basic.py 165 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 functools import wraps
10
11
12
from os import SEEK_END
from os.path import getsize
from pathlib import Path
13
from tempfile import NamedTemporaryFile
14
from typing import Any, Callable, Dict, Iterable, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
15
16
17
18

import numpy as np
import scipy.sparse

19
from .compat import PANDAS_INSTALLED, concat, dt_DataTable, pd_CategoricalDtype, pd_DataFrame, pd_Series
wxchan's avatar
wxchan committed
20
21
from .libpath import find_lib_path

22
23
24
25
26
27
28
29
30
31
32
33
ZERO_THRESHOLD = 1e-35


def _get_sample_count(total_nrow: int, params: str):
    sample_cnt = ctypes.c_int(0)
    _safe_call(_LIB.LGBM_GetSampleCount(
        ctypes.c_int32(total_nrow),
        c_str(params),
        ctypes.byref(sample_cnt),
    ))
    return sample_cnt.value

wxchan's avatar
wxchan committed
34

35
class _DummyLogger:
36
    def info(self, msg: str) -> None:
37
38
        print(msg)

39
    def warning(self, msg: str) -> None:
40
41
42
        warnings.warn(msg, stacklevel=3)


43
44
45
_LOGGER: Any = _DummyLogger()
_INFO_METHOD_NAME = "info"
_WARNING_METHOD_NAME = "warning"
46
47


48
49
50
def register_logger(
    logger: Any, info_method_name: str = "info", warning_method_name: str = "warning"
) -> None:
51
52
53
54
    """Register custom logger.

    Parameters
    ----------
55
    logger : Any
56
        Custom logger.
57
58
59
60
    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.
61
    """
62
63
64
65
66
67
68
69
70
    def _has_method(logger: Any, method_name: str) -> bool:
        return callable(getattr(logger, method_name, None))

    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
71
    _LOGGER = logger
72
73
    _INFO_METHOD_NAME = info_method_name
    _WARNING_METHOD_NAME = warning_method_name
74
75


76
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
77
    """Join log messages from native library which come by chunks."""
78
    msg_normalized: List[str] = []
79
80

    @wraps(func)
81
    def wrapper(msg: str) -> None:
82
83
84
85
86
87
88
89
90
91
92
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


93
def _log_info(msg: str) -> None:
94
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
95
96


97
def _log_warning(msg: str) -> None:
98
    getattr(_LOGGER, _WARNING_METHOD_NAME)(msg)
99
100
101


@_normalize_native_string
102
def _log_native(msg: str) -> None:
103
    getattr(_LOGGER, _INFO_METHOD_NAME)(msg)
104
105


106
def _log_callback(msg: bytes) -> None:
107
    """Redirect logs from native library into Python."""
108
    _log_native(str(msg.decode('utf-8')))
109
110


wxchan's avatar
wxchan committed
111
def _load_lib():
112
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
113
114
    lib_path = find_lib_path()
    if len(lib_path) == 0:
115
        return None
wxchan's avatar
wxchan committed
116
117
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
118
119
120
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
    lib.callback = callback(_log_callback)
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
121
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
122
123
    return lib

wxchan's avatar
wxchan committed
124

wxchan's avatar
wxchan committed
125
126
_LIB = _load_lib()

wxchan's avatar
wxchan committed
127

128
NUMERIC_TYPES = (int, float, bool)
129
_ArrayLike = Union[List, np.ndarray, pd_Series]
130
131


132
def _safe_call(ret: int) -> None:
133
134
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
135
136
137
    Parameters
    ----------
    ret : int
138
        The return value from C API calls.
wxchan's avatar
wxchan committed
139
140
    """
    if ret != 0:
141
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
142

wxchan's avatar
wxchan committed
143

wxchan's avatar
wxchan committed
144
def is_numeric(obj):
145
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
146
147
148
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
149
150
151
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
152
153
        return False

wxchan's avatar
wxchan committed
154

wxchan's avatar
wxchan committed
155
def is_numpy_1d_array(data):
156
    """Check whether data is a numpy 1-D array."""
157
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
158

wxchan's avatar
wxchan committed
159

160
161
162
163
164
165
166
167
def is_numpy_column_array(data):
    """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


168
169
def cast_numpy_array_to_dtype(array, dtype):
    """Cast numpy array to given dtype."""
170
171
172
173
174
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


wxchan's avatar
wxchan committed
175
def is_1d_list(data):
176
177
    """Check whether data is a 1-D list."""
    return isinstance(data, list) and (not data or is_numeric(data[0]))
wxchan's avatar
wxchan committed
178

wxchan's avatar
wxchan committed
179

180
181
182
183
184
185
186
187
188
189
def _is_1d_collection(data: Any) -> bool:
    """Check whether data is a 1-D collection."""
    return (
        is_numpy_1d_array(data)
        or is_numpy_column_array(data)
        or is_1d_list(data)
        or isinstance(data, pd_Series)
    )


190
def list_to_1d_numpy(data, dtype=np.float32, name='list'):
191
    """Convert data to numpy 1-D array."""
wxchan's avatar
wxchan committed
192
    if is_numpy_1d_array(data):
193
        return cast_numpy_array_to_dtype(data, dtype)
194
195
196
    elif is_numpy_column_array(data):
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
197
        return cast_numpy_array_to_dtype(array, dtype)
wxchan's avatar
wxchan committed
198
199
    elif is_1d_list(data):
        return np.array(data, dtype=dtype, copy=False)
200
    elif isinstance(data, pd_Series):
201
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
202
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
203
    else:
204
205
        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
206

wxchan's avatar
wxchan committed
207

208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
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."""
    return isinstance(data, list) and len(data) > 0 and is_1d_list(data[0])


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):
        return cast_numpy_array_to_dtype(data, dtype)
    if _is_2d_list(data):
        return np.array(data, dtype=dtype)
    if isinstance(data, pd_DataFrame):
234
        _check_for_bad_pandas_dtypes(data.dtypes)
235
236
237
238
239
        return cast_numpy_array_to_dtype(data.values, dtype)
    raise TypeError(f"Wrong type({type(data).__name__}) for {name}.\n"
                    "It should be list of lists, numpy 2-D array or pandas DataFrame")


wxchan's avatar
wxchan committed
240
def cfloat32_array_to_numpy(cptr, length):
241
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
242
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
243
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
244
    else:
245
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
246

Guolin Ke's avatar
Guolin Ke committed
247

Guolin Ke's avatar
Guolin Ke committed
248
def cfloat64_array_to_numpy(cptr, length):
249
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
250
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
251
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
252
253
254
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
255

wxchan's avatar
wxchan committed
256
def cint32_array_to_numpy(cptr, length):
257
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
258
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
259
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
260
    else:
261
262
263
264
265
266
        raise RuntimeError('Expected int32 pointer')


def cint64_array_to_numpy(cptr, length):
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
267
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
268
269
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
270

wxchan's avatar
wxchan committed
271

wxchan's avatar
wxchan committed
272
def c_str(string):
273
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
274
275
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
276

wxchan's avatar
wxchan committed
277
def c_array(ctype, values):
278
    """Convert a Python array to C array."""
wxchan's avatar
wxchan committed
279
280
    return (ctype * len(values))(*values)

wxchan's avatar
wxchan committed
281

282
283
284
285
286
287
288
289
290
291
def json_default_with_numpy(obj):
    """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


wxchan's avatar
wxchan committed
292
def param_dict_to_str(data):
293
    """Convert Python dictionary to string, which is passed to C API."""
294
    if data is None or not data:
wxchan's avatar
wxchan committed
295
296
297
        return ""
    pairs = []
    for key, val in data.items():
298
        if isinstance(val, (list, tuple, set)) or is_numpy_1d_array(val):
299
300
            def to_string(x):
                if isinstance(x, list):
301
                    return f"[{','.join(map(str, x))}]"
302
303
                else:
                    return str(x)
304
            pairs.append(f"{key}={','.join(map(to_string, val))}")
305
        elif isinstance(val, (str, Path, NUMERIC_TYPES)) or is_numeric(val):
306
            pairs.append(f"{key}={val}")
307
        elif val is not None:
308
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
309
    return ' '.join(pairs)
310

wxchan's avatar
wxchan committed
311

312
class _TempFile:
313
314
    """Proxy class to workaround errors on Windows."""

315
316
317
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
318
            self.path = Path(self.name)
319
        return self
wxchan's avatar
wxchan committed
320

321
    def __exit__(self, exc_type, exc_val, exc_tb):
322
323
        if self.path.is_file():
            self.path.unlink()
324

wxchan's avatar
wxchan committed
325

326
class LightGBMError(Exception):
327
328
    """Error thrown by LightGBM."""

