predict_type: int,
1629
    ) -> Tuple[np.ndarray, int]:
1630
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1631
        nrow = csc.shape[0]
1632
        if nrow > _MAX_INT32:
1633
1634
1635
1636
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1637
                predict_type=predict_type,
1638
            )
1639
        if predict_type == _C_API_PREDICT_CONTRIB:
1640
1641
1642
1643
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1644
                predict_type=predict_type,
1645
            )
1646
1647
1648
1649
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
1650
            predict_type=predict_type,
1651
        )
1652
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1653
1654
        out_num_preds = ctypes.c_int64(0)

1655
1656
        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
1657

1658
        assert csc.shape[0] <= _MAX_INT32
1659
        csc_indices = csc.indices.astype(np.int32, copy=False)
1660

1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
        _safe_call(
            _LIB.LGBM_BoosterPredictForCSC(
                self._handle,
                ptr_indptr,
                ctypes.c_int(type_ptr_indptr),
                csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csc.indptr)),
                ctypes.c_int64(len(csc.data)),
                ctypes.c_int64(csc.shape[0]),
                ctypes.c_int(predict_type),
                ctypes.c_int(start_iteration),
                ctypes.c_int(num_iteration),
                _c_str(self.pred_parameter),
                ctypes.byref(out_num_preds),
                preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double)),
            )
        )
wxchan's avatar
wxchan committed
1680
        if n_preds != out_num_preds.value:
1681
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1682
        return preds, nrow
1683

1684
1685
1686
1687
1688
    def __pred_for_pyarrow_table(
        self,
        table: pa_Table,
        start_iteration: int,
        num_iteration: int,
1689
        predict_type: int,
1690
1691
1692
1693
1694
1695
    ) -> 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)
1696
        if not all(arrow_is_integer(t) or arrow_is_floating(t) or arrow_is_boolean(t) for t in table.schema.types):
1697
1698
1699
1700
1701
1702
1703
            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,
1704
            predict_type=predict_type,
1705
1706
1707
1708
1709
1710
        )
        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)
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
        _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)),
            )
        )
1725
1726
1727
        if n_preds != out_num_preds.value:
            raise ValueError("Wrong length for predict results")
        return preds, table.num_rows
wxchan's avatar
wxchan committed
1728

1729
    def current_iteration(self) -> int:
1730
1731
1732
1733
1734
1735
1736
1737
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
        out_cur_iter = ctypes.c_int(0)
1738
1739
1740
1741
1742
1743
        _safe_call(
            _LIB.LGBM_BoosterGetCurrentIteration(
                self._handle,
                ctypes.byref(out_cur_iter),
            )
        )
1744
1745
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1746

1747
class Dataset:
1748
1749
1750
1751
1752
1753
1754
1755
1756
    """
    Dataset in LightGBM.

    LightGBM does not train on raw data.
    It discretizes continuous features into histogram bins, tries to combine categorical features,
    and automatically handles missing and infinite values.

    This class handles that preprocessing, and holds that alternative representation of the input data.
    """
1757

1758
1759
    def __init__(
        self,
1760
        data: _LGBM_TrainDataType,
1761
        label: Optional[_LGBM_LabelType] = None,
1762
        reference: Optional["Dataset"] = None,
1763
1764
1765
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1766
1767
        feature_name: _LGBM_FeatureNameConfiguration = "auto",
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = "auto",
1768
        params: Optional[Dict[str, Any]] = None,
1769
1770
        free_raw_data: bool = True,
        position: Optional[_LGBM_PositionType] = None,
1771
    ):
1772
        """Initialize Dataset.
1773

wxchan's avatar
wxchan committed
1774
1775
        Parameters
        ----------
1776
        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
1777
            Data source of Dataset.
1778
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1779
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1780
1781
1782
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1783
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1784
            Weight for each instance. Weights should be non-negative.
1785
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1786
1787
1788
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1789
1790
            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.
1791
        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)
1792
            Init score for Dataset.
1793
        feature_name : list of str, or 'auto', optional (default="auto")
1794
            Feature names.
1795
            If 'auto' and data is pandas DataFrame or pyarrow Table, data columns names are used.
1796
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1797
1798
            Categorical features.
            If list of int, interpreted as indices.
1799
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1800
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1801
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1802
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1803
            All negative values in categorical features will be treated as missing values.
1804
            The output cannot be monotonically constrained with respect to a categorical feature.
1805
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1806
        params : dict or None, optional (default=None)
1807
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1808
        free_raw_data : bool, optional (default=True)
1809
            If True, raw data is freed after constructing inner Dataset.
1810
1811
        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
1812
        """
1813
        self._handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1814
1815
1816
1817
1818
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1819
        self.position = position
1820
        self.init_score = init_score
1821
1822
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1823
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1824
        self.free_raw_data = free_raw_data
1825
        self.used_indices: Optional[List[int]] = None
1826
        self._need_slice = True
1827
        self._predictor: Optional[_InnerPredictor] = None
1828
        self.pandas_categorical: Optional[List[List]] = None
1829
        self._params_back_up: Optional[Dict[str, Any]] = None
1830
        self.version = 0
1831
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1832

1833
    def __del__(self) -> None:
1834
1835
1836
1837
        try:
            self._free_handle()
        except AttributeError:
            pass
1838

1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
    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.
        """
1856
        param_str = _param_dict_to_str(self.get_params())
1857
1858
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1859
        ptr_data, _, _ = _c_int_array(indices)
1860
1861
        actual_sample_cnt = ctypes.c_int32(0)

1862
1863
1864
1865
1866
1867
1868
1869
        _safe_call(
            _LIB.LGBM_SampleIndices(
                ctypes.c_int32(total_nrow),
                _c_str(param_str),
                ptr_data,
                ctypes.byref(actual_sample_cnt),
            )
        )
1870
1871
        assert sample_cnt == actual_sample_cnt.value
        return indices
1872

1873
1874
1875
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
1876
1877
        ref_dataset: _DatasetHandle,
    ) -> "Dataset":
1878
1879
1880
1881
1882
1883
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1884
1885
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1886
1887
1888
1889
1890
1891

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
1892
        self._handle = ctypes.c_void_p()
1893
1894
1895
1896
1897
1898
1899
        _safe_call(
            _LIB.LGBM_DatasetCreateByReference(
                ref_dataset,
                ctypes.c_int64(total_nrow),
                ctypes.byref(self._handle),
            )
        )
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
        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
        ----------
1913
        sample_data : list of numpy array
1914
            Sample data for each column.
1915
        sample_indices : list of numpy array
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
            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.
1938
        sample_col_ptr: _ctypes_float_array = (ctypes.POINTER(ctypes.c_double) * ncol)()
1939
1940
        # c type int**
        # each int* points to start of indices for each column
1941
        indices_col_ptr: _ctypes_int_array = (ctypes.POINTER(ctypes.c_int32) * ncol)()
1942
        for i in range(ncol):
1943
1944
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1945
1946

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1947
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1948

1949
        self._handle = ctypes.c_void_p()
1950
        params_str = _param_dict_to_str(self.get_params())
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
        _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),
                ctypes.c_int64(total_nrow),
                _c_str(params_str),
                ctypes.byref(self._handle),
            )
        )
1964
1965
        return self

1966
    def _push_rows(self, data: np.ndarray) -> "Dataset":
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
        """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)
1981
        data_ptr, data_type, _ = _c_float_array(data)
1982

1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
        _safe_call(
            _LIB.LGBM_DatasetPushRows(
                self._handle,
                data_ptr,
                data_type,
                ctypes.c_int32(nrow),
                ctypes.c_int32(ncol),
                ctypes.c_int32(self._start_row),
            )
        )
1993
1994
1995
        self._start_row += nrow
        return self

1996
    def get_params(self) -> Dict[str, Any]:
1997
1998
1999
2000
        """Get the used parameters in the Dataset.

        Returns
        -------
2001
        params : dict
2002
2003
2004
2005
            The used parameters in this Dataset object.
        """
        if self.params is not None:
            # no min_data, nthreads and verbose in this function
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
            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",
                "linear_tree",
                "max_bin",
                "max_bin_by_feature",
                "min_data_in_bin",
                "pre_partition",
                "precise_float_parser",
                "two_round",
                "use_missing",
                "weight_column",
                "zero_as_missing",
            )
2029
            return {k: v for k, v in self.params.items() if k in dataset_params}
2030
2031
        else:
            return {}
2032

2033
    def _free_handle(self) -> "Dataset":
2034
2035
2036
        if self._handle is not None:
            _safe_call(_LIB.LGBM_DatasetFree(self._handle))
            self._handle = None
2037
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
2038
2039
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
2040
        return self
wxchan's avatar
wxchan committed
2041

2042
2043
2044
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
2045
        data: _LGBM_TrainDataType,
2046
        used_indices: Optional[Union[List[int], np.ndarray]],
2047
    ) -> "Dataset":
Guolin Ke's avatar
Guolin Ke committed
2048
        data_has_header = False
2049
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
2050
            # check data has header or not
2051
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
2052
        num_data = self.num_data()
2053
        if predictor is not None:
2054
2055
2056
            init_score: Union[np.ndarray, scipy.sparse.spmatrix] = predictor.predict(
                data=data,
                raw_score=True,
2057
                data_has_header=data_has_header,
2058
            )
2059
            init_score = init_score.ravel()
2060
            if used_indices is not None:
2061
                assert not self._need_slice
2062
                if isinstance(data, (str, Path)):
2063
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
2064
                    assert num_data == len(used_indices)
2065
2066
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
2067
2068
2069
                            sub_init_score[i * predictor.num_class + j] = init_score[
                                used_indices[i] * predictor.num_class + j
                            ]
2070
2071
2072
                    init_score = sub_init_score
            if predictor.num_class > 1:
                # need to regroup init_score
2073
                new_init_score = np.empty(init_score.size, dtype=np.float64)
2074
2075
                for i in range(num_data):
                    for j in range(predictor.num_class):
2076
2077
2078
                        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:
2079
            init_score = np.full_like(self.init_score, fill_value=0.0, dtype=np.float64)
2080
2081
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
2082
        self.set_init_score(init_score)
2083
        return self
Guolin Ke's avatar
Guolin Ke committed
2084

2085
2086
    def _lazy_init(
        self,
2087
        data: Optional[_LGBM_TrainDataType],
2088
2089
2090
2091
2092
2093
2094
2095
        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,
2096
        params: Optional[Dict[str, Any]],
2097
        position: Optional[_LGBM_PositionType],
2098
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2099
        if data is None:
2100
            self._handle = None
Nikita Titov's avatar
Nikita Titov committed
2101
            return self
Guolin Ke's avatar
Guolin Ke committed
2102
2103
2104
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
2105
2106
2107
2108
2109
        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,
2110
                pandas_categorical=self.pandas_categorical,
2111
            )
Guolin Ke's avatar
Guolin Ke committed
2112

2113
        # process for args
wxchan's avatar
wxchan committed
2114
        params = {} if params is None else params
2115
        args_names = inspect.signature(self.__class__._lazy_init).parameters.keys()
2116
        for key in params.keys():
2117
            if key in args_names:
2118
2119
2120
2121
                _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."
                )
2122
        # get categorical features
2123
        if isinstance(categorical_feature, list):
2124
2125
            categorical_indices = set()
            feature_dict = {}
2126
            if isinstance(feature_name, list):
2127
2128
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
2129
                if isinstance(name, str) and name in feature_dict:
2130
                    categorical_indices.add(feature_dict[name])
2131
                elif isinstance(name, int):
2132
2133
                    categorical_indices.add(name)
                else:
2134
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
2135
            if categorical_indices:
2136
2137
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
2138
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
2139
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
2140
                            _log_warning(f"{cat_alias} in param dict is overridden.")
2141
                        params.pop(cat_alias, None)
2142
                params["categorical_column"] = sorted(categorical_indices)
2143

