basic.py 176 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3
import abc
wxchan's avatar
wxchan committed
4
import ctypes
5
import json
wxchan's avatar
wxchan committed
6
import warnings
7
from collections import OrderedDict
8
from copy import deepcopy
9
from enum import Enum
10
from functools import wraps
11
from os import SEEK_END, environ
12
13
from os.path import getsize
from pathlib import Path
14
from tempfile import NamedTemporaryFile
15
from typing import Any, Callable, Dict, Iterable, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
16
17
18
19

import numpy as np
import scipy.sparse

20
from .compat import PANDAS_INSTALLED, concat, dt_DataTable, pd_CategoricalDtype, pd_DataFrame, pd_Series
wxchan's avatar
wxchan committed
21
22
from .libpath import find_lib_path

23
24
25
26
27
28
29
30
31
__all__ = [
    'Booster',
    'Dataset',
    'LGBMDeprecationWarning',
    'LightGBMError',
    'register_logger',
    'Sequence',
]

32
_DatasetHandle = ctypes.c_void_p
33
_LGBM_EvalFunctionResultType = Tuple[str, float, bool]
34
_LGBM_BoosterBestScoreType = Dict[str, Dict[str, float]]
35
_LGBM_BoosterEvalMethodResultType = Tuple[str, str, float, bool]
36
37
_LGBM_CategoricalFeatureConfiguration = Union[List[str], List[int], str]
_LGBM_FeatureNameConfiguration = Union[List[str], str]
38
39
40
41
42
43
_LGBM_LabelType = Union[
    list,
    np.ndarray,
    pd_Series,
    pd_DataFrame
]
44

45
46
47
ZERO_THRESHOLD = 1e-35


48
49
50
51
def _is_zero(x: float) -> bool:
    return -ZERO_THRESHOLD <= x <= ZERO_THRESHOLD


52
def _get_sample_count(total_nrow: int, params: str) -> int:
53
54
55
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
56
        _c_str(params),
57
58
59
60
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
61

62
63
64
65
66
67
class _MissingType(Enum):
    NONE = 'None'
    NAN = 'NaN'
    ZERO = 'Zero'


68
class _DummyLogger:
69
    def info(self, msg: str) -> None:
70
71
        print(msg)

72
    def warning(self, msg: str) -> None:
73
74
75
        warnings.warn(msg, stacklevel=3)


76
77
78
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
79
80


81
82
83
84
def _has_method(logger: Any, method_name: str) -> bool:
    return callable(getattr(logger, method_name, None))


85
86
87
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
88
89
90
91
    """Register custom logger.

    Parameters
    ----------
92
    logger : Any
93
        Custom logger.
94
95
96
97
    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.
98
    """
99
100
101
102
103
104
    if not _has_method(logger, info_method_name) or not _has_method(logger, warning_method_name):
        raise TypeError(
            f"Logger must provide '{info_method_name}' and '{warning_method_name}' method"
        )

    global _LOGGER, _INFO_METHOD_NAME, _WARNING_METHOD_NAME
105
    _LOGGER = logger
106
107
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
108
109


110
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
111
    """Join log messages from native library which come by chunks."""
112
    msg_normalized: List[str] = []
113
114

    @wraps(func)
115
    def wrapper(msg: str) -> None:
116
117
118
119
120
121
122
123
124
125
126
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


127
def _log_info(msg: str) -> None:
128
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
129
130


131
def _log_warning(msg: str) -> None:
132
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
133
134
135


@_normalize_native_string
136
def _log_native(msg: str) -> None:
137
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
138
139


140
def _log_callback(msg: bytes) -> None:
141
    """Redirect logs from native library into Python."""
142
    _log_native(str(msg.decode('utf-8')))
143
144


145
def _load_lib() -> ctypes.CDLL:
146
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
147
148
149
    lib_path = find_lib_path()
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
150
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
151
    lib.callback = callback(_log_callback)  # type: ignore[attr-defined]
152
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
153
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
154
155
    return lib

wxchan's avatar
wxchan committed
156

157
158
159
160
161
162
163
# we don't need lib_lightgbm while building docs
_LIB: ctypes.CDLL
if environ.get('LIGHTGBM_BUILD_DOC', False):
    from unittest.mock import Mock  # isort: skip
    _LIB = Mock(ctypes.CDLL)  # type: ignore
else:
    _LIB = _load_lib()
wxchan's avatar
wxchan committed
164

wxchan's avatar
wxchan committed
165

166
_NUMERIC_TYPES = (int, float, bool)
167
_ArrayLike = Union[List, np.ndarray, pd_Series]
168
169


170
def _safe_call(ret: int) -> None:
171
172
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
173
174
175
    Parameters
    ----------
    ret : int
176
        The return value from C API calls.
wxchan's avatar
wxchan committed
177
178
    """
    if ret != 0:
179
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
180

wxchan's avatar
wxchan committed
181

182
def _is_numeric(obj: Any) -> bool:
183
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
184
185
186
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
187
188
189
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
190
191
        return False

wxchan's avatar
wxchan committed
192

193
def _is_numpy_1d_array(data: Any) -> bool:
194
    """Check whether data is a numpy 1-D array."""
195
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
196

wxchan's avatar
wxchan committed
197

198
def _is_numpy_column_array(data: Any) -> bool:
199
200
201
202
203
204
205
    """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


206
def _cast_numpy_array_to_dtype(array: np.ndarray, dtype: np.dtype) -> np.ndarray:
207
    """Cast numpy array to given dtype."""
208
209
210
211
212
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


213
def _is_1d_list(data: Any) -> bool:
214
    """Check whether data is a 1-D list."""
215
    return isinstance(data, list) and (not data or _is_numeric(data[0]))
wxchan's avatar
wxchan committed
216

wxchan's avatar
wxchan committed
217

218
219
220
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
221
        _is_numpy_1d_array(data)
222
        or _is_numpy_column_array(data)
223
        or _is_1d_list(data)
224
225
226
227
        or isinstance(data, pd_Series)
    )


228
def _list_to_1d_numpy(data, dtype=np.float32, name='list'):
229
    """Convert data to numpy 1-D array."""
230
    if _is_numpy_1d_array(data):
231
        return _cast_numpy_array_to_dtype(data, dtype)
232
    elif _is_numpy_column_array(data):
233
234
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
235
        return _cast_numpy_array_to_dtype(array, dtype)
236
    elif _is_1d_list(data):
wxchan's avatar
wxchan committed
237
        return np.array(data, dtype=dtype, copy=False)
238
    elif isinstance(data, pd_Series):
239
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
240
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
241
    else:
242
243
        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
244

wxchan's avatar
wxchan committed
245

246
247
248
249
250
251
252
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."""
253
    return isinstance(data, list) and len(data) > 0 and _is_1d_list(data[0])
254
255
256
257
258
259
260
261
262
263
264
265
266
267


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


def _data_to_2d_numpy(data: Any, dtype: type = np.float32, name: str = 'list') -> np.ndarray:
    """Convert data to numpy 2-D array."""
    if _is_numpy_2d_array(data):
268
        return _cast_numpy_array_to_dtype(data, dtype)
269
270
271
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
272
        _check_for_bad_pandas_dtypes(data.dtypes)
273
        return _cast_numpy_array_to_dtype(data.values, dtype)
274
275
276
277
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


278
def _cfloat32_array_to_numpy(cptr: Any, length: int) -> np.ndarray:
279
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
280
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
281
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
282
    else:
283
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
284

Guolin Ke's avatar
Guolin Ke committed
285

286
def _cfloat64_array_to_numpy(cptr: Any, length: int) -> np.ndarray:
287
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
288
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
289
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
290
291
292
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
293

294
def _cint32_array_to_numpy(cptr: Any, length: int) -> np.ndarray:
295
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
296
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
297
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
298
    else:
299
300
301
        raise RuntimeError('Expected int32 pointer')


302
def _cint64_array_to_numpy(cptr: Any, length: int) -> np.ndarray:
303
304
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
305
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
306
307
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
308

wxchan's avatar
wxchan committed
309

310
def _c_str(string: str) -> ctypes.c_char_p:
311
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
312
313
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
314

315
def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
316
    """Convert a Python array to C array."""
wxchan's avatar
wxchan committed
317
318
    return (ctype * len(values))(*values)

wxchan's avatar
wxchan committed
319

320
def _json_default_with_numpy(obj: Any) -> Any:
321
322
323
324
325
326
327
328
329
    """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


330
331
332
333
334
335
336
337
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)


338
def _param_dict_to_str(data: Optional[Dict[str, Any]]) -> str:
339
    """Convert Python dictionary to string, which is passed to C API."""
340
    if data is None or not data:
wxchan's avatar
wxchan committed
341
342
343
        return ""
    pairs = []
    for key, val in data.items():
344
        if isinstance(val, (list, tuple, set)) or _is_numpy_1d_array(val):
345
            pairs.append(f"{key}={','.join(map(_to_string, val))}")
346
        elif isinstance(val, (str, Path, _NUMERIC_TYPES)) or _is_numeric(val):
347
            pairs.append(f"{key}={val}")
348
        elif val is not None:
349
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
350
    return ' '.join(pairs)
351

wxchan's avatar
wxchan committed
352

353
class _TempFile:
354
355
    """Proxy class to workaround errors on Windows."""

356
357
358
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
359
            self.path = Path(self.name)
360
        return self
wxchan's avatar
wxchan committed
361

362
    def __exit__(self, exc_type, exc_val, exc_tb):
363
364
        if self.path.is_file():
            self.path.unlink()
365

wxchan's avatar
wxchan committed
366

367
class LightGBMError(Exception):
368
369
    """Error thrown by LightGBM."""

370
371
372
    pass


373
374
375
376
377
378
379
380
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
381
382
383
384
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
385
    def _get_all_param_aliases() -> Dict[str, List[str]]:
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
        buffer_len = 1 << 20
        tmp_out_len = ctypes.c_int64(0)
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_DumpParamAliases(
            ctypes.c_int64(buffer_len),
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
        # if buffer length is not long enough, re-allocate a buffer
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _safe_call(_LIB.LGBM_DumpParamAliases(
                ctypes.c_int64(actual_len),
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
        aliases = json.loads(
            string_buffer.value.decode('utf-8'),
405
            object_hook=lambda obj: {k: [k] + v for k, v in obj.items()}
406
407
        )
        return aliases
408
409

    @classmethod
410
411
412
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
413
414
        ret = set()
        for i in args:
415
            ret.update(cls.get_sorted(i))
416
417
        return ret

418
419
420
421
422
423
    @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])

424
    @classmethod
425
426
427
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
428
429
430
431
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
432
                    ret.update(aliases)
433
434
435
                    break
        return ret

436

437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
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)

458
459
    aliases = _ConfigAliases.get_sorted(main_param_name)
    aliases = [a for a in aliases if a != main_param_name]
460
461

    # if main_param_name was provided, keep that value and remove all aliases
462
    if main_param_name in params.keys():
463
464
465
        for param in aliases:
            params.pop(param, None)
        return params
466

467
468
469
470
471
    # 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
472

473
474
475
476
477
478
479
    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
480
481
482
483

    return params


484
_MAX_INT32 = (1 << 31) - 1
485

486
"""Macro definition of data type in C API of LightGBM"""
487
488
489
490
_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
491

492
"""Matrix is row major in Python"""
493
_C_API_IS_ROW_MAJOR = 1
wxchan's avatar
wxchan committed
494

495
"""Macro definition of prediction type in C API of LightGBM"""
496
497
498
499
_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
500

501
"""Macro definition of sparse matrix type"""
502
503
_C_API_MATRIX_TYPE_CSR = 0
_C_API_MATRIX_TYPE_CSC = 1
504

505
"""Macro definition of feature importance type"""
506
507
_C_API_FEATURE_IMPORTANCE_SPLIT = 0
_C_API_FEATURE_IMPORTANCE_GAIN = 1
508

509
"""Data type of data field"""
510
511
512
513
514
515
_FIELD_TYPE_MAPPER = {
    "label": _C_API_DTYPE_FLOAT32,
    "weight": _C_API_DTYPE_FLOAT32,
    "init_score": _C_API_DTYPE_FLOAT64,
    "group": _C_API_DTYPE_INT32
}
wxchan's avatar
wxchan committed
516

517
"""String name to int feature importance type mapper"""
518
519
520
521
_FEATURE_IMPORTANCE_TYPE_MAPPER = {
    "split": _C_API_FEATURE_IMPORTANCE_SPLIT,
    "gain": _C_API_FEATURE_IMPORTANCE_GAIN
}
522

wxchan's avatar
wxchan committed
523

524
def _convert_from_sliced_object(data: np.ndarray) -> np.ndarray:
525
    """Fix the memory of multi-dimensional sliced object."""
526
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
527
        if not data.flags.c_contiguous:
528
529
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
530
531
532
533
            return np.copy(data)
    return data