329
330
331
    pass


332
333
334
335
336
337
338
339
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
    # lazy evaluation to allow import without dynamic library, e.g., for docs generation
    aliases = None

    @staticmethod
    def _get_all_param_aliases() -> Dict[str, Set[str]]:
        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'),
            object_hook=lambda obj: {k: set(v) | {k} for k, v in obj.items()}
        )
        return aliases
367
368

    @classmethod
369
370
371
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
372
373
        ret = set()
        for i in args:
374
            ret |= cls.aliases.get(i, {i})
375
376
        return ret

377
    @classmethod
378
379
380
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
381
382
383
384
385
386
387
388
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
                    ret |= aliases
                    break
        return ret

389

390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
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)

411
412
413
    aliases = _ConfigAliases.get(main_param_name) - {main_param_name}

    # if main_param_name was provided, keep that value and remove all aliases
414
    if main_param_name in params.keys():
415
416
417
        for param in aliases:
            params.pop(param, None)
        return params
418

419
420
421
422
423
    # 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
424

425
426
427
428
429
430
431
    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
432
433
434
435

    return params


436
437
MAX_INT32 = (1 << 31) - 1

438
"""Macro definition of data type in C API of LightGBM"""
wxchan's avatar
wxchan committed
439
440
441
442
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
443

444
"""Matrix is row major in Python"""
wxchan's avatar
wxchan committed
445
446
C_API_IS_ROW_MAJOR = 1

447
"""Macro definition of prediction type in C API of LightGBM"""
wxchan's avatar
wxchan committed
448
449
450
C_API_PREDICT_NORMAL = 0
C_API_PREDICT_RAW_SCORE = 1
C_API_PREDICT_LEAF_INDEX = 2
451
C_API_PREDICT_CONTRIB = 3
wxchan's avatar
wxchan committed
452

453
454
455
456
"""Macro definition of sparse matrix type"""
C_API_MATRIX_TYPE_CSR = 0
C_API_MATRIX_TYPE_CSC = 1

457
458
459
460
"""Macro definition of feature importance type"""
C_API_FEATURE_IMPORTANCE_SPLIT = 0
C_API_FEATURE_IMPORTANCE_GAIN = 1

461
"""Data type of data field"""
wxchan's avatar
wxchan committed
462
463
FIELD_TYPE_MAPPER = {"label": C_API_DTYPE_FLOAT32,
                     "weight": C_API_DTYPE_FLOAT32,
Guolin Ke's avatar
Guolin Ke committed
464
                     "init_score": C_API_DTYPE_FLOAT64,
465
                     "group": C_API_DTYPE_INT32}
wxchan's avatar
wxchan committed
466

467
468
469
470
"""String name to int feature importance type mapper"""
FEATURE_IMPORTANCE_TYPE_MAPPER = {"split": C_API_FEATURE_IMPORTANCE_SPLIT,
                                  "gain": C_API_FEATURE_IMPORTANCE_GAIN}

wxchan's avatar
wxchan committed
471

472
def convert_from_sliced_object(data):
473
    """Fix the memory of multi-dimensional sliced object."""
474
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
475
        if not data.flags.c_contiguous:
476
477
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
478
479
480
481
            return np.copy(data)
    return data


wxchan's avatar
wxchan committed
482
def c_float_array(data):
483
    """Get pointer of float numpy array / list."""
wxchan's avatar
wxchan committed
484
485
486
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
487
488
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
489
490
491
492
493
494
495
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
            type_data = C_API_DTYPE_FLOAT32
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
            type_data = C_API_DTYPE_FLOAT64
        else:
496
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
497
    else:
498
        raise TypeError(f"Unknown type({type(data).__name__})")
499
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
500

wxchan's avatar
wxchan committed
501

wxchan's avatar
wxchan committed
502
def c_int_array(data):
503
    """Get pointer of int numpy array / list."""
wxchan's avatar
wxchan committed
504
505
506
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
507
508
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
509
510
511
512
513
514
515
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
            type_data = C_API_DTYPE_INT32
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
            type_data = C_API_DTYPE_INT64
        else:
516
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
517
    else:
518
        raise TypeError(f"Unknown type({type(data).__name__})")
519
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
520

wxchan's avatar
wxchan committed
521

522
def _check_for_bad_pandas_dtypes(pandas_dtypes_series):
523
524
525
526
527
528
529
530
    float128 = getattr(np, 'float128', type(None))

    def is_allowed_numpy_dtype(dtype):
        return (
            issubclass(dtype, (np.integer, np.floating, np.bool_))
            and not issubclass(dtype, (np.timedelta64, float128))
        )

531
532
533
534
535
536
537
538
    bad_pandas_dtypes = [
        f'{column_name}: {pandas_dtype}'
        for column_name, pandas_dtype in pandas_dtypes_series.iteritems()
        if not is_allowed_numpy_dtype(pandas_dtype.type)
    ]
    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)}')
539
540


541
def _data_from_pandas(data, feature_name, categorical_feature, pandas_categorical):
542
    if isinstance(data, pd_DataFrame):
543
544
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
545
        if feature_name == 'auto' or feature_name is None:
546
            data = data.rename(columns=str, copy=False)
547
        cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
548
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
549
550
551
552
553
        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.')
554
            for col, category in zip(cat_cols, pandas_categorical):
555
556
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
557
        if len(cat_cols):  # cat_cols is list
558
            data = data.copy(deep=False)  # not alter origin DataFrame
559
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
560
561
562
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
563
            if categorical_feature == 'auto':  # use cat cols from DataFrame
564
                categorical_feature = cat_cols_not_ordered
565
566
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
567
568
        if feature_name == 'auto':
            feature_name = list(data.columns)
569
        _check_for_bad_pandas_dtypes(data.dtypes)
570
571
572
573
        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, [])
        data = data.astype(target_dtype, copy=False).values
574
575
576
577
578
579
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
580
581
582


def _label_from_pandas(label):
583
    if isinstance(label, pd_DataFrame):
584
585
        if len(label.columns) > 1:
            raise ValueError('DataFrame for label cannot have multiple columns')
586
        _check_for_bad_pandas_dtypes(label.dtypes)
587
        label = np.ravel(label.values.astype(np.float32, copy=False))
588
589
590
    return label


591
def _dump_pandas_categorical(pandas_categorical, file_name=None):
592
593
    categorical_json = json.dumps(pandas_categorical, default=json_default_with_numpy)
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
594
595
596
597
598
599
600
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


def _load_pandas_categorical(file_name=None, model_str=None):
601
602
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
603
    if file_name is not None:
604
        max_offset = -getsize(file_name)
605
606
607
608
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
609
                f.seek(offset, SEEK_END)
610
611
612
613
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
614
        last_line = lines[-1].decode('utf-8').strip()
615
        if not last_line.startswith(pandas_key):
616
            last_line = lines[-2].decode('utf-8').strip()
617
    elif model_str is not None:
618
619
620
621
622
623
        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
624
625


626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
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**.

646
647
    .. versionadded:: 3.3.0

648
649
650
651
652
653
654
655
656
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
657
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
658
659
660
661
662
663
664
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
665
                return self._get_one_line(idx)
666
            elif isinstance(idx, slice):
667
668
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
669
                # Only required if using ``Dataset.subset()``.
670
                return np.array([self._get_one_line(i) for i in idx])
671
            else:
672
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
673
674
675

        Parameters
        ----------
676
        idx : int, slice[int], list[int]
677
678
679
680
            Item index.

        Returns
        -------
681
        result : numpy 1-D array or numpy 2-D array
682
            1-D array if idx is int, 2-D array if idx is slice or list.
683
684
685
686
687
688
689
690
691
        """
        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__()")


692
class _InnerPredictor:
693
694
695
696
697
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
698
699
700
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
701
    """
702

703
    def __init__(self, model_file=None, booster_handle=None, pred_parameter=None):
704
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
705
706
707

        Parameters
        ----------
708
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
709
            Path to the model file.
710
711
712
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
713
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
714
715
716
717
718
        """
        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
719
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
720
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
721
                c_str(str(model_file)),
wxchan's avatar
wxchan committed
722
723
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
724
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
725
726
727
728
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
729
            self.num_total_iteration = out_num_iterations.value
730
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
731
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
732
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
733
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
734
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
735
736
737
738
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
739
            self.num_total_iteration = self.current_iteration()
740
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
741
        else:
742
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
743

744
745
        pred_parameter = {} if pred_parameter is None else pred_parameter
        self.pred_parameter = param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
746

wxchan's avatar
wxchan committed
747
    def __del__(self):
748
749
750
751
752
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
753

754
755
756
757
758
    def __getstate__(self):
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

759
    def predict(self, data, start_iteration=0, num_iteration=-1,
760
                raw_score=False, pred_leaf=False, pred_contrib=False, data_has_header=False, validate_features=False):
761
        """Predict logic.
wxchan's avatar
wxchan committed
762
763
764

        Parameters
        ----------
765
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
766
            Data source for prediction.
767
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
768
769
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
770
771
772
773
774
775
776
777
778
779
780
        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.
781
782
783
        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
784
785
786

        Returns
        -------
787
        result : numpy array, scipy.sparse or list of scipy.sparse
788
            Prediction result.
789
            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
790
        """
wxchan's avatar
wxchan committed
791
        if isinstance(data, Dataset):
792
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
793
794
795
796
797
798
799
800
801
802
803
        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)),
                )
            )
804
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
wxchan's avatar
wxchan committed
805
806
807
808
809
        predict_type = C_API_PREDICT_NORMAL
        if raw_score:
            predict_type = C_API_PREDICT_RAW_SCORE
        if pred_leaf:
            predict_type = C_API_PREDICT_LEAF_INDEX
810
811
        if pred_contrib:
            predict_type = C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
812
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
813

814
        if isinstance(data, (str, Path)):
815
            with _TempFile() as f:
wxchan's avatar
wxchan committed
816
817
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
818
                    c_str(str(data)),
Guolin Ke's avatar
Guolin Ke committed
819
820
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
821
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
822
                    ctypes.c_int(num_iteration),
823
                    c_str(self.pred_parameter),
wxchan's avatar
wxchan committed
824
                    c_str(f.name)))
825
826
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
827
        elif isinstance(data, scipy.sparse.csr_matrix):
828
            preds, nrow = self.__pred_for_csr(data, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
829
        elif isinstance(data, scipy.sparse.csc_matrix):
830
            preds, nrow = self.__pred_for_csc(data, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
831
        elif isinstance(data, np.ndarray):
832
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
833
834
835
        elif isinstance(data, list):
            try:
                data = np.array(data)
836
            except BaseException:
837
                raise ValueError('Cannot convert data list to numpy array.')
838
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
839
        elif isinstance(data, dt_DataTable):
840
            preds, nrow = self.__pred_for_np2d(data.to_numpy(), start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
841
842
        else:
            try:
843
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
844
                csr = scipy.sparse.csr_matrix(data)
845
            except BaseException:
846
                raise TypeError(f'Cannot predict data for type {type(data).__name__}')
847
            preds, nrow = self.__pred_for_csr(csr, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
848
849
        if pred_leaf:
            preds = preds.astype(np.int32)
850
        is_sparse = scipy.sparse.issparse(preds) or isinstance(preds, list)
851
        if not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
852
            if preds.size % nrow == 0:
853
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
854
            else:
855
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
856
857
        return preds