2144
        params_str = _param_dict_to_str(params)
2145
        self.params = params
2146
        # process for reference dataset
wxchan's avatar
wxchan committed
2147
        ref_dataset = None
wxchan's avatar
wxchan committed
2148
        if isinstance(reference, Dataset):
2149
            ref_dataset = reference.construct()._handle
wxchan's avatar
wxchan committed
2150
        elif reference is not None:
2151
            raise TypeError("Reference dataset should be None or dataset instance")
2152
        # start construct data
2153
        if isinstance(data, (str, Path)):
2154
            self._handle = ctypes.c_void_p()
2155
2156
2157
2158
2159
2160
2161
2162
            _safe_call(
                _LIB.LGBM_DatasetCreateFromFile(
                    _c_str(str(data)),
                    _c_str(params_str),
                    ref_dataset,
                    ctypes.byref(self._handle),
                )
            )
wxchan's avatar
wxchan committed
2163
2164
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
2165
2166
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
2167
2168
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
2169
2170
2171
        elif _is_pyarrow_table(data):
            self.__init_from_pyarrow_table(data, params_str, ref_dataset)
            feature_name = data.column_names
2172
        elif isinstance(data, list) and len(data) > 0:
2173
            if _is_list_of_numpy_arrays(data):
2174
                self.__init_from_list_np2d(data, params_str, ref_dataset)
2175
            elif _is_list_of_sequences(data):
2176
2177
                self.__init_from_seqs(data, ref_dataset)
            else:
2178
                raise TypeError("Data list can only be of ndarray or Sequence")
2179
2180
        elif isinstance(data, Sequence):
            self.__init_from_seqs([data], ref_dataset)
2181
        elif isinstance(data, dt_DataTable):
2182
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
2183
2184
2185
2186
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
2187
            except BaseException as err:
2188
                raise TypeError(f"Cannot initialize Dataset from {type(data).__name__}") from err
wxchan's avatar
wxchan committed
2189
2190
2191
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
2192
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
2193
2194
2195
2196
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
2197
2198
        if position is not None:
            self.set_position(position)
2199
2200
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
2201
                _log_warning("The init_score will be overridden by the prediction of init_model.")
2202
            self._set_init_score_by_predictor(predictor=predictor, data=data, used_indices=None)
2203
2204
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
2205
        elif predictor is not None:
2206
            raise TypeError(f"Wrong predictor type {type(predictor).__name__}")
Guolin Ke's avatar
Guolin Ke committed
2207
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
2208
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
2209

2210
    @staticmethod
2211
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]) -> Iterator[np.ndarray]:
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
        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]
2223
            yield row if row.flags["OWNDATA"] else row.copy()
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236

    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.
2237
        sampled = np.array(list(self._yield_row_from_seqlist(seqs, indices)))
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
        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

2253
2254
2255
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
2256
        ref_dataset: Optional[_DatasetHandle],
2257
    ) -> "Dataset":
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
        """
        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:
2272
            param_str = _param_dict_to_str(self.get_params())
2273
2274
2275
2276
2277
2278
2279
            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)
2280
            batch_size = getattr(seq, "batch_size", None) or Sequence.batch_size
2281
2282
2283
2284
2285
            for start in range(0, nrow, batch_size):
                end = min(start + batch_size, nrow)
                self._push_rows(seq[start:end])
        return self

2286
2287
2288
2289
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
2290
        ref_dataset: Optional[_DatasetHandle],
2291
    ) -> "Dataset":
2292
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
2293
        if len(mat.shape) != 2:
2294
            raise ValueError("Input numpy.ndarray must be 2 dimensional")
wxchan's avatar
wxchan committed
2295

2296
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2297
        if mat.dtype == np.float32 or mat.dtype == np.float64:
2298
            data = np.asarray(mat.reshape(mat.size), dtype=mat.dtype)
2299
        else:  # change non-float data to float data, need to copy
2300
            data = np.asarray(mat.reshape(mat.size), dtype=np.float32)
wxchan's avatar
wxchan committed
2301

2302
        ptr_data, type_ptr_data, _ = _c_float_array(data)
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
        _safe_call(
            _LIB.LGBM_DatasetCreateFromMat(
                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),
                _c_str(params_str),
                ref_dataset,
                ctypes.byref(self._handle),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
2315
        return self
wxchan's avatar
wxchan committed
2316

2317
2318
2319
2320
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
2321
        ref_dataset: Optional[_DatasetHandle],
2322
    ) -> "Dataset":
2323
        """Initialize data from a list of 2-D numpy matrices."""
2324
        ncol = mats[0].shape[1]
2325
        nrow = np.empty((len(mats),), np.int32)
2326
        ptr_data: _ctypes_float_array
2327
2328
2329
2330
2331
2332
        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 = []
2333
        type_ptr_data = -1
2334
2335
2336

        for i, mat in enumerate(mats):
            if len(mat.shape) != 2:
2337
                raise ValueError("Input numpy.ndarray must be 2 dimensional")
2338
2339

            if mat.shape[1] != ncol:
2340
                raise ValueError("Input arrays must have same number of columns")
2341
2342
2343
2344

            nrow[i] = mat.shape[0]

            if mat.dtype == np.float32 or mat.dtype == np.float64:
2345
                mats[i] = np.asarray(mat.reshape(mat.size), dtype=mat.dtype)
2346
            else:  # change non-float data to float data, need to copy
2347
2348
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2349
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2350
            if type_ptr_data != -1 and chunk_type_ptr_data != type_ptr_data:
2351
                raise ValueError("Input chunks must have same type")
2352
2353
2354
2355
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

2356
        self._handle = ctypes.c_void_p()
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
        _safe_call(
            _LIB.LGBM_DatasetCreateFromMats(
                ctypes.c_int32(len(mats)),
                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)),
                ctypes.c_int32(ncol),
                ctypes.c_int(_C_API_IS_ROW_MAJOR),
                _c_str(params_str),
                ref_dataset,
                ctypes.byref(self._handle),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
2370
        return self
2371

2372
2373
2374
2375
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
2376
        ref_dataset: Optional[_DatasetHandle],
2377
    ) -> "Dataset":
2378
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2379
        if len(csr.indices) != len(csr.data):
2380
            raise ValueError(f"Length mismatch: {len(csr.indices)} vs {len(csr.data)}")
2381
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2382

2383
2384
        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
2385

2386
        assert csr.shape[1] <= _MAX_INT32
2387
        csr_indices = csr.indices.astype(np.int32, copy=False)
2388

2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
        _safe_call(
            _LIB.LGBM_DatasetCreateFromCSR(
                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]),
                _c_str(params_str),
                ref_dataset,
                ctypes.byref(self._handle),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
2404
        return self
wxchan's avatar
wxchan committed
2405

2406
2407
2408
2409
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
2410
        ref_dataset: Optional[_DatasetHandle],
2411
    ) -> "Dataset":
2412
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2413
        if len(csc.indices) != len(csc.data):
2414
            raise ValueError(f"Length mismatch: {len(csc.indices)} vs {len(csc.data)}")
2415
        self._handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2416

2417
2418
        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
2419

2420
        assert csc.shape[0] <= _MAX_INT32
2421
        csc_indices = csc.indices.astype(np.int32, copy=False)
2422

2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
        _safe_call(
            _LIB.LGBM_DatasetCreateFromCSC(
                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]),
                _c_str(params_str),
                ref_dataset,
                ctypes.byref(self._handle),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
2438
        return self
Guolin Ke's avatar
Guolin Ke committed
2439

2440
2441
2442
2443
    def __init_from_pyarrow_table(
        self,
        table: pa_Table,
        params_str: str,
2444
        ref_dataset: Optional[_DatasetHandle],
2445
2446
2447
2448
2449
2450
    ) -> "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)
2451
        if not all(arrow_is_integer(t) or arrow_is_floating(t) or arrow_is_boolean(t) for t in table.schema.types):
2452
2453
2454
2455
2456
            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()
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
        _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),
            )
        )
2467
2468
        return self

2469
    @staticmethod
2470
    def _compare_params_for_warning(
2471
2472
        params: Dict[str, Any],
        other_params: Dict[str, Any],
2473
        ignore_keys: Set[str],
2474
2475
    ) -> bool:
        """Compare two dictionaries with params ignoring some keys.
2476

2477
2478
2479
2480
        It is only for the warning purpose.

        Parameters
        ----------
2481
        params : dict
2482
            One dictionary with parameters to compare.
2483
        other_params : dict
2484
2485
2486
            Another dictionary with parameters to compare.
        ignore_keys : set
            Keys that should be ignored during comparing two dictionaries.
2487
2488
2489

        Returns
        -------
2490
2491
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
        """
        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

2503
    def construct(self) -> "Dataset":
2504
2505
2506
2507
2508
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2509
            Constructed Dataset object.
2510
        """
2511
        if self._handle is None:
wxchan's avatar
wxchan committed
2512
            if self.reference is not None:
2513
                reference_params = self.reference.get_params()
2514
2515
                params = self.get_params()
                if params != reference_params:
2516
2517
2518
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
2519
                        ignore_keys=_ConfigAliases.get("categorical_feature"),
2520
                    ):
2521
                        _log_warning("Overriding the parameters from Reference Dataset.")
2522
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2523
                if self.used_indices is None:
2524
                    # create valid
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
                    self._lazy_init(
                        data=self.data,
                        label=self.label,
                        reference=self.reference,
                        weight=self.weight,
                        group=self.group,
                        position=self.position,
                        init_score=self.init_score,
                        predictor=self._predictor,
                        feature_name=self.feature_name,
                        categorical_feature="auto",
                        params=self.params,
                    )
wxchan's avatar
wxchan committed
2538
                else:
2539
                    # construct subset
2540
                    used_indices = _list_to_1d_numpy(self.used_indices, dtype=np.int32, name="used_indices")
2541
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2542
                    if self.reference.group is not None:
2543
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2544
2545
2546
                        _, self.group = np.unique(
                            np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices], return_counts=True
                        )
2547
                    self._handle = ctypes.c_void_p()
2548
                    params_str = _param_dict_to_str(self.params)
2549
2550
2551
2552
2553
2554
2555
2556
2557
                    _safe_call(
                        _LIB.LGBM_DatasetGetSubset(
                            self.reference.construct()._handle,
                            used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                            ctypes.c_int32(used_indices.shape[0]),
                            _c_str(params_str),
                            ctypes.byref(self._handle),
                        )
                    )
Guolin Ke's avatar
Guolin Ke committed
2558
2559
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2560
2561
                    if self.group is not None:
                        self.set_group(self.group)
2562
2563
                    if self.position is not None:
                        self.set_position(self.position)
wxchan's avatar
wxchan committed
2564
2565
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
2566
2567
2568
2569
                    if (
                        isinstance(self._predictor, _InnerPredictor)
                        and self._predictor is not self.reference._predictor
                    ):
Guolin Ke's avatar
Guolin Ke committed
2570
                        self.get_data()
2571
                        self._set_init_score_by_predictor(
2572
                            predictor=self._predictor, data=self.data, used_indices=used_indices
2573
                        )
wxchan's avatar
wxchan committed
2574
            else:
2575
                # create train
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
                self._lazy_init(
                    data=self.data,
                    label=self.label,
                    reference=None,
                    weight=self.weight,
                    group=self.group,
                    init_score=self.init_score,
                    predictor=self._predictor,
                    feature_name=self.feature_name,
                    categorical_feature=self.categorical_feature,
                    params=self.params,
                    position=self.position,
                )
wxchan's avatar
wxchan committed
2589
2590
            if self.free_raw_data:
                self.data = None
2591
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2592
        return self
wxchan's avatar
wxchan committed
2593