534
def _c_float_array(data):
535
    """Get pointer of float numpy array / list."""
536
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
537
        data = np.array(data, copy=False)
538
    if _is_numpy_1d_array(data):
539
        data = _convert_from_sliced_object(data)
540
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
541
542
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
543
            type_data = _C_API_DTYPE_FLOAT32
wxchan's avatar
wxchan committed
544
545
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
546
            type_data = _C_API_DTYPE_FLOAT64
wxchan's avatar
wxchan committed
547
        else:
548
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
549
    else:
550
        raise TypeError(f"Unknown type({type(data).__name__})")
551
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
552

wxchan's avatar
wxchan committed
553

554
def _c_int_array(data):
555
    """Get pointer of int numpy array / list."""
556
    if _is_1d_list(data):
wxchan's avatar
wxchan committed
557
        data = np.array(data, copy=False)
558
    if _is_numpy_1d_array(data):
559
        data = _convert_from_sliced_object(data)
560
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
561
562
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
563
            type_data = _C_API_DTYPE_INT32
wxchan's avatar
wxchan committed
564
565
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
566
            type_data = _C_API_DTYPE_INT64
wxchan's avatar
wxchan committed
567
        else:
568
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
569
    else:
570
        raise TypeError(f"Unknown type({type(data).__name__})")
571
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
572

wxchan's avatar
wxchan committed
573

574
def _is_allowed_numpy_dtype(dtype) -> bool:
575
    float128 = getattr(np, 'float128', type(None))
576
577
578
579
    return (
        issubclass(dtype, (np.integer, np.floating, np.bool_))
        and not issubclass(dtype, (np.timedelta64, float128))
    )
580
581


582
def _check_for_bad_pandas_dtypes(pandas_dtypes_series: pd_Series) -> None:
583
584
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
585
        for column_name, pandas_dtype in pandas_dtypes_series.items()
586
        if not _is_allowed_numpy_dtype(pandas_dtype.type)
587
588
589
590
    ]
    if bad_pandas_dtypes:
        raise ValueError('pandas dtypes must be int, float or bool.\n'
                         f'Fields with bad pandas dtypes: {", ".join(bad_pandas_dtypes)}')
591
592


593
594
595
596
597
598
def _data_from_pandas(
    data,
    feature_name: Optional[_LGBM_FeatureNameConfiguration],
    categorical_feature: Optional[_LGBM_CategoricalFeatureConfiguration],
    pandas_categorical: Optional[List[List]]
):
599
    if isinstance(data, pd_DataFrame):
600
601
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
602
        if feature_name == 'auto' or feature_name is None:
603
            data = data.rename(columns=str, copy=False)
604
        cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
605
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
606
607
608
609
610
        if pandas_categorical is None:  # train dataset
            pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
        else:
            if len(cat_cols) != len(pandas_categorical):
                raise ValueError('train and valid dataset categorical_feature do not match.')
611
            for col, category in zip(cat_cols, pandas_categorical):
612
613
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
614
        if len(cat_cols):  # cat_cols is list
615
            data = data.copy(deep=False)  # not alter origin DataFrame
616
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
617
618
619
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
620
            if categorical_feature == 'auto':  # use cat cols from DataFrame
621
                categorical_feature = cat_cols_not_ordered
622
623
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
624
625
        if feature_name == 'auto':
            feature_name = list(data.columns)
626
        _check_for_bad_pandas_dtypes(data.dtypes)
627
628
629
        df_dtypes = [dtype.type for dtype in data.dtypes]
        df_dtypes.append(np.float32)  # so that the target dtype considers floats
        target_dtype = np.find_common_type(df_dtypes, [])
630
631
632
633
634
635
636
637
638
639
        try:
            # most common case (no nullable dtypes)
            data = data.to_numpy(dtype=target_dtype, copy=False)
        except TypeError:
            # 1.0 <= pd version < 1.1 and nullable dtypes, least common case
            # raises error because array is casted to type(pd.NA) and there's no na_value argument
            data = data.astype(target_dtype, copy=False).values
        except ValueError:
            # data has nullable dtypes, but we can specify na_value argument and copy will be made
            data = data.to_numpy(dtype=target_dtype, na_value=np.nan)
640
641
642
643
644
645
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
646
647


648
649
650
651
def _dump_pandas_categorical(
    pandas_categorical: Optional[List[List]],
    file_name: Optional[Union[str, Path]] = None
) -> str:
652
    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
653
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
654
655
656
657
658
659
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


660
661
662
def _load_pandas_categorical(
    file_name: Optional[Union[str, Path]] = None,
    model_str: Optional[str] = None
663
) -> Optional[List[List]]:
664
665
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
666
    if file_name is not None:
667
        max_offset = -getsize(file_name)
668
669
670
671
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
672
                f.seek(offset, SEEK_END)
673
674
675
676
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
677
        last_line = lines[-1].decode('utf-8').strip()
678
        if not last_line.startswith(pandas_key):
679
            last_line = lines[-2].decode('utf-8').strip()
680
    elif model_str is not None:
681
682
683
684
685
686
        idx = model_str.rfind('\n', 0, offset)
        last_line = model_str[idx:].strip()
    if last_line.startswith(pandas_key):
        return json.loads(last_line[len(pandas_key):])
    else:
        return None
687
688


689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
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**.

709
710
    .. versionadded:: 3.3.0

711
712
713
714
715
716
717
718
719
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
720
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
721
722
723
724
725
726
727
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
728
                return self._get_one_line(idx)
729
            elif isinstance(idx, slice):
730
731
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
732
                # Only required if using ``Dataset.subset()``.
733
                return np.array([self._get_one_line(i) for i in idx])
734
            else:
735
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
736
737
738

        Parameters
        ----------
739
        idx : int, slice[int], list[int]
740
741
742
743
            Item index.

        Returns
        -------
744
        result : numpy 1-D array or numpy 2-D array
745
            1-D array if idx is int, 2-D array if idx is slice or list.
746
747
748
749
750
751
752
753
754
        """
        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__()")


755
class _InnerPredictor:
756
757
758
759
760
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
761
762
763
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
764
    """
765

766
767
768
769
770
771
    def __init__(
        self,
        model_file: Optional[Union[str, Path]] = None,
        booster_handle: Optional[ctypes.c_void_p] = None,
        pred_parameter: Optional[Dict[str, Any]] = None
    ):
772
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
773
774
775

        Parameters
        ----------
776
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
777
            Path to the model file.
778
779
780
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
781
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
782
783
784
785
786
        """
        self.handle = ctypes.c_void_p()
        self.__is_manage_handle = True
        if model_file is not None:
            """Prediction task"""
Guolin Ke's avatar
Guolin Ke committed
787
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
788
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
789
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
790
791
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
792
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
793
794
795
796
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
797
            self.num_total_iteration = out_num_iterations.value
798
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
799
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
800
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
801
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
802
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
803
804
805
806
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
807
            self.num_total_iteration = self.current_iteration()
808
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
809
        else:
810
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
811

812
        pred_parameter = {} if pred_parameter is None else pred_parameter
813
        self.pred_parameter = _param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
814

815
    def __del__(self) -> None:
816
817
818
819
820
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
821

822
    def __getstate__(self) -> Dict[str, Any]:
823
824
825
826
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

827
828
829
830
831
832
833
834
835
836
837
    def predict(
        self,
        data,
        start_iteration: int = 0,
        num_iteration: int = -1,
        raw_score: bool = False,
        pred_leaf: bool = False,
        pred_contrib: bool = False,
        data_has_header: bool = False,
        validate_features: bool = False
    ):
838
        """Predict logic.
wxchan's avatar
wxchan committed
839
840
841

        Parameters
        ----------
842
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
843
            Data source for prediction.
844
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
845
846
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
847
848
849
850
851
852
853
854
855
856
857
        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.
858
859
860
        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
861
862
863

        Returns
        -------
864
        result : numpy array, scipy.sparse or list of scipy.sparse
865
            Prediction result.
866
            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
867
        """
wxchan's avatar
wxchan committed
868
        if isinstance(data, Dataset):
869
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
870
871
872
873
874
875
876
877
878
879
880
        elif isinstance(data, pd_DataFrame) and validate_features:
            data_names = [str(x) for x in data.columns]
            ptr_names = (ctypes.c_char_p * len(data_names))()
            ptr_names[:] = [x.encode('utf-8') for x in data_names]
            _safe_call(
                _LIB.LGBM_BoosterValidateFeatureNames(
                    self.handle,
                    ptr_names,
                    ctypes.c_int(len(data_names)),
                )
            )
881
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
882
        predict_type = _C_API_PREDICT_NORMAL
wxchan's avatar
wxchan committed
883
        if raw_score:
884
            predict_type = _C_API_PREDICT_RAW_SCORE
wxchan's avatar
wxchan committed
885
        if pred_leaf:
886
            predict_type = _C_API_PREDICT_LEAF_INDEX
887
        if pred_contrib:
888
            predict_type = _C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
889
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
890

891
        if isinstance(data, (str, Path)):
892
            with _TempFile() as f:
wxchan's avatar
wxchan committed
893
894
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
895
                    _c_str(str(data)),
Guolin Ke's avatar
Guolin Ke committed
896
897
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
898
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
899
                    ctypes.c_int(num_iteration),
900
901
                    _c_str(self.pred_parameter),
                    _c_str(f.name)))
902
903
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
904
        elif isinstance(data, scipy.sparse.csr_matrix):
905
            preds, nrow = self.__pred_for_csr(data, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
906
        elif isinstance(data, scipy.sparse.csc_matrix):
907
            preds, nrow = self.__pred_for_csc(data, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
908
        elif isinstance(data, np.ndarray):
909
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
910
911
912
        elif isinstance(data, list):
            try:
                data = np.array(data)
913
            except BaseException:
914
                raise ValueError('Cannot convert data list to numpy array.')
915
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
916
        elif isinstance(data, dt_DataTable):
917
            preds, nrow = self.__pred_for_np2d(data.to_numpy(), start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
918
919
        else:
            try:
920
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
921
                csr = scipy.sparse.csr_matrix(data)
922
            except BaseException:
923
                raise TypeError(f'Cannot predict data for type {type(data).__name__}')
924
            preds, nrow = self.__pred_for_csr(csr, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
925
926
        if pred_leaf:
            preds = preds.astype(np.int32)
927
        is_sparse = scipy.sparse.issparse(preds) or isinstance(preds, list)
928
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
929
            if preds.size % nrow == 0:
930
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
931
            else:
932
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
933
934
        return preds

935
    def __get_num_preds(self, start_iteration, num_iteration, nrow, predict_type):
936
        """Get size of prediction result."""
937
        if nrow > _MAX_INT32:
938
            raise LightGBMError('LightGBM cannot perform prediction for data '
939
                                f'with number of rows greater than MAX_INT32 ({_MAX_INT32}).\n'
940
                                'You can split your data into chunks '
941
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
942
943
944
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
945
946
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
947
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
948
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
949
950
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
951

952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
    def __inner_predict_np2d(
        self,
        mat: np.ndarray,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
        preds: Optional[np.ndarray]
    ) -> Tuple[np.ndarray, int]:
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
        else:  # change non-float data to float data, need to copy
            data = np.array(mat.reshape(mat.size), dtype=np.float32)
        ptr_data, type_ptr_data, _ = _c_float_array(data)
        n_preds = self.__get_num_preds(start_iteration, num_iteration, mat.shape[0], predict_type)
        if preds is None:
            preds = np.empty(n_preds, dtype=np.float64)
        elif len(preds.shape) != 1 or len(preds) != n_preds:
            raise ValueError("Wrong length of pre-allocated predict array")
        out_num_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterPredictForMat(
            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))))
        if n_preds != out_num_preds.value:
            raise ValueError("Wrong length for predict results")
        return preds, mat.shape[0]

    def __pred_for_np2d(
        self,
        mat: np.ndarray,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ) -> Tuple[np.ndarray, int]:
995
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
996
        if len(mat.shape) != 2:
997
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
998

999
        nrow = mat.shape[0]
1000
1001
        if nrow > _MAX_INT32:
            sections = np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)
1002
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1003
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
1004
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1005
            preds = np.empty(sum(n_preds), dtype=np.float64)
1006
1007
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
1008
                # avoid memory consumption by arrays concatenation operations
1009
1010
1011
1012
1013
1014
1015
                self.__inner_predict_np2d(
                    mat=chunk,
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
                    preds=preds[start_idx_pred:end_idx_pred]
                )
1016
            return preds, nrow
wxchan's avatar
wxchan committed
1017
        else:
1018
1019
1020
1021
1022
1023
1024
            return self.__inner_predict_np2d(
                mat=mat,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
                preds=None
            )
wxchan's avatar
wxchan committed
1025