858
    def __get_num_preds(self, start_iteration, num_iteration, nrow, predict_type):
859
        """Get size of prediction result."""
860
        if nrow > MAX_INT32:
861
            raise LightGBMError('LightGBM cannot perform prediction for data '
862
                                f'with number of rows greater than MAX_INT32 ({MAX_INT32}).\n'
863
                                'You can split your data into chunks '
864
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
865
866
867
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
868
869
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
870
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
871
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
872
873
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
874

875
    def __pred_for_np2d(self, mat, start_iteration, num_iteration, predict_type):
876
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
877
        if len(mat.shape) != 2:
878
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
879

880
        def inner_predict(mat, start_iteration, num_iteration, predict_type, preds=None):
881
882
            if mat.dtype == np.float32 or mat.dtype == np.float64:
                data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
883
            else:  # change non-float data to float data, need to copy
884
885
                data = np.array(mat.reshape(mat.size), dtype=np.float32)
            ptr_data, type_ptr_data, _ = c_float_array(data)
886
            n_preds = self.__get_num_preds(start_iteration, num_iteration, mat.shape[0], predict_type)
887
            if preds is None:
888
                preds = np.empty(n_preds, dtype=np.float64)
889
890
891
892
893
894
895
            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),
896
897
                ctypes.c_int32(mat.shape[0]),
                ctypes.c_int32(mat.shape[1]),
898
899
                ctypes.c_int(C_API_IS_ROW_MAJOR),
                ctypes.c_int(predict_type),
900
                ctypes.c_int(start_iteration),
901
902
903
904
905
906
907
908
909
910
911
912
                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]

        nrow = mat.shape[0]
        if nrow > MAX_INT32:
            sections = np.arange(start=MAX_INT32, stop=nrow, step=MAX_INT32)
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
913
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
914
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
915
            preds = np.empty(sum(n_preds), dtype=np.float64)
916
917
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
918
                # avoid memory consumption by arrays concatenation operations
919
                inner_predict(chunk, start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
920
            return preds, nrow
wxchan's avatar
wxchan committed
921
        else:
922
            return inner_predict(mat, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
923

924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
    def __create_sparse_native(self, cs, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                               indptr_type, data_type, is_csr=True):
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
        if indptr_type == C_API_DTYPE_INT32:
            out_indptr = cint32_array_to_numpy(out_ptr_indptr, indptr_len)
        elif indptr_type == C_API_DTYPE_INT64:
            out_indptr = cint64_array_to_numpy(out_ptr_indptr, indptr_len)
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
        if data_type == C_API_DTYPE_FLOAT32:
            out_data = cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
        elif data_type == C_API_DTYPE_FLOAT64:
            out_data = cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
        else:
            raise TypeError("Expected float32 or float64 type for data")
        out_indices = cint32_array_to_numpy(out_ptr_indices, data_indices_len)
        # 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

970
    def __pred_for_csr(self, csr, start_iteration, num_iteration, predict_type):
971
        """Predict for a CSR data."""
972
        def inner_predict(csr, start_iteration, num_iteration, predict_type, preds=None):
973
            nrow = len(csr.indptr) - 1
974
            n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
975
            if preds is None:
976
                preds = np.empty(n_preds, dtype=np.float64)
977
978
979
980
981
982
983
            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)

            ptr_indptr, type_ptr_indptr, __ = c_int_array(csr.indptr)
            ptr_data, type_ptr_data, _ = c_float_array(csr.data)

984
            assert csr.shape[1] <= MAX_INT32
985
            csr_indices = csr.indices.astype(np.int32, copy=False)
986

987
988
989
            _safe_call(_LIB.LGBM_BoosterPredictForCSR(
                self.handle,
                ptr_indptr,
990
                ctypes.c_int(type_ptr_indptr),
991
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
992
993
994
995
996
997
                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),
998
                ctypes.c_int(start_iteration),
999
1000
1001
1002
1003
1004
1005
                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
wxchan's avatar
wxchan committed
1006

1007
        def inner_predict_sparse(csr, start_iteration, num_iteration, predict_type):
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
            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)()
1021
            out_shape = np.empty(2, dtype=np.int64)
1022
1023
1024
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
1025
                ctypes.c_int(type_ptr_indptr),
1026
1027
1028
1029
1030
1031
1032
                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),
1033
                ctypes.c_int(start_iteration),
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
                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(csr, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                   type_ptr_indptr, type_ptr_data, is_csr=True)
            nrow = len(csr.indptr) - 1
            return matrices, nrow

        if predict_type == C_API_PREDICT_CONTRIB:
1047
            return inner_predict_sparse(csr, start_iteration, num_iteration, predict_type)
1048
1049
1050
1051
        nrow = len(csr.indptr) - 1
        if nrow > MAX_INT32:
            sections = [0] + list(np.arange(start=MAX_INT32, stop=nrow, step=MAX_INT32)) + [nrow]
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
1052
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1053
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1054
            preds = np.empty(sum(n_preds), dtype=np.float64)
1055
1056
            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:])):
1057
                # avoid memory consumption by arrays concatenation operations
1058
                inner_predict(csr[start_idx:end_idx], start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
1059
1060
            return preds, nrow
        else:
1061
            return inner_predict(csr, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
1062

1063
    def __pred_for_csc(self, csc, start_iteration, num_iteration, predict_type):
1064
        """Predict for a CSC data."""
1065
        def inner_predict_sparse(csc, start_iteration, num_iteration, predict_type):
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
            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)()
1079
            out_shape = np.empty(2, dtype=np.int64)
1080
1081
1082
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
1083
                ctypes.c_int(type_ptr_indptr),
1084
1085
1086
1087
1088
1089
1090
                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),
1091
                ctypes.c_int(start_iteration),
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
                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(csc, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                   type_ptr_indptr, type_ptr_data, is_csr=False)
            nrow = csc.shape[0]
            return matrices, nrow

Guolin Ke's avatar
Guolin Ke committed
1104
        nrow = csc.shape[0]
1105
        if nrow > MAX_INT32:
1106
            return self.__pred_for_csr(csc.tocsr(), start_iteration, num_iteration, predict_type)
1107
        if predict_type == C_API_PREDICT_CONTRIB:
1108
1109
            return inner_predict_sparse(csc, start_iteration, num_iteration, predict_type)
        n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
1110
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1111
1112
        out_num_preds = ctypes.c_int64(0)

1113
1114
        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
1115

1116
        assert csc.shape[0] <= MAX_INT32
1117
        csc_indices = csc.indices.astype(np.int32, copy=False)
1118

Guolin Ke's avatar
Guolin Ke committed
1119
1120
1121
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
1122
            ctypes.c_int(type_ptr_indptr),
1123
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1124
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1125
1126
1127
1128
1129
            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),
1130
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1131
            ctypes.c_int(num_iteration),
1132
            c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1133
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1134
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1135
        if n_preds != out_num_preds.value:
1136
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1137
1138
        return preds, nrow

1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
    def current_iteration(self):
        """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
1153

1154
class Dataset:
wxchan's avatar
wxchan committed
1155
    """Dataset in LightGBM."""
1156

1157
    def __init__(self, data, label=None, reference=None,
1158
                 weight=None, group=None, init_score=None,
1159
                 feature_name='auto', categorical_feature='auto', params=None,
wxchan's avatar
wxchan committed
1160
                 free_raw_data=True):
1161
        """Initialize Dataset.
1162

wxchan's avatar
wxchan committed
1163
1164
        Parameters
        ----------
1165
        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
1166
            Data source of Dataset.
1167
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1168
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1169
1170
1171
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1172
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1173
            Weight for each instance. Weights should be non-negative.
1174
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1175
1176
1177
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1178
1179
            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.
1180
        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)
1181
            Init score for Dataset.
1182
        feature_name : list of str, or 'auto', optional (default="auto")
1183
1184
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
1185
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1186
1187
            Categorical features.
            If list of int, interpreted as indices.
1188
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1189
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1190
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1191
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1192
            All negative values in categorical features will be treated as missing values.
1193
            The output cannot be monotonically constrained with respect to a categorical feature.
1194
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1195
        params : dict or None, optional (default=None)
1196
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1197
        free_raw_data : bool, optional (default=True)
1198
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
1199
        """
wxchan's avatar
wxchan committed
1200
1201
1202
1203
1204
1205
        self.handle = None
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1206
        self.init_score = init_score
wxchan's avatar
wxchan committed
1207
        self.feature_name = feature_name
1208
        self.categorical_feature = categorical_feature
1209
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1210
1211
        self.free_raw_data = free_raw_data
        self.used_indices = None
1212
        self.need_slice = True
wxchan's avatar
wxchan committed
1213
        self._predictor = None
1214
        self.pandas_categorical = None
1215
        self.params_back_up = None
1216
1217
        self.feature_penalty = None
        self.monotone_constraints = None
1218
        self.version = 0
1219
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1220
1221

    def __del__(self):
1222
1223
1224
1225
        try:
            self._free_handle()
        except AttributeError:
            pass
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
    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.
        """
        param_str = param_dict_to_str(self.get_params())
        sample_cnt = _get_sample_count(total_nrow, param_str)
        indices = np.empty(sample_cnt, dtype=np.int32)
        ptr_data, _, _ = c_int_array(indices)
        actual_sample_cnt = ctypes.c_int32(0)

        _safe_call(_LIB.LGBM_SampleIndices(
            ctypes.c_int32(total_nrow),
            c_str(param_str),
            ptr_data,
            ctypes.byref(actual_sample_cnt),
        ))
1256
1257
        assert sample_cnt == actual_sample_cnt.value
        return indices
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292

    def _init_from_ref_dataset(self, total_nrow: int, ref_dataset: 'Dataset') -> 'Dataset':
        """Create dataset from a reference dataset.

        Parameters
        ----------
        total_nrow : int
            Number of rows expected to add to dataset.
        ref_dataset : Dataset
            Reference dataset to extract meta from.