2594
2595
    def create_valid(
        self,
2596
        data: _LGBM_TrainDataType,
2597
        label: Optional[_LGBM_LabelType] = None,
2598
2599
2600
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2601
        params: Optional[Dict[str, Any]] = None,
2602
        position: Optional[_LGBM_PositionType] = None,
2603
    ) -> "Dataset":
2604
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2605
2606
2607

        Parameters
        ----------
2608
        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
2609
            Data source of Dataset.
2610
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2611
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2612
            Label of the data.
2613
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2614
            Weight for each instance. Weights should be non-negative.
2615
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2616
2617
2618
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2619
2620
            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.
2621
        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)
2622
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2623
        params : dict or None, optional (default=None)
2624
            Other parameters for validation Dataset.
2625
2626
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
2627
2628
2629

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2630
2631
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2632
        """
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
        ret = Dataset(
            data,
            label=label,
            reference=self,
            weight=weight,
            group=group,
            position=position,
            init_score=init_score,
            params=params,
            free_raw_data=self.free_raw_data,
        )
wxchan's avatar
wxchan committed
2644
        ret._predictor = self._predictor
2645
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2646
        return ret
wxchan's avatar
wxchan committed
2647

2648
2649
2650
    def subset(
        self,
        used_indices: List[int],
2651
        params: Optional[Dict[str, Any]] = None,
2652
    ) -> "Dataset":
2653
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2654
2655
2656
2657

        Parameters
        ----------
        used_indices : list of int
2658
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2659
        params : dict or None, optional (default=None)
2660
            These parameters will be passed to Dataset constructor.
2661
2662
2663
2664
2665

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2666
        """
wxchan's avatar
wxchan committed
2667
2668
        if params is None:
            params = self.params
2669
2670
2671
2672
2673
2674
2675
2676
        ret = Dataset(
            None,
            reference=self,
            feature_name=self.feature_name,
            categorical_feature=self.categorical_feature,
            params=params,
            free_raw_data=self.free_raw_data,
        )
wxchan's avatar
wxchan committed
2677
        ret._predictor = self._predictor
2678
        ret.pandas_categorical = self.pandas_categorical
2679
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2680
2681
        return ret

2682
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2683
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2684

2685
2686
2687
2688
2689
        .. 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
2690
2691
        Parameters
        ----------
2692
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2693
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2694
2695
2696
2697
2698

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2699
        """
2700
2701
2702
2703
2704
2705
        _safe_call(
            _LIB.LGBM_DatasetSaveBinary(
                self.construct()._handle,
                _c_str(str(filename)),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
2706
        return self
wxchan's avatar
wxchan committed
2707

2708
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2709
2710
        if not params:
            return self
2711
        params = deepcopy(params)
2712

2713
        def update() -> None:
2714
2715
2716
            if not self.params:
                self.params = params
            else:
2717
                self._params_back_up = deepcopy(self.params)
2718
2719
                self.params.update(params)

2720
        if self._handle is None:
2721
2722
2723
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2724
                _c_str(_param_dict_to_str(self.params)),
2725
2726
                _c_str(_param_dict_to_str(params)),
            )
2727
2728
2729
2730
2731
2732
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2733
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode("utf-8"))
Nikita Titov's avatar
Nikita Titov committed
2734
        return self
wxchan's avatar
wxchan committed
2735

2736
    def _reverse_update_params(self) -> "Dataset":
2737
        if self._handle is None:
2738
2739
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2740
        return self
2741

2742
2743
2744
    def set_field(
        self,
        field_name: str,
2745
        data: Optional[_LGBM_SetFieldType],
2746
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2747
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2748
2749
2750

        Parameters
        ----------
2751
        field_name : str
2752
            The field name of the information.
2753
        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
2754
            The data to be set.
Nikita Titov's avatar
Nikita Titov committed
2755
2756
2757
2758
2759

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2760
        """
2761
        if self._handle is None:
2762
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2763
        if data is None:
2764
            # set to None
2765
2766
2767
2768
2769
2770
2771
2772
2773
            _safe_call(
                _LIB.LGBM_DatasetSetField(
                    self._handle,
                    _c_str(field_name),
                    None,
                    ctypes.c_int(0),
                    ctypes.c_int(_FIELD_TYPE_MAPPER[field_name]),
                )
            )
Nikita Titov's avatar
Nikita Titov committed
2774
            return self
2775
2776

        # If the data is a arrow data, we can just pass it to C
2777
2778
2779
2780
2781
2782
        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}'")
2783
2784
2785
2786
2787
2788
2789
                data = pa_chunked_array(
                    [
                        chunk
                        for array in data.columns  # type: ignore
                        for chunk in array.chunks
                    ]
                )
2790

2791
            c_array = _export_arrow_to_c(data)
2792
2793
2794
2795
2796
2797
2798
2799
2800
            _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),
                )
            )
2801
2802
2803
            self.version += 1
            return self

2804
        dtype: "np.typing.DTypeLike"
2805
        if field_name == "init_score":
Guolin Ke's avatar
Guolin Ke committed
2806
            dtype = np.float64
2807
            if _is_1d_collection(data):
2808
                data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2809
            elif _is_2d_collection(data):
2810
                data = _data_to_2d_numpy(data, dtype=dtype, name=field_name)
2811
                data = data.ravel(order="F")
2812
2813
            else:
                raise TypeError(
2814
2815
                    "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."
2816
2817
                )
        else:
2818
            if field_name in {"group", "position"}:
2819
2820
2821
                dtype = np.int32
            else:
                dtype = np.float32
2822
            data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2823

2824
        ptr_data: Union[_ctypes_float_ptr, _ctypes_int_ptr]
2825
        if data.dtype == np.float32 or data.dtype == np.float64:
2826
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2827
        elif data.dtype == np.int32:
2828
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2829
        else:
2830
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2831
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2832
            raise TypeError("Input type error for set_field")
2833
2834
2835
2836
2837
2838
2839
2840
2841
        _safe_call(
            _LIB.LGBM_DatasetSetField(
                self._handle,
                _c_str(field_name),
                ptr_data,
                ctypes.c_int(len(data)),
                ctypes.c_int(type_data),
            )
        )
2842
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2843
        return self
wxchan's avatar
wxchan committed
2844

2845
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2846
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2847

2848
2849
2850
2851
2852
2853
        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
2854
2855
        Parameters
        ----------
2856
        field_name : str
2857
            The field name of the information.
wxchan's avatar
wxchan committed
2858
2859
2860

        Returns
        -------
2861
        info : numpy array or None
2862
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2863
        """
2864
        if self._handle is None:
2865
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2866
2867
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2868
        ret = ctypes.POINTER(ctypes.c_void_p)()
2869
2870
2871
2872
2873
2874
2875
2876
2877
        _safe_call(
            _LIB.LGBM_DatasetGetField(
                self._handle,
                _c_str(field_name),
                ctypes.byref(tmp_out_len),
                ctypes.byref(ret),
                ctypes.byref(out_type),
            )
        )
2878
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2879
2880
2881
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2882
        if out_type.value == _C_API_DTYPE_INT32:
2883
2884
            arr = _cint32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)),
2885
                length=tmp_out_len.value,
2886
            )
2887
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2888
2889
            arr = _cfloat32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)),
2890
                length=tmp_out_len.value,
2891
            )
2892
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2893
2894
            arr = _cfloat64_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)),
2895
                length=tmp_out_len.value,
2896
            )
2897
        else:
wxchan's avatar
wxchan committed
2898
            raise TypeError("Unknown type")
2899
        if field_name == "init_score":
2900
2901
2902
            num_data = self.num_data()
            num_classes = arr.size // num_data
            if num_classes > 1:
2903
                arr = arr.reshape((num_data, num_classes), order="F")
2904
        return arr
Guolin Ke's avatar
Guolin Ke committed
2905

2906
2907
    def set_categorical_feature(
        self,
2908
        categorical_feature: _LGBM_CategoricalFeatureConfiguration,
2909
    ) -> "Dataset":
2910
        """Set categorical features.
2911
2912
2913

        Parameters
        ----------
2914
        categorical_feature : list of str or int, or 'auto'
2915
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2916
2917
2918
2919
2920

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2921
2922
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2923
            return self
2924
        if self.data is not None:
2925
2926
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2927
                return self._free_handle()
2928
            elif categorical_feature == "auto":
Nikita Titov's avatar
Nikita Titov committed
2929
                return self
2930
            else:
2931
2932
2933
2934
2935
                if self.categorical_feature != "auto":
                    _log_warning(
                        "categorical_feature in Dataset is overridden.\n"
                        f"New categorical_feature is {list(categorical_feature)}"
                    )
2936
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2937
                return self._free_handle()
2938
        else:
2939
2940
2941
2942
            raise LightGBMError(
                "Cannot set categorical feature after freed raw data, "
                "set free_raw_data=False when construct Dataset to avoid this."
            )
2943

2944
2945
    def _set_predictor(
        self,
2946
        predictor: Optional[_InnerPredictor],
2947
    ) -> "Dataset":
2948
2949
2950
2951
        """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
2952
        """
2953
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2954
            return self
2955
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
2956
2957
2958
            if (predictor == self._predictor) and (
                predictor.current_iteration() == self._predictor.current_iteration()
            ):
2959
                return self
2960
        if self._handle is None:
Guolin Ke's avatar
Guolin Ke committed
2961
            self._predictor = predictor
2962
2963
        elif self.data is not None:
            self._predictor = predictor
2964
2965
2966
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
2967
                used_indices=None,
2968
            )
2969
2970
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2971
2972
2973
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.reference.data,
2974
                used_indices=self.used_indices,
2975
            )
Guolin Ke's avatar
Guolin Ke committed
2976
        else:
2977
2978
2979
2980
            raise LightGBMError(
                "Cannot set predictor after freed raw data, "
                "set free_raw_data=False when construct Dataset to avoid this."
            )
2981
        return self
Guolin Ke's avatar
Guolin Ke committed
2982

2983
    def set_reference(self, reference: "Dataset") -> "Dataset":
2984
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2985
2986
2987
2988

        Parameters
        ----------
        reference : Dataset
2989
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2990
2991
2992
2993
2994

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2995
        """
2996
2997
2998
        self.set_categorical_feature(reference.categorical_feature).set_feature_name(
            reference.feature_name
        )._set_predictor(reference._predictor)
2999
        # we're done if self and reference share a common upstream reference
3000
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
3001
            return self
Guolin Ke's avatar
Guolin Ke committed
3002
3003
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
3004
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
3005
        else:
3006
3007
3008
3009
            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
3010

3011
    def set_feature_name(self, feature_name: _LGBM_FeatureNameConfiguration) -> "Dataset":
3012
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
3013
3014
3015

        Parameters
        ----------
3016
        feature_name : list of str
3017
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
3018
3019
3020
3021
3022

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
3023
        """
3024
        if feature_name != "auto":
3025
            self.feature_name = feature_name
3026
        if self._handle is not None and feature_name is not None and feature_name != "auto":
wxchan's avatar
wxchan committed
3027
            if len(feature_name) != self.num_feature():
3028
3029
3030
                raise ValueError(
                    f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match"
                )
3031
            c_feature_name = [_c_str(name) for name in feature_name]
3032
3033
3034
3035
3036
3037
3038
            _safe_call(
                _LIB.LGBM_DatasetSetFeatureNames(
                    self._handle,
                    _c_array(ctypes.c_char_p, c_feature_name),
                    ctypes.c_int(len(feature_name)),
                )
            )
Nikita Titov's avatar
Nikita Titov committed
3039
        return self
Guolin Ke's avatar
Guolin Ke committed
3040