1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
    def __create_sparse_native(
        self,
        cs: Union[scipy.sparse.csc_matrix, scipy.sparse.csr_matrix],
        out_shape,
        out_ptr_indptr,
        out_ptr_indices,
        out_ptr_data,
        indptr_type,
        data_type,
        is_csr: bool
    ):
1037
1038
1039
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
1040
        if indptr_type == _C_API_DTYPE_INT32:
1041
            out_indptr = _cint32_array_to_numpy(out_ptr_indptr, indptr_len)
1042
        elif indptr_type == _C_API_DTYPE_INT64:
1043
            out_indptr = _cint64_array_to_numpy(out_ptr_indptr, indptr_len)
1044
1045
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
1046
        if data_type == _C_API_DTYPE_FLOAT32:
1047
            out_data = _cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
1048
        elif data_type == _C_API_DTYPE_FLOAT64:
1049
            out_data = _cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
1050
1051
        else:
            raise TypeError("Expected float32 or float64 type for data")
1052
        out_indices = _cint32_array_to_numpy(out_ptr_indices, data_indices_len)
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
        # break up indptr based on number of rows (note more than one matrix in multiclass case)
        per_class_indptr_shape = cs.indptr.shape[0]
        # for CSC there is extra column added
        if not is_csr:
            per_class_indptr_shape += 1
        out_indptr_arrays = np.split(out_indptr, out_indptr.shape[0] / per_class_indptr_shape)
        # reformat output into a csr or csc matrix or list of csr or csc matrices
        cs_output_matrices = []
        offset = 0
        for cs_indptr in out_indptr_arrays:
            matrix_indptr_len = cs_indptr[cs_indptr.shape[0] - 1]
            cs_indices = out_indices[offset + cs_indptr[0]:offset + matrix_indptr_len]
            cs_data = out_data[offset + cs_indptr[0]:offset + matrix_indptr_len]
            offset += matrix_indptr_len
            # same shape as input csr or csc matrix except extra column for expected value
            cs_shape = [cs.shape[0], cs.shape[1] + 1]
            # note: make sure we copy data as it will be deallocated next
            if is_csr:
                cs_output_matrices.append(scipy.sparse.csr_matrix((cs_data, cs_indices, cs_indptr), cs_shape))
            else:
                cs_output_matrices.append(scipy.sparse.csc_matrix((cs_data, cs_indices, cs_indptr), cs_shape))
        # free the temporary native indptr, indices, and data
        _safe_call(_LIB.LGBM_BoosterFreePredictSparse(out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                      ctypes.c_int(indptr_type), ctypes.c_int(data_type)))
        if len(cs_output_matrices) == 1:
            return cs_output_matrices[0]
        return cs_output_matrices

1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
    def __inner_predict_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int,
        preds: Optional[np.ndarray]
    ) -> Tuple[np.ndarray, int]:
        nrow = len(csr.indptr) - 1
        n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
        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
1096

1097
1098
        ptr_indptr, type_ptr_indptr, _ = _c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = _c_float_array(csr.data)
1099

1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
        assert csr.shape[1] <= _MAX_INT32
        csr_indices = csr.indices.astype(np.int32, copy=False)

        _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))))
        if n_preds != out_num_preds.value:
            raise ValueError("Wrong length for predict results")
        return preds, nrow

    def __inner_predict_csr_sparse(
        self,
        csr: scipy.sparse.csr_matrix,
        start_iteration: int,
        num_iteration: int,
        predict_type: int
    ):
        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
        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)()
        if type_ptr_data == _C_API_DTYPE_FLOAT32:
            out_ptr_data = ctypes.POINTER(ctypes.c_float)()
        else:
            out_ptr_data = ctypes.POINTER(ctypes.c_double)()
        out_shape = np.empty(2, dtype=np.int64)
        _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
            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)))
        matrices = self.__create_sparse_native(
            cs=csr,
            out_shape=out_shape,
            out_ptr_indptr=out_ptr_indptr,
            out_ptr_indices=out_ptr_indices,
            out_ptr_data=out_ptr_data,
            indptr_type=type_ptr_indptr,
            data_type=type_ptr_data,
            is_csr=True
        )
        nrow = len(csr.indptr) - 1
        return matrices, nrow

    def __pred_for_csr(self, csr, start_iteration, num_iteration, predict_type):
        """Predict for a CSR data."""
1178
        if predict_type == _C_API_PREDICT_CONTRIB:
1179
1180
1181
1182
1183
1184
            return self.__inner_predict_csr_sparse(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1185
        nrow = len(csr.indptr) - 1
1186
1187
        if nrow > _MAX_INT32:
            sections = [0] + list(np.arange(start=_MAX_INT32, stop=nrow, step=_MAX_INT32)) + [nrow]
1188
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1189
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1190
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1191
            preds = np.empty(sum(n_preds), dtype=np.float64)
1192
1193
            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:])):
1194
                # avoid memory consumption by arrays concatenation operations
1195
1196
1197
1198
1199
1200
1201
                self.__inner_predict_csr(
                    csr=csr[start_idx:end_idx],
                    start_iteration=start_iteration,
                    num_iteration=num_iteration,
                    predict_type=predict_type,
                    preds=preds[start_idx_pred:end_idx_pred]
                )
1202
1203
            return preds, nrow
        else:
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
            return self.__inner_predict_csr(
                csr=csr,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type,
                preds=None
            )

    def __inner_predict_sparse_csc(
        self,
        csc,
        start_iteration,
        num_iteration,
        predict_type
    ):
        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
        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)()
        if type_ptr_data == _C_API_DTYPE_FLOAT32:
            out_ptr_data = ctypes.POINTER(ctypes.c_float)()
        else:
            out_ptr_data = ctypes.POINTER(ctypes.c_double)()
        out_shape = np.empty(2, dtype=np.int64)
        _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
            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)))
        matrices = self.__create_sparse_native(
            cs=csc,
            out_shape=out_shape,
            out_ptr_indptr=out_ptr_indptr,
            out_ptr_indices=out_ptr_indices,
            out_ptr_data=out_ptr_data,
            indptr_type=type_ptr_indptr,
            data_type=type_ptr_data,
            is_csr=False
        )
        nrow = csc.shape[0]
        return matrices, nrow
Guolin Ke's avatar
Guolin Ke committed
1264

1265
    def __pred_for_csc(self, csc, start_iteration, num_iteration, predict_type):
1266
        """Predict for a CSC data."""
Guolin Ke's avatar
Guolin Ke committed
1267
        nrow = csc.shape[0]
1268
        if nrow > _MAX_INT32:
1269
            return self.__pred_for_csr(csc.tocsr(), start_iteration, num_iteration, predict_type)
1270
        if predict_type == _C_API_PREDICT_CONTRIB:
1271
1272
1273
1274
1275
1276
            return self.__inner_predict_sparse_csc(
                csc=csc,
                start_iteration=start_iteration,
                num_iteration=num_iteration,
                predict_type=predict_type
            )
1277
        n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
1278
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1279
1280
        out_num_preds = ctypes.c_int64(0)

1281
1282
        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
1283

1284
        assert csc.shape[0] <= _MAX_INT32
1285
        csc_indices = csc.indices.astype(np.int32, copy=False)
1286

Guolin Ke's avatar
Guolin Ke committed
1287
1288
1289
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
1290
            ctypes.c_int(type_ptr_indptr),
1291
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1292
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1293
1294
1295
1296
1297
            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),
1298
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1299
            ctypes.c_int(num_iteration),
1300
            _c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1301
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1302
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1303
        if n_preds != out_num_preds.value:
1304
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1305
1306
        return preds, nrow

1307
    def current_iteration(self) -> int:
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
        out_cur_iter = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

wxchan's avatar
wxchan committed
1321

1322
class Dataset:
wxchan's avatar
wxchan committed
1323
    """Dataset in LightGBM."""
1324

1325
1326
1327
    def __init__(
        self,
        data,
1328
        label: Optional[_LGBM_LabelType] = None,
1329
1330
1331
1332
        reference: Optional["Dataset"] = None,
        weight=None,
        group=None,
        init_score=None,
1333
1334
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
1335
1336
1337
        params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True
    ):
1338
        """Initialize Dataset.
1339

wxchan's avatar
wxchan committed
1340
1341
        Parameters
        ----------
1342
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence or list of numpy array
wxchan's avatar
wxchan committed
1343
            Data source of Dataset.
1344
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1345
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1346
1347
1348
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1349
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1350
            Weight for each instance. Weights should be non-negative.
1351
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1352
1353
1354
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1355
1356
            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.
1357
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None, optional (default=None)
1358
            Init score for Dataset.
1359
        feature_name : list of str, or 'auto', optional (default="auto")
1360
1361
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
1362
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1363
1364
            Categorical features.
            If list of int, interpreted as indices.
1365
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1366
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1367
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1368
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1369
            All negative values in categorical features will be treated as missing values.
1370
            The output cannot be monotonically constrained with respect to a categorical feature.
1371
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1372
        params : dict or None, optional (default=None)
1373
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1374
        free_raw_data : bool, optional (default=True)
1375
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
1376
        """
1377
        self.handle: Optional[_DatasetHandle] = None
wxchan's avatar
wxchan committed
1378
1379
1380
1381
1382
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1383
        self.init_score = init_score
1384
1385
        self.feature_name: _LGBM_FeatureNameConfiguration = feature_name
        self.categorical_feature: _LGBM_CategoricalFeatureConfiguration = categorical_feature
1386
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1387
        self.free_raw_data = free_raw_data
1388
        self.used_indices: Optional[List[int]] = None
1389
        self._need_slice = True
1390
        self._predictor: Optional[_InnerPredictor] = None
1391
        self.pandas_categorical = None
1392
        self._params_back_up = None
1393
        self.version = 0
1394
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1395

1396
    def __del__(self) -> None:
1397
1398
1399
1400
        try:
            self._free_handle()
        except AttributeError:
            pass
1401

1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
    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.
        """
1419
        param_str = _param_dict_to_str(self.get_params())
1420
1421
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
1422
        ptr_data, _, _ = _c_int_array(indices)
1423
1424
1425
1426
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
1427
            _c_str(param_str),
1428
1429
1430
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1431
1432
        assert sample_cnt == actual_sample_cnt.value
        return indices
1433

1434
1435
1436
1437
1438
    def _init_from_ref_dataset(
        self,
        total_nrow: int,
        ref_dataset: _DatasetHandle
    ) -> 'Dataset':
1439
1440
1441
1442
1443
1444
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
1445
1446
        ref_dataset : object
            Handle of reference dataset to extract metadata from.
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471

        Returns
        -------
        self : Dataset
            Constructed Dataset object.
        """
        self.handle = ctypes.c_void_p()
        _safe_call(_LIB.LGBM_DatasetCreateByReference(
            ref_dataset,
            ctypes.c_int64(total_nrow),
            ctypes.byref(self.handle),
        ))
        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
        ----------
1472
        sample_data : list of numpy array
1473
            Sample data for each column.
1474
        sample_indices : list of numpy array
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
            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.
        sample_col_ptr = (ctypes.POINTER(ctypes.c_double) * ncol)()
        # c type int**
        # each int* points to start of indices for each column
        indices_col_ptr = (ctypes.POINTER(ctypes.c_int32) * ncol)()
        for i in range(ncol):
1502
1503
            sample_col_ptr[i] = _c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = _c_int_array(sample_indices[i])[0]
1504
1505

        num_per_col = np.array([len(d) for d in sample_indices], dtype=np.int32)
1506
        num_per_col_ptr, _, _ = _c_int_array(num_per_col)
1507
1508

        self.handle = ctypes.c_void_p()
1509
        params_str = _param_dict_to_str(self.get_params())
1510
1511
1512
1513
1514
1515
1516
        _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),
1517
            ctypes.c_int64(total_nrow),
1518
            _c_str(params_str),
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
            ctypes.byref(self.handle),
        ))
        return self

    def _push_rows(self, data: np.ndarray) -> 'Dataset':
        """Add rows to Dataset.

        Parameters
        ----------
        data : numpy 1-D array
            New data to add to the Dataset.

        Returns
        -------
        self : Dataset
            Dataset object.
        """
        nrow, ncol = data.shape
        data = data.reshape(data.size)
1538
        data_ptr, data_type, _ = _c_float_array(data)
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550

        _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),
        ))
        self._start_row += nrow
        return self

1551
    def get_params(self) -> Dict[str, Any]:
1552
1553
1554
1555
        """Get the used parameters in the Dataset.

