basic.py 200 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3

4
5
6
7
8
9
10
# This import causes lib_lightgbm.{dll,dylib,so} to be loaded.
# It's intentionally done here, as early as possible, to avoid issues like
# "libgomp.so.1: cannot allocate memory in static TLS block" on aarch64 Linux.
#
# For details, see the "cannot allocate memory in static TLS block" entry in docs/FAQ.rst.
from .libpath import _LIB  # isort: skip

11
import abc
wxchan's avatar
wxchan committed
12
import ctypes
13
import inspect
14
import json
wxchan's avatar
wxchan committed
15
import warnings
16
from collections import OrderedDict
17
from copy import deepcopy
18
from enum import Enum
19
from functools import wraps
20
from os import SEEK_END, environ
21
22
from os.path import getsize
from pathlib import Path
23
from tempfile import NamedTemporaryFile
24
from typing import TYPE_CHECKING, Any, Callable, Dict, Iterable, Iterator, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
25
26
27
28

import numpy as np
import scipy.sparse

29
30
31
32
from .compat import (
    PANDAS_INSTALLED,
    PYARROW_INSTALLED,
    arrow_cffi,
33
    arrow_is_boolean,
34
35
36
37
38
39
40
41
42
43
44
45
46
    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
47

48
49
50
if TYPE_CHECKING:
    from typing import Literal

51
52
53
54
55
56
57
    # typing.TypeGuard was only introduced in Python 3.10
    try:
        from typing import TypeGuard
    except ImportError:
        from typing_extensions import TypeGuard


58
__all__ = [
59
60
61
62
63
64
    "Booster",
    "Dataset",
    "LGBMDeprecationWarning",
    "LightGBMError",
    "register_logger",
    "Sequence",
65
66
]

67
_BoosterHandle = ctypes.c_void_p
68
_DatasetHandle = ctypes.c_void_p
69
70
_ctypes_int_ptr = Union[
    "ctypes._Pointer[ctypes.c_int32]",
71
    "ctypes._Pointer[ctypes.c_int64]",
72
]
73
74
_ctypes_int_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_int32]]",
75
    "ctypes.Array[ctypes._Pointer[ctypes.c_int64]]",
76
]
77
78
_ctypes_float_ptr = Union[
    "ctypes._Pointer[ctypes.c_float]",
79
    "ctypes._Pointer[ctypes.c_double]",
80
81
82
]
_ctypes_float_array = Union[
    "ctypes.Array[ctypes._Pointer[ctypes.c_float]]",
83
    "ctypes.Array[ctypes._Pointer[ctypes.c_double]]",
84
]
85
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
86
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
87
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
88
_LGBM_BoosterEvalMethodResultWithStandardDeviationType = Tuple[str, str, float, bool, float]
89
90
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], "Literal['auto']"]
_LGBM_FeatureNameConfiguration = Union[List[str], "Literal['auto']"]
91
92
93
94
_LGBM_GroupType = Union[
    List[float],
    List[int],
    np.ndarray,
95
96
97
    pd_Series,
    pa_Array,
    pa_ChunkedArray,
98
]
99
100
_LGBM_PositionType = Union[
    np.ndarray,
101
    pd_Series,
102
]
103
104
105
106
107
108
_LGBM_InitScoreType = Union[
    List[float],
    List[List[float]],
    np.ndarray,
    pd_Series,
    pd_DataFrame,
109
110
111
    pa_Table,
    pa_Array,
    pa_ChunkedArray,
112
]
113
114
115
116
117
118
119
120
121
_LGBM_TrainDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
    scipy.sparse.spmatrix,
    "Sequence",
    List["Sequence"],
122
    List[np.ndarray],
123
    pa_Table,
124
]
125
_LGBM_LabelType = Union[
126
127
    List[float],
    List[int],
128
129
    np.ndarray,
    pd_Series,
130
131
132
    pd_DataFrame,
    pa_Array,
    pa_ChunkedArray,
133
]
134
135
136
137
138
139
_LGBM_PredictDataType = Union[
    str,
    Path,
    np.ndarray,
    pd_DataFrame,
    dt_DataTable,
140
141
    scipy.sparse.spmatrix,
    pa_Table,
142
]
143
144
145
146
_LGBM_WeightType = Union[
    List[float],
    List[int],
    np.ndarray,
147
148
149
    pd_Series,
    pa_Array,
    pa_ChunkedArray,
150
]
151
152
153
154
155
156
157
158
159
160
161
162
163
_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,
]

164
165
ZERO_THRESHOLD = 1e-35

166
167
_MULTICLASS_OBJECTIVES = {"multiclass", "multiclassova", "multiclass_ova", "ova", "ovr", "softmax"}

168

169
170
171
172
173
174
class LightGBMError(Exception):
    """Error thrown by LightGBM."""

    pass


175
176
177
178
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


179
def _get_sample_count(total_nrow: int, params: str) -> int:
180
    sample_cnt = ctypes.c_int(0)
181
182
183
184
185
186
187
    _safe_call(
        _LIB.LGBM_GetSampleCount(
            ctypes.c_int32(total_nrow),
            _c_str(params),
            ctypes.byref(sample_cnt),
        )
    )
188
189
    return sample_cnt.value

wxchan's avatar
wxchan committed
190

191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
def _np2d_to_np1d(mat: np.ndarray) -> Tuple[np.ndarray, int]:
    if mat.dtype in (np.float32, np.float64):
        dtype = mat.dtype
    else:
        dtype = np.float32
    if mat.flags["F_CONTIGUOUS"]:
        order = "F"
        layout = _C_API_IS_COL_MAJOR
    else:
        order = "C"
        layout = _C_API_IS_ROW_MAJOR
    # ensure dtype and order, copies if either do not match
    data = np.asarray(mat, dtype=dtype, order=order)
    # flatten array without copying
    return data.ravel(order=order), layout


208
class _MissingType(Enum):
209
210
211
    NONE = "None"
    NAN = "NaN"
    ZERO = "Zero"
212
213


214
class _DummyLogger:
215
    def info(self, msg: str) -> None:
216
        print(msg)  # noqa: T201
217

218
    def warning(self, msg: str) -> None:
219
220
221
        warnings.warn(msg, stacklevel=3)


222
223
224
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
225
226


227
228
229
230
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


231
def register_logger(
232
233
234
    logger: Any,
    info_method_name: str = "info",
    warning_method_name: str = "warning",
235
) -> None:
236
237
238
239
    """Register custom logger.

    Parameters
    ----------
240
    logger : Any
241
        Custom logger.
242
243
244
245
    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.
246
    """
247
    if not _has_method(logger, info_method_name) or not _has_method(logger, warning_method_name):
248
        raise TypeError(f"Logger must provide '{info_method_name}' and '{warning_method_name}' method")
249
250

    global _LOGGER, _INFO_METHOD_NAME, _WARNING_METHOD_NAME
251
    _LOGGER = logger
252
253
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
254
255


256
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
257
    """Join log messages from native library which come by chunks."""
258
    msg_normalized: List[str] = []
259
260

    @wraps(func)
261
    def wrapper(msg: str) -> None:
262
        nonlocal msg_normalized
263
264
        if msg.strip() == "":
            msg = "".join(msg_normalized)
265
266
267
268
269
270
271
272
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


273
def _log_info(msg: str) -> None:
274
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
275
276


277
def _log_warning(msg: str) -> None:
278
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
279
280
281


@_normalize_native_string
282
def _log_native(msg: str) -> None:
283
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
284
285


286
def _log_callback(msg: bytes) -> None:
287
    """Redirect logs from native library into Python."""
288
    _log_native(str(msg.decode("utf-8")))
289
290


291
292
293
# connect the Python logger to logging in lib_lightgbm
if not environ.get("LIGHTGBM_BUILD_DOC", False):
    _LIB.LGBM_GetLastError.restype = ctypes.c_char_p
294
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
295
296
297
    _LIB.callback = callback(_log_callback)  # type: ignore[attr-defined]
    if _LIB.LGBM_RegisterLogCallback(_LIB.callback) != 0:
        raise LightGBMError(_LIB.LGBM_GetLastError().decode("utf-8"))
wxchan's avatar
wxchan committed
298

wxchan's avatar
wxchan committed
299

300
_NUMERIC_TYPES = (int, float, bool)
301
302


303
def _safe_call(ret: int) -> None:
304
305
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
306
307
308
    Parameters
    ----------
    ret : int
309
        The return value from C API calls.
wxchan's avatar
wxchan committed
310
311
    """
    if ret != 0:
312
        raise LightGBMError(_LIB.LGBM_GetLastError().decode("utf-8"))
wxchan's avatar
wxchan committed
313

wxchan's avatar
wxchan committed
314

315
def _is_numeric(obj: Any) -> bool:
316
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
317
318
319
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
320
321
322
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
323
324
        return False

wxchan's avatar
wxchan committed
325

326
def _is_numpy_1d_array(data: Any) -> bool:
327
    """Check whether data is a numpy 1-D array."""
328
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
329

wxchan's avatar
wxchan committed
330

331
def _is_numpy_column_array(data: Any) -> bool:
332
333
334
335
336
337
338
    """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


339
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: "np.typing.DTypeLike") -> np.ndarray:
340
    """Cast numpy array to given dtype."""
341
342
343
344
345
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


346
def _is_1d_list(data: Any) -> bool:
347
    """Check whether data is a 1-D list."""
348
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
349

wxchan's avatar
wxchan committed
350

351
def _is_list_of_numpy_arrays(data: Any) -> "TypeGuard[List[np.ndarray]]":
352
    return isinstance(data, list) and all(isinstance(x, np.ndarray) for x in data)
353
354
355


def _is_list_of_sequences(data: Any) -> "TypeGuard[List[Sequence]]":
356
    return isinstance(data, list) and all(isinstance(x, Sequence) for x in data)
357
358


359
360
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
361
    return _is_numpy_1d_array(data) or _is_numpy_column_array(data) or _is_1d_list(data) or isinstance(data, pd_Series)
362
363


364
365
def _list_to_1d_numpy(
    data: Any,
366
    dtype: "np.typing.DTypeLike",
367
    name: str,
368
) -> np.ndarray:
369
    """Convert data to numpy 1-D array."""
370
    if _is_numpy_1d_array(data):
371
        return _cast_numpy_array_to_dtype(data, dtype)
372
    elif _is_numpy_column_array(data):
373
        _log_warning("Converting column-vector to 1d array")
374
        array = data.ravel()
375
        return _cast_numpy_array_to_dtype(array, dtype)
376
    elif _is_1d_list(data):
377
        return np.asarray(data, dtype=dtype)
378
    elif isinstance(data, pd_Series):
379
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
380
        return np.asarray(data, dtype=dtype)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
381
    else:
382
383
384
        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
385

wxchan's avatar
wxchan committed
386

387
388
389
390
391
392
393
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."""
394
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
395
396
397
398


def _is_2d_collection(data: Any) -> bool:
    """Check whether data is a 2-D collection."""
399
    return _is_numpy_2d_array(data) or _is_2d_list(data) or isinstance(data, pd_DataFrame)
400
401


402
def _is_pyarrow_array(data: Any) -> "TypeGuard[Union[pa_Array, pa_ChunkedArray]]":
403
404
405
406
    """Check whether data is a PyArrow array."""
    return isinstance(data, (pa_Array, pa_ChunkedArray))


407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
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
438
439
440
441
442
    if isinstance(data, pa_Array):
        export_objects = [data]
    elif isinstance(data, pa_ChunkedArray):
        export_objects = data.chunks
    elif isinstance(data, pa_Table):
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
        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)


463
464
def _data_to_2d_numpy(
    data: Any,
465
    dtype: "np.typing.DTypeLike",
466
    name: str,
467
) -> np.ndarray:
468
469
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
470
        return _cast_numpy_array_to_dtype(data, dtype)
471
472
473
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
474
        _check_for_bad_pandas_dtypes(data.dtypes)
475
        return _cast_numpy_array_to_dtype(data.values, dtype)
476
477
478
479
    raise TypeError(
        f"Wrong type({type(data).__name__}) for {name}.\n"
        "It should be list of lists, numpy 2-D array or pandas DataFrame"
    )
480
481


482
def _cfloat32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
483
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
484
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
485
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
486
    else:
487
        raise RuntimeError("Expected float pointer")
wxchan's avatar
wxchan committed
488

Guolin Ke's avatar
Guolin Ke committed
489

490
def _cfloat64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
491
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
492
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
493
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
494
    else:
495
        raise RuntimeError("Expected double pointer")
Guolin Ke's avatar
Guolin Ke committed
496

wxchan's avatar
wxchan committed
497

498
def _cint32_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
499
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
500
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
501
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
502
    else:
503
        raise RuntimeError("Expected int32 pointer")
504
505


506
def _cint64_array_to_numpy(*, cptr: "ctypes._Pointer", length: int) -> np.ndarray:
507
508
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
509
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
510
    else:
511
        raise RuntimeError("Expected int64 pointer")
wxchan's avatar
wxchan committed
512

wxchan's avatar
wxchan committed
513

514
def _c_str(string: str) -> ctypes.c_char_p:
515
    """Convert a Python string to C string."""
516
    return ctypes.c_char_p(string.encode("utf-8"))
wxchan's avatar
wxchan committed
517

wxchan's avatar
wxchan committed
518

519
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
520
    """Convert a Python array to C array."""
521
    return (ctype * len(values))(*values)  # type: ignore[operator]
wxchan's avatar
wxchan committed
522

wxchan's avatar
wxchan committed
523

524
def _json_default_with_numpy(obj: Any) -> Any:
525
526
527
528
529
530
531
532
533
    """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


534
535
536
537
538
539
540
541
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)


542
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
543
    """Convert Python dictionary to string, which is passed to C API."""
544
    if data is None or not data:
wxchan's avatar
wxchan committed
545
546
547
        return ""
    pairs = []
    for key, val in data.items():
548
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
549
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
550
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
551
            pairs.append(f"{key}={val}")
552
        elif val is not None:
553
554
            raise TypeError(f"Unknown type of parameter:{key}, got:{type(val).__name__}")
    return " ".join(pairs)
555

wxchan's avatar
wxchan committed
556

557
class _TempFile:
558
559
    """Proxy class to workaround errors on Windows."""