        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
        ----------
1293
        sample_data : list of numpy array
1294
            Sample data for each column.
1295
        sample_indices : list of numpy array
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
            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):
            sample_col_ptr[i] = c_float_array(sample_data[i])[0]
            indices_col_ptr[i] = c_int_array(sample_indices[i])[0]

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

        self.handle = ctypes.c_void_p()
        params_str = param_dict_to_str(self.get_params())
        _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),
            c_str(params_str),
            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)
        data_ptr, data_type, _ = c_float_array(data)

        _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

1371
    def get_params(self) -> Dict[str, Any]:
1372
1373
1374
1375
        """Get the used parameters in the Dataset.

        Returns
        -------
1376
        params : dict
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
            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",
1392
                                                "linear_tree",
1393
1394
1395
1396
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1397
                                                "precise_float_parser",
1398
1399
1400
1401
1402
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}
1403
1404
        else:
            return {}
1405

1406
    def _free_handle(self):
1407
        if self.handle is not None:
1408
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1409
            self.handle = None
Guolin Ke's avatar
Guolin Ke committed
1410
1411
1412
        self.need_slice = True
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1413
        return self
wxchan's avatar
wxchan committed
1414

Guolin Ke's avatar
Guolin Ke committed
1415
1416
    def _set_init_score_by_predictor(self, predictor, data, used_indices=None):
        data_has_header = False
1417
        if isinstance(data, (str, Path)):
Guolin Ke's avatar
Guolin Ke committed
1418
            # check data has header or not
1419
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1420
        num_data = self.num_data()
1421
1422
1423
        if predictor is not None:
            init_score = predictor.predict(data,
                                           raw_score=True,
1424
1425
                                           data_has_header=data_has_header)
            init_score = init_score.ravel()
1426
1427
            if used_indices is not None:
                assert not self.need_slice
1428
                if isinstance(data, (str, Path)):
1429
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1430
                    assert num_data == len(used_indices)
1431
1432
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1433
1434
1435
1436
                            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
1437
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1438
1439
                for i in range(num_data):
                    for j in range(predictor.num_class):
1440
1441
1442
                        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:
1443
            init_score = np.zeros(self.init_score.shape, dtype=np.float64)
1444
1445
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1446
1447
        self.set_init_score(init_score)

1448
    def _lazy_init(self, data, label=None, reference=None,
1449
                   weight=None, group=None, init_score=None, predictor=None,
1450
                   feature_name='auto', categorical_feature='auto', params=None):
wxchan's avatar
wxchan committed
1451
1452
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1453
            return self
Guolin Ke's avatar
Guolin Ke committed
1454
1455
1456
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1457
1458
1459
1460
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
wxchan's avatar
wxchan committed
1461
        label = _label_from_pandas(label)
Guolin Ke's avatar
Guolin Ke committed
1462

1463
        # process for args
wxchan's avatar
wxchan committed
1464
        params = {} if params is None else params
1465
1466
1467
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1468
        for key in params.keys():
1469
            if key in args_names:
1470
1471
                _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.')
1472
        # get categorical features
1473
1474
1475
1476
1477
1478
        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:
1479
                if isinstance(name, str) and name in feature_dict:
1480
                    categorical_indices.add(feature_dict[name])
1481
                elif isinstance(name, int):
1482
1483
                    categorical_indices.add(name)
                else:
1484
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1485
            if categorical_indices:
1486
1487
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1488
1489
1490
                        # If the params[cat_alias] is equal to categorical_indices, do not report the warning.
                        if not(isinstance(params[cat_alias], list) and set(params[cat_alias]) == categorical_indices):
                            _log_warning(f'{cat_alias} in param dict is overridden.')
1491
                        params.pop(cat_alias, None)
1492
                params['categorical_column'] = sorted(categorical_indices)
1493

wxchan's avatar
wxchan committed
1494
        params_str = param_dict_to_str(params)
1495
        self.params = params
1496
        # process for reference dataset
wxchan's avatar
wxchan committed
1497
        ref_dataset = None
wxchan's avatar
wxchan committed
1498
        if isinstance(reference, Dataset):
1499
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1500
1501
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1502
        # start construct data
1503
        if isinstance(data, (str, Path)):
wxchan's avatar
wxchan committed
1504
1505
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1506
                c_str(str(data)),
wxchan's avatar
wxchan committed
1507
1508
1509
1510
1511
                c_str(params_str),
                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
1512
1513
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1514
1515
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1516
1517
1518
1519
1520
1521
1522
1523
1524
        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)
1525
        elif isinstance(data, dt_DataTable):
1526
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1527
1528
1529
1530
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1531
            except BaseException:
1532
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}')
wxchan's avatar
wxchan committed
1533
1534
1535
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1536
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1537
1538
1539
1540
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1541
1542
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1543
                _log_warning("The init_score will be overridden by the prediction of init_model.")
Guolin Ke's avatar
Guolin Ke committed
1544
            self._set_init_score_by_predictor(predictor, data)
1545
1546
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1547
        elif predictor is not None:
1548
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1549
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1550
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1551

1552
1553
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
        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.
1579
        sampled = np.array([row for row in self._yield_row_from_seqlist(seqs, indices)])
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
        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

    def __init_from_seqs(self, seqs: List[Sequence], ref_dataset: Optional['Dataset'] = None):
        """
        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:
            param_str = param_dict_to_str(self.get_params())
            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

wxchan's avatar
wxchan committed
1624
    def __init_from_np2d(self, mat, params_str, ref_dataset):
1625
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1626
1627
1628
1629
1630
1631
        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)
1632
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1633
1634
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1635
        ptr_data, type_ptr_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1636
1637
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1638
            ctypes.c_int(type_ptr_data),
1639
1640
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
Guolin Ke's avatar
Guolin Ke committed
1641
            ctypes.c_int(C_API_IS_ROW_MAJOR),
wxchan's avatar
wxchan committed
1642
1643
1644
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1645
        return self
wxchan's avatar
wxchan committed
1646

1647
    def __init_from_list_np2d(self, mats, params_str, ref_dataset):
1648
        """Initialize data from a list of 2-D numpy matrices."""
1649
        ncol = mats[0].shape[1]
1650
        nrow = np.empty((len(mats),), np.int32)
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
        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)
1670
            else:  # change non-float data to float data, need to copy
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

            chunk_ptr_data, chunk_type_ptr_data, holder = c_float_array(mats[i])
            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(
1682
            ctypes.c_int32(len(mats)),
1683
1684
1685
            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)),
1686
            ctypes.c_int32(ncol),
1687
1688
1689
1690
            ctypes.c_int(C_API_IS_ROW_MAJOR),
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1691
        return self
1692

wxchan's avatar
wxchan committed
1693
    def __init_from_csr(self, csr, params_str, ref_dataset):
1694
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
1695
        if len(csr.indices) != len(csr.data):
1696
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
wxchan's avatar
wxchan committed
1697
1698
        self.handle = ctypes.c_void_p()

1699
1700
        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
1701

1702
        assert csr.shape[1] <= MAX_INT32
1703
        csr_indices = csr.indices.astype(np.int32, copy=False)
1704

wxchan's avatar
wxchan committed
1705
1706
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1707
            ctypes.c_int(type_ptr_indptr),
1708
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
1709
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1710
1711
1712
1713
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
wxchan's avatar
wxchan committed
1714
1715
1716
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1717
        return self
wxchan's avatar
wxchan committed
1718

Guolin Ke's avatar
Guolin Ke committed
1719
    def __init_from_csc(self, csc, params_str, ref_dataset):
1720
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
1721
        if len(csc.indices) != len(csc.data):
1722
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
Guolin Ke's avatar
Guolin Ke committed
1723
1724
        self.handle = ctypes.c_void_p()

1725
1726
        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
1727

1728
        assert csc.shape[0] <= MAX_INT32
1729
        csc_indices = csc.indices.astype(np.int32, copy=False)
1730

Guolin Ke's avatar
Guolin Ke committed
1731
1732
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1733
            ctypes.c_int(type_ptr_indptr),
1734
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1735
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1736
1737
1738
1739
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
Guolin Ke's avatar
Guolin Ke committed
1740
1741
1742
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1743
        return self
Guolin Ke's avatar
Guolin Ke committed
1744

1745
    @staticmethod
1746
1747
1748
1749
1750
1751
    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.
1752

1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
        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.
1763
1764
1765

        Returns
        -------
1766
1767
        compare_result : bool
          Returns whether two dictionaries with params are equal.
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
        """
        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

wxchan's avatar
wxchan committed
1783
    def construct(self):
1784
1785
1786
1787
1788
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
1789
            Constructed Dataset object.
1790
        """
1791
        if self.handle is None:
wxchan's avatar
wxchan committed
1792
            if self.reference is not None:
1793
                reference_params = self.reference.get_params()
1794
1795
                params = self.get_params()
                if params != reference_params:
1796
1797
1798
1799
1800
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
1801
                        _log_warning('Overriding the parameters from Reference Dataset.')
1802
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
1803
                if self.used_indices is None:
1804
                    # create valid
1805
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
1806
1807
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
1808
                                    feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
1809
                else:
1810
                    # construct subset
wxchan's avatar
wxchan committed
1811
                    used_indices = list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
1812
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
1813
                    if self.reference.group is not None:
1814
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
1815
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
1816
                                                  return_counts=True)
1817
                    self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1818
1819
                    params_str = param_dict_to_str(self.params)
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
1820
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
1821
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
1822
                        ctypes.c_int32(used_indices.shape[0]),
wxchan's avatar
wxchan committed
1823
1824
                        c_str(params_str),
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
1825
1826
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
1827
1828
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
1829
1830
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
1831
1832
1833
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
                        self._set_init_score_by_predictor(self._predictor, self.data, used_indices)
wxchan's avatar
wxchan committed
1834
            else:
1835
                # create train
1836
                self._lazy_init(self.data, label=self.label,
1837
1838
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
1839
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
1840
1841
            if self.free_raw_data:
                self.data = None
1842
            self.feature_name = self.get_feature_name()
wxchan's avatar
wxchan committed
1843
        return self
wxchan's avatar
wxchan committed
1844

1845
    def create_valid(self, data, label=None, weight=None, group=None, init_score=None, params=None):
1846
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
1847
1848
1849