        Returns
        -------
1556
        params : dict
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
            The used parameters in this Dataset object.
        """
        if self.params is not None:
            # no min_data, nthreads and verbose in this function
            dataset_params = _ConfigAliases.get("bin_construct_sample_cnt",
                                                "categorical_feature",
                                                "data_random_seed",
                                                "enable_bundle",
                                                "feature_pre_filter",
                                                "forcedbins_filename",
                                                "group_column",
                                                "header",
                                                "ignore_column",
                                                "is_enable_sparse",
                                                "label_column",
1572
                                                "linear_tree",
1573
1574
1575
1576
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1577
                                                "precise_float_parser",
1578
1579
1580
1581
1582
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1583
1584
        else:
            return {}
1585

1586
    def _free_handle(self) -> "Dataset":
1587
        if self.handle is not None:
1588
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1589
            self.handle = None
1590
        self._need_slice = True
Guolin Ke's avatar
Guolin Ke committed
1591
1592
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1593
        return self
wxchan's avatar
wxchan committed
1594

1595
1596
1597
1598
1599
1600
    def _set_init_score_by_predictor(
        self,
        predictor: Optional[_InnerPredictor],
        data,
        used_indices: Optional[List[int]]
    ):
Guolin Ke's avatar
Guolin Ke committed
1601
        data_has_header = False
1602
        if isinstance(data, (str, Path)):
Guolin Ke's avatar
Guolin Ke committed
1603
            # check data has header or not
1604
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1605
        num_data = self.num_data()
1606
1607
1608
        if predictor is not None:
            init_score = predictor.predict(data,
                                           raw_score=True,
1609
1610
                                           data_has_header=data_has_header)
            init_score = init_score.ravel()
1611
            if used_indices is not None:
1612
                assert not self._need_slice
1613
                if isinstance(data, (str, Path)):
1614
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1615
                    assert num_data == len(used_indices)
1616
1617
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1618
1619
1620
1621
                            sub_init_score[i * predictor.num_class + j] = init_score[used_indices[i] * predictor.num_class + j]
                    init_score = sub_init_score
            if predictor.num_class > 1:
                # need to regroup init_score
1622
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1623
1624
                for i in range(num_data):
                    for j in range(predictor.num_class):
1625
1626
1627
                        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:
1628
            init_score = np.zeros(self.init_score.shape, dtype=np.float64)
1629
1630
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1631
1632
        self.set_init_score(init_score)

1633
1634
1635
    def _lazy_init(
        self,
        data,
1636
        label: Optional[_LGBM_LabelType] = None,
1637
1638
1639
1640
1641
1642
1643
1644
1645
        reference: Optional["Dataset"] = None,
        weight=None,
        group=None,
        init_score=None,
        predictor=None,
        feature_name='auto',
        categorical_feature='auto',
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
wxchan's avatar
wxchan committed
1646
1647
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1648
            return self
Guolin Ke's avatar
Guolin Ke committed
1649
1650
1651
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1652
1653
1654
1655
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
Guolin Ke's avatar
Guolin Ke committed
1656

1657
        # process for args
wxchan's avatar
wxchan committed
1658
        params = {} if params is None else params
1659
1660
1661
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1662
        for key in params.keys():
1663
            if key in args_names:
1664
1665
                _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.')
1666
        # get categorical features
1667
1668
1669
1670
1671
1672
        if categorical_feature is not None:
            categorical_indices = set()
            feature_dict = {}
            if feature_name is not None:
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
1673
                if isinstance(name, str) and name in feature_dict:
1674
                    categorical_indices.add(feature_dict[name])
1675
                elif isinstance(name, int):
1676
1677
                    categorical_indices.add(name)
                else:
1678
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1679
            if categorical_indices:
1680
1681
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1682
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
1683
                        if not (isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
1684
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1685
                        params.pop(cat_alias, None)
1686
                params['categorical_column'] = sorted(categorical_indices)
1687

1688
        params_str = _param_dict_to_str(params)
1689
        self.params = params
1690
        # process for reference dataset
wxchan's avatar
wxchan committed
1691
        ref_dataset = None
wxchan's avatar
wxchan committed
1692
        if isinstance(reference, Dataset):
1693
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1694
1695
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1696
        # start construct data
1697
        if isinstance(data, (str, Path)):
wxchan's avatar
wxchan committed
1698
1699
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1700
1701
                _c_str(str(data)),
                _c_str(params_str),
wxchan's avatar
wxchan committed
1702
1703
1704
1705
                ref_dataset,
                ctypes.byref(self.handle)))
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
1706
1707
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1708
1709
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1710
1711
1712
1713
1714
1715
1716
1717
1718
        elif isinstance(data, list) and len(data) > 0:
            if all(isinstance(x, np.ndarray) for x in data):
                self.__init_from_list_np2d(data, params_str, ref_dataset)
            elif all(isinstance(x, Sequence) for x in data):
                self.__init_from_seqs(data, ref_dataset)
            else:
                raise TypeError('Data list can only be of ndarray or Sequence')
        elif isinstance(data, Sequence):
            self.__init_from_seqs([data], ref_dataset)
1719
        elif isinstance(data, dt_DataTable):
1720
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1721
1722
1723
1724
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1725
            except BaseException:
1726
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}')
wxchan's avatar
wxchan committed
1727
1728
1729
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1730
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1731
1732
1733
1734
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1735
1736
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1737
                _log_warning("The init_score will be overridden by the prediction of init_model.")
1738
1739
1740
1741
1742
            self._set_init_score_by_predictor(
                predictor=predictor,
                data=data,
                used_indices=None
            )
1743
1744
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1745
        elif predictor is not None:
1746
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1747
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1748
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1749

1750
1751
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
        offset = 0
        seq_id = 0
        seq = seqs[seq_id]
        for row_id in indices:
            assert row_id >= offset, "sample indices are expected to be monotonic"
            while row_id >= offset + len(seq):
                offset += len(seq)
                seq_id += 1
                seq = seqs[seq_id]
            id_in_seq = row_id - offset
            row = seq[id_in_seq]
            yield row if row.flags['OWNDATA'] else row.copy()

    def __sample(self, seqs: List[Sequence], total_nrow: int) -> Tuple[List[np.ndarray], List[np.ndarray]]:
        """Sample data from seqs.

        Mimics behavior in c_api.cpp:LGBM_DatasetCreateFromMats()

        Returns
        -------
            sampled_rows, sampled_row_indices
        """
        indices = self._create_sample_indices(total_nrow)

        # Select sampled rows, transpose to column order.
1777
        sampled = np.array([row for row in self._yield_row_from_seqlist(seqs, indices)])
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
        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

1793
1794
1795
    def __init_from_seqs(
        self,
        seqs: List[Sequence],
1796
        ref_dataset: Optional[_DatasetHandle]
1797
    ) -> "Dataset":
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
        """
        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:
1812
            param_str = _param_dict_to_str(self.get_params())
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
            sample_cnt = _get_sample_count(total_nrow, param_str)

            sample_data, col_indices = self.__sample(seqs, total_nrow)
            self._init_from_sample(sample_data, col_indices, sample_cnt, total_nrow)

        for seq in seqs:
            nrow = len(seq)
            batch_size = getattr(seq, 'batch_size', None) or Sequence.batch_size
            for start in range(0, nrow, batch_size):
                end = min(start + batch_size, nrow)
                self._push_rows(seq[start:end])
        return self

1826
1827
1828
1829
1830
1831
    def __init_from_np2d(
        self,
        mat: np.ndarray,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1832
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1833
1834
1835
1836
1837
1838
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

        self.handle = ctypes.c_void_p()
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
1839
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1840
1841
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1842
        ptr_data, type_ptr_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
1843
1844
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1845
            ctypes.c_int(type_ptr_data),
1846
1847
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
1848
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
1849
            _c_str(params_str),
wxchan's avatar
wxchan committed
1850
1851
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1852
        return self
wxchan's avatar
wxchan committed
1853

1854
1855
1856
1857
1858
1859
    def __init_from_list_np2d(
        self,
        mats: List[np.ndarray],
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1860
        """Initialize data from a list of 2-D numpy matrices."""
1861
        ncol = mats[0].shape[1]
1862
        nrow = np.empty((len(mats),), np.int32)
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
        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 = []
        type_ptr_data = None

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

            if mat.shape[1] != ncol:
                raise ValueError('Input arrays must have same number of columns')

            nrow[i] = mat.shape[0]

            if mat.dtype == np.float32 or mat.dtype == np.float64:
                mats[i] = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
1882
            else:  # change non-float data to float data, need to copy
1883
1884
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

1885
            chunk_ptr_data, chunk_type_ptr_data, holder = _c_float_array(mats[i])
1886
1887
1888
1889
1890
1891
1892
1893
            if type_ptr_data is not None and chunk_type_ptr_data != type_ptr_data:
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

        self.handle = ctypes.c_void_p()
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
1894
            ctypes.c_int32(len(mats)),
1895
1896
1897
            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)),
1898
            ctypes.c_int32(ncol),
1899
            ctypes.c_int(_C_API_IS_ROW_MAJOR),
1900
            _c_str(params_str),
1901
1902
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1903
        return self
1904

1905
1906
1907
1908
1909
1910
    def __init_from_csr(
        self,
        csr: scipy.sparse.csr_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1911
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
1912
        if len(csr.indices) != len(csr.data):
1913
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
wxchan's avatar
wxchan committed
1914
1915
        self.handle = ctypes.c_void_p()

1916
1917
        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
1918

1919
        assert csr.shape[1] <= _MAX_INT32
1920
        csr_indices = csr.indices.astype(np.int32, copy=False)
1921

wxchan's avatar
wxchan committed
1922
1923
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1924
            ctypes.c_int(type_ptr_indptr),
1925
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
1926
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1927
1928
1929
1930
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
1931
            _c_str(params_str),
wxchan's avatar
wxchan committed
1932
1933
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1934
        return self
wxchan's avatar
wxchan committed
1935

1936
1937
1938
1939
1940
1941
    def __init_from_csc(
        self,
        csc: scipy.sparse.csc_matrix,
        params_str: str,
        ref_dataset: Optional[_DatasetHandle]
    ) -> "Dataset":
1942
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
1943
        if len(csc.indices) != len(csc.data):
1944
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
Guolin Ke's avatar
Guolin Ke committed
1945
1946
        self.handle = ctypes.c_void_p()

1947
1948
        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
1949

1950
        assert csc.shape[0] <= _MAX_INT32
1951
        csc_indices = csc.indices.astype(np.int32, copy=False)
1952

Guolin Ke's avatar
Guolin Ke committed
1953
1954
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1955
            ctypes.c_int(type_ptr_indptr),
1956
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1957
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1958
1959
1960
1961
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
1962
            _c_str(params_str),
Guolin Ke's avatar
Guolin Ke committed
1963
1964
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1965
        return self
Guolin Ke's avatar
Guolin Ke committed
1966

1967
    @staticmethod
1968
1969
1970
1971
1972
1973
    def _compare_params_for_warning(
        params: Optional[Dict[str, Any]],
        other_params: Optional[Dict[str, Any]],
        ignore_keys: Set[str]
    ) -> bool:
        """Compare two dictionaries with params ignoring some keys.
1974

1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
        It is only for the warning purpose.

        Parameters
        ----------
        params : dict or None
            One dictionary with parameters to compare.
        other_params : dict or None
            Another dictionary with parameters to compare.
        ignore_keys : set
            Keys that should be ignored during comparing two dictionaries.
1985
1986
1987

        Returns
        -------
1988
1989
        compare_result : bool
          Returns whether two dictionaries with params are equal.
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
        """
        if params is None:
            params = {}
        if other_params is None:
            other_params = {}
        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

2005
    def construct(self) -> "Dataset":
2006
2007
2008
2009
2010
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
2011
            Constructed Dataset object.
2012
        """
2013
        if self.handle is None:
wxchan's avatar
wxchan committed
2014
            if self.reference is not None:
2015
                reference_params = self.reference.get_params()
2016
2017
                params = self.get_params()
                if params != reference_params:
2018
2019
2020
2021
2022
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
2023
                        _log_warning('Overriding the parameters from Reference Dataset.')
2024
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
2025
                if self.used_indices is None:
2026
                    # create valid
2027
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
2028
2029
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
2030
                                    feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
2031
                else:
2032
                    # construct subset
2033
                    used_indices = _list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
2034
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
2035
                    if self.reference.group is not None:
2036
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
2037
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
2038
                                                  return_counts=True)
2039
                    self.handle = ctypes.c_void_p()
2040
                    params_str = _param_dict_to_str(self.params)
wxchan's avatar
wxchan committed
2041
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
2042
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
2043
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
2044
                        ctypes.c_int32(used_indices.shape[0]),
2045
                        _c_str(params_str),
wxchan's avatar
wxchan committed
2046
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2047
2048
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
2049
2050
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
2051
2052
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
2053
2054
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
2055
2056
2057
2058
2059
                        self._set_init_score_by_predictor(
                            predictor=self._predictor,
                            data=self.data,
                            used_indices=used_indices
                        )
wxchan's avatar
wxchan committed
2060
            else:
2061
                # create train
2062
                self._lazy_init(self.data, label=self.label,
2063
2064
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
2065
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
2066
2067
            if self.free_raw_data:
                self.data = None
2068
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
2069
        return self
wxchan's avatar
wxchan committed
2070

2071
2072
2073
    def create_valid(
        self,
        data,
2074
        label: Optional[_LGBM_LabelType] = None,
2075
2076
2077
2078
2079
        weight=None,
        group=None,
        init_score=None,
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2080
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
2081
2082
2083

        Parameters
        ----------
2084
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence or list of numpy array
wxchan's avatar
wxchan committed
2085
            Data source of Dataset.
2086
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
2087
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
2088
2089
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
2090
            Weight for each instance. Weights should be non-negative.
2091
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
2092
2093
2094
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2095
2096
            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.
2097
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None, optional (default=None)
2098
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
2099
        params : dict or None, optional (default=None)
2100
            Other parameters for validation Dataset.
2101
2102
2103

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2104
2105
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
2106
        """
2107
        ret = Dataset(data, label=label, reference=self,
2108
                      weight=weight, group=group, init_score=init_score,
2109
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2110
        ret._predictor = self._predictor
2111
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2112
        return ret
wxchan's avatar
wxchan committed
2113

2114
2115
2116
2117
2118
    def subset(
        self,
        used_indices: List[int],
        params: Optional[Dict[str, Any]] = None
    ) -> "Dataset":
2119
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
2120
2121
2122
2123