560
    def __enter__(self) -> "_TempFile":
561
562
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
563
            self.path = Path(self.name)
564
        return self
wxchan's avatar
wxchan committed
565

566
    def __exit__(self, exc_type: Any, exc_val: Any, exc_tb: Any) -> None:
567
568
        if self.path.is_file():
            self.path.unlink()
569

wxchan's avatar
wxchan committed
570

571
# DeprecationWarning is not shown by default, so let's create our own with higher level
572
573
# ref: https://peps.python.org/pep-0565/#additional-use-case-for-futurewarning
class LGBMDeprecationWarning(FutureWarning):
574
575
576
577
578
    """Custom deprecation warning."""

    pass


579
580
581
582
583
584
585
586
587
def _emit_datatable_deprecation_warning() -> None:
    msg = (
        "Support for 'datatable' in LightGBM is deprecated, and will be removed in a future release. "
        "To avoid this warning, convert 'datatable' inputs to a supported format "
        "(for example, use the 'to_numpy()' method)."
    )
    warnings.warn(msg, category=LGBMDeprecationWarning, stacklevel=2)


588
class _ConfigAliases:
589
590
591
592
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
593
    def _get_all_param_aliases() -> Dict[str, List[str]]:
594
595
596
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
597
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
598
599
600
601
602
603
604
        _safe_call(
            _LIB.LGBM_DumpParamAliases(
                ctypes.c_int64(buffer_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer,
            )
        )
605
606
607
608
        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)
609
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
610
611
612
613
614
615
616
            _safe_call(
                _LIB.LGBM_DumpParamAliases(
                    ctypes.c_int64(actual_len),
                    ctypes.byref(tmp_out_len),
                    ptr_string_buffer,
                )
            )
617
        return json.loads(
618
            string_buffer.value.decode("utf-8"), object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
619
        )
620
621

    @classmethod
622
    def get(cls, *args: str) -> Set[str]:
623
624
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
625
626
        ret = set()
        for i in args:
627
            ret.update(cls.get_sorted(i))
628
629
        return ret

630
631
632
633
634
635
    @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])

636
    @classmethod
637
    def get_by_alias(cls, *args: str) -> Set[str]:
638
639
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
640
641
642
643
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
644
                    ret.update(aliases)
645
646
647
                    break
        return ret

648

649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
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)

670
671
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
672
673

    # if main_param_name was provided, keep that value and remove all aliases
674
    if main_param_name in params.keys():
675
676
677
        for param in aliases:
            params.pop(param, None)
        return params
678

679
680
681
682
683
    # 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
684

685
686
687
688
689
690
691
    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
692
693
694
695

    return params


696
_MAX_INT32 = (1 << 31) - 1
697

698
"""Macro definition of data type in C API of LightGBM"""
699
700
701
702
_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
703

704
705
"""Macro definition of data order in matrix"""
_C_API_IS_COL_MAJOR = 0
706
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
707

708
"""Macro definition of prediction type in C API of LightGBM"""
709
710
711
712
_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
713

714
"""Macro definition of sparse matrix type"""
715
716
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
717

718
"""Macro definition of feature importance type"""
719
720
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
721

722
"""Data type of data field"""
723
724
725
726
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
727
    "group": _C_API_DTYPE_INT32,
728
    "position": _C_API_DTYPE_INT32,
729
}
wxchan's avatar
wxchan committed
730

731
"""String name to int feature importance type mapper"""
732
733
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
734
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN,
735
}
736

wxchan's avatar
wxchan committed
737

738
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
739
    """Fix the memory of multi-dimensional sliced object."""
740
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
741
        if not data.flags.c_contiguous:
742
743
744
745
            _log_warning(
                "Usage of np.ndarray subset (sliced data) is not recommended "
                "due to it will double the peak memory cost in LightGBM."
            )
746
747
748
749
            return np.copy(data)
    return data


750
def _c_float_array(data: np.ndarray) -> Tuple[_ctypes_float_ptr, int, np.ndarray]:
751
    """Get pointer of float numpy array / list."""
752
    if _is_1d_list(data):
753
        data = np.asarray(data)
754
    if _is_numpy_1d_array(data):
755
        data = _convert_from_sliced_object(data)
756
        assert data.flags.c_contiguous
757
        ptr_data: _ctypes_float_ptr
wxchan's avatar
wxchan committed
758
759
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
760
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
761
762
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
763
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
764
        else:
765
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
766
    else:
767
        raise TypeError(f"Unknown type({type(data).__name__})")
768
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
769

wxchan's avatar
wxchan committed
770

771
def _c_int_array(data: np.ndarray) -> Tuple[_ctypes_int_ptr, int, np.ndarray]:
772
    """Get pointer of int numpy array / list."""
773
    if _is_1d_list(data):
774
        data = np.asarray(data)
775
    if _is_numpy_1d_array(data):
776
        data = _convert_from_sliced_object(data)
777
        assert data.flags.c_contiguous
778
        ptr_data: _ctypes_int_ptr
wxchan's avatar
wxchan committed
779
780
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
781
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
782
783
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
784
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
785
        else:
786
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
787
    else:
788
        raise TypeError(f"Unknown type({type(data).__name__})")
789
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
790

wxchan's avatar
wxchan committed
791

792
def _is_allowed_numpy_dtype(dtype: type) -> bool:
793
794
    float128 = getattr(np, "float128", type(None))
    return issubclass(dtype, (np.integer, np.floating, np.bool_)) and not issubclass(dtype, (np.timedelta64, float128))
795
796


797
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
798
    bad_pandas_dtypes = [
799
        f"{column_name}: {pandas_dtype}"
800
        for column_name, pandas_dtype in pandas_dtypes_series.items()
801
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
802
803
    ]
    if bad_pandas_dtypes:
804
805
806
807
        raise ValueError(
            'pandas dtypes must be int, float or bool.\n'
            f'Fields with bad pandas dtypes: {", ".join(bad_pandas_dtypes)}'
        )
808
809


810
811
def _pandas_to_numpy(
    data: pd_DataFrame,
812
    target_dtype: "np.typing.DTypeLike",
813
814
815
816
817
818
819
820
821
822
823
824
825
826
) -> 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)


827
def _data_from_pandas(
828
829
830
    data: pd_DataFrame,
    feature_name: _LGBM_FeatureNameConfiguration,
    categorical_feature: _LGBM_CategoricalFeatureConfiguration,
831
    pandas_categorical: Optional[List[List]],
832
) -> Tuple[np.ndarray, List[str], Union[List[str], List[int]], List[List]]:
833
    if len(data.shape) != 2 or data.shape[0] < 1:
834
        raise ValueError("Input data must be 2 dimensional and non empty.")
835

836
837
838
839
    # take shallow copy in case we modify categorical columns
    # whole column modifications don't change the original df
    data = data.copy(deep=False)

840
    # determine feature names
841
    if feature_name == "auto":
842
843
844
845
        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)]
846
    cat_cols_not_ordered: List[str] = [col for col in cat_cols if not data[col].cat.ordered]
847
848
    if pandas_categorical is None:  # train dataset
        pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
849
    else:
850
        if len(cat_cols) != len(pandas_categorical):
851
            raise ValueError("train and valid dataset categorical_feature do not match.")
852
853
854
855
856
        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})
857
858

    # use cat cols from DataFrame
859
    if categorical_feature == "auto":
860
861
862
        categorical_feature = cat_cols_not_ordered

    df_dtypes = [dtype.type for dtype in data.dtypes]
863
864
    # so that the target dtype considers floats
    df_dtypes.append(np.float32)
865
    target_dtype = np.result_type(*df_dtypes)
866
867
868
869
870

    return (
        _pandas_to_numpy(data, target_dtype=target_dtype),
        feature_name,
        categorical_feature,
871
        pandas_categorical,
872
    )
873
874


875
876
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
877
    file_name: Optional[Union[str, Path]] = None,
878
) -> str:
879
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
880
    pandas_str = f"\npandas_categorical:{categorical_json}\n"
881
    if file_name is not None:
882
        with open(file_name, "a") as f:
883
884
885
886
            f.write(pandas_str)
    return pandas_str


887
888
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
889
    model_str: Optional[str] = None,
890
) -> Optional[List[List]]:
891
    pandas_key = "pandas_categorical:"
892
    offset = -len(pandas_key)
893
    if file_name is not None:
894
        max_offset = -getsize(file_name)
895
        with open(file_name, "rb") as f:
896
            while True:
897
                offset = max(offset, max_offset)
898
                f.seek(offset, SEEK_END)
899
900
901
902
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
903
        last_line = lines[-1].decode("utf-8").strip()
904
        if not last_line.startswith(pandas_key):
905
            last_line = lines[-2].decode("utf-8").strip()
906
    elif model_str is not None:
907
        idx = model_str.rfind("\n", 0, offset)
908
909
        last_line = model_str[idx:].strip()
    if last_line.startswith(pandas_key):
910
        return json.loads(last_line[len(pandas_key) :])
911
912
    else:
        return None
913
914


915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
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**.

935
936
    .. versionadded:: 3.3.0

937
938
939
940
941
942
943
944
945
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
946
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
947
948
949
950
951
952
953
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
954
                return self._get_one_line(idx)
955
            elif isinstance(idx, slice):
956
957
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
958
                # Only required if using ``Dataset.subset()``.
959
                return np.array([self._get_one_line(i) for i in idx])
960
            else:
961
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
962
963
964

        Parameters
        ----------
965
        idx : int, slice[int], list[int]
966
967
968
969
            Item index.

        Returns
        -------
970
        result : numpy 1-D array or numpy 2-D array
971
            1-D array if idx is int, 2-D array if idx is slice or list.
972
973
974
975
976
977
978
979
980
        """
        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__()")


981
class _InnerPredictor:
982
983
984
985
986
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
987
988
989
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
990
    """
991

992
993
    def __init__(
        self,
994
995
996
        booster_handle: _BoosterHandle,
        pandas_categorical: Optional[List[List]],
        pred_parameter: Dict[str, Any],
997
        manage_handle: bool,
998
    ):
999
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
1000
1001
1002

        Parameters
        ----------
1003
        booster_handle : object
1004
            Handle of Booster.
1005
1006
1007
1008
        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
1009
            Other parameters for the prediction.
1010
1011
        manage_handle : bool
            If ``True``, free the corresponding Booster on the C++ side when this Python object is deleted.
wxchan's avatar
wxchan committed
1012
        """
1013
1014
1015
1016
1017
1018
1019
1020
        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(
1021
                self._handle,
1022
                ctypes.byref(out_num_class),
1023
1024
1025
            )
        )
        self.num_class = out_num_class.value
wxchan's avatar
wxchan committed
1026

1027
1028
1029
1030
    @classmethod
    def from_booster(
        cls,
        booster: "Booster",
1031
        pred_parameter: Dict[str, Any],
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
    ) -> "_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,
1046
                ctypes.byref(out_cur_iter),
1047
1048
1049
1050
1051
1052
            )
        )
        return cls(
            booster_handle=booster._handle,
            pandas_categorical=booster.pandas_categorical,
            pred_parameter=pred_parameter,
1053
            manage_handle=False,
1054
1055
1056
1057
1058
1059
        )

    @classmethod
    def from_model_file(
        cls,
        model_file: Union[str, Path],
1060
        pred_parameter: Dict[str, Any],
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
    ) -> "_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),
1077
                ctypes.byref(booster_handle),
1078
1079
1080
1081
1082
1083
            )
        )
        return cls(
            booster_handle=booster_handle,
            pandas_categorical=_load_pandas_categorical(file_name=model_file),
            pred_parameter=pred_parameter,
1084
            manage_handle=True,
1085
        )
cbecker's avatar
cbecker committed
1086

1087
    def __del__(self) -> None:
1088
1089
        try:
            if self.__is_manage_handle:
1090
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
1091
1092
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
1093

1094
    def __getstate__(self) -> Dict[str, Any]:
1095
        this = self.__dict__.copy()
1096
1097
        this.pop("handle", None)
        this.pop("_handle", None)
1098
1099
        return this

1100
1101
    def predict(
        self,
1102
        data: _LGBM_PredictDataType,
1103
1104
1105
1106
1107
1108
        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,
1109
        validate_features: bool = False,
1110
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
1111
        """Predict logic.
wxchan's avatar
wxchan committed
1112
1113
1114

        Parameters
        ----------
1115
        data : str, pathlib.Path, numpy array, pandas DataFrame, pyarrow Table, H2O DataTable's Frame (deprecated) or scipy.sparse
1116
            Data source for prediction.
1117
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
1118
1119
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
        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.
1131
1132
1133
        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
1134

1135
1136
            .. versionadded:: 4.0.0

wxchan's avatar
wxchan committed
1137
1138
        Returns
        -------
1139
        result : numpy array, scipy.sparse or list of scipy.sparse
1140
            Prediction result.
1141
            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
1142
        """
wxchan's avatar
wxchan committed
1143
        if isinstance(data, Dataset):
1144
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
1145
1146
1147
        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))()
1148
            ptr_names[:] = [x.encode("utf-8") for x in data_names]
1149
1150
            _safe_call(
                _LIB.LGBM_BoosterValidateFeatureNames(
1151
                    self._handle,
1152
1153
1154
1155
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
1156
1157
1158
1159
1160
1161

        if isinstance(data, pd_DataFrame):
            data = _data_from_pandas(
                data=data,
                feature_name="auto",
                categorical_feature="auto",
1162
                pandas_categorical=self.pandas_categorical,
1163
1164
            )[0]

1165
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
1166
        if raw_score:
1167
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
1168
        if pred_leaf:
1169
            predict_type = _C_API_PREDICT_LEAF_INDEX
1170
        if pred_contrib:
1171
            predict_type = _C_API_PREDICT_CONTRIB
cbecker's avatar
cbecker committed
1172

1173
        if isinstance(data, (str, Path)):
1174
            with _TempFile() as f:
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
                _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),
                    )
                )
1187
1188
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
1189
        elif isinstance(data, scipy.sparse.csr_matrix):
1190
1191
1192
1193
            preds, nrow = self.__pred_for_csr(
                csr=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1194
                predict_type=predict_type,
1195
            )
Guolin Ke's avatar
Guolin Ke committed
1196
        elif isinstance(data, scipy.sparse.csc_matrix):
1197
1198
1199
1200
            preds, nrow = self.__pred_for_csc(
                csc=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1201
                predict_type=predict_type,
1202
            )
wxchan's avatar
wxchan committed
1203
        elif isinstance(data, np.ndarray):
1204
1205
1206
1207
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1208
                predict_type=predict_type,
1209
            )
1210
1211
1212
1213
1214
        elif _is_pyarrow_table(data):
            preds, nrow = self.__pred_for_pyarrow_table(
                table=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1215
                predict_type=predict_type,
1216
            )
1217
1218
1219
        elif isinstance(data, list):
            try:
                data = np.array(data)
1220
            except BaseException as err:
1221
                raise ValueError("Cannot convert data list to numpy array.") from err
1222
1223
1224
1225
            preds, nrow = self.__pred_for_np2d(
                mat=data,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1226
                predict_type=predict_type,
1227
            )
1228
        elif isinstance(data, dt_DataTable):
1229
            _emit_datatable_deprecation_warning()
1230
1231
1232
1233
            preds, nrow = self.__pred_for_np2d(
                mat=data.to_numpy(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1234
                predict_type=predict_type,
1235
            )
wxchan's avatar
wxchan committed
1236
1237
        else:
            try:
1238
                _log_warning("Converting data to scipy sparse matrix.")
wxchan's avatar
wxchan committed
1239
                csr = scipy.sparse.csr_matrix(data)
1240
            except BaseException as err:
1241
                raise TypeError(f"Cannot predict data for type {type(data).__name__}") from err
1242
1243
1244
1245
            preds, nrow = self.__pred_for_csr(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1246
                predict_type=predict_type,
1247
            )
wxchan's avatar
wxchan committed
1248
1249
        if pred_leaf:
            preds = preds.astype(np.int32)
1250
        is_sparse = isinstance(preds, (list, scipy.sparse.spmatrix))
1251
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
1252
            if preds.size % nrow == 0:
1253
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
1254
            else:
1255
                raise ValueError(f"Length of predict result ({preds.size}) cannot be divide nrow ({nrow})")
wxchan's avatar
wxchan committed
1256
1257
        return preds