3041
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
3042
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
3043
3044
3045

        Parameters
        ----------
3046
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None
3047
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
3048
3049
3050
3051
3052

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
3053
3054
        """
        self.label = label
3055
        if self._handle is not None:
3056
3057
            if isinstance(label, pd_DataFrame):
                if len(label.columns) > 1:
3058
                    raise ValueError("DataFrame for label cannot have multiple columns")
3059
                label_array = np.ravel(_pandas_to_numpy(label, target_dtype=np.float32))
3060
3061
            elif _is_pyarrow_array(label):
                label_array = label
3062
            else:
3063
3064
3065
                label_array = _list_to_1d_numpy(label, dtype=np.float32, name="label")
            self.set_field("label", label_array)
            self.label = self.get_field("label")  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
3066
        return self
Guolin Ke's avatar
Guolin Ke committed
3067

3068
3069
    def set_weight(
        self,
3070
        weight: Optional[_LGBM_WeightType],
3071
    ) -> "Dataset":
3072
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
3073
3074
3075

        Parameters
        ----------
3076
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None
3077
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
3078
3079
3080
3081
3082

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
3083
        """
3084
3085
3086
3087
3088
3089
3090
        # 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
3091
        self.weight = weight
3092
3093

        # Set field
3094
        if self._handle is not None and weight is not None:
3095
            if not _is_pyarrow_array(weight):
3096
3097
3098
                weight = _list_to_1d_numpy(weight, dtype=np.float32, name="weight")
            self.set_field("weight", weight)
            self.weight = self.get_field("weight")  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
3099
        return self
Guolin Ke's avatar
Guolin Ke committed
3100

3101
3102
    def set_init_score(
        self,
3103
        init_score: Optional[_LGBM_InitScoreType],
3104
    ) -> "Dataset":
3105
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
3106
3107
3108

        Parameters
        ----------
3109
        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
3110
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
3111
3112
3113
3114
3115

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
3116
3117
        """
        self.init_score = init_score
3118
        if self._handle is not None and init_score is not None:
3119
3120
            self.set_field("init_score", init_score)
            self.init_score = self.get_field("init_score")  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
3121
        return self
Guolin Ke's avatar
Guolin Ke committed
3122

3123
3124
    def set_group(
        self,
3125
        group: Optional[_LGBM_GroupType],
3126
    ) -> "Dataset":
3127
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
3128
3129
3130

        Parameters
        ----------
3131
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None
3132
3133
3134
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
3135
3136
            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
3137
3138
3139
3140
3141

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
3142
3143
        """
        self.group = group
3144
        if self._handle is not None and group is not None:
3145
            if not _is_pyarrow_array(group):
3146
3147
                group = _list_to_1d_numpy(group, dtype=np.int32, name="group")
            self.set_field("group", group)
3148
            # original values can be modified at cpp side
3149
            constructed_group = self.get_field("group")
3150
3151
            if constructed_group is not None:
                self.group = np.diff(constructed_group)
Nikita Titov's avatar
Nikita Titov committed
3152
        return self
Guolin Ke's avatar
Guolin Ke committed
3153

3154
3155
    def set_position(
        self,
3156
        position: Optional[_LGBM_PositionType],
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
    ) -> "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:
3172
3173
            position = _list_to_1d_numpy(position, dtype=np.int32, name="position")
            self.set_field("position", position)
3174
3175
        return self

3176
    def get_feature_name(self) -> List[str]:
3177
3178
3179
3180
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
3181
        feature_names : list of str
3182
3183
            The names of columns (features) in the Dataset.
        """
3184
        if self._handle is None:
3185
3186
3187
3188
3189
            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)
3190
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
3191
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
        _safe_call(
            _LIB.LGBM_DatasetGetFeatureNames(
                self._handle,
                ctypes.c_int(num_feature),
                ctypes.byref(tmp_out_len),
                ctypes.c_size_t(reserved_string_buffer_size),
                ctypes.byref(required_string_buffer_size),
                ptr_string_buffers,
            )
        )
3202
3203
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
3204
3205
3206
3207
        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)]
3208
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
            _safe_call(
                _LIB.LGBM_DatasetGetFeatureNames(
                    self._handle,
                    ctypes.c_int(num_feature),
                    ctypes.byref(tmp_out_len),
                    ctypes.c_size_t(actual_string_buffer_size),
                    ctypes.byref(required_string_buffer_size),
                    ptr_string_buffers,
                )
            )
        return [string_buffers[i].value.decode("utf-8") for i in range(num_feature)]
3220

3221
    def get_label(self) -> Optional[_LGBM_LabelType]:
3222
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3223
3224
3225

        Returns
        -------
3226
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
3227
            The label information from the Dataset.
3228
            For a constructed ``Dataset``, this will only return a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3229
        """
3230
        if self.label is None:
3231
            self.label = self.get_field("label")
Guolin Ke's avatar
Guolin Ke committed
3232
3233
        return self.label

3234
    def get_weight(self) -> Optional[_LGBM_WeightType]:
3235
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3236
3237
3238

        Returns
        -------
3239
        weight : list, numpy 1-D array, pandas Series or None
3240
            Weight for each data point from the Dataset. Weights should be non-negative.
3241
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3242
        """
3243
        if self.weight is None:
3244
            self.weight = self.get_field("weight")
Guolin Ke's avatar
Guolin Ke committed
3245
3246
        return self.weight

3247
    def get_init_score(self) -> Optional[_LGBM_InitScoreType]:
3248
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3249
3250
3251

        Returns
        -------
3252
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
3253
            Init score of Booster.
3254
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3255
        """
3256
        if self.init_score is None:
3257
            self.init_score = self.get_field("init_score")
Guolin Ke's avatar
Guolin Ke committed
3258
3259
        return self.init_score

3260
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
3261
3262
3263
3264
        """Get the raw data of the Dataset.

        Returns
        -------
3265
        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
3266
3267
            Raw data used in the Dataset construction.
        """
3268
        if self._handle is None:
3269
            raise Exception("Cannot get data before construct Dataset")
3270
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
3271
3272
            self.data = self.reference.data
            if self.data is not None:
3273
                if isinstance(self.data, (np.ndarray, scipy.sparse.spmatrix)):
Guolin Ke's avatar
Guolin Ke committed
3274
                    self.data = self.data[self.used_indices, :]
3275
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3276
                    self.data = self.data.iloc[self.used_indices].copy()
3277
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3278
                    self.data = self.data[self.used_indices, :]
3279
3280
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
3281
                elif _is_list_of_sequences(self.data) and len(self.data) > 0:
3282
                    self.data = np.array(list(self._yield_row_from_seqlist(self.data, self.used_indices)))
Guolin Ke's avatar
Guolin Ke committed
3283
                else:
3284
3285
3286
                    _log_warning(
                        f"Cannot subset {type(self.data).__name__} type of raw data.\n" "Returning original raw data"
                    )
3287
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
3288
        if self.data is None:
3289
3290
3291
3292
            raise LightGBMError(
                "Cannot call `get_data` after freed raw data, "
                "set free_raw_data=False when construct Dataset to avoid this."
            )
3293
3294
        return self.data

3295
    def get_group(self) -> Optional[_LGBM_GroupType]:
3296
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3297
3298
3299

        Returns
        -------
3300
        group : list, numpy 1-D array, pandas Series or None
3301
3302
3303
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
3304
3305
            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.
3306
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3307
        """
3308
        if self.group is None:
3309
            self.group = self.get_field("group")
Guolin Ke's avatar
Guolin Ke committed
3310
3311
            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
3312
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
3313
3314
        return self.group

3315
    def get_position(self) -> Optional[_LGBM_PositionType]:
3316
3317
3318
3319
        """Get the position of the Dataset.

        Returns
        -------
3320
        position : numpy 1-D array, pandas Series or None
3321
            Position of items used in unbiased learning-to-rank task.
3322
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
3323
3324
        """
        if self.position is None:
3325
            self.position = self.get_field("position")
3326
3327
        return self.position

3328
    def num_data(self) -> int:
3329
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3330
3331
3332

        Returns
        -------
3333
3334
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3335
        """
3336
        if self._handle is not None:
3337
            ret = ctypes.c_int(0)
3338
3339
3340
3341
3342
3343
            _safe_call(
                _LIB.LGBM_DatasetGetNumData(
                    self._handle,
                    ctypes.byref(ret),
                )
            )
wxchan's avatar
wxchan committed
3344
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3345
        else:
3346
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3347

3348
    def num_feature(self) -> int:
3349
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3350
3351
3352

        Returns
        -------
3353
3354
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3355
        """
3356
        if self._handle is not None:
3357
            ret = ctypes.c_int(0)
3358
3359
3360
3361
3362
3363
            _safe_call(
                _LIB.LGBM_DatasetGetNumFeature(
                    self._handle,
                    ctypes.byref(ret),
                )
            )
wxchan's avatar
wxchan committed
3364
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3365
        else:
3366
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3367

3368
    def feature_num_bin(self, feature: Union[int, str]) -> int:
3369
3370
        """Get the number of bins for a feature.

3371
3372
        .. versionadded:: 4.0.0

3373
3374
        Parameters
        ----------
3375
3376
        feature : int or str
            Index or name of the feature.
3377
3378
3379
3380
3381
3382

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
3383
        if self._handle is not None:
3384
            if isinstance(feature, str):
3385
3386
3387
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
3388
            ret = ctypes.c_int(0)
3389
3390
3391
3392
3393
3394
3395
            _safe_call(
                _LIB.LGBM_DatasetGetFeatureNumBin(
                    self._handle,
                    ctypes.c_int(feature_index),
                    ctypes.byref(ret),
                )
            )
3396
3397
3398
3399
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

3400
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
3401
3402
3403
3404
3405
        """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.
3406
3407
3408
3409
3410

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
3411
3412
3413

        Returns
        -------
3414
3415
3416
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
3417
        head = self
3418
        ref_chain: Set[Dataset] = set()
3419
3420
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
3421
                ref_chain.add(head)
3422
3423
3424
3425
3426
3427
                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
3428
        return ref_chain
3429

3430
    def add_features_from(self, other: "Dataset") -> "Dataset":
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
        """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.
        """
3445
        if self._handle is None or other._handle is None:
3446
3447
3448
3449
3450
3451
3452
            raise ValueError("Both source and target Datasets must be constructed before adding features")
        _safe_call(
            _LIB.LGBM_DatasetAddFeaturesFrom(
                self._handle,
                other._handle,
            )
        )
Guolin Ke's avatar
Guolin Ke committed
3453
3454
3455
3456
3457
3458
3459
3460
        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))
3461
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3462
                    self.data = np.hstack((self.data, other.data.toarray()))
3463
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3464
                    self.data = np.hstack((self.data, other.data.values))
3465
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3466
3467
3468
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
3469
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3470
                sparse_format = self.data.getformat()
3471
                if isinstance(other.data, (np.ndarray, scipy.sparse.spmatrix)):
Guolin Ke's avatar
Guolin Ke committed
3472
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
3473
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3474
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
3475
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3476
3477
3478
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
3479
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3480
                if not PANDAS_INSTALLED:
3481
3482
3483
3484
3485
                    raise LightGBMError(
                        "Cannot add features to DataFrame type of raw data "
                        "without pandas installed. "
                        "Install pandas and restart your session."
                    )
Guolin Ke's avatar
Guolin Ke committed
3486
                if isinstance(other.data, np.ndarray):
3487
                    self.data = concat((self.data, pd_DataFrame(other.data)), axis=1, ignore_index=True)
3488
                elif isinstance(other.data, scipy.sparse.spmatrix):
3489
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())), axis=1, ignore_index=True)
3490
                elif isinstance(other.data, pd_DataFrame):
3491
                    self.data = concat((self.data, other.data), axis=1, ignore_index=True)
3492
                elif isinstance(other.data, dt_DataTable):
3493
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())), axis=1, ignore_index=True)
Guolin Ke's avatar
Guolin Ke committed
3494
3495
                else:
                    self.data = None
3496
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
3497
                if isinstance(other.data, np.ndarray):
3498
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
3499
                elif isinstance(other.data, scipy.sparse.spmatrix):
3500
3501
3502
3503
3504
                    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
3505
3506
3507
3508
3509
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
3510
3511
3512
3513
3514
3515
3516
            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"
            )
            err_msg += (
                "Set free_raw_data=False when construct Dataset to avoid this" if was_none else "Freeing raw data"
            )
3517
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
3518
        self.feature_name = self.get_feature_name()
3519
3520
3521
3522
        _log_warning(
            "Reseting categorical features.\n"
            "You can set new categorical features via ``set_categorical_feature`` method"
        )
Guolin Ke's avatar
Guolin Ke committed
3523
3524
        self.categorical_feature = "auto"
        self.pandas_categorical = None
3525
3526
        return self

3527
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
3528
3529
3530
3531
3532
3533
        """Save Dataset to a text file.