        Parameters
        ----------
        used_indices : list of int
2124
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
2125
        params : dict or None, optional (default=None)
2126
            These parameters will be passed to Dataset constructor.
2127
2128
2129
2130
2131

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
2132
        """
wxchan's avatar
wxchan committed
2133
2134
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
2135
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
2136
2137
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
2138
        ret._predictor = self._predictor
2139
        ret.pandas_categorical = self.pandas_categorical
2140
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
2141
2142
        return ret

2143
    def save_binary(self, filename: Union[str, Path]) -> "Dataset":
2144
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
2145

2146
2147
2148
2149
2150
        .. 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
2151
2152
        Parameters
        ----------
2153
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
2154
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
2155
2156
2157
2158
2159

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
2160
2161
2162
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
2163
            _c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
2164
        return self
wxchan's avatar
wxchan committed
2165

2166
    def _update_params(self, params: Optional[Dict[str, Any]]) -> "Dataset":
2167
2168
        if not params:
            return self
2169
        params = deepcopy(params)
2170
2171
2172
2173
2174

        def update():
            if not self.params:
                self.params = params
            else:
2175
                self._params_back_up = deepcopy(self.params)
2176
2177
2178
2179
2180
2181
                self.params.update(params)

        if self.handle is None:
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
2182
2183
                _c_str(_param_dict_to_str(self.params)),
                _c_str(_param_dict_to_str(params)))
2184
2185
2186
2187
2188
2189
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
2190
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
2191
        return self
wxchan's avatar
wxchan committed
2192

2193
    def _reverse_update_params(self) -> "Dataset":
2194
        if self.handle is None:
2195
2196
            self.params = deepcopy(self._params_back_up)
            self._params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
2197
        return self
2198

2199
2200
2201
2202
2203
    def set_field(
        self,
        field_name: str,
        data
    ) -> "Dataset":
wxchan's avatar
wxchan committed
2204
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
2205
2206
2207

        Parameters
        ----------
2208
        field_name : str
2209
            The field name of the information.
2210
2211
        data : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
            The data to be set.
Nikita Titov's avatar
Nikita Titov committed
2212
2213
2214
2215
2216

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
2217
        """
2218
        if self.handle is None:
2219
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
2220
        if data is None:
2221
            # set to None
wxchan's avatar
wxchan committed
2222
2223
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
2224
                _c_str(field_name),
wxchan's avatar
wxchan committed
2225
                None,
Guolin Ke's avatar
Guolin Ke committed
2226
                ctypes.c_int(0),
2227
                ctypes.c_int(_FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
2228
            return self
2229
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
2230
            dtype = np.float64
2231
            if _is_1d_collection(data):
2232
                data = _list_to_1d_numpy(data, dtype, name=field_name)
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
            elif _is_2d_collection(data):
                data = _data_to_2d_numpy(data, dtype, name=field_name)
                data = data.ravel(order='F')
            else:
                raise TypeError(
                    'init_score must be list, numpy 1-D array or pandas Series.\n'
                    'In multiclass classification init_score can also be a list of lists, numpy 2-D array or pandas DataFrame.'
                )
        else:
            dtype = np.int32 if field_name == 'group' else np.float32
2243
            data = _list_to_1d_numpy(data, dtype, name=field_name)
2244

2245
        if data.dtype == np.float32 or data.dtype == np.float64:
2246
            ptr_data, type_data, _ = _c_float_array(data)
wxchan's avatar
wxchan committed
2247
        elif data.dtype == np.int32:
2248
            ptr_data, type_data, _ = _c_int_array(data)
wxchan's avatar
wxchan committed
2249
        else:
2250
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
2251
        if type_data != _FIELD_TYPE_MAPPER[field_name]:
2252
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2253
2254
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
2255
            _c_str(field_name),
wxchan's avatar
wxchan committed
2256
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2257
2258
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2259
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2260
        return self
wxchan's avatar
wxchan committed
2261

2262
    def get_field(self, field_name: str) -> Optional[np.ndarray]:
wxchan's avatar
wxchan committed
2263
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2264
2265
2266

        Parameters
        ----------
2267
        field_name : str
2268
            The field name of the information.
wxchan's avatar
wxchan committed
2269
2270
2271

        Returns
        -------
2272
        info : numpy array or None
2273
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2274
        """
2275
        if self.handle is None:
2276
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2277
2278
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2279
2280
2281
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
            self.handle,
2282
            _c_str(field_name),
wxchan's avatar
wxchan committed
2283
2284
2285
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
2286
        if out_type.value != _FIELD_TYPE_MAPPER[field_name]:
wxchan's avatar
wxchan committed
2287
2288
2289
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
2290
        if out_type.value == _C_API_DTYPE_INT32:
2291
            arr = _cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
2292
        elif out_type.value == _C_API_DTYPE_FLOAT32:
2293
            arr = _cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
2294
        elif out_type.value == _C_API_DTYPE_FLOAT64:
2295
            arr = _cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2296
        else:
wxchan's avatar
wxchan committed
2297
            raise TypeError("Unknown type")
2298
2299
2300
2301
2302
2303
        if field_name == 'init_score':
            num_data = self.num_data()
            num_classes = arr.size // num_data
            if num_classes > 1:
                arr = arr.reshape((num_data, num_classes), order='F')
        return arr
Guolin Ke's avatar
Guolin Ke committed
2304

2305
2306
    def set_categorical_feature(
        self,
2307
        categorical_feature: _LGBM_CategoricalFeatureConfiguration
2308
    ) -> "Dataset":
2309
        """Set categorical features.
2310
2311
2312

        Parameters
        ----------
2313
        categorical_feature : list of str or int, or 'auto'
2314
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2315
2316
2317
2318
2319

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2320
2321
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2322
            return self
2323
        if self.data is not None:
2324
2325
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2326
                return self._free_handle()
2327
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2328
                return self
2329
            else:
2330
2331
2332
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
                                 f'New categorical_feature is {sorted(list(categorical_feature))}')
2333
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2334
                return self._free_handle()
2335
        else:
2336
2337
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2338

2339
2340
2341
2342
    def _set_predictor(
        self,
        predictor: Optional[_InnerPredictor]
    ) -> "Dataset":
2343
2344
2345
2346
        """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
2347
        """
2348
        if predictor is None and self._predictor is None:
Nikita Titov's avatar
Nikita Titov committed
2349
            return self
2350
2351
2352
        elif isinstance(predictor, _InnerPredictor) and isinstance(self._predictor, _InnerPredictor):
            if (predictor == self._predictor) and (predictor.current_iteration() == self._predictor.current_iteration()):
                return self
2353
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
2354
            self._predictor = predictor
2355
2356
        elif self.data is not None:
            self._predictor = predictor
2357
2358
2359
2360
2361
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.data,
                used_indices=None
            )
2362
2363
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
2364
2365
2366
2367
2368
            self._set_init_score_by_predictor(
                predictor=self._predictor,
                data=self.reference.data,
                used_indices=self.used_indices
            )
Guolin Ke's avatar
Guolin Ke committed
2369
        else:
2370
2371
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2372
        return self
Guolin Ke's avatar
Guolin Ke committed
2373

2374
    def set_reference(self, reference: "Dataset") -> "Dataset":
2375
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2376
2377
2378
2379

        Parameters
        ----------
        reference : Dataset
2380
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2381
2382
2383
2384
2385

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2386
        """
2387
2388
2389
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2390
        # we're done if self and reference share a common upstream reference
2391
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2392
            return self
Guolin Ke's avatar
Guolin Ke committed
2393
2394
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2395
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2396
        else:
2397
2398
            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
2399

2400
    def set_feature_name(self, feature_name: Union[List[str], str]) -> "Dataset":
2401
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2402
2403
2404

        Parameters
        ----------
2405
        feature_name : list of str
2406
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2407
2408
2409
2410
2411

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2412
        """
2413
2414
        if feature_name != 'auto':
            self.feature_name = feature_name
2415
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2416
            if len(feature_name) != self.num_feature():
2417
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2418
            c_feature_name = [_c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2419
2420
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
2421
                _c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2422
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2423
        return self
Guolin Ke's avatar
Guolin Ke committed
2424

2425
    def set_label(self, label: Optional[_LGBM_LabelType]) -> "Dataset":
2426
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2427
2428
2429

        Parameters
        ----------
2430
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2431
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2432
2433
2434
2435
2436

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2437
2438
        """
        self.label = label
2439
        if self.handle is not None:
2440
2441
2442
2443
            if isinstance(label, pd_DataFrame):
                if len(label.columns) > 1:
                    raise ValueError('DataFrame for label cannot have multiple columns')
                _check_for_bad_pandas_dtypes(label.dtypes)
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
                try:
                    # most common case (no nullable dtypes)
                    label = label.to_numpy(dtype=np.float32, copy=False)
                except TypeError:
                    # 1.0 <= pd version < 1.1 and nullable dtypes, least common case
                    # raises error because array is casted to type(pd.NA) and there's no na_value argument
                    label = label.astype(np.float32, copy=False).values
                except ValueError:
                    # data has nullable dtypes, but we can specify na_value argument and copy will be made
                    label = label.to_numpy(dtype=np.float32, na_value=np.nan)
                label_array = np.ravel(label)
2455
            else:
2456
                label_array = _list_to_1d_numpy(label, name='label')
2457
            self.set_field('label', label_array)
2458
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2459
        return self
Guolin Ke's avatar
Guolin Ke committed
2460

2461
    def set_weight(self, weight) -> "Dataset":
2462
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2463
2464
2465

        Parameters
        ----------
2466
        weight : list, numpy 1-D array, pandas Series or None
2467
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2468
2469
2470
2471
2472

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2473
        """
2474
2475
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2476
        self.weight = weight
2477
        if self.handle is not None and weight is not None:
2478
            weight = _list_to_1d_numpy(weight, name='weight')
wxchan's avatar
wxchan committed
2479
            self.set_field('weight', weight)
2480
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2481
        return self
Guolin Ke's avatar
Guolin Ke committed
2482

2483
    def set_init_score(self, init_score) -> "Dataset":
2484
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2485
2486
2487

        Parameters
        ----------
2488
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2489
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2490
2491
2492
2493
2494

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2495
2496
        """
        self.init_score = init_score
2497
        if self.handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2498
            self.set_field('init_score', init_score)
2499
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2500
        return self
Guolin Ke's avatar
Guolin Ke committed
2501

2502
    def set_group(self, group) -> "Dataset":
2503
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2504
2505
2506

        Parameters
        ----------
2507
        group : list, numpy 1-D array, pandas Series or None
2508
2509
2510
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2511
2512
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
Nikita Titov's avatar
Nikita Titov committed
2513
2514
2515
2516
2517

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2518
2519
        """
        self.group = group
2520
        if self.handle is not None and group is not None:
2521
            group = _list_to_1d_numpy(group, np.int32, name='group')
wxchan's avatar
wxchan committed
2522
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2523
        return self
Guolin Ke's avatar
Guolin Ke committed
2524

2525
    def get_feature_name(self) -> List[str]:
2526
2527
2528
2529
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2530
        feature_names : list of str
2531
2532
2533
2534
2535
2536
2537
2538
            The names of columns (features) in the Dataset.
        """
        if self.handle is None:
            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)
2539
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2540
2541
2542
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
2543
            ctypes.c_int(num_feature),
2544
            ctypes.byref(tmp_out_len),
2545
            ctypes.c_size_t(reserved_string_buffer_size),
2546
2547
2548
2549
            ctypes.byref(required_string_buffer_size),
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
                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))
2562
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2563

2564
    def get_label(self) -> Optional[np.ndarray]:
2565
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2566
2567
2568

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2569
        label : numpy array or None
2570
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2571
        """
2572
        if self.label is None:
wxchan's avatar
wxchan committed
2573
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2574
2575
        return self.label

2576
    def get_weight(self) -> Optional[np.ndarray]:
2577
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2578
2579
2580

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2581
        weight : numpy array or None
2582
            Weight for each data point from the Dataset. Weights should be non-negative.
Guolin Ke's avatar
Guolin Ke committed
2583
        """
2584
        if self.weight is None:
wxchan's avatar
wxchan committed
2585
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2586
2587
        return self.weight

2588
    def get_init_score(self) -> Optional[np.ndarray]:
2589
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2590
2591
2592

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2593
        init_score : numpy array or None
2594
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2595
        """
2596
        if self.init_score is None:
wxchan's avatar
wxchan committed
2597
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2598
2599
        return self.init_score

2600
2601
2602
2603
2604
    def get_data(self):
        """Get the raw data of the Dataset.