1258
1259
1260
1261
1262
    def __get_num_preds(
        self,
        start_iteration: int,
        num_iteration: int,
        nrow: int,
1263
        predict_type: int,
1264
    ) -> int:
1265
        """Get size of prediction result."""
1266
        if nrow > _MAX_INT32:
1267
1268
1269
1270
1271
1272
            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
1273
        n_preds = ctypes.c_int64(0)
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
        _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
1284
        return n_preds.value
wxchan's avatar
wxchan committed
1285

1286
1287
1288
1289
1290
1291
    def __inner_predict_np2d(
        self,
        mat: np.ndarray,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
1292
        preds: Optional[np.ndarray],
1293
1294
    ) -> Tuple[np.ndarray, int]:
        if mat.dtype == np.float32 or mat.dtype == np.float64:
1295
            data = np.asarray(mat.reshape(mat.size), dtype=mat.dtype)
1296
1297
1298
        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)
1299
1300
1301
1302
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=mat.shape[0],
1303
            predict_type=predict_type,
1304
        )
1305
1306
1307
1308
1309
        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)
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
        _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)),
            )
        )
1326
1327
1328
1329
1330
1331
1332
1333
1334
        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,
1335
        predict_type: int,
1336
    ) -> Tuple[np.ndarray, int]:
1337
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1338
        if len(mat.shape) != 2:
1339
            raise ValueError("Input numpy.ndarray or list must be 2 dimensional")
wxchan's avatar
wxchan committed
1340

1341
        nrow = mat.shape[0]
1342
1343
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1344
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1345
1346
1347
1348
            n_preds = [
                self.__get_num_preds(start_iteration, num_iteration, i, predict_type)
                for i in np.diff([0] + list(sections) + [nrow])
            ]
1349
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1350
            preds = np.empty(sum(n_preds), dtype=np.float64)
1351
1352
1353
            for chunk, (start_idx_pred, end_idx_pred) in zip(
                np.array_split(mat, sections), zip(n_preds_sections, n_preds_sections[1:])
            ):
1354
                # avoid memory consumption by arrays concatenation operations
1355
1356
1357
1358
1359
                self.__inner_predict_np2d(
                    mat=chunk,
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
1360
                    preds=preds[start_idx_pred:end_idx_pred],
1361
                )
1362
            return preds, nrow
wxchan's avatar
wxchan committed
1363
        else:
1364
1365
1366
1367
1368
            return self.__inner_predict_np2d(
                mat=mat,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
1369
                preds=None,
1370
            )
wxchan's avatar
wxchan committed
1371

1372
1373
1374
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
1375
1376
1377
1378
1379
1380
        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,
1381
        is_csr: bool,
1382
    ) -> Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]]:
1383
1384
1385
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1386
        if indptr_type == _C_API_DTYPE_INT32:
1387
            out_indptr = _cint32_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1388
        elif indptr_type == _C_API_DTYPE_INT64:
1389
            out_indptr = _cint64_array_to_numpy(cptr=out_ptr_indptr, length=indptr_len)
1390
1391
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1392
        if data_type == _C_API_DTYPE_FLOAT32:
1393
            out_data = _cfloat32_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1394
        elif data_type == _C_API_DTYPE_FLOAT64:
1395
            out_data = _cfloat64_array_to_numpy(cptr=out_ptr_data, length=data_indices_len)
1396
1397
        else:
            raise TypeError("Expected float32 or float64 type for data")
1398
        out_indices = _cint32_array_to_numpy(cptr=out_ptr_indices, length=data_indices_len)
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
        # 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]
1410
1411
            cs_indices = out_indices[offset + cs_indptr[0] : offset + matrix_indptr_len]
            cs_data = out_data[offset + cs_indptr[0] : offset + matrix_indptr_len]
1412
1413
1414
1415
1416
1417
1418
1419
1420
            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
1421
1422
1423
1424
1425
1426
1427
1428
1429
        _safe_call(
            _LIB.LGBM_BoosterFreePredictSparse(
                out_ptr_indptr,
                out_ptr_indices,
                out_ptr_data,
                ctypes.c_int(indptr_type),
                ctypes.c_int(data_type),
            )
        )
1430
1431
1432
1433
        if len(cs_output_matrices) == 1:
            return cs_output_matrices[0]
        return cs_output_matrices

1434
1435
1436
1437
1438
1439
    def __inner_predict_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
1440
        preds: Optional[np.ndarray],
1441
1442
    ) -> Tuple[np.ndarray, int]:
        nrow = len(csr.indptr) - 1
1443
1444
1445
1446
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
1447
            predict_type=predict_type,
1448
        )
1449
1450
1451
1452
1453
        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
1454

1455
1456
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1457

1458
1459
1460
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
        _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)),
            )
        )
1480
1481
1482
1483
1484
1485
1486
1487
1488
        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,
1489
        predict_type: int,
1490
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1491
1492
1493
1494
        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
1495
        out_ptr_indptr: _ctypes_int_ptr
1496
1497
1498
1499
1500
        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)()
1501
        out_ptr_data: _ctypes_float_ptr
1502
1503
1504
1505
1506
        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)
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
        _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),
            )
        )
1529
1530
1531
1532
1533
1534
1535
1536
        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,
1537
            is_csr=True,
1538
1539
1540
1541
        )
        nrow = len(csr.indptr) - 1
        return matrices, nrow

1542
1543
1544
1545
1546
    def __pred_for_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
1547
        predict_type: int,
1548
    ) -> Tuple[np.ndarray, int]:
1549
        """Predict for a CSR data."""
1550
        if predict_type == _C_API_PREDICT_CONTRIB:
1551
1552
1553
1554
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1555
                predict_type=predict_type,
1556
            )
1557
        nrow = len(csr.indptr) - 1
1558
1559
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1560
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1561
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1562
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1563
            preds = np.empty(sum(n_preds), dtype=np.float64)
1564
1565
1566
            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:])
            ):
1567
                # avoid memory consumption by arrays concatenation operations
1568
1569
1570
1571
1572
                self.__inner_predict_csr(
                    csr=csr[start_idx:end_idx],
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
1573
                    preds=preds[start_idx_pred:end_idx_pred],
1574
                )
1575
1576
            return preds, nrow
        else:
1577
1578
1579
1580
1581
            return self.__inner_predict_csr(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
1582
                preds=None,
1583
1584
1585
1586
            )

    def __inner_predict_sparse_csc(
        self,
1587
1588
1589
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
1590
        predict_type: int,
1591
    ) -> Tuple[Union[List[scipy.sparse.csc_matrix], List[scipy.sparse.csr_matrix]], int]:
1592
1593
1594
1595
        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
1596
        out_ptr_indptr: _ctypes_int_ptr
1597
1598
1599
1600
1601
        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)()
1602
        out_ptr_data: _ctypes_float_ptr
1603
1604
1605
1606
1607
        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)
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
        _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),
            )
        )
1630
1631
1632
1633
1634
1635
1636
1637
        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,
1638
            is_csr=False,
1639
1640
1641
        )
        nrow = csc.shape[0]
        return matrices, nrow
Guolin Ke's avatar
Guolin Ke committed
1642

1643
1644
1645
1646
1647
    def __pred_for_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        start_iteration: int,
        num_iteration: int,
1648
        predict_type: int,
1649
    ) -> Tuple[np.ndarray, int]:
1650
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1651
        nrow = csc.shape[0]
1652
        if nrow > _MAX_INT32:
1653
1654
1655
1656
            return self.__pred_for_csr(
                csr=csc.tocsr(),
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1657
                predict_type=predict_type,
1658
            )
1659
        if predict_type == _C_API_PREDICT_CONTRIB:
1660
1661
1662
1663
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
1664
                predict_type=predict_type,
1665
            )
1666
1667
1668
1669
        n_preds = self.__get_num_preds(
            start_iteration=start_iteration,
            num_iteration=num_iteration,
            nrow=nrow,
1670
            predict_type=predict_type,
1671
        )
1672
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1673
1674
        out_num_preds = ctypes.c_int64(0)

1675
1676
        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
1677

1678
        assert csc.shape[0] <= _MAX_INT32
1679
        csc_indices = csc.indices.astype(np.int32, copy=False)
1680

1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
        _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
1700
        if n_preds != out_num_preds.value:
1701
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1702
        return preds, nrow
1703

1704
1705
1706
1707
1708
    def __pred_for_pyarrow_table(
        self,
        table: pa_Table,
        start_iteration: int,
        num_iteration: int,
1709
        predict_type: int,
1710
1711
1712
1713
1714
1715
    ) -> 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)
1716
        if not all(arrow_is_integer(t) or arrow_is_floating(t) or arrow_is_boolean(t) for t in table.schema.types):
1717
1718
1719
1720
1721
1722
1723
            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,
1724
            predict_type=predict_type,
1725
1726
1727
1728
1729
1730
        )
        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)
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
        _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)),
            )
        )
1745
1746
1747
        if n_preds != out_num_preds.value:
            raise ValueError("Wrong length for predict results")
        return preds, table.num_rows
wxchan's avatar
wxchan committed
1748

1749
    def current_iteration(self) -> int:
1750
1751
1752
1753
1754
1755
1756
1757
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
        out_cur_iter = ctypes.c_int(0)
1758
1759
1760
1761
1762
1763
        _safe_call(
            _LIB.LGBM_BoosterGetCurrentIteration(
                self._handle,
                ctypes.byref(out_cur_iter),
            )
        )
1764
1765
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1766

1767
class Dataset:
1768
1769
1770
1771
1772
1773
1774
1775
1776
    """
    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.
    """
1777

1778
1779
    def __init__(
        self,
1780
        data: _LGBM_TrainDataType,
1781
        label: Optional[_LGBM_LabelType] = None,
1782
        reference: Optional["Dataset"] = None,
1783
1784
1785
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
1786
1787
        feature_name: _LGBM_FeatureNameConfiguration = "auto",
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = "auto",
1788
        params: Optional[Dict[str, Any]] = None,
1789
1790
        free_raw_data: bool = True,
        position: Optional[_LGBM_PositionType] = None,
1791
    ):
1792
        """Initialize Dataset.
1793

wxchan's avatar
wxchan committed
1794
1795
        Parameters
        ----------
1796
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame (deprecated), scipy.sparse, Sequence, list of Sequence, list of numpy array or pyarrow Table
wxchan's avatar
wxchan committed
1797
            Data source of Dataset.
1798
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1799
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1800
1801
1802
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1803
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1804
            Weight for each instance. Weights should be non-negative.
1805
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
1806
1807
1808
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1809
1810
            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.
1811
        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)
1812
            Init score for Dataset.
1813
        feature_name : list of str, or 'auto', optional (default="auto")
1814
            Feature names.
1815
            If 'auto' and data is pandas DataFrame or pyarrow Table, data columns names are used.
1816
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1817
1818
            Categorical features.
            If list of int, interpreted as indices.
1819
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1820
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1821
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1822
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1823
            All negative values in categorical features will be treated as missing values.
1824
            The output cannot be monotonically constrained with respect to a categorical feature.
1825
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1826
        params : dict or None, optional (default=None)
1827
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1828
        free_raw_data : bool, optional (default=True)
1829
            If True, raw data is freed after constructing inner Dataset.
1830
1831
        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
1832
        """
1833
        self._handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1834
1835
1836
1837
1838
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1839
        self.position = position
1840
        self.init_score = init_score
1841
1842
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1843
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1844
        self.free_raw_data = free_raw_data
1845
        self.used_indices: Optional[List[int]] = None
1846
        self._need_slice = True
1847
        self._predictor: Optional[_InnerPredictor] = None
1848
        self.pandas_categorical: Optional[List[List]] = None
1849
        self._params_back_up: Optional[Dict[str, Any]] = None
1850
        self.version = 0
1851
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1852

1853
    def __del__(self) -> None:
1854
1855
1856
1857
        try:
            self._free_handle()
        except AttributeError:
            pass
1858

1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
    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.
        """
1876
        param_str = _param_dict_to_str(self.get_params())
1877
1878
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1879
        ptr_data, _, _ = _c_int_array(indices)
1880
1881
        actual_sample_cnt = ctypes.c_int32(0)

1882
1883
1884
1885
1886
1887
1888
1889
        _safe_call(
            _LIB.LGBM_SampleIndices(
                ctypes.c_int32(total_nrow),
                _c_str(param_str),
                ptr_data,
                ctypes.byref(actual_sample_cnt),
            )
        )
1890
1891
        assert sample_cnt == actual_sample_cnt.value
        return indices
1892

1893
1894
1895
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
1896
1897
        ref_dataset: _DatasetHandle,
    ) -> "Dataset":
1898
1899
1900
1901
1902
1903
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1904
1905
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1906
1907
1908
1909
1910
1911

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
1912
        self._handle = ctypes.c_void_p()
1913
1914
1915
1916
1917
1918
1919
        _safe_call(
            _LIB.LGBM_DatasetCreateByReference(
                ref_dataset,
                ctypes.c_int64(total_nrow),
                ctypes.byref(self._handle),
            )
        )
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
        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
        ----------
1933
        sample_data : list of numpy array
1934
            Sample data for each column.
1935
        sample_indices : list of numpy array
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
            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.
1958
        sample_col_ptr: _ctypes_float_array = (ctypes.POINTER(ctypes.c_double) * ncol)()
1959
1960
        # c type int**
        # each int* points to start of indices for each column
1961
        indices_col_ptr: _ctypes_int_array = (ctypes.POINTER(ctypes.c_int32) * ncol)()
1962
        for i in range(ncol):
1963
1964
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1965
1966

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1967
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1968

1969
        self._handle = ctypes.c_void_p()
1970
        params_str = _param_dict_to_str(self.get_params())
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
        _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),
            )
        )
1984
1985
        return self

1986
    def _push_rows(self, data: np.ndarray) -> "Dataset":
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
        """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)
2001
        data_ptr, data_type, _ = _c_float_array(data)
2002

2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
        _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),
            )
        )
2013
2014
2015
        self._start_row += nrow
        return self

2016
    def get_params(self) -> Dict[str, Any]:
2017
2018
2019
2020
        """Get the used parameters in the Dataset.