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

        Parameters
        ----------
3534
        filename : str or pathlib.Path
3535
3536
3537
3538
3539
3540
3541
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
3542
3543
3544
3545
3546
3547
        _safe_call(
            _LIB.LGBM_DatasetDumpText(
                self.construct()._handle,
                _c_str(str(filename)),
            )
        )
3548
3549
        return self

wxchan's avatar
wxchan committed
3550

3551
3552
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
3553
    Tuple[np.ndarray, np.ndarray],
3554
]
3555
3556
3557
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
3558
        _LGBM_EvalFunctionResultType,
3559
3560
3561
    ],
    Callable[
        [np.ndarray, Dataset],
3562
3563
        List[_LGBM_EvalFunctionResultType],
    ],
3564
]
3565
3566


3567
class Booster:
3568
    """Booster in LightGBM."""
3569

3570
3571
3572
3573
3574
    def __init__(
        self,
        params: Optional[Dict[str, Any]] = None,
        train_set: Optional[Dataset] = None,
        model_file: Optional[Union[str, Path]] = None,
3575
        model_str: Optional[str] = None,
3576
    ):
3577
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3578
3579
3580

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3581
        params : dict or None, optional (default=None)
3582
3583
3584
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3585
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3586
            Path to the model file.
3587
        model_str : str or None, optional (default=None)
3588
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3589
        """
3590
        self._handle = ctypes.c_void_p()
3591
        self._network = False
wxchan's avatar
wxchan committed
3592
        self.__need_reload_eval_info = True
3593
        self._train_data_name = "training"
3594
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3595
        self.best_iteration = -1
3596
        self.best_score: _LGBM_BoosterBestScoreType = {}
3597
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3598
        if train_set is not None:
3599
            # Training task
wxchan's avatar
wxchan committed
3600
            if not isinstance(train_set, Dataset):
3601
                raise TypeError(f"Training data should be Dataset instance, met {type(train_set).__name__}")
3602
3603
3604
            params = _choose_param_value(
                main_param_name="machines",
                params=params,
3605
                default_value=None,
3606
3607
3608
3609
3610
3611
3612
            )
            # 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):
3613
                    num_machines_from_machine_list = len(machines.split(","))
3614
3615
                elif isinstance(machines, (list, set)):
                    num_machines_from_machine_list = len(machines)
3616
                    machines = ",".join(machines)
3617
3618
3619
3620
3621
3622
                else:
                    raise ValueError("Invalid machines in params.")

                params = _choose_param_value(
                    main_param_name="num_machines",
                    params=params,
3623
                    default_value=num_machines_from_machine_list,
3624
3625
3626
3627
                )
                params = _choose_param_value(
                    main_param_name="local_listen_port",
                    params=params,
3628
                    default_value=12400,
3629
3630
3631
3632
3633
                )
                self.set_network(
                    machines=machines,
                    local_listen_port=params["local_listen_port"],
                    listen_time_out=params.get("time_out", 120),
3634
                    num_machines=params["num_machines"],
3635
                )
3636
            # construct booster object
3637
3638
3639
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3640
            params_str = _param_dict_to_str(params)
3641
3642
3643
3644
3645
3646
3647
            _safe_call(
                _LIB.LGBM_BoosterCreate(
                    train_set._handle,
                    _c_str(params_str),
                    ctypes.byref(self._handle),
                )
            )
3648
            # save reference to data
wxchan's avatar
wxchan committed
3649
            self.train_set = train_set
3650
3651
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3652
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3653
3654
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
3655
3656
3657
3658
3659
3660
                _safe_call(
                    _LIB.LGBM_BoosterMerge(
                        self._handle,
                        self.__init_predictor._handle,
                    )
                )
Guolin Ke's avatar
Guolin Ke committed
3661
            out_num_class = ctypes.c_int(0)
3662
3663
3664
3665
3666
3667
            _safe_call(
                _LIB.LGBM_BoosterGetNumClasses(
                    self._handle,
                    ctypes.byref(out_num_class),
                )
            )
wxchan's avatar
wxchan committed
3668
            self.__num_class = out_num_class.value
3669
            # buffer for inner predict
3670
            self.__inner_predict_buffer: List[Optional[np.ndarray]] = [None]
wxchan's avatar
wxchan committed
3671
3672
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3673
            self.pandas_categorical = train_set.pandas_categorical
3674
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3675
        elif model_file is not None:
3676
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3677
            out_num_iterations = ctypes.c_int(0)
3678
3679
3680
3681
3682
3683
3684
            _safe_call(
                _LIB.LGBM_BoosterCreateFromModelfile(
                    _c_str(str(model_file)),
                    ctypes.byref(out_num_iterations),
                    ctypes.byref(self._handle),
                )
            )
Guolin Ke's avatar
Guolin Ke committed
3685
            out_num_class = ctypes.c_int(0)
3686
3687
3688
3689
3690
3691
            _safe_call(
                _LIB.LGBM_BoosterGetNumClasses(
                    self._handle,
                    ctypes.byref(out_num_class),
                )
            )
wxchan's avatar
wxchan committed
3692
            self.__num_class = out_num_class.value
3693
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3694
            if params:
3695
                _log_warning("Ignoring params argument, using parameters from model file.")
3696
            params = self._get_loaded_param()
3697
        elif model_str is not None:
3698
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3699
        else:
3700
3701
3702
            raise TypeError(
                "Need at least one training dataset or model file or model string " "to create Booster instance"
            )
3703
        self.params = params
wxchan's avatar
wxchan committed
3704

3705
    def __del__(self) -> None:
3706
        try:
3707
            if self._network:
3708
3709
3710
3711
                self.free_network()
        except AttributeError:
            pass
        try:
3712
3713
            if self._handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
3714
3715
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3716

3717
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3718
3719
        return self.__deepcopy__(None)

3720
    def __deepcopy__(self, *args: Any, **kwargs: Any) -> "Booster":
3721
        model_str = self.model_to_string(num_iteration=-1)
3722
        return Booster(model_str=model_str)
wxchan's avatar
wxchan committed
3723

3724
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3725
        this = self.__dict__.copy()
3726
3727
3728
        handle = this["_handle"]
        this.pop("train_set", None)
        this.pop("valid_sets", None)
wxchan's avatar
wxchan committed
3729
        if handle is not None:
3730
            this["_handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3731
3732
        return this

3733
    def __setstate__(self, state: Dict[str, Any]) -> None:
3734
        model_str = state.get("_handle", state.get("handle", None))
3735
        if model_str is not None:
wxchan's avatar
wxchan committed
3736
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3737
            out_num_iterations = ctypes.c_int(0)
3738
3739
3740
3741
3742
3743
3744
3745
            _safe_call(
                _LIB.LGBM_BoosterLoadModelFromString(
                    _c_str(model_str),
                    ctypes.byref(out_num_iterations),
                    ctypes.byref(handle),
                )
            )
            state["_handle"] = handle
wxchan's avatar
wxchan committed
3746
3747
        self.__dict__.update(state)

3748
3749
3750
3751
    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)
3752
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3753
3754
3755
3756
3757
3758
3759
3760
        _safe_call(
            _LIB.LGBM_BoosterGetLoadedParam(
                self._handle,
                ctypes.c_int64(buffer_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer,
            )
        )
3761
3762
3763
3764
        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)
3765
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3766
3767
3768
3769
3770
3771
3772
3773
3774
            _safe_call(
                _LIB.LGBM_BoosterGetLoadedParam(
                    self._handle,
                    ctypes.c_int64(actual_len),
                    ctypes.byref(tmp_out_len),
                    ptr_string_buffer,
                )
            )
        return json.loads(string_buffer.value.decode("utf-8"))
3775

3776
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3777
3778
3779
3780
3781
3782
3783
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
3784
3785
        self.__dict__.pop("train_set", None)
        self.__dict__.pop("valid_sets", None)
3786
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3787
        return self
wxchan's avatar
wxchan committed
3788

3789
    def _free_buffer(self) -> "Booster":
3790
3791
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3792
        return self
3793

3794
3795
3796
3797
3798
    def set_network(
        self,
        machines: Union[List[str], Set[str], str],
        local_listen_port: int = 12400,
        listen_time_out: int = 120,
3799
        num_machines: int = 1,
3800
    ) -> "Booster":
3801
3802
3803
3804
        """Set the network configuration.

        Parameters
        ----------
3805
        machines : list, set or str
3806
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3807
        local_listen_port : int, optional (default=12400)
3808
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3809
        listen_time_out : int, optional (default=120)
3810
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3811
        num_machines : int, optional (default=1)
3812
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3813
3814
3815
3816
3817

        Returns
        -------
        self : Booster
            Booster with set network.
3818
        """
3819
        if isinstance(machines, (list, set)):
3820
3821
3822
3823
3824
3825
3826
3827
3828
            machines = ",".join(machines)
        _safe_call(
            _LIB.LGBM_NetworkInit(
                _c_str(machines),
                ctypes.c_int(local_listen_port),
                ctypes.c_int(listen_time_out),
                ctypes.c_int(num_machines),
            )
        )
3829
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3830
        return self
3831

3832
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3833
3834
3835
3836
3837
3838
3839
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3840
        _safe_call(_LIB.LGBM_NetworkFree())
3841
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3842
        return self
3843

3844
    def trees_to_dataframe(self) -> pd_DataFrame:
3845
3846
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3847
3848
3849
3850
        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.
3851
3852
3853
3854
3855
            - ``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.
3856
3857
            - ``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.
3858
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3859
3860
              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.
3861
3862
            - ``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.
3863
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3864
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3865
3866
            - ``count`` : int64, number of records in the training data that fall into this node.