        Returns
        -------
2605
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, Sequence, list of Sequence or list of numpy array or None
2606
2607
2608
2609
            Raw data used in the Dataset construction.
        """
        if self.handle is None:
            raise Exception("Cannot get data before construct Dataset")
2610
        if self._need_slice and self.used_indices is not None and self.reference is not None:
Guolin Ke's avatar
Guolin Ke committed
2611
2612
2613
2614
            self.data = self.reference.data
            if self.data is not None:
                if isinstance(self.data, np.ndarray) or scipy.sparse.issparse(self.data):
                    self.data = self.data[self.used_indices, :]
2615
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2616
                    self.data = self.data.iloc[self.used_indices].copy()
2617
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2618
                    self.data = self.data[self.used_indices, :]
2619
2620
2621
2622
                elif isinstance(self.data, Sequence):
                    self.data = self.data[self.used_indices]
                elif isinstance(self.data, list) and len(self.data) > 0 and all(isinstance(x, Sequence) for x in self.data):
                    self.data = np.array([row for row in self._yield_row_from_seqlist(self.data, self.used_indices)])
Guolin Ke's avatar
Guolin Ke committed
2623
                else:
2624
2625
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
2626
            self._need_slice = False
Guolin Ke's avatar
Guolin Ke committed
2627
2628
2629
        if self.data is None:
            raise LightGBMError("Cannot call `get_data` after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2630
2631
        return self.data

Guolin Ke's avatar
Guolin Ke committed
2632
    def get_group(self):
2633
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2634
2635
2636

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2637
        group : numpy array or None
2638
2639
2640
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2641
2642
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
Guolin Ke's avatar
Guolin Ke committed
2643
        """
2644
        if self.group is None:
wxchan's avatar
wxchan committed
2645
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
2646
2647
            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
2648
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
2649
2650
        return self.group

2651
    def num_data(self) -> int:
2652
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2653
2654
2655

        Returns
        -------
2656
2657
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2658
        """
2659
        if self.handle is not None:
2660
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2661
2662
2663
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2664
        else:
2665
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2666

2667
    def num_feature(self) -> int:
2668
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2669
2670
2671

        Returns
        -------
2672
2673
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2674
        """
2675
        if self.handle is not None:
2676
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2677
2678
2679
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2680
        else:
2681
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2682

2683
    def feature_num_bin(self, feature: Union[int, str]) -> int:
2684
2685
2686
2687
        """Get the number of bins for a feature.

        Parameters
        ----------
2688
2689
        feature : int or str
            Index or name of the feature.
2690
2691
2692
2693
2694
2695
2696

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
        if self.handle is not None:
2697
            if isinstance(feature, str):
2698
2699
2700
                feature_index = self.feature_name.index(feature)
            else:
                feature_index = feature
2701
2702
            ret = ctypes.c_int(0)
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self.handle,
2703
                                                         ctypes.c_int(feature_index),
2704
2705
2706
2707
2708
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

2709
    def get_ref_chain(self, ref_limit: int = 100) -> Set["Dataset"]:
2710
2711
2712
2713
2714
        """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.
2715
2716
2717
2718
2719

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2720
2721
2722

        Returns
        -------
2723
2724
2725
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2726
        head = self
2727
        ref_chain: Set[Dataset] = set()
2728
2729
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2730
                ref_chain.add(head)
2731
2732
2733
2734
2735
2736
                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
2737
        return ref_chain
2738

2739
    def add_features_from(self, other: "Dataset") -> "Dataset":
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
        """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.
        """
        if self.handle is None or other.handle is None:
            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
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
        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))
                elif scipy.sparse.issparse(other.data):
                    self.data = np.hstack((self.data, other.data.toarray()))
2767
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2768
                    self.data = np.hstack((self.data, other.data.values))
2769
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2770
2771
2772
2773
2774
2775
2776
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
            elif scipy.sparse.issparse(self.data):
                sparse_format = self.data.getformat()
                if isinstance(other.data, np.ndarray) or scipy.sparse.issparse(other.data):
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
2777
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2778
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
2779
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2780
2781
2782
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
2783
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2784
2785
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
2786
2787
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
2788
                if isinstance(other.data, np.ndarray):
2789
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
2790
2791
                                       axis=1, ignore_index=True)
                elif scipy.sparse.issparse(other.data):
2792
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
2793
                                       axis=1, ignore_index=True)
2794
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2795
2796
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
2797
2798
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
2799
2800
2801
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
2802
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2803
                if isinstance(other.data, np.ndarray):
2804
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
Guolin Ke's avatar
Guolin Ke committed
2805
                elif scipy.sparse.issparse(other.data):
2806
2807
2808
2809
2810
                    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
2811
2812
2813
2814
2815
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
2816
2817
            err_msg = (f"Cannot add features from {type(other.data).__name__} type of raw data to "
                       f"{old_self_data_type} type of raw data.\n")
Guolin Ke's avatar
Guolin Ke committed
2818
2819
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
2820
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
2821
        self.feature_name = self.get_feature_name()
2822
2823
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
2824
2825
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2826
2827
        return self

2828
    def _dump_text(self, filename: Union[str, Path]) -> "Dataset":
2829
2830
2831
2832
2833
2834
        """Save Dataset to a text file.

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

        Parameters
        ----------
2835
        filename : str or pathlib.Path
2836
2837
2838
2839
2840
2841
2842
2843
2844
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
            self.construct().handle,
2845
            _c_str(str(filename))))
2846
2847
        return self

wxchan's avatar
wxchan committed
2848

2849
2850
2851
2852
_LGBM_CustomObjectiveFunction = Callable[
    [np.ndarray, Dataset],
    Tuple[np.ndarray, np.ndarray]
]
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
_LGBM_CustomEvalFunction = Union[
    Callable[
        [np.ndarray, Dataset],
        _LGBM_EvalFunctionResultType
    ],
    Callable[
        [np.ndarray, Dataset],
        List[_LGBM_EvalFunctionResultType]
    ]
]
2863
2864


2865
class Booster:
2866
    """Booster in LightGBM."""
2867

2868
2869
2870
2871
2872
2873
2874
    def __init__(
        self,
        params: Optional[Dict[str, Any]] = None,
        train_set: Optional[Dataset] = None,
        model_file: Optional[Union[str, Path]] = None,
        model_str: Optional[str] = None
    ):
2875
        """Initialize the Booster.
wxchan's avatar
wxchan committed
2876
2877
2878

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2879
        params : dict or None, optional (default=None)
2880
2881
2882
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
2883
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
2884
            Path to the model file.
2885
        model_str : str or None, optional (default=None)
2886
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
2887
        """
2888
        self.handle = None
2889
        self._network = False
wxchan's avatar
wxchan committed
2890
        self.__need_reload_eval_info = True
2891
        self._train_data_name = "training"
2892
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
2893
        self.best_iteration = -1
2894
        self.best_score: _LGBM_BoosterBestScoreType = {}
2895
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
2896
        if train_set is not None:
2897
            # Training task
wxchan's avatar
wxchan committed
2898
            if not isinstance(train_set, Dataset):
2899
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
            params = _choose_param_value(
                main_param_name="machines",
                params=params,
                default_value=None
            )
            # if "machines" is given, assume user wants to do distributed learning, and set up network
            if params["machines"] is None:
                params.pop("machines", None)
            else:
                machines = params["machines"]
                if isinstance(machines, str):
                    num_machines_from_machine_list = len(machines.split(','))
                elif isinstance(machines, (list, set)):
                    num_machines_from_machine_list = len(machines)
                    machines = ','.join(machines)
                else:
                    raise ValueError("Invalid machines in params.")

                params = _choose_param_value(
                    main_param_name="num_machines",
                    params=params,
                    default_value=num_machines_from_machine_list
                )
                params = _choose_param_value(
                    main_param_name="local_listen_port",
                    params=params,
                    default_value=12400
                )
                self.set_network(
                    machines=machines,
                    local_listen_port=params["local_listen_port"],
                    listen_time_out=params.get("time_out", 120),
                    num_machines=params["num_machines"]
                )
2934
            # construct booster object
2935
2936
2937
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
2938
            params_str = _param_dict_to_str(params)
2939
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2940
            _safe_call(_LIB.LGBM_BoosterCreate(
2941
                train_set.handle,
2942
                _c_str(params_str),
wxchan's avatar
wxchan committed
2943
                ctypes.byref(self.handle)))
2944
            # save reference to data
wxchan's avatar
wxchan committed
2945
            self.train_set = train_set
2946
2947
            self.valid_sets: List[Dataset] = []
            self.name_valid_sets: List[str] = []
wxchan's avatar
wxchan committed
2948
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
2949
2950
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
2951
2952
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
2953
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
2954
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2955
2956
2957
2958
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2959
            # buffer for inner predict
wxchan's avatar
wxchan committed
2960
2961
2962
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
2963
            self.pandas_categorical = train_set.pandas_categorical
2964
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
2965
        elif model_file is not None:
2966
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
2967
            out_num_iterations = ctypes.c_int(0)
2968
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2969
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
2970
                _c_str(str(model_file)),
wxchan's avatar
wxchan committed
2971
2972
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2973
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2974
2975
2976
2977
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2978
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
2979
2980
2981
            if params:
                _log_warning('Ignoring params argument, using parameters from model file.')
            params = self._get_loaded_param()
2982
        elif model_str is not None:
2983
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
2984
        else:
2985
2986
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
2987
        self.params = params
wxchan's avatar
wxchan committed
2988

2989
    def __del__(self) -> None:
2990
        try:
2991
            if self._network:
2992
2993
2994
2995
2996
2997
2998
2999
                self.free_network()
        except AttributeError:
            pass
        try:
            if self.handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
3000

3001
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
3002
3003
        return self.__deepcopy__(None)

3004
    def __deepcopy__(self, _) -> "Booster":
3005
        model_str = self.model_to_string(num_iteration=-1)
3006
        booster = Booster(model_str=model_str)
3007
        return booster
wxchan's avatar
wxchan committed
3008

3009
    def __getstate__(self) -> Dict[str, Any]:
wxchan's avatar
wxchan committed
3010
3011
3012
3013
3014
        this = self.__dict__.copy()
        handle = this['handle']
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
3015
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
3016
3017
        return this

3018
    def __setstate__(self, state: Dict[str, Any]) -> None:
3019
3020
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
3021
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
3022
            out_num_iterations = ctypes.c_int(0)
3023
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3024
                _c_str(model_str),
3025
3026
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
3027
3028
3029
            state['handle'] = handle
        self.__dict__.update(state)

3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
    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)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterGetLoadedParam(
            self.handle,
            ctypes.c_int64(buffer_len),
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
        # if buffer length is not long enough, re-allocate a buffer
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _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'))

3052
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3053
3054
3055
3056
3057
3058
3059
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
3060
3061
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
3062
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
3063
        return self
wxchan's avatar
wxchan committed
3064

3065
    def _free_buffer(self) -> "Booster":
3066
3067
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
3068
        return self
3069

3070
3071
3072
3073
3074
3075
3076
    def set_network(
        self,
        machines: Union[List[str], Set[str], str],
        local_listen_port: int = 12400,
        listen_time_out: int = 120,
        num_machines: int = 1
    ) -> "Booster":
3077
3078
3079
3080
        """Set the network configuration.

        Parameters
        ----------
3081
        machines : list, set or str
3082
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
3083
        local_listen_port : int, optional (default=12400)
3084
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
3085
        listen_time_out : int, optional (default=120)
3086
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
3087
        num_machines : int, optional (default=1)
3088
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
3089
3090
3091
3092
3093

        Returns
        -------
        self : Booster
            Booster with set network.
3094
        """
3095
3096
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
3097
        _safe_call(_LIB.LGBM_NetworkInit(_c_str(machines),
3098
3099
3100
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
3101
        self._network = True
Nikita Titov's avatar
Nikita Titov committed
3102
        return self
3103

3104
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3105
3106
3107
3108
3109
3110
3111
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
3112
        _safe_call(_LIB.LGBM_NetworkFree())
3113
        self._network = False
Nikita Titov's avatar
Nikita Titov committed
3114
        return self
3115

3116
    def trees_to_dataframe(self) -> pd_DataFrame:
3117
3118
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

3119
3120
3121
3122
        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.
3123
3124
3125
3126
3127
            - ``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.
3128
3129
            - ``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.
3130
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
3131
3132
              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.
3133
3134
            - ``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.
3135
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
3136
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
3137
3138
            - ``count`` : int64, number of records in the training data that fall into this node.