        Returns
        -------
2021
        params : dict
2022
2023
2024
2025
            The used parameters in this Dataset object.
        """
        if self.params is not None:
            # no min_data, nthreads and verbose in this function
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
            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",
            )
2049
            return {k: v for k, v in self.params.items() if k in dataset_params}
2050
2051
        else:
            return {}
2052

2053
    def _free_handle(self) -> "Dataset":
2054
2055
2056
        if self._handle is not None:
            _safe_call(_LIB.LGBM_DatasetFree(self._handle))
            self._handle = None
2057
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
2058
2059
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
2060
        return self
wxchan's avatar
wxchan committed
2061

2062
2063
2064
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
2065
        data: _LGBM_TrainDataType,
2066
        used_indices: Optional[Union[List[int], np.ndarray]],
2067
    ) -> "Dataset":
Guolin Ke's avatar
Guolin Ke committed
2068
        data_has_header = False
2069
        if isinstance(data, (str, Path)) and self.params is not None:
Guolin Ke's avatar
Guolin Ke committed
2070
            # check data has header or not
2071
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
2072
        num_data = self.num_data()
2073
        if predictor is not None:
2074
2075
2076
            init_score: Union[np.ndarray, scipy.sparse.spmatrix] = predictor.predict(
                data=data,
                raw_score=True,
2077
                data_has_header=data_has_header,
2078
            )
2079
            init_score = init_score.ravel()
2080
            if used_indices is not None:
2081
                assert not self._need_slice
2082
                if isinstance(data, (str, Path)):
2083
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
2084
                    assert num_data == len(used_indices)
2085
2086
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
2087
2088
2089
                            sub_init_score[i * predictor.num_class + j] = init_score[
                                used_indices[i] * predictor.num_class + j
                            ]
2090
2091
2092
                    init_score = sub_init_score
            if predictor.num_class > 1:
                # need to regroup init_score
2093
                new_init_score = np.empty(init_score.size, dtype=np.float64)
2094
2095
                for i in range(num_data):
                    for j in range(predictor.num_class):
2096
2097
2098
                        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:
2099
            init_score = np.full_like(self.init_score, fill_value=0.0, dtype=np.float64)
2100
2101
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
2102
        self.set_init_score(init_score)
2103
        return self
Guolin Ke's avatar
Guolin Ke committed
2104

2105
2106
    def _lazy_init(
        self,
2107
        data: Optional[_LGBM_TrainDataType],
2108
2109
2110
2111
2112
2113
2114
2115
        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,
2116
        params: Optional[Dict[str, Any]],
2117
        position: Optional[_LGBM_PositionType],
2118
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2119
        if data is None:
2120
            self._handle = None
Nikita Titov's avatar
Nikita Titov committed
2121
            return self
Guolin Ke's avatar
Guolin Ke committed
2122
2123
2124
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
2125
2126
2127
2128
2129
        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,
2130
                pandas_categorical=self.pandas_categorical,
2131
            )
Guolin Ke's avatar
Guolin Ke committed
2132

2133
        # process for args
wxchan's avatar
wxchan committed
2134
        params = {} if params is None else params
2135
        args_names = inspect.signature(self.__class__._lazy_init).parameters.keys()
2136
        for key in params.keys():
2137
            if key in args_names:
2138
2139
2140
2141
                _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."
                )
2142
        # get categorical features
2143
        if isinstance(categorical_feature, list):
2144
2145
            categorical_indices = set()
            feature_dict = {}
2146
            if isinstance(feature_name, list):
2147
2148
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
2149
                if isinstance(name, str) and name in feature_dict:
2150
                    categorical_indices.add(feature_dict[name])
2151
                elif isinstance(name, int):
2152
2153
                    categorical_indices.add(name)
                else:
2154
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
2155
            if categorical_indices:
2156
2157
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
2158
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
2159
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
2160
                            _log_warning(f"{cat_alias} in param dict is overridden.")
2161
                        params.pop(cat_alias, None)
2162
                params["categorical_column"] = sorted(categorical_indices)
2163

2164
        params_str = _param_dict_to_str(params)
2165
        self.params = params
2166
        # process for reference dataset
wxchan's avatar
wxchan committed
2167
        ref_dataset = None
wxchan's avatar
wxchan committed
2168
        if isinstance(reference, Dataset):
2169
            ref_dataset = reference.construct()._handle
wxchan's avatar
wxchan committed
2170
        elif reference is not None:
2171
            raise TypeError("Reference dataset should be None or dataset instance")
2172
        # start construct data
2173
        if isinstance(data, (str, Path)):
2174
            self._handle = ctypes.c_void_p()
2175
2176
2177
2178
2179
2180
2181
2182
            _safe_call(
                _LIB.LGBM_DatasetCreateFromFile(
                    _c_str(str(data)),
                    _c_str(params_str),
                    ref_dataset,
                    ctypes.byref(self._handle),
                )
            )
wxchan's avatar
wxchan committed
2183
2184
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
2185
2186
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
2187
2188
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
2189
2190
2191
        elif _is_pyarrow_table(data):
            self.__init_from_pyarrow_table(data, params_str, ref_dataset)
            feature_name = data.column_names
2192
        elif isinstance(data, list) and len(data) > 0:
2193
            if _is_list_of_numpy_arrays(data):
2194
                self.__init_from_list_np2d(data, params_str, ref_dataset)
2195
            elif _is_list_of_sequences(data):
2196
2197
                self.__init_from_seqs(data, ref_dataset)
            else:
2198
                raise TypeError("Data list can only be of ndarray or Sequence")
2199
2200
        elif isinstance(data, Sequence):
            self.__init_from_seqs([data], ref_dataset)
2201
        elif isinstance(data, dt_DataTable):
2202
            _emit_datatable_deprecation_warning()
2203
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
2204
2205
2206
2207
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
2208
            except BaseException as err:
2209
                raise TypeError(f"Cannot initialize Dataset from {type(data).__name__}") from err
wxchan's avatar
wxchan committed
2210
2211
2212
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
2213
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
2214
2215
2216
2217
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
2218
2219
        if position is not None:
            self.set_position(position)
2220
2221
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
2222
                _log_warning("The init_score will be overridden by the prediction of init_model.")
2223
            self._set_init_score_by_predictor(predictor=predictor, data=data, used_indices=None)
2224
2225
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
2226
        elif predictor is not None:
2227
            raise TypeError(f"Wrong predictor type {type(predictor).__name__}")
Guolin Ke's avatar
Guolin Ke committed
2228
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
2229
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
2230

2231
    @staticmethod
2232
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]) -> Iterator[np.ndarray]:
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
        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]
2244
            yield row if row.flags["OWNDATA"] else row.copy()
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257

    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.
2258
        sampled = np.array(list(self._yield_row_from_seqlist(seqs, indices)))
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
        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

2274
2275
2276
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
2277
        ref_dataset: Optional[_DatasetHandle],
2278
    ) -> "Dataset":
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
        """
        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:
2293
            param_str = _param_dict_to_str(self.get_params())
2294
2295
2296
2297
2298
2299
2300
            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)
2301
            batch_size = getattr(seq, "batch_size", None) or Sequence.batch_size
2302
2303
2304
2305
2306
            for start in range(0, nrow, batch_size):
                end = min(start + batch_size, nrow)
                self._push_rows(seq[start:end])
        return self

2307
2308
2309
2310
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
2311
        ref_dataset: Optional[_DatasetHandle],
2312
    ) -> "Dataset":
2313
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
2314
        if len(mat.shape) != 2:
2315
            raise ValueError("Input numpy.ndarray must be 2 dimensional")
wxchan's avatar
wxchan committed
2316

2317
        self._handle = ctypes.c_void_p()
2318
        data, layout = _np2d_to_np1d(mat)
2319
        ptr_data, type_ptr_data, _ = _c_float_array(data)
2320
2321
2322
2323
2324
2325
        _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]),
2326
                ctypes.c_int(layout),
2327
2328
2329
2330
2331
                _c_str(params_str),
                ref_dataset,
                ctypes.byref(self._handle),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
2332
        return self
wxchan's avatar
wxchan committed
2333

2334
2335
2336
2337
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
2338
        ref_dataset: Optional[_DatasetHandle],
2339
    ) -> "Dataset":
2340
        """Initialize data from a list of 2-D numpy matrices."""
2341
        ncol = mats[0].shape[1]
2342
        nrow = np.empty((len(mats),), np.int32)
2343
        ptr_data: _ctypes_float_array
2344
2345
2346
2347
2348
2349
        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 = []
2350
        type_ptr_data = -1
2351
2352
2353

        for i, mat in enumerate(mats):
            if len(mat.shape) != 2:
2354
                raise ValueError("Input numpy.ndarray must be 2 dimensional")
2355
2356

            if mat.shape[1] != ncol:
2357
                raise ValueError("Input arrays must have same number of columns")
2358
2359
2360
2361

            nrow[i] = mat.shape[0]

            if mat.dtype == np.float32 or mat.dtype == np.float64:
2362
                mats[i] = np.asarray(mat.reshape(mat.size), dtype=mat.dtype)
2363
            else:  # change non-float data to float data, need to copy
2364
2365
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

2366
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
2367
            if type_ptr_data != -1 and chunk_type_ptr_data != type_ptr_data:
2368
                raise ValueError("Input chunks must have same type")
2369
2370
2371
2372
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

2373
        self._handle = ctypes.c_void_p()
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
        _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
2387
        return self
2388

2389
2390
2391
2392
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
2393
        ref_dataset: Optional[_DatasetHandle],
2394
    ) -> "Dataset":
2395
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
2396
        if len(csr.indices) != len(csr.data):
2397
            raise ValueError(f"Length mismatch: {len(csr.indices)} vs {len(csr.data)}")
2398
        self._handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2399

2400
2401
        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
2402

2403
        assert csr.shape[1] <= _MAX_INT32
2404
        csr_indices = csr.indices.astype(np.int32, copy=False)
2405

2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
        _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
2421
        return self
wxchan's avatar
wxchan committed
2422

2423
2424
2425
2426
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
2427
        ref_dataset: Optional[_DatasetHandle],
2428
    ) -> "Dataset":
2429
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
2430
        if len(csc.indices) != len(csc.data):
2431
            raise ValueError(f"Length mismatch: {len(csc.indices)} vs {len(csc.data)}")
2432
        self._handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2433

2434
2435
        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
2436

2437
        assert csc.shape[0] <= _MAX_INT32
2438
        csc_indices = csc.indices.astype(np.int32, copy=False)
2439

2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
        _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
2455
        return self
Guolin Ke's avatar
Guolin Ke committed
2456

2457
2458
2459
2460
    def __init_from_pyarrow_table(
        self,
        table: pa_Table,
        params_str: str,
2461
        ref_dataset: Optional[_DatasetHandle],
2462
2463
2464
2465
2466
2467
    ) -> "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)
2468
        if not all(arrow_is_integer(t) or arrow_is_floating(t) or arrow_is_boolean(t) for t in table.schema.types):
2469
2470
2471
2472
2473
            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()
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
        _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),
            )
        )
2484
2485
        return self

2486
    @staticmethod
2487
    def _compare_params_for_warning(
2488
2489
        params: Dict[str, Any],
        other_params: Dict[str, Any],
2490
        ignore_keys: Set[str],
2491
2492
    ) -> bool:
        """Compare two dictionaries with params ignoring some keys.
2493

2494
2495
2496
2497
        It is only for the warning purpose.

        Parameters
        ----------
2498
        params : dict
2499
            One dictionary with parameters to compare.
2500
        other_params : dict
2501
2502
2503
            Another dictionary with parameters to compare.
        ignore_keys : set
            Keys that should be ignored during comparing two dictionaries.
2504
2505
2506

        Returns
        -------
2507
2508
        compare_result : bool
          Returns whether two dictionaries with params are equal.
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
        """
        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

2520
    def construct(self) -> "Dataset":
2521
2522
2523
2524
2525
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2526
            Constructed Dataset object.
2527
        """
2528
        if self._handle is None:
wxchan's avatar
wxchan committed
2529
            if self.reference is not None:
2530
                reference_params = self.reference.get_params()
2531
2532
                params = self.get_params()
                if params != reference_params:
2533
2534
2535
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
2536
                        ignore_keys=_ConfigAliases.get("categorical_feature"),
2537
                    ):
2538
                        _log_warning("Overriding the parameters from Reference Dataset.")
2539
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2540
                if self.used_indices is None:
2541
                    # create valid
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
                    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
2555
                else:
2556
                    # construct subset
2557
                    used_indices = _list_to_1d_numpy(self.used_indices, dtype=np.int32, name="used_indices")
2558
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2559
                    if self.reference.group is not None:
2560
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2561
2562
2563
                        _, self.group = np.unique(
                            np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices], return_counts=True
                        )
2564
                    self._handle = ctypes.c_void_p()
2565
                    params_str = _param_dict_to_str(self.params)
2566
2567
2568
2569
2570
2571
2572
2573
2574
                    _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
2575
2576
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2577
2578
                    if self.group is not None:
                        self.set_group(self.group)
2579
2580
                    if self.position is not None:
                        self.set_position(self.position)
wxchan's avatar
wxchan committed
2581
2582
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
2583
2584
2585
2586
                    if (
                        isinstance(self._predictor, _InnerPredictor)
                        and self._predictor is not self.reference._predictor
                    ):
Guolin Ke's avatar
Guolin Ke committed
2587
                        self.get_data()
2588
                        self._set_init_score_by_predictor(
2589
                            predictor=self._predictor, data=self.data, used_indices=used_indices
2590
                        )
wxchan's avatar
wxchan committed
2591
            else:
2592
                # create train
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
                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
2606
2607
            if self.free_raw_data:
                self.data = None
2608
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2609
        return self
wxchan's avatar
wxchan committed
2610

2611
2612
    def create_valid(
        self,
2613
        data: _LGBM_TrainDataType,
2614
        label: Optional[_LGBM_LabelType] = None,
2615
2616
2617
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
2618
        params: Optional[Dict[str, Any]] = None,
2619
        position: Optional[_LGBM_PositionType] = None,
2620
    ) -> "Dataset":
2621
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2622
2623
2624

        Parameters
        ----------
2625
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame (deprecated), scipy.sparse, Sequence, list of Sequence or list of numpy array
wxchan's avatar
wxchan committed
2626
            Data source of Dataset.
2627
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2628
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2629
            Label of the data.
2630
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2631
            Weight for each instance. Weights should be non-negative.
2632
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
2633
2634
2635
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2636
2637
            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.
2638
        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)
2639
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2640
        params : dict or None, optional (default=None)
2641
            Other parameters for validation Dataset.
2642
2643
        position : numpy 1-D array, pandas Series or None, optional (default=None)
            Position of items used in unbiased learning-to-rank task.
2644
2645
2646

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2647
2648
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2649
        """
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
        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
2661
        ret._predictor = self._predictor
2662
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2663
        return ret
wxchan's avatar
wxchan committed
2664

2665
2666
2667
    def subset(
        self,
        used_indices: List[int],
2668
        params: Optional[Dict[str, Any]] = None,
2669
    ) -> "Dataset":
2670
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2671
2672
2673
2674

        Parameters
        ----------
        used_indices : list of int
2675
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2676
        params : dict or None, optional (default=None)
2677
            These parameters will be passed to Dataset constructor.
2678
2679
2680
2681
2682

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2683
        """
wxchan's avatar
wxchan committed
2684
2685
        if params is None:
            params = self.params
2686
2687
2688
2689
2690
2691
2692
2693
        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
2694
        ret._predictor = self._predictor
2695
        ret.pandas_categorical = self.pandas_categorical
2696
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2697
2698
        return ret

2699
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2700
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2701

2702
2703
2704
2705
2706
        .. 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
2707
2708
        Parameters
        ----------
2709
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2710
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2711
2712
2713
2714
2715

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2716
        """
2717
2718
2719
2720
2721
2722
        _safe_call(
            _LIB.LGBM_DatasetSaveBinary(
                self.construct()._handle,
                _c_str(str(filename)),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
2723
        return self
wxchan's avatar
wxchan committed
2724

2725
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2726
2727
        if not params:
            return self
2728
        params = deepcopy(params)
2729

2730
        def update() -> None:
2731
2732
2733
            if not self.params:
                self.params = params
            else:
2734
                self._params_back_up = deepcopy(self.params)
2735
2736
                self.params.update(params)