        Parameters
        ----------
1850
        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
1851
            Data source of Dataset.
1852
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1853
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1854
1855
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1856
            Weight for each instance. Weights should be non-negative.
1857
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1858
1859
1860
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1861
1862
            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.
1863
        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)
1864
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1865
        params : dict or None, optional (default=None)
1866
            Other parameters for validation Dataset.
1867
1868
1869

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1870
1871
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
1872
        """
1873
        ret = Dataset(data, label=label, reference=self,
1874
                      weight=weight, group=group, init_score=init_score,
1875
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1876
        ret._predictor = self._predictor
1877
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
1878
        return ret
wxchan's avatar
wxchan committed
1879

wxchan's avatar
wxchan committed
1880
    def subset(self, used_indices, params=None):
1881
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
1882
1883
1884
1885

        Parameters
        ----------
        used_indices : list of int
1886
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
1887
        params : dict or None, optional (default=None)
1888
            These parameters will be passed to Dataset constructor.
1889
1890
1891
1892
1893

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
1894
        """
wxchan's avatar
wxchan committed
1895
1896
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
1897
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
1898
1899
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1900
        ret._predictor = self._predictor
1901
        ret.pandas_categorical = self.pandas_categorical
1902
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
1903
1904
1905
        return ret

    def save_binary(self, filename):
1906
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
1907

1908
1909
1910
1911
1912
        .. 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
1913
1914
        Parameters
        ----------
1915
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
1916
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
1917
1918
1919
1920
1921

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
1922
1923
1924
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
1925
            c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
1926
        return self
wxchan's avatar
wxchan committed
1927
1928

    def _update_params(self, params):
1929
1930
        if not params:
            return self
1931
        params = deepcopy(params)
1932
1933
1934
1935
1936

        def update():
            if not self.params:
                self.params = params
            else:
1937
                self.params_back_up = deepcopy(self.params)
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
                self.params.update(params)

        if self.handle is None:
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
                c_str(param_dict_to_str(self.params)),
                c_str(param_dict_to_str(params)))
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
1952
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
1953
        return self
wxchan's avatar
wxchan committed
1954

1955
    def _reverse_update_params(self):
1956
        if self.handle is None:
1957
            self.params = deepcopy(self.params_back_up)
1958
            self.params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
1959
        return self
1960

wxchan's avatar
wxchan committed
1961
    def set_field(self, field_name, data):
wxchan's avatar
wxchan committed
1962
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
1963
1964
1965

        Parameters
        ----------
1966
        field_name : str
1967
            The field name of the information.
1968
1969
        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
1970
1971
1972
1973
1974

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
1975
        """
1976
        if self.handle is None:
1977
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
1978
        if data is None:
1979
            # set to None
wxchan's avatar
wxchan committed
1980
1981
1982
1983
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
                c_str(field_name),
                None,
Guolin Ke's avatar
Guolin Ke committed
1984
1985
                ctypes.c_int(0),
                ctypes.c_int(FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
1986
            return self
1987
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
1988
            dtype = np.float64
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
            if _is_1d_collection(data):
                data = list_to_1d_numpy(data, dtype, name=field_name)
            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
            data = list_to_1d_numpy(data, dtype, name=field_name)

2003
2004
        if data.dtype == np.float32 or data.dtype == np.float64:
            ptr_data, type_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
2005
        elif data.dtype == np.int32:
2006
            ptr_data, type_data, _ = c_int_array(data)
wxchan's avatar
wxchan committed
2007
        else:
2008
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
wxchan's avatar
wxchan committed
2009
        if type_data != FIELD_TYPE_MAPPER[field_name]:
2010
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
2011
2012
2013
2014
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
            c_str(field_name),
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
2015
2016
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
2017
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
2018
        return self
wxchan's avatar
wxchan committed
2019

wxchan's avatar
wxchan committed
2020
2021
    def get_field(self, field_name):
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
2022
2023
2024

        Parameters
        ----------
2025
        field_name : str
2026
            The field name of the information.
wxchan's avatar
wxchan committed
2027
2028
2029

        Returns
        -------
2030
        info : numpy array or None
2031
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2032
        """
2033
        if self.handle is None:
2034
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2035
2036
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
            self.handle,
            c_str(field_name),
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
        if out_type.value != FIELD_TYPE_MAPPER[field_name]:
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
        if out_type.value == C_API_DTYPE_INT32:
2049
            arr = cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
wxchan's avatar
wxchan committed
2050
        elif out_type.value == C_API_DTYPE_FLOAT32:
2051
            arr = cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
Guolin Ke's avatar
Guolin Ke committed
2052
        elif out_type.value == C_API_DTYPE_FLOAT64:
2053
            arr = cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2054
        else:
wxchan's avatar
wxchan committed
2055
            raise TypeError("Unknown type")
2056
2057
2058
2059
2060
2061
        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
2062

2063
    def set_categorical_feature(self, categorical_feature):
2064
        """Set categorical features.
2065
2066
2067

        Parameters
        ----------
2068
        categorical_feature : list of int or str
2069
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2070
2071
2072
2073
2074

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2075
2076
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2077
            return self
2078
        if self.data is not None:
2079
2080
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2081
                return self._free_handle()
2082
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2083
                return self
2084
            else:
2085
2086
2087
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
                                 f'New categorical_feature is {sorted(list(categorical_feature))}')
2088
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2089
                return self._free_handle()
2090
        else:
2091
2092
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2093

Guolin Ke's avatar
Guolin Ke committed
2094
    def _set_predictor(self, predictor):
2095
2096
2097
2098
        """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
2099
        """
2100
        if predictor is self._predictor and (predictor is None or predictor.current_iteration() == self._predictor.current_iteration()):
Nikita Titov's avatar
Nikita Titov committed
2101
            return self
2102
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
2103
            self._predictor = predictor
2104
2105
2106
2107
2108
2109
        elif self.data is not None:
            self._predictor = predictor
            self._set_init_score_by_predictor(self._predictor, self.data)
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
            self._set_init_score_by_predictor(self._predictor, self.reference.data, self.used_indices)
Guolin Ke's avatar
Guolin Ke committed
2110
        else:
2111
2112
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2113
        return self
Guolin Ke's avatar
Guolin Ke committed
2114
2115

    def set_reference(self, reference):
2116
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2117
2118
2119
2120

        Parameters
        ----------
        reference : Dataset
2121
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2122
2123
2124
2125
2126

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2127
        """
2128
2129
2130
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2131
        # we're done if self and reference share a common upstream reference
2132
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2133
            return self
Guolin Ke's avatar
Guolin Ke committed
2134
2135
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2136
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2137
        else:
2138
2139
            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
2140
2141

    def set_feature_name(self, feature_name):
2142
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2143
2144
2145

        Parameters
        ----------
2146
        feature_name : list of str
2147
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2148
2149
2150
2151
2152

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2153
        """
2154
2155
        if feature_name != 'auto':
            self.feature_name = feature_name
2156
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2157
            if len(feature_name) != self.num_feature():
2158
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2159
            c_feature_name = [c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2160
2161
2162
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
                c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2163
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2164
        return self
Guolin Ke's avatar
Guolin Ke committed
2165
2166

    def set_label(self, label):
2167
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2168
2169
2170

        Parameters
        ----------
2171
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2172
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2173
2174
2175
2176
2177

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2178
2179
        """
        self.label = label
2180
        if self.handle is not None:
2181
            label = list_to_1d_numpy(_label_from_pandas(label), name='label')
wxchan's avatar
wxchan committed
2182
            self.set_field('label', label)
2183
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2184
        return self
Guolin Ke's avatar
Guolin Ke committed
2185
2186

    def set_weight(self, weight):
2187
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2188
2189
2190

        Parameters
        ----------
2191
        weight : list, numpy 1-D array, pandas Series or None
2192
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2193
2194
2195
2196
2197

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2198
        """
2199
2200
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2201
        self.weight = weight
2202
        if self.handle is not None and weight is not None:
wxchan's avatar
wxchan committed
2203
2204
            weight = list_to_1d_numpy(weight, name='weight')
            self.set_field('weight', weight)
2205
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2206
        return self
Guolin Ke's avatar
Guolin Ke committed
2207
2208

    def set_init_score(self, init_score):
2209
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2210
2211
2212

        Parameters
        ----------
2213
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2214
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2215
2216
2217
2218
2219

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2220
2221
        """
        self.init_score = init_score
2222
        if self.handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2223
            self.set_field('init_score', init_score)
2224
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2225
        return self
Guolin Ke's avatar
Guolin Ke committed
2226
2227

    def set_group(self, group):
2228
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2229
2230
2231

        Parameters
        ----------
2232
        group : list, numpy 1-D array, pandas Series or None
2233
2234
2235
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2236
2237
            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
2238
2239
2240
2241
2242

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2243
2244
        """
        self.group = group
2245
        if self.handle is not None and group is not None:
wxchan's avatar
wxchan committed
2246
2247
            group = list_to_1d_numpy(group, np.int32, name='group')
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2248
        return self
Guolin Ke's avatar
Guolin Ke committed
2249

2250
2251
2252
2253
2254
    def get_feature_name(self):
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2255
        feature_names : list of str
2256
2257
2258
2259
2260
2261
2262
2263
            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)
2264
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2265
2266
2267
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
2268
            ctypes.c_int(num_feature),
2269
            ctypes.byref(tmp_out_len),
2270
            ctypes.c_size_t(reserved_string_buffer_size),
2271
2272
2273
2274
            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")
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
        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))
2287
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2288

Guolin Ke's avatar
Guolin Ke committed
2289
    def get_label(self):
2290
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2291
2292
2293

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2294
        label : numpy array or None
2295
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2296
        """
2297
        if self.label is None:
wxchan's avatar
wxchan committed
2298
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2299
2300
2301
        return self.label

    def get_weight(self):
2302
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2303
2304
2305

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2306
        weight : numpy array or None
2307
            Weight for each data point from the Dataset. Weights should be non-negative.
Guolin Ke's avatar
Guolin Ke committed
2308
        """
2309
        if self.weight is None:
wxchan's avatar
wxchan committed
2310
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2311
2312
2313
        return self.weight

    def get_init_score(self):
2314
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2315
2316
2317

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2318
        init_score : numpy array or None
2319
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2320
        """
2321
        if self.init_score is None:
wxchan's avatar
wxchan committed
2322
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2323
2324
        return self.init_score

2325
2326
2327
2328
2329
    def get_data(self):
        """Get the raw data of the Dataset.