3867
3868
3869
3870
3871
3872
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3873
3874
3875
3876
            raise LightGBMError(
                "This method cannot be run without pandas installed. "
                "You must install pandas and restart your session to use this method."
            )
3877
3878

        if self.num_trees() == 0:
3879
            raise LightGBMError("There are no trees in this Booster and thus nothing to parse")
3880

3881
        def _is_split_node(tree: Dict[str, Any]) -> bool:
3882
            return "split_index" in tree.keys()
3883

3884
3885
3886
3887
3888
        def create_node_record(
            tree: Dict[str, Any],
            node_depth: int = 1,
            tree_index: Optional[int] = None,
            feature_names: Optional[List[str]] = None,
3889
            parent_node: Optional[str] = None,
3890
3891
3892
        ) -> Dict[str, Any]:
            def _get_node_index(
                tree: Dict[str, Any],
3893
                tree_index: Optional[int],
3894
            ) -> str:
3895
                tree_num = f"{tree_index}-" if tree_index is not None else ""
3896
                is_split = _is_split_node(tree)
3897
                node_type = "S" if is_split else "L"
3898
                # if a single node tree it won't have `leaf_index` so return 0
3899
                node_num = tree.get("split_index" if is_split else "leaf_index", 0)
3900
                return f"{tree_num}{node_type}{node_num}"
3901

3902
3903
            def _get_split_feature(
                tree: Dict[str, Any],
3904
                feature_names: Optional[List[str]],
3905
            ) -> Optional[str]:
3906
3907
                if _is_split_node(tree):
                    if feature_names is not None:
3908
                        feature_name = feature_names[tree["split_feature"]]
3909
                    else:
3910
                        feature_name = tree["split_feature"]
3911
3912
3913
3914
                else:
                    feature_name = None
                return feature_name

3915
            def _is_single_node_tree(tree: Dict[str, Any]) -> bool:
3916
                return set(tree.keys()) == {"leaf_value"}
3917
3918

            # Create the node record, and populate universal data members
3919
            node: Dict[str, Union[int, str, None]] = OrderedDict()
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
            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
3935
3936
3937

            # Update values to reflect node type (leaf or split)
            if _is_split_node(tree):
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
                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"]
3948
            else:
3949
                node["value"] = tree["leaf_value"]
3950
                if not _is_single_node_tree(tree):
3951
3952
                    node["weight"] = tree["leaf_weight"]
                    node["count"] = tree["leaf_count"]
3953
3954
3955

            return node

3956
3957
3958
3959
3960
        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,
3961
            parent_node: Optional[str] = None,
3962
        ) -> List[Dict[str, Any]]:
3963
3964
3965
3966
3967
3968
3969
            node = create_node_record(
                tree=tree,
                node_depth=node_depth,
                tree_index=tree_index,
                feature_names=feature_names,
                parent_node=parent_node,
            )
3970
3971
3972
3973
3974

            res = [node]

            if _is_split_node(tree):
                # traverse the next level of the tree
3975
                children = ["left_child", "right_child"]
3976
3977
                for child in children:
                    subtree_list = tree_dict_to_node_list(
3978
                        tree=tree[child],
3979
3980
3981
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
3982
                        parent_node=node["node_index"],
3983
                    )
3984
3985
3986
3987
3988
3989
                    # 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()
3990
        feature_names = model_dict["feature_names"]
3991
        model_list = []
3992
3993
3994
3995
3996
3997
        for tree in model_dict["tree_info"]:
            model_list.extend(
                tree_dict_to_node_list(
                    tree=tree["tree_structure"], tree_index=tree["tree_index"], feature_names=feature_names
                )
            )
3998

3999
        return pd_DataFrame(model_list, columns=model_list[0].keys())
4000

4001
    def set_train_data_name(self, name: str) -> "Booster":
4002
4003
4004
4005
        """Set the name to the training Dataset.

        Parameters
        ----------
4006
        name : str
Nikita Titov's avatar
Nikita Titov committed
4007
4008
4009
4010
4011
4012
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
4013
        """
4014
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
4015
        return self
wxchan's avatar
wxchan committed
4016

4017
    def add_valid(self, data: Dataset, name: str) -> "Booster":
4018
        """Add validation data.
wxchan's avatar
wxchan committed
4019
4020
4021
4022

        Parameters
        ----------
        data : Dataset
4023
            Validation data.
4024
        name : str
4025
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
4026
4027
4028
4029
4030

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
4031
        """
Guolin Ke's avatar
Guolin Ke committed
4032
        if not isinstance(data, Dataset):
4033
            raise TypeError(f"Validation data should be Dataset instance, met {type(data).__name__}")
Guolin Ke's avatar
Guolin Ke committed
4034
        if data._predictor is not self.__init_predictor:
4035
4036
4037
4038
4039
4040
4041
            raise LightGBMError("Add validation data failed, " "you should use same predictor for these data")
        _safe_call(
            _LIB.LGBM_BoosterAddValidData(
                self._handle,
                data.construct()._handle,
            )
        )
wxchan's avatar
wxchan committed
4042
4043
4044
4045
4046
        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
4047
        return self
wxchan's avatar
wxchan committed
4048

4049
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
4050
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
4051
4052
4053
4054

        Parameters
        ----------
        params : dict
4055
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
4056
4057
4058
4059
4060

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
4061
        """
4062
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
4063
        if params_str:
4064
4065
4066
4067
4068
4069
            _safe_call(
                _LIB.LGBM_BoosterResetParameter(
                    self._handle,
                    _c_str(params_str),
                )
            )
Guolin Ke's avatar
Guolin Ke committed
4070
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
4071
        return self
wxchan's avatar
wxchan committed
4072

4073
4074
4075
    def update(
        self,
        train_set: Optional[Dataset] = None,
4076
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None,
4077
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
4078
        """Update Booster for one iteration.
4079

wxchan's avatar
wxchan committed
4080
4081
        Parameters
        ----------
4082
4083
4084
4085
        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
4086
            Customized objective function.
4087
4088
4089
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

4090
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4091
                    The predicted values.
4092
4093
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
4094
4095
                train_data : Dataset
                    The training dataset.
4096
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
4097
4098
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
4099
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
4100
4101
                    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
4102

4103
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
4104
            and grad and hess should be returned in the same format.
4105

wxchan's avatar
wxchan committed
4106
4107
        Returns
        -------
4108
4109
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
4110
        """
4111
        # need reset training data
4112
4113
4114
4115
4116
4117
        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
4118
            if not isinstance(train_set, Dataset):
4119
                raise TypeError(f"Training data should be Dataset instance, met {type(train_set).__name__}")
Guolin Ke's avatar
Guolin Ke committed
4120
            if train_set._predictor is not self.__init_predictor:
4121
                raise LightGBMError("Replace training data failed, " "you should use same predictor for these data")
wxchan's avatar
wxchan committed
4122
            self.train_set = train_set
4123
4124
4125
4126
4127
4128
            _safe_call(
                _LIB.LGBM_BoosterResetTrainingData(
                    self._handle,
                    self.train_set.construct()._handle,
                )
            )
wxchan's avatar
wxchan committed
4129
            self.__inner_predict_buffer[0] = None
4130
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
4131
4132
        is_finished = ctypes.c_int(0)
        if fobj is None:
4133
            if self.__set_objective_to_none:
4134
4135
4136
4137
4138
4139
4140
                raise LightGBMError("Cannot update due to null objective function.")
            _safe_call(
                _LIB.LGBM_BoosterUpdateOneIter(
                    self._handle,
                    ctypes.byref(is_finished),
                )
            )
4141
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
4142
4143
            return is_finished.value == 1
        else:
4144
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
4145
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
4146
4147
4148
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

4149
4150
4151
    def __boost(
        self,
        grad: np.ndarray,
4152
        hess: np.ndarray,
4153
    ) -> bool:
4154
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
4155

Nikita Titov's avatar
Nikita Titov committed
4156
4157
        .. note::

4158
4159
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
4160
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
4161
            and grad and hess should be returned in the same format.
4162

wxchan's avatar
wxchan committed
4163
4164
        Parameters
        ----------
4165
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
4166
4167
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
4168
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
4169
4170
            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
4171
4172
4173

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4174
4175
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
4176
        """
4177
        if self.__num_class > 1:
4178
4179
4180
4181
            grad = grad.ravel(order="F")
            hess = hess.ravel(order="F")
        grad = _list_to_1d_numpy(grad, dtype=np.float32, name="gradient")
        hess = _list_to_1d_numpy(hess, dtype=np.float32, name="hessian")
4182
4183
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
4184
        if len(grad) != len(hess):
4185
4186
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
4187
        if len(grad) != num_train_data * self.__num_class:
4188
4189
4190
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
4191
                f"number of models per one iteration ({self.__num_class})"
4192
            )
wxchan's avatar
wxchan committed
4193
        is_finished = ctypes.c_int(0)
4194
4195
4196
4197
4198
4199
4200
4201
        _safe_call(
            _LIB.LGBM_BoosterUpdateOneIterCustom(
                self._handle,
                grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
                hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
                ctypes.byref(is_finished),
            )
        )
4202
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
4203
4204
        return is_finished.value == 1

4205
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
4206
4207
4208
4209
4210
4211
4212
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
4213
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(self._handle))
4214
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
4215
        return self
wxchan's avatar
wxchan committed
4216

4217
    def current_iteration(self) -> int:
4218
4219
4220
4221
4222
4223
4224
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
4225
        out_cur_iter = ctypes.c_int(0)
4226
4227
4228
4229
4230
4231
        _safe_call(
            _LIB.LGBM_BoosterGetCurrentIteration(
                self._handle,
                ctypes.byref(out_cur_iter),
            )
        )
wxchan's avatar
wxchan committed
4232
4233
        return out_cur_iter.value

4234
    def num_model_per_iteration(self) -> int:
4235
4236
4237
4238
4239
4240
4241
4242
        """Get number of models per iteration.

        Returns
        -------
        model_per_iter : int
            The number of models per iteration.
        """
        model_per_iter = ctypes.c_int(0)
4243
4244
4245
4246
4247
4248
        _safe_call(
            _LIB.LGBM_BoosterNumModelPerIteration(
                self._handle,
                ctypes.byref(model_per_iter),
            )
        )
4249
4250
        return model_per_iter.value

4251
    def num_trees(self) -> int:
4252
4253
4254
4255
4256
4257
4258
4259
        """Get number of weak sub-models.

        Returns
        -------
        num_trees : int
            The number of weak sub-models.
        """
        num_trees = ctypes.c_int(0)
4260
4261
4262
4263
4264
4265
        _safe_call(
            _LIB.LGBM_BoosterNumberOfTotalModel(
                self._handle,
                ctypes.byref(num_trees),
            )
        )
4266
4267
        return num_trees.value

4268
    def upper_bound(self) -> float:
4269
4270
4271
4272
        """Get upper bound value of a model.

        Returns
        -------
4273
        upper_bound : float
4274
4275
4276
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
4277
4278
4279
4280
4281
4282
        _safe_call(
            _LIB.LGBM_BoosterGetUpperBoundValue(
                self._handle,
                ctypes.byref(ret),
            )
        )
4283
4284
        return ret.value

4285
    def lower_bound(self) -> float:
4286
4287
4288
4289
        """Get lower bound value of a model.