2737
        if self._handle is None:
2738
2739
2740
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2741
                _c_str(_param_dict_to_str(self.params)),
2742
2743
                _c_str(_param_dict_to_str(params)),
            )
2744
2745
2746
2747
2748
2749
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2750
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode("utf-8"))
Nikita Titov's avatar
Nikita Titov committed
2751
        return self
wxchan's avatar
wxchan committed
2752

2753
    def _reverse_update_params(self) -> "Dataset":
2754
        if self._handle is None:
2755
2756
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2757
        return self
2758

2759
2760
2761
    def set_field(
        self,
        field_name: str,
2762
        data: Optional[_LGBM_SetFieldType],
2763
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2764
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2765
2766
2767

        Parameters
        ----------
2768
        field_name : str
2769
            The field name of the information.
2770
        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
2771
            The data to be set.
Nikita Titov's avatar
Nikita Titov committed
2772
2773
2774
2775
2776

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2777
        """
2778
        if self._handle is None:
2779
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2780
        if data is None:
2781
            # set to None
2782
2783
2784
2785
2786
2787
2788
2789
2790
            _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
2791
            return self
2792
2793

        # If the data is a arrow data, we can just pass it to C
2794
2795
2796
2797
2798
2799
        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}'")
2800
2801
2802
2803
2804
2805
2806
                data = pa_chunked_array(
                    [
                        chunk
                        for array in data.columns  # type: ignore
                        for chunk in array.chunks
                    ]
                )
2807

2808
            c_array = _export_arrow_to_c(data)
2809
2810
2811
2812
2813
2814
2815
2816
2817
            _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),
                )
            )
2818
2819
2820
            self.version += 1
            return self

2821
        dtype: "np.typing.DTypeLike"
2822
        if field_name == "init_score":
Guolin Ke's avatar
Guolin Ke committed
2823
            dtype = np.float64
2824
            if _is_1d_collection(data):
2825
                data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2826
            elif _is_2d_collection(data):
2827
                data = _data_to_2d_numpy(data, dtype=dtype, name=field_name)
2828
                data = data.ravel(order="F")
2829
2830
            else:
                raise TypeError(
2831
2832
                    "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."
2833
2834
                )
        else:
2835
            if field_name in {"group", "position"}:
2836
2837
2838
                dtype = np.int32
            else:
                dtype = np.float32
2839
            data = _list_to_1d_numpy(data, dtype=dtype, name=field_name)
2840

2841
        ptr_data: Union[_ctypes_float_ptr, _ctypes_int_ptr]
2842
        if data.dtype == np.float32 or data.dtype == np.float64:
2843
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2844
        elif data.dtype == np.int32:
2845
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2846
        else:
2847
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2848
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2849
            raise TypeError("Input type error for set_field")
2850
2851
2852
2853
2854
2855
2856
2857
2858
        _safe_call(
            _LIB.LGBM_DatasetSetField(
                self._handle,
                _c_str(field_name),
                ptr_data,
                ctypes.c_int(len(data)),
                ctypes.c_int(type_data),
            )
        )
2859
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2860
        return self
wxchan's avatar
wxchan committed
2861

2862
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2863
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2864

2865
2866
2867
2868
2869
2870
        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
2871
2872
        Parameters
        ----------
2873
        field_name : str
2874
            The field name of the information.
wxchan's avatar
wxchan committed
2875
2876
2877

        Returns
        -------
2878
        info : numpy array or None
2879
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2880
        """
2881
        if self._handle is None:
2882
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2883
2884
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2885
        ret = ctypes.POINTER(ctypes.c_void_p)()
2886
2887
2888
2889
2890
2891
2892
2893
2894
        _safe_call(
            _LIB.LGBM_DatasetGetField(
                self._handle,
                _c_str(field_name),
                ctypes.byref(tmp_out_len),
                ctypes.byref(ret),
                ctypes.byref(out_type),
            )
        )
2895
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2896
2897
2898
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2899
        if out_type.value == _C_API_DTYPE_INT32:
2900
2901
            arr = _cint32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)),
2902
                length=tmp_out_len.value,
2903
            )
2904
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2905
2906
            arr = _cfloat32_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)),
2907
                length=tmp_out_len.value,
2908
            )
2909
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2910
2911
            arr = _cfloat64_array_to_numpy(
                cptr=ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)),
2912
                length=tmp_out_len.value,
2913
            )
2914
        else:
wxchan's avatar
wxchan committed
2915
            raise TypeError("Unknown type")
2916
        if field_name == "init_score":
2917
2918
2919
            num_data = self.num_data()
            num_classes = arr.size // num_data
            if num_classes > 1:
2920
                arr = arr.reshape((num_data, num_classes), order="F")
2921
        return arr
Guolin Ke's avatar
Guolin Ke committed
2922

2923
2924
    def set_categorical_feature(
        self,
2925
        categorical_feature: _LGBM_CategoricalFeatureConfiguration,
2926
    ) -> "Dataset":
2927
        """Set categorical features.
2928
2929
2930

        Parameters
        ----------
2931
        categorical_feature : list of str or int, or 'auto'
2932
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2933
2934
2935
2936
2937

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2938
2939
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2940
            return self
2941
        if self.data is not None:
2942
2943
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2944
                return self._free_handle()
2945
            elif categorical_feature == "auto":
Nikita Titov's avatar
Nikita Titov committed
2946
                return self
2947
            else:
2948
2949
2950
2951
2952
                if self.categorical_feature != "auto":
                    _log_warning(
                        "categorical_feature in Dataset is overridden.\n"
                        f"New categorical_feature is {list(categorical_feature)}"
                    )
2953
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2954
                return self._free_handle()
2955
        else:
2956
2957
2958
2959
            raise LightGBMError(
                "Cannot set categorical feature after freed raw data, "
                "set free_raw_data=False when construct Dataset to avoid this."
            )
2960

2961
2962
    def _set_predictor(
        self,
2963
        predictor: Optional[_InnerPredictor],
2964
    ) -> "Dataset":
2965
2966
2967
2968
        """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
2969
        """
2970
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2971
            return self
2972
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
2973
2974
2975
            if (predictor == self._predictor) and (
                predictor.current_iteration() == self._predictor.current_iteration()
            ):
2976
                return self
2977
        if self._handle is None:
Guolin Ke's avatar
Guolin Ke committed
2978
            self._predictor = predictor
2979
2980
        elif self.data is not None:
            self._predictor = predictor
2981
2982
2983
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
2984
                used_indices=None,
2985
            )
2986
2987
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2988
2989
2990
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.reference.data,
2991
                used_indices=self.used_indices,
2992
            )
Guolin Ke's avatar
Guolin Ke committed
2993
        else:
2994
2995
2996
2997
            raise LightGBMError(
                "Cannot set predictor after freed raw data, "
                "set free_raw_data=False when construct Dataset to avoid this."
            )
2998
        return self
Guolin Ke's avatar
Guolin Ke committed
2999

3000
    def set_reference(self, reference: "Dataset") -> "Dataset":
3001
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
3002
3003
3004
3005

        Parameters
        ----------
        reference : Dataset
3006
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
3007
3008
3009
3010
3011

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
3012
        """
3013
3014
3015
        self.set_categorical_feature(reference.categorical_feature).set_feature_name(
            reference.feature_name
        )._set_predictor(reference._predictor)
3016
        # we're done if self and reference share a common upstream reference
3017
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
3018
            return self
Guolin Ke's avatar
Guolin Ke committed
3019
3020
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
3021
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
3022
        else:
3023
3024
3025
3026
            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
3027

3028
    def set_feature_name(self, feature_name: _LGBM_FeatureNameConfiguration) -> "Dataset":
3029
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
3030
3031
3032

        Parameters
        ----------
3033
        feature_name : list of str
3034
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
3035
3036
3037
3038
3039

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
3040
        """
3041
        if feature_name != "auto":
3042
            self.feature_name = feature_name
3043
        if self._handle is not None and feature_name is not None and feature_name != "auto":
wxchan's avatar
wxchan committed
3044
            if len(feature_name) != self.num_feature():
3045
3046
3047
                raise ValueError(
                    f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match"
                )
3048
            c_feature_name = [_c_str(name) for name in feature_name]
3049
3050
3051
3052
3053
3054
3055
            _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
3056
        return self
Guolin Ke's avatar
Guolin Ke committed
3057

3058
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
3059
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
3060
3061
3062

        Parameters
        ----------
3063
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array, pyarrow ChunkedArray or None
3064
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
3065
3066
3067
3068
3069

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
3070
3071
        """
        self.label = label
3072
        if self._handle is not None:
3073
3074
            if isinstance(label, pd_DataFrame):
                if len(label.columns) > 1:
3075
                    raise ValueError("DataFrame for label cannot have multiple columns")
3076
                label_array = np.ravel(_pandas_to_numpy(label, target_dtype=np.float32))
3077
3078
            elif _is_pyarrow_array(label):
                label_array = label
3079
            else:
3080
3081
3082
                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
3083
        return self
Guolin Ke's avatar
Guolin Ke committed
3084

3085
3086
    def set_weight(
        self,
3087
        weight: Optional[_LGBM_WeightType],
3088
    ) -> "Dataset":
3089
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
3090
3091
3092

        Parameters
        ----------
3093
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None
3094
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
3095
3096
3097
3098
3099

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
3100
        """
3101
3102
3103
3104
3105
3106
3107
        # 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
3108
        self.weight = weight
3109
3110

        # Set field
3111
        if self._handle is not None and weight is not None:
3112
            if not _is_pyarrow_array(weight):
3113
3114
3115
                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
3116
        return self
Guolin Ke's avatar
Guolin Ke committed
3117

3118
3119
    def set_init_score(
        self,
3120
        init_score: Optional[_LGBM_InitScoreType],
3121
    ) -> "Dataset":
3122
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
3123
3124
3125

        Parameters
        ----------
3126
        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
3127
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
3128
3129
3130
3131
3132

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
3133
3134
        """
        self.init_score = init_score
3135
        if self._handle is not None and init_score is not None:
3136
3137
            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
3138
        return self
Guolin Ke's avatar
Guolin Ke committed
3139

3140
3141
    def set_group(
        self,
3142
        group: Optional[_LGBM_GroupType],
3143
    ) -> "Dataset":
3144
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
3145
3146
3147

        Parameters
        ----------
3148
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None
3149
3150
3151
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
3152
3153
            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
3154
3155
3156
3157
3158

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
3159
3160
        """
        self.group = group
3161
        if self._handle is not None and group is not None:
3162
            if not _is_pyarrow_array(group):
3163
3164
                group = _list_to_1d_numpy(group, dtype=np.int32, name="group")
            self.set_field("group", group)
3165
            # original values can be modified at cpp side
3166
            constructed_group = self.get_field("group")
3167
3168
            if constructed_group is not None:
                self.group = np.diff(constructed_group)
Nikita Titov's avatar
Nikita Titov committed
3169
        return self
Guolin Ke's avatar
Guolin Ke committed
3170

3171
3172
    def set_position(
        self,
3173
        position: Optional[_LGBM_PositionType],
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
    ) -> "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:
3189
3190
            position = _list_to_1d_numpy(position, dtype=np.int32, name="position")
            self.set_field("position", position)
3191
3192
        return self

3193
    def get_feature_name(self) -> List[str]:
3194
3195
3196
3197
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
3198
        feature_names : list of str
3199
3200
            The names of columns (features) in the Dataset.
        """
3201
        if self._handle is None:
3202
3203
3204
3205
3206
            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)
3207
        string_buffers = [ctypes.create_string_buffer(reserved_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
        _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,
            )
        )
3219
3220
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
3221
3222
3223
3224
        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)]
3225
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
            _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)]
3237

3238
    def get_label(self) -> Optional[_LGBM_LabelType]:
3239
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3240
3241
3242

        Returns
        -------
3243
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
3244
            The label information from the Dataset.
3245
            For a constructed ``Dataset``, this will only return a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3246
        """
3247
        if self.label is None:
3248
            self.label = self.get_field("label")
Guolin Ke's avatar
Guolin Ke committed
3249
3250
        return self.label

3251
    def get_weight(self) -> Optional[_LGBM_WeightType]:
3252
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3253
3254
3255

        Returns
        -------
3256
        weight : list, numpy 1-D array, pandas Series or None
3257
            Weight for each data point from the Dataset. Weights should be non-negative.
3258
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3259
        """
3260
        if self.weight is None:
3261
            self.weight = self.get_field("weight")
Guolin Ke's avatar
Guolin Ke committed
3262
3263
        return self.weight

3264
    def get_init_score(self) -> Optional[_LGBM_InitScoreType]:
3265
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3266
3267
3268

        Returns
        -------
3269
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
3270
            Init score of Booster.
3271
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3272
        """
3273
        if self.init_score is None:
3274
            self.init_score = self.get_field("init_score")
Guolin Ke's avatar
Guolin Ke committed
3275
3276
        return self.init_score

3277
    def get_data(self) -> Optional[_LGBM_TrainDataType]:
3278
3279
3280
3281
        """Get the raw data of the Dataset.

        Returns
        -------
3282
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame (deprecated), scipy.sparse, Sequence, list of Sequence or list of numpy array or None
3283
3284
            Raw data used in the Dataset construction.
        """
3285
        if self._handle is None:
3286
            raise Exception("Cannot get data before construct Dataset")
3287
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
3288
3289
            self.data = self.reference.data
            if self.data is not None:
3290
                if isinstance(self.data, (np.ndarray, scipy.sparse.spmatrix)):
Guolin Ke's avatar
Guolin Ke committed
3291
                    self.data = self.data[self.used_indices, :]
3292
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3293
                    self.data = self.data.iloc[self.used_indices].copy()
3294
                elif isinstance(self.data, dt_DataTable):
3295
                    _emit_datatable_deprecation_warning()
Guolin Ke's avatar
Guolin Ke committed
3296
                    self.data = self.data[self.used_indices, :]
3297
3298
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
3299
                elif _is_list_of_sequences(self.data) and len(self.data) > 0:
3300
                    self.data = np.array(list(self._yield_row_from_seqlist(self.data, self.used_indices)))
Guolin Ke's avatar
Guolin Ke committed
3301
                else:
3302
3303
3304
                    _log_warning(
                        f"Cannot subset {type(self.data).__name__} type of raw data.\n" "Returning original raw data"
                    )
3305
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
3306
        if self.data is None:
3307
3308
3309
3310
            raise LightGBMError(
                "Cannot call `get_data` after freed raw data, "
                "set free_raw_data=False when construct Dataset to avoid this."
            )
3311
3312
        return self.data

3313
    def get_group(self) -> Optional[_LGBM_GroupType]:
3314
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3315
3316
3317

        Returns
        -------
3318
        group : list, numpy 1-D array, pandas Series or None
3319
3320
3321
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
3322
3323
            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.
3324
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
Guolin Ke's avatar
Guolin Ke committed
3325
        """
3326
        if self.group is None:
3327
            self.group = self.get_field("group")
Guolin Ke's avatar
Guolin Ke committed
3328
3329
            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
3330
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
3331
3332
        return self.group

3333
    def get_position(self) -> Optional[_LGBM_PositionType]:
3334
3335
3336
3337
        """Get the position of the Dataset.