        Returns
        -------
2330
        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
2331
2332
2333
2334
            Raw data used in the Dataset construction.
        """
        if self.handle is None:
            raise Exception("Cannot get data before construct Dataset")
Guolin Ke's avatar
Guolin Ke committed
2335
2336
2337
2338
2339
        if self.need_slice and self.used_indices is not None and self.reference is not None:
            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, :]
2340
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2341
                    self.data = self.data.iloc[self.used_indices].copy()
2342
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2343
                    self.data = self.data[self.used_indices, :]
2344
2345
2346
2347
                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
2348
                else:
2349
2350
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
2351
            self.need_slice = False
Guolin Ke's avatar
Guolin Ke committed
2352
2353
2354
        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.")
2355
2356
        return self.data

Guolin Ke's avatar
Guolin Ke committed
2357
    def get_group(self):
2358
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2359
2360
2361

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2362
        group : numpy array or None
2363
2364
2365
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2366
2367
            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
2368
        """
2369
        if self.group is None:
wxchan's avatar
wxchan committed
2370
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
2371
2372
            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
2373
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
2374
2375
2376
        return self.group

    def num_data(self):
2377
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2378
2379
2380

        Returns
        -------
2381
2382
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2383
        """
2384
        if self.handle is not None:
2385
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2386
2387
2388
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2389
        else:
2390
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2391
2392

    def num_feature(self):
2393
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2394
2395
2396

        Returns
        -------
2397
2398
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2399
        """
2400
        if self.handle is not None:
2401
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2402
2403
2404
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2405
        else:
2406
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2407

2408
    def feature_num_bin(self, feature: Union[int, str]) -> int:
2409
2410
2411
2412
        """Get the number of bins for a feature.

        Parameters
        ----------
2413
2414
        feature : int or str
            Index or name of the feature.
2415
2416
2417
2418
2419
2420
2421

        Returns
        -------
        number_of_bins : int
            The number of constructed bins for the feature in the Dataset.
        """
        if self.handle is not None:
2422
2423
            if isinstance(feature, str):
                feature = self.feature_name.index(feature)
2424
2425
2426
2427
2428
2429
2430
2431
            ret = ctypes.c_int(0)
            _safe_call(_LIB.LGBM_DatasetGetFeatureNumBin(self.handle,
                                                         ctypes.c_int(feature),
                                                         ctypes.byref(ret)))
            return ret.value
        else:
            raise LightGBMError("Cannot get feature_num_bin before construct dataset")

2432
    def get_ref_chain(self, ref_limit=100):
2433
2434
2435
2436
2437
        """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.
2438
2439
2440
2441
2442

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2443
2444
2445

        Returns
        -------
2446
2447
2448
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2449
        head = self
2450
        ref_chain = set()
2451
2452
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2453
                ref_chain.add(head)
2454
2455
2456
2457
2458
2459
                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
2460
        return ref_chain
2461

2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
    def add_features_from(self, other):
        """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
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
        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()))
2490
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2491
                    self.data = np.hstack((self.data, other.data.values))
2492
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2493
2494
2495
2496
2497
2498
2499
                    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)
2500
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2501
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
2502
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2503
2504
2505
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
2506
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2507
2508
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
2509
2510
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
2511
                if isinstance(other.data, np.ndarray):
2512
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
2513
2514
                                       axis=1, ignore_index=True)
                elif scipy.sparse.issparse(other.data):
2515
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
2516
                                       axis=1, ignore_index=True)
2517
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2518
2519
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
2520
2521
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
2522
2523
2524
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
2525
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2526
                if isinstance(other.data, np.ndarray):
2527
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
Guolin Ke's avatar
Guolin Ke committed
2528
                elif scipy.sparse.issparse(other.data):
2529
2530
2531
2532
2533
                    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
2534
2535
2536
2537
2538
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
2539
2540
            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
2541
2542
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
2543
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
2544
        self.feature_name = self.get_feature_name()
2545
2546
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
2547
2548
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2549
2550
        return self

2551
    def _dump_text(self, filename):
2552
2553
2554
2555
2556
2557
        """Save Dataset to a text file.

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

        Parameters
        ----------
2558
        filename : str or pathlib.Path
2559
2560
2561
2562
2563
2564
2565
2566
2567
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
            self.construct().handle,
2568
            c_str(str(filename))))
2569
2570
        return self

wxchan's avatar
wxchan committed
2571

2572
class Booster:
2573
    """Booster in LightGBM."""
2574

2575
2576
2577
2578
2579
2580
2581
    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
    ):
2582
        """Initialize the Booster.
wxchan's avatar
wxchan committed
2583
2584
2585

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2586
        params : dict or None, optional (default=None)
2587
2588
2589
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
2590
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
2591
            Path to the model file.
2592
        model_str : str or None, optional (default=None)
2593
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
2594
        """
2595
        self.handle = None
2596
        self.network = False
wxchan's avatar
wxchan committed
2597
        self.__need_reload_eval_info = True
2598
        self._train_data_name = "training"
2599
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
2600
        self.best_iteration = -1
wxchan's avatar
wxchan committed
2601
        self.best_score = {}
2602
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
2603
        if train_set is not None:
2604
            # Training task
wxchan's avatar
wxchan committed
2605
            if not isinstance(train_set, Dataset):
2606
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
            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"]
                )
2641
            # construct booster object
2642
2643
2644
2645
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
            params_str = param_dict_to_str(params)
2646
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2647
            _safe_call(_LIB.LGBM_BoosterCreate(
2648
                train_set.handle,
wxchan's avatar
wxchan committed
2649
2650
                c_str(params_str),
                ctypes.byref(self.handle)))
2651
            # save reference to data
wxchan's avatar
wxchan committed
2652
2653
2654
2655
            self.train_set = train_set
            self.valid_sets = []
            self.name_valid_sets = []
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
2656
2657
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
2658
2659
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
2660
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
2661
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2662
2663
2664
2665
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2666
            # buffer for inner predict
wxchan's avatar
wxchan committed
2667
2668
2669
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
2670
            self.pandas_categorical = train_set.pandas_categorical
2671
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
2672
        elif model_file is not None:
2673
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
2674
            out_num_iterations = ctypes.c_int(0)
2675
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2676
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
2677
                c_str(str(model_file)),
wxchan's avatar
wxchan committed
2678
2679
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2680
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2681
2682
2683
2684
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2685
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
2686
        elif model_str is not None:
2687
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
2688
        else:
2689
2690
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
2691
        self.params = params
wxchan's avatar
wxchan committed
2692

2693
    def __del__(self) -> None:
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
        try:
            if self.network:
                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
2704

2705
    def __copy__(self) -> "Booster":
wxchan's avatar
wxchan committed
2706
2707
        return self.__deepcopy__(None)

2708
    def __deepcopy__(self, _) -> "Booster":
2709
        model_str = self.model_to_string(num_iteration=-1)
2710
        booster = Booster(model_str=model_str)
2711
        return booster
wxchan's avatar
wxchan committed
2712
2713
2714
2715
2716
2717
2718

    def __getstate__(self):
        this = self.__dict__.copy()
        handle = this['handle']
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
2719
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
2720
2721
2722
        return this

    def __setstate__(self, state):
2723
2724
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
2725
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2726
            out_num_iterations = ctypes.c_int(0)
2727
2728
2729
2730
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
                c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
2731
2732
2733
            state['handle'] = handle
        self.__dict__.update(state)

2734
    def free_dataset(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
2735
2736
2737
2738
2739
2740
2741
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
2742
2743
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
2744
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
2745
        return self
wxchan's avatar
wxchan committed
2746

2747
    def _free_buffer(self) -> "Booster":
2748
2749
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
2750
        return self
2751

2752
2753
2754
2755
2756
2757
2758
    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":
2759
2760
2761
2762
        """Set the network configuration.

        Parameters
        ----------
2763
        machines : list, set or str
2764
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
2765
        local_listen_port : int, optional (default=12400)
2766
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
2767
        listen_time_out : int, optional (default=120)
2768
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
2769
        num_machines : int, optional (default=1)
2770
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
2771
2772
2773
2774
2775

        Returns
        -------
        self : Booster
            Booster with set network.
2776
        """
2777
2778
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
2779
2780
2781
2782
2783
        _safe_call(_LIB.LGBM_NetworkInit(c_str(machines),
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
        self.network = True
Nikita Titov's avatar
Nikita Titov committed
2784
        return self
2785

2786
    def free_network(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
2787
2788
2789
2790
2791
2792
2793
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
2794
2795
        _safe_call(_LIB.LGBM_NetworkFree())
        self.network = False
Nikita Titov's avatar
Nikita Titov committed
2796
        return self
2797

2798
    def trees_to_dataframe(self) -> pd_DataFrame:
2799
2800
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

2801
2802
2803
2804
        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.
2805
2806
2807
2808
2809
            - ``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.
2810
2811
            - ``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.
2812
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
2813
2814
              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.
2815
2816
            - ``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.
2817
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
2818
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
2819
2820
            - ``count`` : int64, number of records in the training data that fall into this node.

2821
2822
2823
2824
2825
2826
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
2827
2828
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839

        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):
2840
                tree_num = f'{tree_index}-' if tree_index is not None else ''
2841
2842
2843
                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
2844
2845
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857

            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):
2858
                return set(tree.keys()) == {'leaf_value'}
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
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

            # 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))

2932
        return pd_DataFrame(model_list, columns=model_list[0].keys())
2933

2934
    def set_train_data_name(self, name: str) -> "Booster":
2935
2936
2937
2938
        """Set the name to the training Dataset.