3139
3140
3141
3142
3143
3144
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
3145
3146
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157

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

        def _is_split_node(tree):
            return 'split_index' in tree.keys()

        def create_node_record(tree, node_depth=1, tree_index=None,
                               feature_names=None, parent_node=None):

            def _get_node_index(tree, tree_index):
3158
                tree_num = f'{tree_index}-' if tree_index is not None else ''
3159
3160
3161
                is_split = _is_split_node(tree)
                node_type = 'S' if is_split else 'L'
                # if a single node tree it won't have `leaf_index` so return 0
3162
3163
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175

            def _get_split_feature(tree, feature_names):
                if _is_split_node(tree):
                    if feature_names is not None:
                        feature_name = feature_names[tree['split_feature']]
                    else:
                        feature_name = tree['split_feature']
                else:
                    feature_name = None
                return feature_name

            def _is_single_node_tree(tree):
3176
                return set(tree.keys()) == {'leaf_value'}
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249

            # Create the node record, and populate universal data members
            node = OrderedDict()
            node['tree_index'] = tree_index
            node['node_depth'] = node_depth
            node['node_index'] = _get_node_index(tree, tree_index)
            node['left_child'] = None
            node['right_child'] = None
            node['parent_index'] = parent_node
            node['split_feature'] = _get_split_feature(tree, feature_names)
            node['split_gain'] = None
            node['threshold'] = None
            node['decision_type'] = None
            node['missing_direction'] = None
            node['missing_type'] = None
            node['value'] = None
            node['weight'] = None
            node['count'] = None

            # Update values to reflect node type (leaf or split)
            if _is_split_node(tree):
                node['left_child'] = _get_node_index(tree['left_child'], tree_index)
                node['right_child'] = _get_node_index(tree['right_child'], tree_index)
                node['split_gain'] = tree['split_gain']
                node['threshold'] = tree['threshold']
                node['decision_type'] = tree['decision_type']
                node['missing_direction'] = 'left' if tree['default_left'] else 'right'
                node['missing_type'] = tree['missing_type']
                node['value'] = tree['internal_value']
                node['weight'] = tree['internal_weight']
                node['count'] = tree['internal_count']
            else:
                node['value'] = tree['leaf_value']
                if not _is_single_node_tree(tree):
                    node['weight'] = tree['leaf_weight']
                    node['count'] = tree['leaf_count']

            return node

        def tree_dict_to_node_list(tree, node_depth=1, tree_index=None,
                                   feature_names=None, parent_node=None):

            node = create_node_record(tree,
                                      node_depth=node_depth,
                                      tree_index=tree_index,
                                      feature_names=feature_names,
                                      parent_node=parent_node)

            res = [node]

            if _is_split_node(tree):
                # traverse the next level of the tree
                children = ['left_child', 'right_child']
                for child in children:
                    subtree_list = tree_dict_to_node_list(
                        tree[child],
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
                        parent_node=node['node_index'])
                    # In tree format, "subtree_list" is a list of node records (dicts),
                    # and we add node to the list.
                    res.extend(subtree_list)
            return res

        model_dict = self.dump_model()
        feature_names = model_dict['feature_names']
        model_list = []
        for tree in model_dict['tree_info']:
            model_list.extend(tree_dict_to_node_list(tree['tree_structure'],
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

3250
        return pd_DataFrame(model_list, columns=model_list[0].keys())
3251

3252
    def set_train_data_name(self, name: str) -> "Booster":
3253
3254
3255
3256
        """Set the name to the training Dataset.

        Parameters
        ----------
3257
        name : str
Nikita Titov's avatar
Nikita Titov committed
3258
3259
3260
3261
3262
3263
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
3264
        """
3265
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
3266
        return self
wxchan's avatar
wxchan committed
3267

3268
    def add_valid(self, data: Dataset, name: str) -> "Booster":
3269
        """Add validation data.
wxchan's avatar
wxchan committed
3270
3271
3272
3273

        Parameters
        ----------
        data : Dataset
3274
            Validation data.
3275
        name : str
3276
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
3277
3278
3279
3280
3281

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
3282
        """
Guolin Ke's avatar
Guolin Ke committed
3283
        if not isinstance(data, Dataset):
3284
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3285
        if data._predictor is not self.__init_predictor:
3286
3287
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3288
3289
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
3290
            data.construct().handle))
wxchan's avatar
wxchan committed
3291
3292
3293
3294
3295
        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
3296
        return self
wxchan's avatar
wxchan committed
3297

3298
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
3299
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
3300
3301
3302
3303

        Parameters
        ----------
        params : dict
3304
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
3305
3306
3307
3308
3309

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
3310
        """
3311
        params_str = _param_dict_to_str(params)
wxchan's avatar
wxchan committed
3312
3313
3314
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
                self.handle,
3315
                _c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
3316
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
3317
        return self
wxchan's avatar
wxchan committed
3318

3319
3320
3321
3322
3323
    def update(
        self,
        train_set: Optional[Dataset] = None,
        fobj: Optional[_LGBM_CustomObjectiveFunction] = None
    ) -> bool:
Nikita Titov's avatar
Nikita Titov committed
3324
        """Update Booster for one iteration.
3325

wxchan's avatar
wxchan committed
3326
3327
        Parameters
        ----------
3328
3329
3330
3331
        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
3332
            Customized objective function.
3333
3334
3335
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3336
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3337
                    The predicted values.
3338
3339
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3340
3341
                train_data : Dataset
                    The training dataset.
3342
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3343
3344
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3345
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3346
3347
                    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
3348

3349
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3350
            and grad and hess should be returned in the same format.
3351

wxchan's avatar
wxchan committed
3352
3353
        Returns
        -------
3354
3355
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3356
        """
3357
        # need reset training data
3358
3359
3360
3361
3362
3363
        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
3364
            if not isinstance(train_set, Dataset):
3365
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3366
            if train_set._predictor is not self.__init_predictor:
3367
3368
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3369
3370
3371
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
3372
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
3373
            self.__inner_predict_buffer[0] = None
3374
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3375
3376
        is_finished = ctypes.c_int(0)
        if fobj is None:
3377
            if self.__set_objective_to_none:
3378
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3379
3380
3381
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
3382
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3383
3384
            return is_finished.value == 1
        else:
3385
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3386
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3387
3388
3389
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

3390
3391
3392
3393
3394
    def __boost(
        self,
        grad: np.ndarray,
        hess: np.ndarray
    ) -> bool:
3395
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3396

Nikita Titov's avatar
Nikita Titov committed
3397
3398
        .. note::

3399
3400
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3401
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3402
            and grad and hess should be returned in the same format.
3403

wxchan's avatar
wxchan committed
3404
3405
        Parameters
        ----------
3406
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3407
3408
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3409
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3410
3411
            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
3412
3413
3414

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3415
3416
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3417
        """
3418
3419
3420
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3421
3422
        grad = _list_to_1d_numpy(grad, name='gradient')
        hess = _list_to_1d_numpy(hess, name='hessian')
3423
3424
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3425
        if len(grad) != len(hess):
3426
3427
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3428
        if len(grad) != num_train_data * self.__num_class:
3429
3430
3431
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3432
                f"number of models per one iteration ({self.__num_class})"
3433
            )
wxchan's avatar
wxchan committed
3434
3435
3436
3437
3438
3439
        is_finished = ctypes.c_int(0)
        _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)))
3440
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3441
3442
        return is_finished.value == 1

3443
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3444
3445
3446
3447
3448
3449
3450
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3451
3452
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
3453
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3454
        return self
wxchan's avatar
wxchan committed
3455

3456
    def current_iteration(self) -> int:
3457
3458
3459
3460
3461
3462
3463
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3464
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3465
3466
3467
3468
3469
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3470
    def num_model_per_iteration(self) -> int:
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
        """Get number of models per iteration.

        Returns
        -------
        model_per_iter : int
            The number of models per iteration.
        """
        model_per_iter = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterNumModelPerIteration(
            self.handle,
            ctypes.byref(model_per_iter)))
        return model_per_iter.value

3484
    def num_trees(self) -> int:
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
        """Get number of weak sub-models.

        Returns
        -------
        num_trees : int
            The number of weak sub-models.
        """
        num_trees = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterNumberOfTotalModel(
            self.handle,
            ctypes.byref(num_trees)))
        return num_trees.value

3498
    def upper_bound(self) -> float:
3499
3500
3501
3502
        """Get upper bound value of a model.

        Returns
        -------
3503
        upper_bound : float
3504
3505
3506
3507
3508
3509
3510
3511
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

3512
    def lower_bound(self) -> float:
3513
3514
3515
3516
        """Get lower bound value of a model.

        Returns
        -------
3517
        lower_bound : float
3518
3519
3520
3521
3522
3523
3524
3525
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

3526
3527
3528
3529
3530
3531
    def eval(
        self,
        data: Dataset,
        name: str,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3532
        """Evaluate for data.
wxchan's avatar
wxchan committed
3533
3534
3535

        Parameters
        ----------
3536
3537
        data : Dataset
            Data for the evaluating.
3538
        name : str
3539
            Name of the data.
3540
        feval : callable, list of callable, or None, optional (default=None)
3541
            Customized evaluation function.
3542
            Each evaluation function should accept two parameters: preds, eval_data,
3543
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3544

3545
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3546
                    The predicted values.
3547
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3548
                    If custom objective function is used, predicted values are returned before any transformation,
3549
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3550
                eval_data : Dataset
3551
                    A ``Dataset`` to evaluate.
3552
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3553
                    The name of evaluation function (without whitespace).
3554
3555
3556
3557
3558
                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
3559
3560
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3561
        result : list
3562
            List with (dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3563
        """
Guolin Ke's avatar
Guolin Ke committed
3564
3565
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3566
3567
3568
3569
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3570
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3571
3572
3573
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3574
        # need to push new valid data
wxchan's avatar
wxchan committed
3575
3576
3577
3578
3579
3580
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

3581
3582
3583
3584
    def eval_train(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3585
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3586
3587
3588

        Parameters
        ----------
3589
        feval : callable, list of callable, or None, optional (default=None)
3590
            Customized evaluation function.
3591
            Each evaluation function should accept two parameters: preds, eval_data,
3592
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3593

3594
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3595
                    The predicted values.
3596
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3597
                    If custom objective function is used, predicted values are returned before any transformation,
3598
                    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
3599
                eval_data : Dataset
3600
                    The training dataset.
3601
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3602
                    The name of evaluation function (without whitespace).
3603
3604
3605
3606
3607
                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
3608
3609
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3610
        result : list
3611
            List with (train_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3612
        """
3613
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
3614

3615
3616
3617
3618
    def eval_valid(
        self,
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]] = None
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
3619
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
3620
3621
3622

        Parameters
        ----------
3623
        feval : callable, list of callable, or None, optional (default=None)
3624
            Customized evaluation function.
3625
            Each evaluation function should accept two parameters: preds, eval_data,
3626
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3627

3628
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3629
                    The predicted values.
3630
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3631
                    If custom objective function is used, predicted values are returned before any transformation,
3632
                    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
3633
                eval_data : Dataset
3634
                    The validation dataset.
3635
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3636
                    The name of evaluation function (without whitespace).
3637
3638
3639
3640
3641
                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
3642
3643
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3644
        result : list
3645
            List with (validation_dataset_name, eval_name, eval_result, is_higher_better) tuples.
wxchan's avatar
wxchan committed
3646
        """
3647
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
3648
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
3649

3650
3651
3652
3653
3654
3655
3656
    def save_model(
        self,
        filename: Union[str, Path],
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> "Booster":
3657
        """Save Booster to file.
wxchan's avatar
wxchan committed
3658
3659
3660

        Parameters
        ----------
3661
        filename : str or pathlib.Path
3662
            Filename to save Booster.
3663
3664
3665
3666
        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
3667
        start_iteration : int, optional (default=0)
3668
            Start index of the iteration that should be saved.
3669
        importance_type : str, optional (default="split")
3670
3671
3672
            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
3673
3674
3675
3676
3677

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
3678
        """
3679
        if num_iteration is None:
3680
            num_iteration = self.best_iteration
3681
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3682
3683
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
3684
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3685
            ctypes.c_int(num_iteration),
3686
            ctypes.c_int(importance_type_int),
3687
            _c_str(str(filename))))
3688
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
3689
        return self
wxchan's avatar
wxchan committed
3690

3691
3692
3693
3694
3695
    def shuffle_models(
        self,
        start_iteration: int = 0,
        end_iteration: int = -1
    ) -> "Booster":
3696
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
3697

3698
3699
3700
        Parameters
        ----------
        start_iteration : int, optional (default=0)
3701
            The first iteration that will be shuffled.
3702
3703
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
3704
            If <= 0, means the last available iteration.
3705

Nikita Titov's avatar
Nikita Titov committed
3706
3707
3708
3709
        Returns
        -------
        self : Booster
            Booster with shuffled models.
3710
        """
3711
3712
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
3713
3714
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
3715
        return self
3716

3717
    def model_from_string(self, model_str: str) -> "Booster":
3718
3719
3720
3721
        """Load Booster from a string.