        Returns
        -------
4290
        lower_bound : float
4291
4292
4293
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
4294
4295
4296
4297
4298
4299
        _safe_call(
            _LIB.LGBM_BoosterGetLowerBoundValue(
                self._handle,
                ctypes.byref(ret),
            )
        )
4300
4301
        return ret.value

4302
4303
4304
4305
    def eval(
        self,
        data: Dataset,
        name: str,
4306
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None,
4307
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4308
        """Evaluate for data.
wxchan's avatar
wxchan committed
4309
4310
4311

        Parameters
        ----------
4312
4313
        data : Dataset
            Data for the evaluating.
4314
        name : str
4315
            Name of the data.
4316
        feval : callable, list of callable, or None, optional (default=None)
4317
            Customized evaluation function.
4318
            Each evaluation function should accept two parameters: preds, eval_data,
4319
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4320

4321
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4322
                    The predicted values.
4323
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4324
                    If custom objective function is used, predicted values are returned before any transformation,
4325
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
4326
                eval_data : Dataset
4327
                    A ``Dataset`` to evaluate.
4328
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4329
                    The name of evaluation function (without whitespace).
4330
4331
4332
4333
4334
                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
4335
4336
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4337
        result : list
4338
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4339
        """
Guolin Ke's avatar
Guolin Ke committed
4340
4341
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
4342
4343
4344
4345
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
4346
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
4347
4348
4349
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
4350
        # need to push new valid data
wxchan's avatar
wxchan committed
4351
4352
4353
4354
4355
4356
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

4357
4358
    def eval_train(
        self,
4359
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None,
4360
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4361
        """Evaluate for training data.
wxchan's avatar
wxchan committed
4362
4363
4364

        Parameters
        ----------
4365
        feval : callable, list of callable, or None, optional (default=None)
4366
            Customized evaluation function.
4367
            Each evaluation function should accept two parameters: preds, eval_data,
4368
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4369

4370
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4371
                    The predicted values.
4372
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4373
                    If custom objective function is used, predicted values are returned before any transformation,
4374
                    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
4375
                eval_data : Dataset
4376
                    The training dataset.
4377
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4378
                    The name of evaluation function (without whitespace).
4379
4380
4381
4382
4383
                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
4384
4385
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4386
        result : list
4387
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4388
        """
4389
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
4390

4391
4392
    def eval_valid(
        self,
4393
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None,
4394
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4395
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
4396
4397
4398

        Parameters
        ----------
4399
        feval : callable, list of callable, or None, optional (default=None)
4400
            Customized evaluation function.
4401
            Each evaluation function should accept two parameters: preds, eval_data,
4402
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4403

4404
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4405
                    The predicted values.
4406
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4407
                    If custom objective function is used, predicted values are returned before any transformation,
4408
                    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
4409
                eval_data : Dataset
4410
                    The validation dataset.
4411
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4412
                    The name of evaluation function (without whitespace).
4413
4414
4415
4416
4417
                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
4418
4419
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4420
        result : list
4421
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4422
        """
4423
4424
4425
4426
4427
        return [
            item
            for i in range(1, self.__num_dataset)
            for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)
        ]
wxchan's avatar
wxchan committed
4428

4429
4430
4431
4432
4433
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
4434
        importance_type: str = "split",
4435
    ) -> "Booster":
4436
        """Save Booster to file.
wxchan's avatar
wxchan committed
4437
4438
4439

        Parameters
        ----------
4440
        filename : str or pathlib.Path
4441
            Filename to save Booster.
4442
4443
4444
4445
        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
4446
        start_iteration : int, optional (default=0)
4447
            Start index of the iteration that should be saved.
4448
        importance_type : str, optional (default="split")
4449
4450
4451
            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
4452
4453
4454
4455
4456

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
4457
        """
4458
        if num_iteration is None:
4459
            num_iteration = self.best_iteration
4460
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4461
4462
4463
4464
4465
4466
4467
4468
4469
        _safe_call(
            _LIB.LGBM_BoosterSaveModel(
                self._handle,
                ctypes.c_int(start_iteration),
                ctypes.c_int(num_iteration),
                ctypes.c_int(importance_type_int),
                _c_str(str(filename)),
            )
        )
4470
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
4471
        return self
wxchan's avatar
wxchan committed
4472

4473
4474
4475
    def shuffle_models(
        self,
        start_iteration: int = 0,
4476
        end_iteration: int = -1,
4477
    ) -> "Booster":
4478
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
4479

4480
4481
4482
        Parameters
        ----------
        start_iteration : int, optional (default=0)
4483
            The first iteration that will be shuffled.
4484
4485
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
4486
            If <= 0, means the last available iteration.
4487

Nikita Titov's avatar
Nikita Titov committed
4488
4489
4490
4491
        Returns
        -------
        self : Booster
            Booster with shuffled models.
4492
        """
4493
4494
4495
4496
4497
4498
4499
        _safe_call(
            _LIB.LGBM_BoosterShuffleModels(
                self._handle,
                ctypes.c_int(start_iteration),
                ctypes.c_int(end_iteration),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
4500
        return self
4501

4502
    def model_from_string(self, model_str: str) -> "Booster":
4503
4504
4505
4506
        """Load Booster from a string.

        Parameters
        ----------
4507
        model_str : str
4508
4509
4510
4511
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4512
        self : Booster
4513
4514
            Loaded Booster object.
        """
4515
4516
4517
        # ensure that existing Booster is freed before replacing it
        # with a new one createdfrom file
        _safe_call(_LIB.LGBM_BoosterFree(self._handle))
4518
        self._free_buffer()
4519
        self._handle = ctypes.c_void_p()
4520
        out_num_iterations = ctypes.c_int(0)
4521
4522
4523
4524
4525
4526
4527
        _safe_call(
            _LIB.LGBM_BoosterLoadModelFromString(
                _c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(self._handle),
            )
        )
4528
        out_num_class = ctypes.c_int(0)
4529
4530
4531
4532
4533
4534
        _safe_call(
            _LIB.LGBM_BoosterGetNumClasses(
                self._handle,
                ctypes.byref(out_num_class),
            )
        )
4535
        self.__num_class = out_num_class.value
4536
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
4537
4538
        return self

4539
4540
4541
4542
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
4543
        importance_type: str = "split",
4544
    ) -> str:
4545
        """Save Booster to string.
4546

4547
4548
4549
4550
4551
4552
        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
4553
        start_iteration : int, optional (default=0)
4554
            Start index of the iteration that should be saved.
4555
        importance_type : str, optional (default="split")
4556
4557
4558
            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.
4559
4560
4561

        Returns
        -------
4562
        str_repr : str
4563
4564
            String representation of Booster.
        """
4565
        if num_iteration is None:
4566
            num_iteration = self.best_iteration
4567
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4568
        buffer_len = 1 << 20
4569
        tmp_out_len = ctypes.c_int64(0)
4570
        string_buffer = ctypes.create_string_buffer(buffer_len)
4571
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4572
4573
        _safe_call(
            _LIB.LGBM_BoosterSaveModelToString(
4574
                self._handle,
4575
                ctypes.c_int(start_iteration),
4576
                ctypes.c_int(num_iteration),
4577
                ctypes.c_int(importance_type_int),
4578
                ctypes.c_int64(buffer_len),
4579
                ctypes.byref(tmp_out_len),
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
                ptr_string_buffer,
            )
        )
        actual_len = tmp_out_len.value
        # if buffer length is not long enough, re-allocate a buffer
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
            _safe_call(
                _LIB.LGBM_BoosterSaveModelToString(
                    self._handle,
                    ctypes.c_int(start_iteration),
                    ctypes.c_int(num_iteration),
                    ctypes.c_int(importance_type_int),
                    ctypes.c_int64(actual_len),
                    ctypes.byref(tmp_out_len),
                    ptr_string_buffer,
                )
            )
        ret = string_buffer.value.decode("utf-8")
4600
4601
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
4602

4603
4604
4605
4606
    def dump_model(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
4607
4608
        importance_type: str = "split",
        object_hook: Optional[Callable[[Dict[str, Any]], Dict[str, Any]]] = None,
4609
    ) -> Dict[str, Any]:
Nikita Titov's avatar
Nikita Titov committed
4610
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
4611

4612
4613
        Parameters
        ----------
4614
4615
4616
4617
        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
4618
        start_iteration : int, optional (default=0)
4619
            Start index of the iteration that should be dumped.
4620
        importance_type : str, optional (default="split")
4621
4622
4623
            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.
4624
4625
4626
4627
4628
4629
4630
4631
4632
        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.
4633

wxchan's avatar
wxchan committed
4634
4635
        Returns
        -------
4636
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
4637
            JSON format of Booster.
wxchan's avatar
wxchan committed
4638
        """
4639
        if num_iteration is None:
4640
            num_iteration = self.best_iteration
4641
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4642
        buffer_len = 1 << 20
4643
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
4644
        string_buffer = ctypes.create_string_buffer(buffer_len)
4645
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4646
4647
        _safe_call(
            _LIB.LGBM_BoosterDumpModel(
4648
                self._handle,
4649
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4650
                ctypes.c_int(num_iteration),
4651
                ctypes.c_int(importance_type_int),
4652
                ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
4653
                ctypes.byref(tmp_out_len),
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
                ptr_string_buffer,
            )
        )
        actual_len = tmp_out_len.value
        # if buffer length is not long enough, reallocate a buffer
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
            _safe_call(
                _LIB.LGBM_BoosterDumpModel(
                    self._handle,
                    ctypes.c_int(start_iteration),
                    ctypes.c_int(num_iteration),
                    ctypes.c_int(importance_type_int),
                    ctypes.c_int64(actual_len),
                    ctypes.byref(tmp_out_len),
                    ptr_string_buffer,
                )
            )
        ret = json.loads(string_buffer.value.decode("utf-8"), object_hook=object_hook)
        ret["pandas_categorical"] = json.loads(
            json.dumps(
                self.pandas_categorical,
                default=_json_default_with_numpy,
            )
        )
4680
        return ret
wxchan's avatar
wxchan committed
4681

4682
4683
    def predict(
        self,
4684
        data: _LGBM_PredictDataType,
4685
4686
4687
4688
4689
4690
4691
        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,
4692
        **kwargs: Any,
4693
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4694
        """Make a prediction.
wxchan's avatar
wxchan committed
4695
4696
4697

        Parameters
        ----------
4698
        data : str, pathlib.Path, numpy array, pandas DataFrame, pyarrow Table, H2O DataTable's Frame or scipy.sparse
4699
            Data source for prediction.
4700
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4701
        start_iteration : int, optional (default=0)
4702
            Start index of the iteration to predict.
4703
            If <= 0, starts from the first iteration.
4704
        num_iteration : int or None, optional (default=None)
4705
4706
4707
4708
            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).
4709
4710
4711
4712
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4713
4714
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4715

Nikita Titov's avatar
Nikita Titov committed
4716
4717
4718
4719
4720
4721
4722
            .. 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.
4723

4724
4725
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4726
            Used only if data is str.
4727
4728
4729
        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.
4730
4731
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4732
4733
4734

        Returns
        -------
4735
        result : numpy array, scipy.sparse or list of scipy.sparse
4736
            Prediction result.
4737
            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
4738
        """
4739
4740
4741
4742
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs),
        )
4743
        if num_iteration is None:
4744
            if start_iteration <= 0:
4745
4746
4747
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4748
4749
4750
4751
4752
4753
4754
4755
        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,
4756
            validate_features=validate_features,
4757
        )
wxchan's avatar
wxchan committed
4758

4759
4760
    def refit(
        self,
4761
        data: _LGBM_TrainDataType,
4762
        label: _LGBM_LabelType,
4763
4764
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4765
4766
4767
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4768
4769
        feature_name: _LGBM_FeatureNameConfiguration = "auto",
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = "auto",
4770
4771
4772
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4773
        **kwargs: Any,
4774
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4775
4776
4777
4778
        """Refit the existing Booster by new data.