        Returns
        -------
3338
        position : numpy 1-D array, pandas Series or None
3339
            Position of items used in unbiased learning-to-rank task.
3340
            For a constructed ``Dataset``, this will only return ``None`` or a numpy array.
3341
3342
        """
        if self.position is None:
3343
            self.position = self.get_field("position")
3344
3345
        return self.position

3346
    def num_data(self) -> int:
3347
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3348
3349
3350

        Returns
        -------
3351
3352
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3353
        """
3354
        if self._handle is not None:
3355
            ret = ctypes.c_int(0)
3356
3357
3358
3359
3360
3361
            _safe_call(
                _LIB.LGBM_DatasetGetNumData(
                    self._handle,
                    ctypes.byref(ret),
                )
            )
wxchan's avatar
wxchan committed
3362
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3363
        else:
3364
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3365

3366
    def num_feature(self) -> int:
3367
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3368
3369
3370

        Returns
        -------
3371
3372
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
3373
        """
3374
        if self._handle is not None:
3375
            ret = ctypes.c_int(0)
3376
3377
3378
3379
3380
3381
            _safe_call(
                _LIB.LGBM_DatasetGetNumFeature(
                    self._handle,
                    ctypes.byref(ret),
                )
            )
wxchan's avatar
wxchan committed
3382
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
3383
        else:
3384
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
3385

3386
    def feature_num_bin(self, feature: Union[int, str]) -> int:
3387
3388
        """Get the number of bins for a feature.

3389
3390
        .. versionadded:: 4.0.0

3391
3392
        Parameters
        ----------
3393
3394
        feature : int or str
            Index or name of the feature.
3395
3396
3397
3398
3399
3400

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
3401
        if self._handle is not None:
3402
            if isinstance(feature, str):
3403
3404
3405
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
3406
            ret = ctypes.c_int(0)
3407
3408
3409
3410
3411
3412
3413
            _safe_call(
                _LIB.LGBM_DatasetGetFeatureNumBin(
                    self._handle,
                    ctypes.c_int(feature_index),
                    ctypes.byref(ret),
                )
            )
3414
3415
3416
3417
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

3418
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
3419
3420
3421
3422
3423
        """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.
3424
3425
3426
3427
3428

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
3429
3430
3431

        Returns
        -------
3432
3433
3434
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
3435
        head = self
3436
        ref_chain: Set[Dataset] = set()
3437
3438
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
3439
                ref_chain.add(head)
3440
3441
3442
3443
3444
3445
                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
3446
        return ref_chain
3447

3448
    def add_features_from(self, other: "Dataset") -> "Dataset":
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
        """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.
        """
3463
        if self._handle is None or other._handle is None:
3464
3465
3466
3467
3468
3469
3470
            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
3471
3472
3473
3474
3475
3476
3477
3478
        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))
3479
                elif isinstance(other.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3480
                    self.data = np.hstack((self.data, other.data.toarray()))
3481
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3482
                    self.data = np.hstack((self.data, other.data.values))
3483
                elif isinstance(other.data, dt_DataTable):
3484
                    _emit_datatable_deprecation_warning()
Guolin Ke's avatar
Guolin Ke committed
3485
3486
3487
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
3488
            elif isinstance(self.data, scipy.sparse.spmatrix):
Guolin Ke's avatar
Guolin Ke committed
3489
                sparse_format = self.data.getformat()
3490
                if isinstance(other.data, (np.ndarray, scipy.sparse.spmatrix)):
Guolin Ke's avatar
Guolin Ke committed
3491
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
3492
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3493
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
3494
                elif isinstance(other.data, dt_DataTable):
3495
                    _emit_datatable_deprecation_warning()
Guolin Ke's avatar
Guolin Ke committed
3496
3497
3498
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
3499
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
3500
                if not PANDAS_INSTALLED:
3501
3502
3503
3504
3505
                    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
3506
                if isinstance(other.data, np.ndarray):
3507
                    self.data = concat((self.data, pd_DataFrame(other.data)), axis=1, ignore_index=True)
3508
                elif isinstance(other.data, scipy.sparse.spmatrix):
3509
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())), axis=1, ignore_index=True)
3510
                elif isinstance(other.data, pd_DataFrame):
3511
                    self.data = concat((self.data, other.data), axis=1, ignore_index=True)
3512
                elif isinstance(other.data, dt_DataTable):
3513
                    _emit_datatable_deprecation_warning()
3514
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())), axis=1, ignore_index=True)
Guolin Ke's avatar
Guolin Ke committed
3515
3516
                else:
                    self.data = None
3517
            elif isinstance(self.data, dt_DataTable):
3518
                _emit_datatable_deprecation_warning()
Guolin Ke's avatar
Guolin Ke committed
3519
                if isinstance(other.data, np.ndarray):
3520
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
3521
                elif isinstance(other.data, scipy.sparse.spmatrix):
3522
3523
3524
3525
3526
                    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
3527
3528
3529
3530
3531
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
3532
3533
3534
3535
3536
3537
3538
            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"
            )
3539
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
3540
        self.feature_name = self.get_feature_name()
3541
        _log_warning(
3542
            "Resetting categorical features.\n"
3543
3544
            "You can set new categorical features via ``set_categorical_feature`` method"
        )
Guolin Ke's avatar
Guolin Ke committed
3545
3546
        self.categorical_feature = "auto"
        self.pandas_categorical = None
3547
3548
        return self

3549
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
3550
3551
3552
3553
3554
3555
        """Save Dataset to a text file.

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

        Parameters
        ----------
3556
        filename : str or pathlib.Path
3557
3558
3559
3560
3561
3562
3563
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
3564
3565
3566
3567
3568
3569
        _safe_call(
            _LIB.LGBM_DatasetDumpText(
                self.construct()._handle,
                _c_str(str(filename)),
            )
        )
3570
3571
        return self

wxchan's avatar
wxchan committed
3572

3573
3574
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
3575
    Tuple[np.ndarray, np.ndarray],
3576
]
3577
3578
3579
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
3580
        _LGBM_EvalFunctionResultType,
3581
3582
3583
    ],
    Callable[
        [np.ndarray, Dataset],
3584
3585
        List[_LGBM_EvalFunctionResultType],
    ],
3586
]
3587
3588


3589
class Booster:
3590
    """Booster in LightGBM."""
3591

3592
3593
3594
3595
3596
    def __init__(
        self,
        params: Optional[Dict[str, Any]] = None,
        train_set: Optional[Dataset] = None,
        model_file: Optional[Union[str, Path]] = None,
3597
        model_str: Optional[str] = None,
3598
    ):
3599
        """Initialize the Booster.
wxchan's avatar
wxchan committed
3600
3601
3602

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
3603
        params : dict or None, optional (default=None)
3604
3605
3606
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
3607
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
3608
            Path to the model file.
3609
        model_str : str or None, optional (default=None)
3610
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
3611
        """
3612
        self._handle = ctypes.c_void_p()
3613
        self._network = False
wxchan's avatar
wxchan committed
3614
        self.__need_reload_eval_info = True
3615
        self._train_data_name = "training"
3616
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
3617
        self.best_iteration = -1
3618
        self.best_score: _LGBM_BoosterBestScoreType = {}
3619
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
3620
        if train_set is not None:
3621
            # Training task
wxchan's avatar
wxchan committed
3622
            if not isinstance(train_set, Dataset):
3623
                raise TypeError(f"Training data should be Dataset instance, met {type(train_set).__name__}")
3624
3625
3626
            params = _choose_param_value(
                main_param_name="machines",
                params=params,
3627
                default_value=None,
3628
3629
3630
3631
3632
3633
3634
            )
            # 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):
3635
                    num_machines_from_machine_list = len(machines.split(","))
3636
3637
                elif isinstance(machines, (list, set)):
                    num_machines_from_machine_list = len(machines)
3638
                    machines = ",".join(machines)
3639
3640
3641
3642
3643
3644
                else:
                    raise ValueError("Invalid machines in params.")

                params = _choose_param_value(
                    main_param_name="num_machines",
                    params=params,
3645
                    default_value=num_machines_from_machine_list,
3646
3647
3648
3649
                )
                params = _choose_param_value(
                    main_param_name="local_listen_port",
                    params=params,
3650
                    default_value=12400,
3651
3652
3653
3654
3655
                )
                self.set_network(
                    machines=machines,
                    local_listen_port=params["local_listen_port"],
                    listen_time_out=params.get("time_out", 120),
3656
                    num_machines=params["num_machines"],
3657
                )
3658
            # construct booster object
3659
3660
3661
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
3662
            params_str = _param_dict_to_str(params)
3663
3664
3665
3666
3667
3668
3669
            _safe_call(
                _LIB.LGBM_BoosterCreate(
                    train_set._handle,
                    _c_str(params_str),
                    ctypes.byref(self._handle),
                )
            )
3670
            # save reference to data
wxchan's avatar
wxchan committed
3671
            self.train_set = train_set
3672
3673
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
3674
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
3675
3676
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
3677
3678
3679
3680
3681
3682
                _safe_call(
                    _LIB.LGBM_BoosterMerge(
                        self._handle,
                        self.__init_predictor._handle,
                    )
                )
Guolin Ke's avatar
Guolin Ke committed
3683
            out_num_class = ctypes.c_int(0)
3684
3685
3686
3687
3688
3689
            _safe_call(
                _LIB.LGBM_BoosterGetNumClasses(
                    self._handle,
                    ctypes.byref(out_num_class),
                )
            )
wxchan's avatar
wxchan committed
3690
            self.__num_class = out_num_class.value
3691
            # buffer for inner predict
3692
            self.__inner_predict_buffer: List[Optional[np.ndarray]] = [None]
wxchan's avatar
wxchan committed
3693
3694
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
3695
            self.pandas_categorical = train_set.pandas_categorical
3696
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
3697
        elif model_file is not None:
3698
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
3699
            out_num_iterations = ctypes.c_int(0)
3700
3701
3702
3703
3704
3705
3706
            _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
3707
            out_num_class = ctypes.c_int(0)
3708
3709
3710
3711
3712
3713
            _safe_call(
                _LIB.LGBM_BoosterGetNumClasses(
                    self._handle,
                    ctypes.byref(out_num_class),
                )
            )
wxchan's avatar
wxchan committed
3714
            self.__num_class = out_num_class.value
3715
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
3716
            if params:
3717
                _log_warning("Ignoring params argument, using parameters from model file.")
3718
            params = self._get_loaded_param()
3719
        elif model_str is not None:
3720
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
3721
        else:
3722
3723
3724
            raise TypeError(
                "Need at least one training dataset or model file or model string " "to create Booster instance"
            )
3725
        self.params = params
wxchan's avatar
wxchan committed
3726

3727
    def __del__(self) -> None:
3728
        try:
3729
            if self._network:
3730
3731
3732
3733
                self.free_network()
        except AttributeError:
            pass
        try:
3734
3735
            if self._handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self._handle))
3736
3737
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3738

3739
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3740
3741
        return self.__deepcopy__(None)

3742
    def __deepcopy__(self, *args: Any, **kwargs: Any) -> "Booster":
3743
        model_str = self.model_to_string(num_iteration=-1)
3744
        return Booster(model_str=model_str)
wxchan's avatar
wxchan committed
3745

3746
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3747
        this = self.__dict__.copy()
3748
3749
3750
        handle = this["_handle"]
        this.pop("train_set", None)
        this.pop("valid_sets", None)
wxchan's avatar
wxchan committed
3751
        if handle is not None:
3752
            this["_handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3753
3754
        return this

3755
    def __setstate__(self, state: Dict[str, Any]) -> None:
3756
        model_str = state.get("_handle", state.get("handle", None))
3757
        if model_str is not None:
wxchan's avatar
wxchan committed
3758
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3759
            out_num_iterations = ctypes.c_int(0)
3760
3761
3762
3763
3764
3765
3766
3767
            _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
3768
3769
        self.__dict__.update(state)

3770
3771
3772
3773
    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)
3774
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3775
3776
3777
3778
3779
3780
3781
3782
        _safe_call(
            _LIB.LGBM_BoosterGetLoadedParam(
                self._handle,
                ctypes.c_int64(buffer_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer,
            )
        )
3783
3784
3785
3786
        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)
3787
            ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
3788
3789
3790
3791
3792
3793
3794
3795
3796
            _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"))
3797

3798
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3799
3800
3801
3802
3803
3804
3805
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
3806
3807
        self.__dict__.pop("train_set", None)
        self.__dict__.pop("valid_sets", None)
3808
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3809
        return self
wxchan's avatar
wxchan committed
3810

3811
    def _free_buffer(self) -> "Booster":
3812
3813
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3814
        return self
3815

3816
3817
3818
3819
3820
    def set_network(
        self,
        machines: Union[List[str], Set[str], str],
        local_listen_port: int = 12400,
        listen_time_out: int = 120,
3821
        num_machines: int = 1,
3822
    ) -> "Booster":
3823
3824
3825
3826
        """Set the network configuration.

        Parameters
        ----------
3827
        machines : list, set or str
3828
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3829
        local_listen_port : int, optional (default=12400)
3830
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3831
        listen_time_out : int, optional (default=120)
3832
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3833
        num_machines : int, optional (default=1)
3834
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3835
3836
3837
3838
3839

        Returns
        -------
        self : Booster
            Booster with set network.
3840
        """
3841
        if isinstance(machines, (list, set)):
3842
3843
3844
3845
3846
3847
3848
3849
3850
            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),
            )
        )
3851
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3852
        return self
3853

3854
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3855
3856
3857
3858
3859
3860
3861
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3862
        _safe_call(_LIB.LGBM_NetworkFree())
3863
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3864
        return self
3865

3866
    def trees_to_dataframe(self) -> pd_DataFrame:
3867
3868
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3869
3870
3871
3872
        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.
3873
3874
3875
3876
3877
            - ``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.
3878
3879
            - ``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.
3880
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3881
3882
              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.
3883
3884
            - ``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.
3885
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3886
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3887
3888
            - ``count`` : int64, number of records in the training data that fall into this node.