        Parameters
        ----------
3722
        model_str : str
3723
3724
3725
3726
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3727
        self : Booster
3728
3729
            Loaded Booster object.
        """
3730
3731
3732
3733
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
3734
3735
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
3736
            _c_str(model_str),
3737
3738
3739
3740
3741
3742
3743
            ctypes.byref(out_num_iterations),
            ctypes.byref(self.handle)))
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
            self.handle,
            ctypes.byref(out_num_class)))
        self.__num_class = out_num_class.value
3744
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
3745
3746
        return self

3747
3748
3749
3750
3751
3752
    def model_to_string(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split'
    ) -> str:
3753
        """Save Booster to string.
3754

3755
3756
3757
3758
3759
3760
        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
3761
        start_iteration : int, optional (default=0)
3762
            Start index of the iteration that should be saved.
3763
        importance_type : str, optional (default="split")
3764
3765
3766
            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.
3767
3768
3769

        Returns
        -------
3770
        str_repr : str
3771
3772
            String representation of Booster.
        """
3773
        if num_iteration is None:
3774
            num_iteration = self.best_iteration
3775
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3776
        buffer_len = 1 << 20
3777
        tmp_out_len = ctypes.c_int64(0)
3778
3779
3780
3781
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
            self.handle,
3782
            ctypes.c_int(start_iteration),
3783
            ctypes.c_int(num_iteration),
3784
            ctypes.c_int(importance_type_int),
3785
            ctypes.c_int64(buffer_len),
3786
3787
3788
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
3789
        # if buffer length is not long enough, re-allocate a buffer
3790
3791
3792
3793
3794
        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,
3795
                ctypes.c_int(start_iteration),
3796
                ctypes.c_int(num_iteration),
3797
                ctypes.c_int(importance_type_int),
3798
                ctypes.c_int64(actual_len),
3799
3800
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
3801
        ret = string_buffer.value.decode('utf-8')
3802
3803
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
3804

3805
3806
3807
3808
3809
3810
3811
    def dump_model(
        self,
        num_iteration: Optional[int] = None,
        start_iteration: int = 0,
        importance_type: str = 'split',
        object_hook: Optional[Callable[[Dict[str, Any]], Dict[str, Any]]] = None
    ) -> Dict[str, Any]:
Nikita Titov's avatar
Nikita Titov committed
3812
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
3813

3814
3815
        Parameters
        ----------
3816
3817
3818
3819
        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
3820
        start_iteration : int, optional (default=0)
3821
            Start index of the iteration that should be dumped.
3822
        importance_type : str, optional (default="split")
3823
3824
3825
            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.
3826
3827
3828
3829
3830
3831
3832
3833
3834
        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.
3835

wxchan's avatar
wxchan committed
3836
3837
        Returns
        -------
3838
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
3839
            JSON format of Booster.
wxchan's avatar
wxchan committed
3840
        """
3841
        if num_iteration is None:
3842
            num_iteration = self.best_iteration
3843
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3844
        buffer_len = 1 << 20
3845
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
3846
3847
3848
3849
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterDumpModel(
            self.handle,
3850
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3851
            ctypes.c_int(num_iteration),
3852
            ctypes.c_int(importance_type_int),
3853
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
3854
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3855
            ptr_string_buffer))
wxchan's avatar
wxchan committed
3856
        actual_len = tmp_out_len.value
3857
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
3858
3859
3860
3861
3862
        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,
3863
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3864
                ctypes.c_int(num_iteration),
3865
                ctypes.c_int(importance_type_int),
3866
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
3867
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3868
                ptr_string_buffer))
3869
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
3870
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
3871
                                                          default=_json_default_with_numpy))
3872
        return ret
wxchan's avatar
wxchan committed
3873

3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
    def predict(
        self,
        data,
        start_iteration: int = 0,
        num_iteration: Optional[int] = None,
        raw_score: bool = False,
        pred_leaf: bool = False,
        pred_contrib: bool = False,
        data_has_header: bool = False,
        validate_features: bool = False,
        **kwargs: Any
    ):
3886
        """Make a prediction.
wxchan's avatar
wxchan committed
3887
3888
3889

        Parameters
        ----------
3890
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
3891
            Data source for prediction.
3892
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3893
        start_iteration : int, optional (default=0)
3894
            Start index of the iteration to predict.
3895
            If <= 0, starts from the first iteration.
3896
        num_iteration : int or None, optional (default=None)
3897
3898
3899
3900
            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).
3901
3902
3903
3904
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
3905
3906
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
3907

Nikita Titov's avatar
Nikita Titov committed
3908
3909
3910
3911
3912
3913
3914
            .. 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.
3915

3916
3917
        data_has_header : bool, optional (default=False)
            Whether the data has header.
3918
            Used only if data is str.
3919
3920
3921
        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.
3922
3923
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
3924
3925
3926

        Returns
        -------
3927
        result : numpy array, scipy.sparse or list of scipy.sparse
3928
            Prediction result.
3929
            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
3930
        """
3931
        predictor = self._to_predictor(deepcopy(kwargs))
3932
        if num_iteration is None:
3933
            if start_iteration <= 0:
3934
3935
3936
3937
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
        return predictor.predict(data, start_iteration, num_iteration,
3938
                                 raw_score, pred_leaf, pred_contrib,
3939
                                 data_has_header, validate_features)
wxchan's avatar
wxchan committed
3940

3941
3942
3943
3944
    def refit(
        self,
        data,
        label,
3945
3946
        decay_rate: float = 0.9,
        reference: Optional[Dataset] = None,
3947
3948
3949
        weight=None,
        group=None,
        init_score=None,
3950
3951
        feature_name: _LGBM_FeatureNameConfiguration = 'auto',
        categorical_feature: _LGBM_CategoricalFeatureConfiguration = 'auto',
3952
3953
3954
        dataset_params: Optional[Dict[str, Any]] = None,
        free_raw_data: bool = True,
        validate_features: bool = False,
3955
3956
        **kwargs
    ):
Guolin Ke's avatar
Guolin Ke committed
3957
3958
3959
3960
        """Refit the existing Booster by new data.

        Parameters
        ----------
3961
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
Guolin Ke's avatar
Guolin Ke committed
3962
            Data source for refit.
3963
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3964
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
3965
3966
            Label for refit.
        decay_rate : float, optional (default=0.9)
3967
3968
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
3969
3970
3971
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
3972
            Weight for each ``data`` instance. Weights should be non-negative.
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
            Group/query size for ``data``.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups,
            where the first 10 records are in the first group, records 11-30 are in the second group, records 31-70 are in the third group, etc.
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None, optional (default=None)
            Init score for ``data``.
        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.
        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.
3989
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
3990
3991
3992
            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.
3993
            Floating point numbers in categorical features will be rounded towards 0.
3994
3995
3996
3997
        dataset_params : dict or None, optional (default=None)
            Other parameters for Dataset ``data``.
        free_raw_data : bool, optional (default=True)
            If True, raw data is freed after constructing inner Dataset for ``data``.
3998
3999
4000
        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.
4001
4002
        **kwargs
            Other parameters for refit.
4003
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
4004
4005
4006
4007
4008
4009

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
4010
4011
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
4012
4013
        if dataset_params is None:
            dataset_params = {}
4014
        predictor = self._to_predictor(deepcopy(kwargs))
4015
        leaf_preds = predictor.predict(data, -1, pred_leaf=True, validate_features=validate_features)
4016
        nrow, ncol = leaf_preds.shape
4017
        out_is_linear = ctypes.c_int(0)
4018
4019
4020
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
4021
4022
4023
4024
4025
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
4026
        new_params["linear_tree"] = bool(out_is_linear.value)
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
        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,
        )
4040
        new_params['refit_decay_rate'] = decay_rate
4041
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
4042
4043
4044
4045
4046
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
4047
        ptr_data, _, _ = _c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
4048
4049
4050
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
4051
4052
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
4053
        new_booster._network = self._network
Guolin Ke's avatar
Guolin Ke committed
4054
4055
        return new_booster

4056
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
        """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.
        """
4071
4072
4073
4074
4075
4076
4077
4078
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
            self.handle,
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
    def set_leaf_output(
        self,
        tree_id: int,
        leaf_id: int,
        value: float,
    ) -> 'Booster':
        """Set 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.
        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(
                self.handle,
                ctypes.c_int(tree_id),
                ctypes.c_int(leaf_id),
                ctypes.c_double(value)
            )
        )
        return self

4111
4112
4113
4114
    def _to_predictor(
        self,
        pred_parameter: Optional[Dict[str, Any]] = None
    ) -> _InnerPredictor:
4115
        """Convert to predictor."""
4116
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
4117
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
4118
4119
        return predictor

4120
    def num_feature(self) -> int:
4121
4122
4123
4124
4125
4126
4127
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
4128
4129
4130
4131
4132
4133
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
            self.handle,
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

4134
    def feature_name(self) -> List[str]:
4135
        """Get names of features.
wxchan's avatar
wxchan committed
4136
4137
4138

        Returns
        -------
4139
        result : list of str
4140
            List with names of features.
wxchan's avatar
wxchan committed
4141
        """
4142
        num_feature = self.num_feature()
4143
        # Get name of features
wxchan's avatar
wxchan committed
4144
        tmp_out_len = ctypes.c_int(0)
4145
4146
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
4147
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
4148
4149
4150
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
4151
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
4152
            ctypes.byref(tmp_out_len),
4153
            ctypes.c_size_t(reserved_string_buffer_size),
4154
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4155
4156
4157
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
        actual_string_buffer_size = required_string_buffer_size.value
        # if buffer length is not long enough, reallocate buffers
        if reserved_string_buffer_size < actual_string_buffer_size:
            string_buffers = [ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(num_feature)]
            ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
            _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
                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))
4170
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
4171

4172
4173
4174
4175
4176
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
4177
        """Get feature importances.
4178

4179
4180
        Parameters
        ----------
4181
        importance_type : str, optional (default="split")
4182
4183
4184
            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.
4185
4186
4187
4188
        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).
4189

4190
4191
        Returns
        -------
4192
4193
        result : numpy array
            Array with feature importances.
4194
        """
4195
4196
        if iteration is None:
            iteration = self.best_iteration
4197
        importance_type_int = _FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
4198
        result = np.empty(self.num_feature(), dtype=np.float64)
4199
4200
4201
4202
4203
        _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))))
4204
        if importance_type_int == _C_API_FEATURE_IMPORTANCE_SPLIT:
4205
            return result.astype(np.int32)
4206
4207
        else:
            return result
4208

4209
4210
4211
4212
4213
4214
    def get_split_value_histogram(
        self,
        feature: Union[int, str],
        bins: Optional[Union[int, str]] = None,
        xgboost_style: bool = False
    ) -> Union[Tuple[np.ndarray, np.ndarray], np.ndarray, pd_DataFrame]:
4215
4216
4217
4218
        """Get split value histogram for the specified feature.

        Parameters
        ----------
4219
        feature : int or str
4220
4221
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
4222
            If str, interpreted as name.
4223

Nikita Titov's avatar
Nikita Titov committed
4224
4225
4226
            .. warning::

                Categorical features are not supported.
4227

4228
        bins : int, str or None, optional (default=None)
4229
4230
4231
            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.
4232
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
        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.
        """
4247
        def add(root: Dict[str, Any]) -> None:
4248
4249
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
4250
                if feature_names is not None and isinstance(feature, str):
4251
4252
4253
4254
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
4255
                    if isinstance(root['threshold'], str):
4256
4257
4258
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
        values = []
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

4269
        if bins is None or isinstance(bins, int) and xgboost_style:
4270
4271
4272
4273
4274
4275
4276
            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:
4277
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
4278
4279
4280
4281
4282
            else:
                return ret
        else:
            return hist, bin_edges

4283
4284
4285
4286
    def __inner_eval(
        self,
        data_name: str,
        data_idx: int,
4287
        feval: Optional[Union[_LGBM_CustomEvalFunction, List[_LGBM_CustomEvalFunction]]]
4288
    ) -> List[_LGBM_BoosterEvalMethodResultType]:
4289
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
4290
        if data_idx >= self.__num_dataset:
4291
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4292
4293
4294
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
4295
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
4296
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
4297
4298
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
4299
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4300
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
4301
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
4302
            if tmp_out_len.value != self.__num_inner_eval:
4303
                raise ValueError("Wrong length of eval results")
4304
            for i in range(self.__num_inner_eval):
4305
4306
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
4307
4308
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
4309
4310
4311
4312
4313
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
4314
4315
4316
4317
4318
4319
4320
4321
4322
            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
4323
4324
4325
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

4326
    def __inner_predict(self, data_idx: int) -> np.ndarray:
4327
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
4328
        if data_idx >= self.__num_dataset:
4329
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
4330
4331
4332
4333
4334
        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
4335
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
4336
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
4337
4338
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
4339
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
4340
4341
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
4342
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
4343
4344
4345
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
4346
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
4347
            self.__is_predicted_cur_iter[data_idx] = True
4348
4349
4350
4351
4352
        result = self.__inner_predict_buffer[data_idx]
        if self.__num_class > 1:
            num_data = result.size // self.__num_class
            result = result.reshape(num_data, self.__num_class, order='F')
        return result
wxchan's avatar
wxchan committed
4353

4354
    def __get_eval_info(self) -> None:
4355
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
4356
4357
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
4358
            out_num_eval = ctypes.c_int(0)
4359
            # Get num of inner evals
wxchan's avatar
wxchan committed
4360
4361
4362
4363
4364
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
                self.handle,
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
4365
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
4366
                tmp_out_len = ctypes.c_int(0)
4367
4368
4369
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
4370
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
4371
                ]
wxchan's avatar
wxchan committed
4372
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
4373
4374
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
4375
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
4376
                    ctypes.byref(tmp_out_len),
4377
                    ctypes.c_size_t(reserved_string_buffer_size),
4378
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
4379
4380
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
4381
                    raise ValueError("Length of eval names doesn't equal with num_evals")
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
                actual_string_buffer_size = required_string_buffer_size.value
                # if buffer length is not long enough, reallocate buffers
                if reserved_string_buffer_size < actual_string_buffer_size:
                    string_buffers = [
                        ctypes.create_string_buffer(actual_string_buffer_size) for _ in range(self.__num_inner_eval)
                    ]
                    ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
                    _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                        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)
                ]
                self.__higher_better_inner_eval = [
                    name.startswith(('auc', 'ndcg@', 'map@', 'average_precision')) for name in self.__name_inner_eval
                ]