        Parameters
        ----------
2939
        name : str
Nikita Titov's avatar
Nikita Titov committed
2940
2941
2942
2943
2944
2945
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
2946
        """
2947
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
2948
        return self
wxchan's avatar
wxchan committed
2949

2950
    def add_valid(self, data: Dataset, name: str) -> "Booster":
2951
        """Add validation data.
wxchan's avatar
wxchan committed
2952
2953
2954
2955

        Parameters
        ----------
        data : Dataset
2956
            Validation data.
2957
        name : str
2958
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
2959
2960
2961
2962
2963

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
2964
        """
Guolin Ke's avatar
Guolin Ke committed
2965
        if not isinstance(data, Dataset):
2966
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2967
        if data._predictor is not self.__init_predictor:
2968
2969
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2970
2971
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
2972
            data.construct().handle))
wxchan's avatar
wxchan committed
2973
2974
2975
2976
2977
        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
2978
        return self
wxchan's avatar
wxchan committed
2979

2980
    def reset_parameter(self, params: Dict[str, Any]) -> "Booster":
2981
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
2982
2983
2984
2985

        Parameters
        ----------
        params : dict
2986
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
2987
2988
2989
2990
2991

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
2992
2993
2994
2995
2996
2997
        """
        params_str = param_dict_to_str(params)
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
                self.handle,
                c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
2998
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
2999
        return self
wxchan's avatar
wxchan committed
3000
3001

    def update(self, train_set=None, fobj=None):
Nikita Titov's avatar
Nikita Titov committed
3002
        """Update Booster for one iteration.
3003

wxchan's avatar
wxchan committed
3004
3005
        Parameters
        ----------
3006
3007
3008
3009
        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
3010
            Customized objective function.
3011
3012
3013
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

3014
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3015
                    The predicted values.
3016
3017
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
3018
3019
                train_data : Dataset
                    The training dataset.
3020
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3021
3022
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
3023
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3024
3025
                    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
3026

3027
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
3028
            and grad and hess should be returned in the same format.
3029

wxchan's avatar
wxchan committed
3030
3031
        Returns
        -------
3032
3033
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
3034
        """
3035
        # need reset training data
3036
3037
3038
3039
3040
3041
        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
3042
            if not isinstance(train_set, Dataset):
3043
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
3044
            if train_set._predictor is not self.__init_predictor:
3045
3046
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
3047
3048
3049
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
3050
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
3051
            self.__inner_predict_buffer[0] = None
3052
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
3053
3054
        is_finished = ctypes.c_int(0)
        if fobj is None:
3055
            if self.__set_objective_to_none:
3056
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
3057
3058
3059
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
3060
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3061
3062
            return is_finished.value == 1
        else:
3063
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3064
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3065
3066
3067
3068
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

    def __boost(self, grad, hess):
3069
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3070

Nikita Titov's avatar
Nikita Titov committed
3071
3072
        .. note::

3073
3074
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3075
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3076
            and grad and hess should be returned in the same format.
3077

wxchan's avatar
wxchan committed
3078
3079
        Parameters
        ----------
3080
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3081
3082
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3083
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3084
3085
            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
3086
3087
3088

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3089
3090
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3091
        """
3092
3093
3094
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3095
3096
        grad = list_to_1d_numpy(grad, name='gradient')
        hess = list_to_1d_numpy(hess, name='hessian')
3097
3098
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3099
        if len(grad) != len(hess):
3100
3101
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3102
        if len(grad) != num_train_data * self.__num_class:
3103
3104
3105
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3106
                f"number of models per one iteration ({self.__num_class})"
3107
            )
wxchan's avatar
wxchan committed
3108
3109
3110
3111
3112
3113
        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)))
3114
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3115
3116
        return is_finished.value == 1

3117
    def rollback_one_iter(self) -> "Booster":
Nikita Titov's avatar
Nikita Titov committed
3118
3119
3120
3121
3122
3123
3124
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3125
3126
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
3127
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3128
        return self
wxchan's avatar
wxchan committed
3129

3130
    def current_iteration(self) -> int:
3131
3132
3133
3134
3135
3136
3137
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3138
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3139
3140
3141
3142
3143
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3144
    def num_model_per_iteration(self) -> int:
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
        """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

3158
    def num_trees(self) -> int:
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
        """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

3172
    def upper_bound(self) -> float:
3173
3174
3175
3176
        """Get upper bound value of a model.

        Returns
        -------
3177
        upper_bound : float
3178
3179
3180
3181
3182
3183
3184
3185
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

3186
    def lower_bound(self) -> float:
3187
3188
3189
3190
        """Get lower bound value of a model.

        Returns
        -------
3191
        lower_bound : float
3192
3193
3194
3195
3196
3197
3198
3199
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

wxchan's avatar
wxchan committed
3200
    def eval(self, data, name, feval=None):
3201
        """Evaluate for data.
wxchan's avatar
wxchan committed
3202
3203
3204

        Parameters
        ----------
3205
3206
        data : Dataset
            Data for the evaluating.
3207
        name : str
3208
            Name of the data.
3209
        feval : callable, list of callable, or None, optional (default=None)
3210
            Customized evaluation function.
3211
            Each evaluation function should accept two parameters: preds, eval_data,
3212
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3213

3214
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3215
                    The predicted values.
3216
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3217
                    If custom objective function is used, predicted values are returned before any transformation,
3218
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3219
                eval_data : Dataset
3220
                    A ``Dataset`` to evaluate.
3221
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3222
                    The name of evaluation function (without whitespace).
3223
3224
3225
3226
3227
                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
3228
3229
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3230
        result : list
3231
            List with evaluation results.
wxchan's avatar
wxchan committed
3232
        """
Guolin Ke's avatar
Guolin Ke committed
3233
3234
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3235
3236
3237
3238
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3239
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3240
3241
3242
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3243
        # need to push new valid data
wxchan's avatar
wxchan committed
3244
3245
3246
3247
3248
3249
3250
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

    def eval_train(self, feval=None):
3251
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3252
3253
3254

        Parameters
        ----------
3255
        feval : callable, list of callable, or None, optional (default=None)
3256
            Customized evaluation function.
3257
            Each evaluation function should accept two parameters: preds, eval_data,
3258
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3259

3260
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3261
                    The predicted values.
3262
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3263
                    If custom objective function is used, predicted values are returned before any transformation,
3264
                    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
3265
                eval_data : Dataset
3266
                    The training dataset.
3267
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3268
                    The name of evaluation function (without whitespace).
3269
3270
3271
3272
3273
                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
3274
3275
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3276
        result : list
3277
            List with evaluation results.
wxchan's avatar
wxchan committed
3278
        """
3279
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
3280
3281

    def eval_valid(self, feval=None):
3282
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
3283
3284
3285

        Parameters
        ----------
3286
        feval : callable, list of callable, or None, optional (default=None)
3287
            Customized evaluation function.
3288
            Each evaluation function should accept two parameters: preds, eval_data,
3289
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3290

3291
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3292
                    The predicted values.
3293
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3294
                    If custom objective function is used, predicted values are returned before any transformation,
3295
                    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
3296
                eval_data : Dataset
3297
                    The validation dataset.
3298
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3299
                    The name of evaluation function (without whitespace).
3300
3301
3302
3303
3304
                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
3305
3306
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3307
        result : list
3308
            List with evaluation results.
wxchan's avatar
wxchan committed
3309
        """
3310
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
3311
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
3312

3313
    def save_model(self, filename, num_iteration=None, start_iteration=0, importance_type='split'):
3314
        """Save Booster to file.
wxchan's avatar
wxchan committed
3315
3316
3317

        Parameters
        ----------
3318
        filename : str or pathlib.Path
3319
            Filename to save Booster.
3320
3321
3322
3323
        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
3324
        start_iteration : int, optional (default=0)
3325
            Start index of the iteration that should be saved.
3326
        importance_type : str, optional (default="split")
3327
3328
3329
            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
3330
3331
3332
3333
3334

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
3335
        """
3336
        if num_iteration is None:
3337
            num_iteration = self.best_iteration
3338
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3339
3340
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
3341
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3342
            ctypes.c_int(num_iteration),
3343
            ctypes.c_int(importance_type_int),
3344
            c_str(str(filename))))
3345
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
3346
        return self
wxchan's avatar
wxchan committed
3347

3348
    def shuffle_models(self, start_iteration=0, end_iteration=-1):
3349
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
3350

3351
3352
3353
        Parameters
        ----------
        start_iteration : int, optional (default=0)
3354
            The first iteration that will be shuffled.
3355
3356
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
3357
            If <= 0, means the last available iteration.
3358

Nikita Titov's avatar
Nikita Titov committed
3359
3360
3361
3362
        Returns
        -------
        self : Booster
            Booster with shuffled models.
3363
        """
3364
3365
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
3366
3367
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
3368
        return self
3369

3370
    def model_from_string(self, model_str: str) -> "Booster":
3371
3372
3373
3374
        """Load Booster from a string.