        Parameters
        ----------
4779
        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
4780
            Data source for refit.
4781
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4782
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array or pyarrow ChunkedArray
Guolin Ke's avatar
Guolin Ke committed
4783
4784
            Label for refit.
        decay_rate : float, optional (default=0.9)
4785
4786
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4787
4788
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
4789
4790
4791

            .. versionadded:: 4.0.0

4792
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
4793
            Weight for each ``data`` instance. Weights should be non-negative.
4794
4795
4796

            .. versionadded:: 4.0.0

4797
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
4798
4799
4800
4801
4802
            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.
4803
4804
4805

            .. versionadded:: 4.0.0

4806
        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)
4807
            Init score for ``data``.
4808
4809
4810

            .. versionadded:: 4.0.0

4811
4812
4813
        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.
4814
4815
4816

            .. versionadded:: 4.0.0

4817
4818
4819
4820
4821
        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.
4822
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4823
4824
4825
            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.
4826
            Floating point numbers in categorical features will be rounded towards 0.
4827
4828
4829

            .. versionadded:: 4.0.0

4830
4831
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
4832
4833
4834

            .. versionadded:: 4.0.0

4835
4836
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
4837
4838
4839

            .. versionadded:: 4.0.0

4840
4841
4842
        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.
4843
4844
4845

            .. versionadded:: 4.0.0

4846
4847
        **kwargs
            Other parameters for refit.
4848
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4849
4850
4851
4852
4853
4854

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4855
        if self.__set_objective_to_none:
4856
            raise LightGBMError("Cannot refit due to null objective function.")
4857
4858
        if dataset_params is None:
            dataset_params = {}
4859
        predictor = _InnerPredictor.from_booster(booster=self, pred_parameter=deepcopy(kwargs))
4860
        leaf_preds: np.ndarray = predictor.predict(  # type: ignore[assignment]
4861
4862
4863
            data=data,
            start_iteration=-1,
            pred_leaf=True,
4864
            validate_features=validate_features,
4865
        )
4866
        nrow, ncol = leaf_preds.shape
4867
        out_is_linear = ctypes.c_int(0)
4868
4869
4870
4871
4872
4873
        _safe_call(
            _LIB.LGBM_BoosterGetLinear(
                self._handle,
                ctypes.byref(out_is_linear),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
4874
4875
4876
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
4877
            default_value=None,
Nikita Titov's avatar
Nikita Titov committed
4878
        )
4879
        new_params["linear_tree"] = bool(out_is_linear.value)
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
        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,
        )
4893
        new_params["refit_decay_rate"] = decay_rate
4894
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4895
        # Copy models
4896
4897
4898
4899
4900
4901
        _safe_call(
            _LIB.LGBM_BoosterMerge(
                new_booster._handle,
                predictor._handle,
            )
        )
Guolin Ke's avatar
Guolin Ke committed
4902
        leaf_preds = leaf_preds.reshape(-1)
4903
        ptr_data, _, _ = _c_int_array(leaf_preds)
4904
4905
4906
4907
4908
4909
4910
4911
        _safe_call(
            _LIB.LGBM_BoosterRefit(
                new_booster._handle,
                ptr_data,
                ctypes.c_int32(nrow),
                ctypes.c_int32(ncol),
            )
        )
4912
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4913
4914
        return new_booster

4915
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
        """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.
        """
4930
        ret = ctypes.c_double(0)
4931
4932
4933
4934
4935
4936
4937
4938
        _safe_call(
            _LIB.LGBM_BoosterGetLeafValue(
                self._handle,
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.byref(ret),
            )
        )
4939
4940
        return ret.value

4941
4942
4943
4944
4945
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
4946
    ) -> "Booster":
4947
4948
        """Set the output of a leaf.

4949
4950
        .. versionadded:: 4.0.0

4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
        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(
4967
                self._handle,
4968
4969
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
4970
                ctypes.c_double(value),
4971
4972
4973
4974
            )
        )
        return self

4975
    def num_feature(self) -> int:
4976
4977
4978
4979
4980
4981
4982
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4983
        out_num_feature = ctypes.c_int(0)
4984
4985
4986
4987
4988
4989
        _safe_call(
            _LIB.LGBM_BoosterGetNumFeature(
                self._handle,
                ctypes.byref(out_num_feature),
            )
        )
4990
4991
        return out_num_feature.value

4992
    def feature_name(self) -> List[str]:
4993
        """Get names of features.
wxchan's avatar
wxchan committed
4994
4995
4996

        Returns
        -------
4997
        result : list of str
4998
            List with names of features.
wxchan's avatar
wxchan committed
4999
        """
5000
        num_feature = self.num_feature()
5001
        # Get name of features
wxchan's avatar
wxchan committed
5002
        tmp_out_len = ctypes.c_int(0)
5003
5004
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
5005
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
5006
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
        _safe_call(
            _LIB.LGBM_BoosterGetFeatureNames(
                self._handle,
                ctypes.c_int(num_feature),
                ctypes.byref(tmp_out_len),
                ctypes.c_size_t(reserved_string_buffer_size),
                ctypes.byref(required_string_buffer_size),
                ptr_string_buffers,
            )
        )
wxchan's avatar
wxchan committed
5017
5018
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
5019
5020
5021
5022
        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)]
5023
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
            _safe_call(
                _LIB.LGBM_BoosterGetFeatureNames(
                    self._handle,
                    ctypes.c_int(num_feature),
                    ctypes.byref(tmp_out_len),
                    ctypes.c_size_t(actual_string_buffer_size),
                    ctypes.byref(required_string_buffer_size),
                    ptr_string_buffers,
                )
            )
        return [string_buffers[i].value.decode("utf-8") for i in range(num_feature)]
wxchan's avatar
wxchan committed
5035

5036
5037
    def feature_importance(
        self,
5038
5039
        importance_type: str = "split",
        iteration: Optional[int] = None,
5040
    ) -> np.ndarray:
5041
        """Get feature importances.
5042

5043
5044
        Parameters
        ----------
5045
        importance_type : str, optional (default="split")
5046
5047
5048
            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.
5049
5050
5051
5052
        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).
5053

5054
5055
        Returns
        -------
5056
5057
        result : numpy array
            Array with feature importances.
5058
        """
5059
5060
        if iteration is None:
            iteration = self.best_iteration
5061
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
5062
        result = np.empty(self.num_feature(), dtype=np.float64)
5063
5064
5065
5066
5067
5068
5069
5070
        _safe_call(
            _LIB.LGBM_BoosterFeatureImportance(
                self._handle,
                ctypes.c_int(iteration),
                ctypes.c_int(importance_type_int),
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double)),
            )
        )
5071
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
5072
            return result.astype(np.int32)
5073
5074
        else:
            return result
5075

5076
5077
5078
5079
    def get_split_value_histogram(
        self,
        feature: Union[int, str],
        bins: Optional[Union[int, str]] = None,
5080
        xgboost_style: bool = False,
5081
    ) -> Union[Tuple[np.ndarray, np.ndarray], np.ndarray, pd_DataFrame]:
5082
5083
5084
5085
        """Get split value histogram for the specified feature.

        Parameters
        ----------
5086
        feature : int or str
5087
5088
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
5089
            If str, interpreted as name.
5090

Nikita Titov's avatar
Nikita Titov committed
5091
5092
5093
            .. warning::

                Categorical features are not supported.
5094

5095
        bins : int, str or None, optional (default=None)
5096
5097
5098
            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.
5099
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
        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.
        """
5114

5115
        def add(root: Dict[str, Any]) -> None:
5116
            """Recursively add thresholds."""
5117
            if "split_index" in root:  # non-leaf
5118
                if feature_names is not None and isinstance(feature, str):
5119
                    split_feature = feature_names[root["split_feature"]]
5120
                else:
5121
                    split_feature = root["split_feature"]
5122
                if split_feature == feature:
5123
5124
                    if isinstance(root["threshold"], str):
                        raise LightGBMError("Cannot compute split value histogram for the categorical feature")
5125
                    else:
5126
5127
5128
                        values.append(root["threshold"])
                add(root["left_child"])
                add(root["right_child"])
5129
5130

        model = self.dump_model()
5131
5132
        feature_names = model.get("feature_names")
        tree_infos = model["tree_info"]
5133
        values: List[float] = []
5134
        for tree_info in tree_infos:
5135
            add(tree_info["tree_structure"])
5136

5137
        if bins is None or isinstance(bins, int) and xgboost_style:
5138
5139
5140
5141
5142
5143
5144
            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:
5145
                return pd_DataFrame(ret, columns=["SplitValue", "Count"])
5146
5147
5148
5149
5150
            else:
                return ret
        else:
            return hist, bin_edges

5151
5152
5153
5154
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
5155
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]],
5156
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
5157
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
5158
        if data_idx >= self.__num_dataset:
5159
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
5160
5161
5162
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
5163
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
5164
            tmp_out_len = ctypes.c_int(0)
5165
5166
5167
5168
5169
5170
5171
5172
            _safe_call(
                _LIB.LGBM_BoosterGetEval(
                    self._handle,
                    ctypes.c_int(data_idx),
                    ctypes.byref(tmp_out_len),
                    result.ctypes.data_as(ctypes.POINTER(ctypes.c_double)),
                )
            )
wxchan's avatar
wxchan committed
5173
            if tmp_out_len.value != self.__num_inner_eval:
5174
                raise ValueError("Wrong length of eval results")
5175
            for i in range(self.__num_inner_eval):
5176
                ret.append((data_name, self.__name_inner_eval[i], result[i], self.__higher_better_inner_eval[i]))
5177
5178
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
5179
5180
5181
5182
5183
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
5184
5185
5186
5187
5188
5189
5190
5191
5192
            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
5193
5194
5195
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

5196
    def __inner_predict(self, data_idx: int) -> np.ndarray:
5197
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
5198
        if data_idx >= self.__num_dataset:
5199
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
5200
5201
5202
5203
5204
        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
5205
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
5206
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
5207
5208
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
5209
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))  # type: ignore[union-attr]
5210
5211
5212
5213
5214
5215
5216
5217
            _safe_call(
                _LIB.LGBM_BoosterGetPredict(
                    self._handle,
                    ctypes.c_int(data_idx),
                    ctypes.byref(tmp_out_len),
                    data_ptr,
                )
            )
5218
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):  # type: ignore[arg-type]
5219
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
5220
            self.__is_predicted_cur_iter[data_idx] = True
5221
        result: np.ndarray = self.__inner_predict_buffer[data_idx]  # type: ignore[assignment]
5222
5223
        if self.__num_class > 1:
            num_data = result.size // self.__num_class
5224
            result = result.reshape(num_data, self.__num_class, order="F")
5225
        return result
wxchan's avatar
wxchan committed
5226

5227
    def __get_eval_info(self) -> None:
5228
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
5229
5230
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
5231
            out_num_eval = ctypes.c_int(0)
5232
            # Get num of inner evals
5233
5234
5235
5236
5237
5238
            _safe_call(
                _LIB.LGBM_BoosterGetEvalCounts(
                    self._handle,
                    ctypes.byref(out_num_eval),
                )
            )
wxchan's avatar
wxchan committed
5239
5240
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
5241
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
5242
                tmp_out_len = ctypes.c_int(0)
5243
5244
5245
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
5246
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
5247
                ]
5248
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
                _safe_call(
                    _LIB.LGBM_BoosterGetEvalNames(
                        self._handle,
                        ctypes.c_int(self.__num_inner_eval),
                        ctypes.byref(tmp_out_len),
                        ctypes.c_size_t(reserved_string_buffer_size),
                        ctypes.byref(required_string_buffer_size),
                        ptr_string_buffers,
                    )
                )
wxchan's avatar
wxchan committed
5259
                if self.__num_inner_eval != tmp_out_len.value:
5260
                    raise ValueError("Length of eval names doesn't equal with num_evals")
5261
5262
5263
5264
5265
5266
                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)
                    ]
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
                    ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(
                        *map(ctypes.addressof, string_buffers)
                    )  # type: ignore[misc]
                    _safe_call(
                        _LIB.LGBM_BoosterGetEvalNames(
                            self._handle,
                            ctypes.c_int(self.__num_inner_eval),
                            ctypes.byref(tmp_out_len),
                            ctypes.c_size_t(actual_string_buffer_size),
                            ctypes.byref(required_string_buffer_size),
                            ptr_string_buffers,
                        )
                    )
                self.__name_inner_eval = [string_buffers[i].value.decode("utf-8") for i in range(self.__num_inner_eval)]
5281
                self.__higher_better_inner_eval = [
5282
                    name.startswith(("auc", "ndcg@", "map@", "average_precision")) for name in self.__name_inner_eval
5283
                ]