3889
3890
3891
3892
3893
3894
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3895
3896
3897
3898
            raise LightGBMError(
                "This method cannot be run without pandas installed. "
                "You must install pandas and restart your session to use this method."
            )
3899
3900

        if self.num_trees() == 0:
3901
            raise LightGBMError("There are no trees in this Booster and thus nothing to parse")
3902

3903
        def _is_split_node(tree: Dict[str, Any]) -> bool:
3904
            return "split_index" in tree.keys()
3905

3906
3907
3908
3909
3910
        def create_node_record(
            tree: Dict[str, Any],
            node_depth: int = 1,
            tree_index: Optional[int] = None,
            feature_names: Optional[List[str]] = None,
3911
            parent_node: Optional[str] = None,
3912
3913
3914
        ) -> Dict[str, Any]:
            def _get_node_index(
                tree: Dict[str, Any],
3915
                tree_index: Optional[int],
3916
            ) -> str:
3917
                tree_num = f"{tree_index}-" if tree_index is not None else ""
3918
                is_split = _is_split_node(tree)
3919
                node_type = "S" if is_split else "L"
3920
                # if a single node tree it won't have `leaf_index` so return 0
3921
                node_num = tree.get("split_index" if is_split else "leaf_index", 0)
3922
                return f"{tree_num}{node_type}{node_num}"
3923

3924
3925
            def _get_split_feature(
                tree: Dict[str, Any],
3926
                feature_names: Optional[List[str]],
3927
            ) -> Optional[str]:
3928
3929
                if _is_split_node(tree):
                    if feature_names is not None:
3930
                        feature_name = feature_names[tree["split_feature"]]
3931
                    else:
3932
                        feature_name = tree["split_feature"]
3933
3934
3935
3936
                else:
                    feature_name = None
                return feature_name

3937
            def _is_single_node_tree(tree: Dict[str, Any]) -> bool:
3938
                return set(tree.keys()) == {"leaf_value", "leaf_count"}
3939
3940

            # Create the node record, and populate universal data members
3941
            node: Dict[str, Union[int, str, None]] = OrderedDict()
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
            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
3957
3958
3959

            # Update values to reflect node type (leaf or split)
            if _is_split_node(tree):
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
                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"]
3970
            else:
3971
                node["value"] = tree["leaf_value"]
3972
                if not _is_single_node_tree(tree):
3973
3974
                    node["weight"] = tree["leaf_weight"]
                    node["count"] = tree["leaf_count"]
3975
3976
3977

            return node

3978
3979
3980
3981
3982
        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,
3983
            parent_node: Optional[str] = None,
3984
        ) -> List[Dict[str, Any]]:
3985
3986
3987
3988
3989
3990
3991
            node = create_node_record(
                tree=tree,
                node_depth=node_depth,
                tree_index=tree_index,
                feature_names=feature_names,
                parent_node=parent_node,
            )
3992
3993
3994
3995
3996

            res = [node]

            if _is_split_node(tree):
                # traverse the next level of the tree
3997
                children = ["left_child", "right_child"]
3998
3999
                for child in children:
                    subtree_list = tree_dict_to_node_list(
4000
                        tree=tree[child],
4001
4002
4003
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
4004
                        parent_node=node["node_index"],
4005
                    )
4006
4007
4008
4009
4010
4011
                    # 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()
4012
        feature_names = model_dict["feature_names"]
4013
        model_list = []
4014
4015
4016
4017
4018
4019
        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
                )
            )
4020

4021
        return pd_DataFrame(model_list, columns=model_list[0].keys())
4022

4023
    def set_train_data_name(self, name: str) -> "Booster":
4024
4025
4026
4027
        """Set the name to the training Dataset.

        Parameters
        ----------
4028
        name : str
Nikita Titov's avatar
Nikita Titov committed
4029
4030
4031
4032
4033
4034
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
4035
        """
4036
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
4037
        return self
wxchan's avatar
wxchan committed
4038

4039
    def add_valid(self, data: Dataset, name: str) -> "Booster":
4040
        """Add validation data.
wxchan's avatar
wxchan committed
4041
4042
4043
4044

        Parameters
        ----------
        data : Dataset
4045
            Validation data.
4046
        name : str
4047
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
4048
4049
4050
4051
4052

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
4053
        """
Guolin Ke's avatar
Guolin Ke committed
4054
        if not isinstance(data, Dataset):
4055
            raise TypeError(f"Validation data should be Dataset instance, met {type(data).__name__}")
Guolin Ke's avatar
Guolin Ke committed
4056
        if data._predictor is not self.__init_predictor:
4057
4058
4059
4060
4061
4062
4063
            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
4064
4065
4066
4067
4068
        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
4069
        return self
wxchan's avatar
wxchan committed
4070

4071
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
4072
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
4073
4074
4075
4076

        Parameters
        ----------
        params : dict
4077
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
4078
4079
4080
4081
4082

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
4083
        """
4084
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
4085
        if params_str:
4086
4087
4088
4089
4090
4091
            _safe_call(
                _LIB.LGBM_BoosterResetParameter(
                    self._handle,
                    _c_str(params_str),
                )
            )
Guolin Ke's avatar
Guolin Ke committed
4092
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
4093
        return self
wxchan's avatar
wxchan committed
4094

4095
4096
4097
    def update(
        self,
        train_set: Optional[Dataset] = None,
4098
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None,
4099
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
4100
        """Update Booster for one iteration.
4101

wxchan's avatar
wxchan committed
4102
4103
        Parameters
        ----------
4104
4105
4106
4107
        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
4108
            Customized objective function.
4109
4110
4111
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

4112
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4113
                    The predicted values.
4114
4115
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
4116
4117
                train_data : Dataset
                    The training dataset.
4118
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
4119
4120
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
4121
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
4122
4123
                    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
4124

4125
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
4126
            and grad and hess should be returned in the same format.
4127

wxchan's avatar
wxchan committed
4128
4129
        Returns
        -------
4130
4131
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
4132
        """
4133
        # need reset training data
4134
4135
4136
4137
4138
4139
        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
4140
            if not isinstance(train_set, Dataset):
4141
                raise TypeError(f"Training data should be Dataset instance, met {type(train_set).__name__}")
Guolin Ke's avatar
Guolin Ke committed
4142
            if train_set._predictor is not self.__init_predictor:
4143
                raise LightGBMError("Replace training data failed, " "you should use same predictor for these data")
wxchan's avatar
wxchan committed
4144
            self.train_set = train_set
4145
4146
4147
4148
4149
4150
            _safe_call(
                _LIB.LGBM_BoosterResetTrainingData(
                    self._handle,
                    self.train_set.construct()._handle,
                )
            )
wxchan's avatar
wxchan committed
4151
            self.__inner_predict_buffer[0] = None
4152
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
4153
4154
        is_finished = ctypes.c_int(0)
        if fobj is None:
4155
            if self.__set_objective_to_none:
4156
4157
4158
4159
4160
4161
4162
                raise LightGBMError("Cannot update due to null objective function.")
            _safe_call(
                _LIB.LGBM_BoosterUpdateOneIter(
                    self._handle,
                    ctypes.byref(is_finished),
                )
            )
4163
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
4164
4165
            return is_finished.value == 1
        else:
4166
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
4167
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
4168
4169
4170
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

4171
4172
4173
    def __boost(
        self,
        grad: np.ndarray,
4174
        hess: np.ndarray,
4175
    ) -> bool:
4176
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
4177

Nikita Titov's avatar
Nikita Titov committed
4178
4179
        .. note::

4180
4181
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
4182
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
4183
            and grad and hess should be returned in the same format.
4184

wxchan's avatar
wxchan committed
4185
4186
        Parameters
        ----------
4187
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
4188
4189
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
4190
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
4191
4192
            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
4193
4194
4195

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4196
4197
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
4198
        """
4199
        if self.__num_class > 1:
4200
4201
4202
4203
            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")
4204
4205
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
4206
        if len(grad) != len(hess):
4207
4208
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
4209
        if len(grad) != num_train_data * self.__num_class:
4210
4211
4212
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
4213
                f"number of models per one iteration ({self.__num_class})"
4214
            )
wxchan's avatar
wxchan committed
4215
        is_finished = ctypes.c_int(0)
4216
4217
4218
4219
4220
4221
4222
4223
        _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),
            )
        )
4224
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
4225
4226
        return is_finished.value == 1

4227
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
4228
4229
4230
4231
4232
4233
4234
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
4235
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(self._handle))
4236
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
4237
        return self
wxchan's avatar
wxchan committed
4238

4239
    def current_iteration(self) -> int:
4240
4241
4242
4243
4244
4245
4246
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
4247
        out_cur_iter = ctypes.c_int(0)
4248
4249
4250
4251
4252
4253
        _safe_call(
            _LIB.LGBM_BoosterGetCurrentIteration(
                self._handle,
                ctypes.byref(out_cur_iter),
            )
        )
wxchan's avatar
wxchan committed
4254
4255
        return out_cur_iter.value

4256
    def num_model_per_iteration(self) -> int:
4257
4258
4259
4260
4261
4262
4263
4264
        """Get number of models per iteration.

        Returns
        -------
        model_per_iter : int
            The number of models per iteration.
        """
        model_per_iter = ctypes.c_int(0)
4265
4266
4267
4268
4269
4270
        _safe_call(
            _LIB.LGBM_BoosterNumModelPerIteration(
                self._handle,
                ctypes.byref(model_per_iter),
            )
        )
4271
4272
        return model_per_iter.value

4273
    def num_trees(self) -> int:
4274
4275
4276
4277
4278
4279
4280
4281
        """Get number of weak sub-models.

        Returns
        -------
        num_trees : int
            The number of weak sub-models.
        """
        num_trees = ctypes.c_int(0)
4282
4283
4284
4285
4286
4287
        _safe_call(
            _LIB.LGBM_BoosterNumberOfTotalModel(
                self._handle,
                ctypes.byref(num_trees),
            )
        )
4288
4289
        return num_trees.value

4290
    def upper_bound(self) -> float:
4291
4292
4293
4294
        """Get upper bound value of a model.

        Returns
        -------
4295
        upper_bound : float
4296
4297
4298
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
4299
4300
4301
4302
4303
4304
        _safe_call(
            _LIB.LGBM_BoosterGetUpperBoundValue(
                self._handle,
                ctypes.byref(ret),
            )
        )
4305
4306
        return ret.value

4307
    def lower_bound(self) -> float:
4308
4309
4310
4311
        """Get lower bound value of a model.

        Returns
        -------
4312
        lower_bound : float
4313
4314
4315
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
4316
4317
4318
4319
4320
4321
        _safe_call(
            _LIB.LGBM_BoosterGetLowerBoundValue(
                self._handle,
                ctypes.byref(ret),
            )
        )
4322
4323
        return ret.value

4324
4325
4326
4327
    def eval(
        self,
        data: Dataset,
        name: str,
4328
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None,
4329
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4330
        """Evaluate for data.
wxchan's avatar
wxchan committed
4331
4332
4333

        Parameters
        ----------
4334
4335
        data : Dataset
            Data for the evaluating.
4336
        name : str
4337
            Name of the data.
4338
        feval : callable, list of callable, or None, optional (default=None)
4339
            Customized evaluation function.
4340
            Each evaluation function should accept two parameters: preds, eval_data,
4341
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4342

4343
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4344
                    The predicted values.
4345
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4346
                    If custom objective function is used, predicted values are returned before any transformation,
4347
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
4348
                eval_data : Dataset
4349
                    A ``Dataset`` to evaluate.
4350
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4351
                    The name of evaluation function (without whitespace).
4352
4353
4354
4355
4356
                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
4357
4358
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4359
        result : list
4360
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4361
        """
Guolin Ke's avatar
Guolin Ke committed
4362
4363
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
4364
4365
4366
4367
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
4368
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
4369
4370
4371
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
4372
        # need to push new valid data
wxchan's avatar
wxchan committed
4373
4374
4375
4376
4377
4378
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

4379
4380
    def eval_train(
        self,
4381
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None,
4382
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4383
        """Evaluate for training data.
wxchan's avatar
wxchan committed
4384
4385
4386

        Parameters
        ----------
4387
        feval : callable, list of callable, or None, optional (default=None)
4388
            Customized evaluation function.
4389
            Each evaluation function should accept two parameters: preds, eval_data,
4390
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4391

4392
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4393
                    The predicted values.
4394
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4395
                    If custom objective function is used, predicted values are returned before any transformation,
4396
                    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
4397
                eval_data : Dataset
4398
                    The training dataset.
4399
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4400
                    The name of evaluation function (without whitespace).
4401
4402
4403
4404
4405
                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
4406
4407
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4408
        result : list
4409
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4410
        """
4411
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
4412

4413
4414
    def eval_valid(
        self,
4415
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None,
4416
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4417
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
4418
4419
4420

        Parameters
        ----------
4421
        feval : callable, list of callable, or None, optional (default=None)
4422
            Customized evaluation function.
4423
            Each evaluation function should accept two parameters: preds, eval_data,
4424
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
4425

4426
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
4427
                    The predicted values.
4428
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
4429
                    If custom objective function is used, predicted values are returned before any transformation,
4430
                    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
4431
                eval_data : Dataset
4432
                    The validation dataset.
4433
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
4434
                    The name of evaluation function (without whitespace).
4435
4436
4437
4438
4439
                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
4440
4441
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4442
        result : list
4443
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
4444
        """
4445
4446
4447
4448
4449
        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
4450

4451
4452
4453
4454
4455
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
4456
        importance_type: str = "split",
4457
    ) -> "Booster":
4458
        """Save Booster to file.
wxchan's avatar
wxchan committed
4459
4460
4461

        Parameters
        ----------
4462
        filename : str or pathlib.Path
4463
            Filename to save Booster.
4464
4465
4466
4467
        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
4468
        start_iteration : int, optional (default=0)
4469
            Start index of the iteration that should be saved.
4470
        importance_type : str, optional (default="split")
4471
4472
4473
            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
4474
4475
4476
4477
4478

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
4479
        """
4480
        if num_iteration is None:
4481
            num_iteration = self.best_iteration
4482
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4483
4484
4485
4486
4487
4488
4489
4490
4491
        _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)),
            )
        )
4492
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
4493
        return self
wxchan's avatar
wxchan committed
4494

4495
4496
4497
    def shuffle_models(
        self,
        start_iteration: int = 0,
4498
        end_iteration: int = -1,
4499
    ) -> "Booster":
4500
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
4501

4502
4503
4504
        Parameters
        ----------
        start_iteration : int, optional (default=0)
4505
            The first iteration that will be shuffled.
4506
4507
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
4508
            If <= 0, means the last available iteration.
4509

Nikita Titov's avatar
Nikita Titov committed
4510
4511
4512
4513
        Returns
        -------
        self : Booster
            Booster with shuffled models.
4514
        """
4515
4516
4517
4518
4519
4520
4521
        _safe_call(
            _LIB.LGBM_BoosterShuffleModels(
                self._handle,
                ctypes.c_int(start_iteration),
                ctypes.c_int(end_iteration),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
4522
        return self
4523

4524
    def model_from_string(self, model_str: str) -> "Booster":
4525
4526
4527
4528
        """Load Booster from a string.