        Parameters
        ----------
3375
        model_str : str
3376
3377
3378
3379
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3380
        self : Booster
3381
3382
            Loaded Booster object.
        """
3383
3384
3385
3386
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
            c_str(model_str),
            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
3397
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
3398
3399
        return self

3400
    def model_to_string(self, num_iteration=None, start_iteration=0, importance_type='split'):
3401
        """Save Booster to string.
3402

3403
3404
3405
3406
3407
3408
        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
3409
        start_iteration : int, optional (default=0)
3410
            Start index of the iteration that should be saved.
3411
        importance_type : str, optional (default="split")
3412
3413
3414
            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.
3415
3416
3417

        Returns
        -------
3418
        str_repr : str
3419
3420
            String representation of Booster.
        """
3421
        if num_iteration is None:
3422
            num_iteration = self.best_iteration
3423
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3424
        buffer_len = 1 << 20
3425
        tmp_out_len = ctypes.c_int64(0)
3426
3427
3428
3429
        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,
3430
            ctypes.c_int(start_iteration),
3431
            ctypes.c_int(num_iteration),
3432
            ctypes.c_int(importance_type_int),
3433
            ctypes.c_int64(buffer_len),
3434
3435
3436
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
3437
        # if buffer length is not long enough, re-allocate a buffer
3438
3439
3440
3441
3442
        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,
3443
                ctypes.c_int(start_iteration),
3444
                ctypes.c_int(num_iteration),
3445
                ctypes.c_int(importance_type_int),
3446
                ctypes.c_int64(actual_len),
3447
3448
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
3449
        ret = string_buffer.value.decode('utf-8')
3450
3451
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
3452

3453
    def dump_model(self, num_iteration=None, start_iteration=0, importance_type='split', object_hook=None):
Nikita Titov's avatar
Nikita Titov committed
3454
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
3455

3456
3457
        Parameters
        ----------
3458
3459
3460
3461
        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
3462
        start_iteration : int, optional (default=0)
3463
            Start index of the iteration that should be dumped.
3464
        importance_type : str, optional (default="split")
3465
3466
3467
            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.
3468
3469
3470
3471
3472
3473
3474
3475
3476
        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.
3477

wxchan's avatar
wxchan committed
3478
3479
        Returns
        -------
3480
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
3481
            JSON format of Booster.
wxchan's avatar
wxchan committed
3482
        """
3483
        if num_iteration is None:
3484
            num_iteration = self.best_iteration
3485
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3486
        buffer_len = 1 << 20
3487
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
3488
3489
3490
3491
        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,
3492
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3493
            ctypes.c_int(num_iteration),
3494
            ctypes.c_int(importance_type_int),
3495
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
3496
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3497
            ptr_string_buffer))
wxchan's avatar
wxchan committed
3498
        actual_len = tmp_out_len.value
3499
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
3500
3501
3502
3503
3504
        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,
3505
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3506
                ctypes.c_int(num_iteration),
3507
                ctypes.c_int(importance_type_int),
3508
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
3509
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3510
                ptr_string_buffer))
3511
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
3512
3513
3514
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
                                                          default=json_default_with_numpy))
        return ret
wxchan's avatar
wxchan committed
3515

3516
    def predict(self, data, start_iteration=0, num_iteration=None,
3517
                raw_score=False, pred_leaf=False, pred_contrib=False,
3518
3519
                data_has_header=False, validate_features=False,
                **kwargs):
3520
        """Make a prediction.
wxchan's avatar
wxchan committed
3521
3522
3523

        Parameters
        ----------
3524
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
3525
            Data source for prediction.
3526
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3527
        start_iteration : int, optional (default=0)
3528
            Start index of the iteration to predict.
3529
            If <= 0, starts from the first iteration.
3530
        num_iteration : int or None, optional (default=None)
3531
3532
3533
3534
            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).
3535
3536
3537
3538
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
3539
3540
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
3541

Nikita Titov's avatar
Nikita Titov committed
3542
3543
3544
3545
3546
3547
3548
            .. 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.
3549

3550
3551
        data_has_header : bool, optional (default=False)
            Whether the data has header.
3552
            Used only if data is str.
3553
3554
3555
        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.
3556
3557
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
3558
3559
3560

        Returns
        -------
3561
        result : numpy array, scipy.sparse or list of scipy.sparse
3562
            Prediction result.
3563
            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
3564
        """
3565
        predictor = self._to_predictor(deepcopy(kwargs))
3566
        if num_iteration is None:
3567
            if start_iteration <= 0:
3568
3569
3570
3571
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
        return predictor.predict(data, start_iteration, num_iteration,
3572
                                 raw_score, pred_leaf, pred_contrib,
3573
                                 data_has_header, validate_features)
wxchan's avatar
wxchan committed
3574

3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
    def refit(
        self,
        data,
        label,
        decay_rate=0.9,
        reference=None,
        weight=None,
        group=None,
        init_score=None,
        feature_name='auto',
        categorical_feature='auto',
        dataset_params=None,
        free_raw_data=True,
        **kwargs
    ):
Guolin Ke's avatar
Guolin Ke committed
3590
3591
3592
3593
        """Refit the existing Booster by new data.

        Parameters
        ----------
3594
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
Guolin Ke's avatar
Guolin Ke committed
3595
            Data source for refit.
3596
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3597
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
3598
3599
            Label for refit.
        decay_rate : float, optional (default=0.9)
3600
3601
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
3602
3603
3604
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
3605
            Weight for each ``data`` instance. Weights should be non-negative.
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
        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.
3622
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
3623
3624
3625
            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.
3626
            Floating point numbers in categorical features will be rounded towards 0.
3627
3628
3629
3630
        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``.
3631
3632
        **kwargs
            Other parameters for refit.
3633
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
3634
3635
3636
3637
3638
3639

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
3640
3641
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
3642
3643
        if dataset_params is None:
            dataset_params = {}
3644
        predictor = self._to_predictor(deepcopy(kwargs))
3645
        leaf_preds = predictor.predict(data, -1, pred_leaf=True)
3646
        nrow, ncol = leaf_preds.shape
3647
        out_is_linear = ctypes.c_int(0)
3648
3649
3650
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
3651
3652
3653
3654
3655
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
3656
        new_params["linear_tree"] = bool(out_is_linear.value)
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
        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,
        )
3670
        new_params['refit_decay_rate'] = decay_rate
3671
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
3672
3673
3674
3675
3676
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
3677
        ptr_data, _, _ = c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
3678
3679
3680
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
3681
3682
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
3683
        new_booster.network = self.network
Guolin Ke's avatar
Guolin Ke committed
3684
3685
        return new_booster

3686
    def get_leaf_output(self, tree_id: int, leaf_id: int) -> float:
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
        """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.
        """
3701
3702
3703
3704
3705
3706
3707
3708
        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

3709
    def _to_predictor(self, pred_parameter=None):
3710
        """Convert to predictor."""
3711
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
3712
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
3713
3714
        return predictor

3715
    def num_feature(self) -> int:
3716
3717
3718
3719
3720
3721
3722
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
3723
3724
3725
3726
3727
3728
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
            self.handle,
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

3729
    def feature_name(self) -> List[str]:
3730
        """Get names of features.
wxchan's avatar
wxchan committed
3731
3732
3733

        Returns
        -------
3734
        result : list of str
3735
            List with names of features.
wxchan's avatar
wxchan committed
3736
        """
3737
        num_feature = self.num_feature()
3738
        # Get name of features
wxchan's avatar
wxchan committed
3739
        tmp_out_len = ctypes.c_int(0)
3740
3741
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
3742
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
3743
3744
3745
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
3746
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
3747
            ctypes.byref(tmp_out_len),
3748
            ctypes.c_size_t(reserved_string_buffer_size),
3749
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3750
3751
3752
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
        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))
3765
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
3766

3767
3768
3769
3770
3771
    def feature_importance(
        self,
        importance_type: str = 'split',
        iteration: Optional[int] = None
    ) -> np.ndarray:
3772
        """Get feature importances.
3773

3774
3775
        Parameters
        ----------
3776
        importance_type : str, optional (default="split")
3777
3778
3779
            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.
3780
3781
3782
3783
        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).
3784

3785
3786
        Returns
        -------
3787
3788
        result : numpy array
            Array with feature importances.
3789
        """
3790
3791
        if iteration is None:
            iteration = self.best_iteration
3792
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3793
        result = np.empty(self.num_feature(), dtype=np.float64)
3794
3795
3796
3797
3798
        _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))))
3799
        if importance_type_int == C_API_FEATURE_IMPORTANCE_SPLIT:
3800
            return result.astype(np.int32)
3801
3802
        else:
            return result
3803

3804
3805
3806
3807
3808
    def get_split_value_histogram(self, feature, bins=None, xgboost_style=False):
        """Get split value histogram for the specified feature.

        Parameters
        ----------
3809
        feature : int or str
3810
3811
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
3812
            If str, interpreted as name.
3813

Nikita Titov's avatar
Nikita Titov committed
3814
3815
3816
            .. warning::

                Categorical features are not supported.
3817

3818
        bins : int, str or None, optional (default=None)
3819
3820
3821
            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.
3822
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
        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.
        """
        def add(root):
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
3840
                if feature_names is not None and isinstance(feature, str):
3841
3842
3843
3844
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
3845
                    if isinstance(root['threshold'], str):
3846
3847
3848
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
                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'])

3859
        if bins is None or isinstance(bins, int) and xgboost_style:
3860
3861
3862
3863
3864
3865
3866
            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:
3867
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
3868
3869
3870
3871
3872
            else:
                return ret
        else:
            return hist, bin_edges

wxchan's avatar
wxchan committed
3873
    def __inner_eval(self, data_name, data_idx, feval=None):
3874
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
3875
        if data_idx >= self.__num_dataset:
3876
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3877
3878
3879
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
3880
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
3881
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3882
3883
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3884
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3885
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3886
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
3887
            if tmp_out_len.value != self.__num_inner_eval:
3888
                raise ValueError("Wrong length of eval results")
3889
            for i in range(self.__num_inner_eval):
3890
3891
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
3892
3893
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
3894
3895
3896
3897
3898
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
3899
3900
3901
3902
3903
3904
3905
3906
3907
            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
3908
3909
3910
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

3911
    def __inner_predict(self, data_idx: int):
3912
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
3913
        if data_idx >= self.__num_dataset:
3914
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3915
3916
3917
3918
3919
        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
3920
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
3921
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
3922
3923
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
3924
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
3925
3926
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3927
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3928
3929
3930
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
3931
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
3932
            self.__is_predicted_cur_iter[data_idx] = True
3933
3934
3935
3936
3937
        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
3938

3939
    def __get_eval_info(self) -> None:
3940
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
3941
3942
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
3943
            out_num_eval = ctypes.c_int(0)
3944
            # Get num of inner evals
wxchan's avatar
wxchan committed
3945
3946
3947
3948
3949
            _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:
3950
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
3951
                tmp_out_len = ctypes.c_int(0)
3952
3953
3954
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
3955
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
3956
                ]
wxchan's avatar
wxchan committed
3957
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
3958
3959
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
3960
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
3961
                    ctypes.byref(tmp_out_len),
3962
                    ctypes.c_size_t(reserved_string_buffer_size),
3963
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3964
3965
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
3966
                    raise ValueError("Length of eval names doesn't equal with num_evals")
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
                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
                ]