        Parameters
        ----------
4529
        model_str : str
4530
4531
4532
4533
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
4534
        self : Booster
4535
4536
            Loaded Booster object.
        """
4537
4538
4539
        # ensure that existing Booster is freed before replacing it
        # with a new one createdfrom file
        _safe_call(_LIB.LGBM_BoosterFree(self._handle))
4540
        self._free_buffer()
4541
        self._handle = ctypes.c_void_p()
4542
        out_num_iterations = ctypes.c_int(0)
4543
4544
4545
4546
4547
4548
4549
        _safe_call(
            _LIB.LGBM_BoosterLoadModelFromString(
                _c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(self._handle),
            )
        )
4550
        out_num_class = ctypes.c_int(0)
4551
4552
4553
4554
4555
4556
        _safe_call(
            _LIB.LGBM_BoosterGetNumClasses(
                self._handle,
                ctypes.byref(out_num_class),
            )
        )
4557
        self.__num_class = out_num_class.value
4558
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
4559
4560
        return self

4561
4562
4563
4564
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
4565
        importance_type: str = "split",
4566
    ) -> str:
4567
        """Save Booster to string.
4568

4569
4570
4571
4572
4573
4574
        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
4575
        start_iteration : int, optional (default=0)
4576
            Start index of the iteration that should be saved.
4577
        importance_type : str, optional (default="split")
4578
4579
4580
            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.
4581
4582
4583

        Returns
        -------
4584
        str_repr : str
4585
4586
            String representation of Booster.
        """
4587
        if num_iteration is None:
4588
            num_iteration = self.best_iteration
4589
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4590
        buffer_len = 1 << 20
4591
        tmp_out_len = ctypes.c_int64(0)
4592
        string_buffer = ctypes.create_string_buffer(buffer_len)
4593
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4594
4595
        _safe_call(
            _LIB.LGBM_BoosterSaveModelToString(
4596
                self._handle,
4597
                ctypes.c_int(start_iteration),
4598
                ctypes.c_int(num_iteration),
4599
                ctypes.c_int(importance_type_int),
4600
                ctypes.c_int64(buffer_len),
4601
                ctypes.byref(tmp_out_len),
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
                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")
4622
4623
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
4624

4625
4626
4627
4628
    def dump_model(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
4629
4630
        importance_type: str = "split",
        object_hook: Optional[Callable[[Dict[str, Any]], Dict[str, Any]]] = None,
4631
    ) -> Dict[str, Any]:
Nikita Titov's avatar
Nikita Titov committed
4632
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
4633

4634
4635
        Parameters
        ----------
4636
4637
4638
4639
        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
4640
        start_iteration : int, optional (default=0)
4641
            Start index of the iteration that should be dumped.
4642
        importance_type : str, optional (default="split")
4643
4644
4645
            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.
4646
4647
4648
4649
4650
4651
4652
4653
4654
        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.
4655

wxchan's avatar
wxchan committed
4656
4657
        Returns
        -------
4658
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
4659
            JSON format of Booster.
wxchan's avatar
wxchan committed
4660
        """
4661
        if num_iteration is None:
4662
            num_iteration = self.best_iteration
4663
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
4664
        buffer_len = 1 << 20
4665
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
4666
        string_buffer = ctypes.create_string_buffer(buffer_len)
4667
        ptr_string_buffer = ctypes.c_char_p(ctypes.addressof(string_buffer))
4668
4669
        _safe_call(
            _LIB.LGBM_BoosterDumpModel(
4670
                self._handle,
4671
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
4672
                ctypes.c_int(num_iteration),
4673
                ctypes.c_int(importance_type_int),
4674
                ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
4675
                ctypes.byref(tmp_out_len),
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
                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,
            )
        )
4702
        return ret
wxchan's avatar
wxchan committed
4703

4704
4705
    def predict(
        self,
4706
        data: _LGBM_PredictDataType,
4707
4708
4709
4710
4711
4712
4713
        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,
4714
        **kwargs: Any,
4715
    ) -> Union[np.ndarray, scipy.sparse.spmatrix, List[scipy.sparse.spmatrix]]:
4716
        """Make a prediction.
wxchan's avatar
wxchan committed
4717
4718
4719

        Parameters
        ----------
4720
        data : str, pathlib.Path, numpy array, pandas DataFrame, pyarrow Table, H2O DataTable's Frame (deprecated) or scipy.sparse
4721
            Data source for prediction.
4722
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4723
        start_iteration : int, optional (default=0)
4724
            Start index of the iteration to predict.
4725
            If <= 0, starts from the first iteration.
4726
        num_iteration : int or None, optional (default=None)
4727
4728
4729
4730
            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).
4731
4732
4733
4734
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
4735
4736
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
4737

Nikita Titov's avatar
Nikita Titov committed
4738
4739
4740
4741
4742
4743
4744
            .. 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.
4745

4746
4747
        data_has_header : bool, optional (default=False)
            Whether the data has header.
4748
            Used only if data is str.
4749
4750
4751
        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.
4752
4753
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
4754
4755
4756

        Returns
        -------
4757
        result : numpy array, scipy.sparse or list of scipy.sparse
4758
            Prediction result.
4759
            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
4760
        """
4761
4762
4763
4764
        predictor = _InnerPredictor.from_booster(
            booster=self,
            pred_parameter=deepcopy(kwargs),
        )
4765
        if num_iteration is None:
4766
            if start_iteration <= 0:
4767
4768
4769
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
4770
4771
4772
4773
4774
4775
4776
4777
        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,
4778
            validate_features=validate_features,
4779
        )
wxchan's avatar
wxchan committed
4780

4781
4782
    def refit(
        self,
4783
        data: _LGBM_TrainDataType,
4784
        label: _LGBM_LabelType,
4785
4786
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
4787
4788
4789
        weight: Optional[_LGBM_WeightType] = None,
        group: Optional[_LGBM_GroupType] = None,
        init_score: Optional[_LGBM_InitScoreType] = None,
4790
4791
        feature_name: _LGBM_FeatureNameConfiguration = "auto",
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = "auto",
4792
4793
4794
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
4795
        **kwargs: Any,
4796
    ) -> "Booster":
Guolin Ke's avatar
Guolin Ke committed
4797
4798
4799
4800
        """Refit the existing Booster by new data.

        Parameters
        ----------
4801
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame (deprecated), scipy.sparse, Sequence, list of Sequence or list of numpy array
Guolin Ke's avatar
Guolin Ke committed
4802
            Data source for refit.
4803
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
4804
        label : list, numpy 1-D array, pandas Series / one-column DataFrame, pyarrow Array or pyarrow ChunkedArray
Guolin Ke's avatar
Guolin Ke committed
4805
4806
            Label for refit.
        decay_rate : float, optional (default=0.9)
4807
4808
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
4809
4810
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
4811
4812
4813

            .. versionadded:: 4.0.0

4814
        weight : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
4815
            Weight for each ``data`` instance. Weights should be non-negative.
4816
4817
4818

            .. versionadded:: 4.0.0

4819
        group : list, numpy 1-D array, pandas Series, pyarrow Array, pyarrow ChunkedArray or None, optional (default=None)
4820
4821
4822
4823
4824
            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.
4825
4826
4827

            .. versionadded:: 4.0.0

4828
        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)
4829
            Init score for ``data``.
4830
4831
4832

            .. versionadded:: 4.0.0

4833
4834
4835
        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.
4836
4837
4838

            .. versionadded:: 4.0.0

4839
4840
4841
4842
4843
        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.
4844
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
4845
4846
4847
            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.
4848
            Floating point numbers in categorical features will be rounded towards 0.
4849
4850
4851

            .. versionadded:: 4.0.0

4852
4853
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
4854
4855
4856

            .. versionadded:: 4.0.0

4857
4858
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
4859
4860
4861

            .. versionadded:: 4.0.0

4862
4863
4864
        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.
4865
4866
4867

            .. versionadded:: 4.0.0

4868
4869
        **kwargs
            Other parameters for refit.
4870
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4871
4872
4873
4874
4875
4876

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4877
        if self.__set_objective_to_none:
4878
            raise LightGBMError("Cannot refit due to null objective function.")
4879
4880
        if dataset_params is None:
            dataset_params = {}
4881
        predictor = _InnerPredictor.from_booster(booster=self, pred_parameter=deepcopy(kwargs))
4882
        leaf_preds: np.ndarray = predictor.predict(  # type: ignore[assignment]
4883
4884
4885
            data=data,
            start_iteration=-1,
            pred_leaf=True,
4886
            validate_features=validate_features,
4887
        )
4888
        nrow, ncol = leaf_preds.shape
4889
        out_is_linear = ctypes.c_int(0)
4890
4891
4892
4893
4894
4895
        _safe_call(
            _LIB.LGBM_BoosterGetLinear(
                self._handle,
                ctypes.byref(out_is_linear),
            )
        )
Nikita Titov's avatar
Nikita Titov committed
4896
4897
4898
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
4899
            default_value=None,
Nikita Titov's avatar
Nikita Titov committed
4900
        )
4901
        new_params["linear_tree"] = bool(out_is_linear.value)
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
        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,
        )
4915
        new_params["refit_decay_rate"] = decay_rate
4916
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4917
        # Copy models
4918
4919
4920
4921
4922
4923
        _safe_call(
            _LIB.LGBM_BoosterMerge(
                new_booster._handle,
                predictor._handle,
            )
        )
Guolin Ke's avatar
Guolin Ke committed
4924
        leaf_preds = leaf_preds.reshape(-1)
4925
        ptr_data, _, _ = _c_int_array(leaf_preds)
4926
4927
4928
4929
4930
4931
4932
4933
        _safe_call(
            _LIB.LGBM_BoosterRefit(
                new_booster._handle,
                ptr_data,
                ctypes.c_int32(nrow),
                ctypes.c_int32(ncol),
            )
        )
4934
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4935
4936
        return new_booster

4937
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
        """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.
        """
4952
        ret = ctypes.c_double(0)
4953
4954
4955
4956
4957
4958
4959
4960
        _safe_call(
            _LIB.LGBM_BoosterGetLeafValue(
                self._handle,
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.byref(ret),
            )
        )
4961
4962
        return ret.value

4963
4964
4965
4966
4967
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
4968
    ) -> "Booster":
4969
4970
        """Set the output of a leaf.

4971
4972
        .. versionadded:: 4.0.0

4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
        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(
4989
                self._handle,
4990
4991
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
4992
                ctypes.c_double(value),
4993
4994
4995
4996
            )
        )
        return self

4997
    def num_feature(self) -> int:
4998
4999
5000
5001
5002
5003
5004
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
5005
        out_num_feature = ctypes.c_int(0)
5006
5007
5008
5009
5010
5011
        _safe_call(
            _LIB.LGBM_BoosterGetNumFeature(
                self._handle,
                ctypes.byref(out_num_feature),
            )
        )
5012
5013
        return out_num_feature.value

5014
    def feature_name(self) -> List[str]:
5015
        """Get names of features.
wxchan's avatar
wxchan committed
5016
5017
5018

        Returns
        -------
5019
        result : list of str
5020
            List with names of features.
wxchan's avatar
wxchan committed
5021
        """
5022
        num_feature = self.num_feature()
5023
        # Get name of features
wxchan's avatar
wxchan committed
5024
        tmp_out_len = ctypes.c_int(0)
5025
5026
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
5027
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
5028
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
        _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
5039
5040
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
5041
5042
5043
5044
        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)]
5045
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
            _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
5057

5058
5059
    def feature_importance(
        self,
5060
5061
        importance_type: str = "split",
        iteration: Optional[int] = None,
5062
    ) -> np.ndarray:
5063
        """Get feature importances.
5064

5065
5066
        Parameters
        ----------
5067
        importance_type : str, optional (default="split")
5068
5069
5070
            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.
5071
5072
5073
5074
        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).
5075

5076
5077
        Returns
        -------
5078
5079
        result : numpy array
            Array with feature importances.
5080
        """
5081
5082
        if iteration is None:
            iteration = self.best_iteration
5083
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
5084
        result = np.empty(self.num_feature(), dtype=np.float64)
5085
5086
5087
5088
5089
5090
5091
5092
        _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)),
            )
        )
5093
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
5094
            return result.astype(np.int32)
5095
5096
        else:
            return result
5097

5098
5099
5100
5101
    def get_split_value_histogram(
        self,
        feature: Union[int, str],
        bins: Optional[Union[int, str]] = None,
5102
        xgboost_style: bool = False,
5103
    ) -> Union[Tuple[np.ndarray, np.ndarray], np.ndarray, pd_DataFrame]:
5104
5105
5106
5107
        """Get split value histogram for the specified feature.

        Parameters
        ----------
5108
        feature : int or str
5109
5110
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
5111
            If str, interpreted as name.
5112

Nikita Titov's avatar
Nikita Titov committed
5113
5114
5115
            .. warning::

                Categorical features are not supported.
5116

5117
        bins : int, str or None, optional (default=None)
5118
5119
5120
            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.
5121
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
        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.
        """
5136

5137
        def add(root: Dict[str, Any]) -> None:
5138
            """Recursively add thresholds."""
5139
            if "split_index" in root:  # non-leaf
5140
                if feature_names is not None and isinstance(feature, str):
5141
                    split_feature = feature_names[root["split_feature"]]
5142
                else:
5143
                    split_feature = root["split_feature"]
5144
                if split_feature == feature:
5145
5146
                    if isinstance(root["threshold"], str):
                        raise LightGBMError("Cannot compute split value histogram for the categorical feature")
5147
                    else:
5148
5149
5150
                        values.append(root["threshold"])
                add(root["left_child"])
                add(root["right_child"])
5151
5152

        model = self.dump_model()
5153
5154
        feature_names = model.get("feature_names")
        tree_infos = model["tree_info"]
5155
        values: List[float] = []
5156
        for tree_info in tree_infos:
5157
            add(tree_info["tree_structure"])
5158

5159
        if bins is None or isinstance(bins, int) and xgboost_style:
5160
5161
5162
5163
5164
5165
5166
            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:
5167
                return pd_DataFrame(ret, columns=["SplitValue", "Count"])
5168
5169
5170
5171
5172
            else:
                return ret
        else:
            return hist, bin_edges

5173
5174
5175
5176
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
5177
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]],
5178
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
5179
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
5180
        if data_idx >= self.__num_dataset:
5181
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
5182
5183
5184
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
5185
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
5186
            tmp_out_len = ctypes.c_int(0)
5187
5188
5189
5190
5191
5192
5193
5194
            _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
5195
            if tmp_out_len.value != self.__num_inner_eval:
5196
                raise ValueError("Wrong length of eval results")
5197
            for i in range(self.__num_inner_eval):
5198
                ret.append((data_name, self.__name_inner_eval[i], result[i], self.__higher_better_inner_eval[i]))
5199
5200
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
5201
5202
5203
5204
5205
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
5206
5207
5208
5209
5210
5211
5212
5213
5214
            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
5215
5216
5217
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

5218
    def __inner_predict(self, data_idx: int) -> np.ndarray:
5219
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
5220
        if data_idx >= self.__num_dataset:
5221
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
5222
5223
5224
5225
5226
        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
5227
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
5228
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
5229
5230
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
5231
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))  # type: ignore[union-attr]
5232
5233
5234
5235
5236
5237
5238
5239
            _safe_call(
                _LIB.LGBM_BoosterGetPredict(
                    self._handle,
                    ctypes.c_int(data_idx),
                    ctypes.byref(tmp_out_len),
                    data_ptr,
                )
            )
5240
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):  # type: ignore[arg-type]
5241
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
5242
            self.__is_predicted_cur_iter[data_idx] = True
5243
        result: np.ndarray = self.__inner_predict_buffer[data_idx]  # type: ignore[assignment]
5244
5245
        if self.__num_class > 1:
            num_data = result.size // self.__num_class
5246
            result = result.reshape(num_data, self.__num_class, order="F")
5247
        return result
wxchan's avatar
wxchan committed
5248

5249
    def __get_eval_info(self) -> None:
5250
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
5251
5252
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
5253
            out_num_eval = ctypes.c_int(0)
5254
            # Get num of inner evals
5255
5256
5257
5258
5259
5260
            _safe_call(
                _LIB.LGBM_BoosterGetEvalCounts(
                    self._handle,
                    ctypes.byref(out_num_eval),
                )
            )
wxchan's avatar
wxchan committed
5261
5262
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
5263
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
5264
                tmp_out_len = ctypes.c_int(0)
5265
5266
5267
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
5268
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
5269
                ]
5270
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))  # type: ignore[misc]
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
                _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
5281
                if self.__num_inner_eval != tmp_out_len.value:
5282
                    raise ValueError("Length of eval names doesn't equal with num_evals")
5283
5284
5285
5286
5287
5288
                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)
                    ]
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
                    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)]
5303
                self.__higher_better_inner_eval = [
5304
                    name.startswith(("auc", "ndcg@", "map@", "average_precision")) for name in self.__name_inner_eval
5305
                ]