basic.py 163 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
10
from functools import wraps
from logging import Logger
11
12
13
from os import SEEK_END
from os.path import getsize
from pathlib import Path
14
from tempfile import NamedTemporaryFile
15
from typing import Any, Callable, Dict, Iterable, List, Optional, Set, Tuple, Union
wxchan's avatar
wxchan committed
16
17
18
19

import numpy as np
import scipy.sparse

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

23
24
25
26
27
28
29
30
31
32
33
34
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
35

36
class _DummyLogger:
37
    def info(self, msg: str) -> None:
38
39
        print(msg)

40
    def warning(self, msg: str) -> None:
41
42
43
        warnings.warn(msg, stacklevel=3)


44
_LOGGER: Union[_DummyLogger, Logger] = _DummyLogger()
45
46


47
def register_logger(logger: Logger) -> None:
48
49
50
51
52
53
54
55
56
57
58
59
60
    """Register custom logger.

    Parameters
    ----------
    logger : logging.Logger
        Custom logger.
    """
    if not isinstance(logger, Logger):
        raise TypeError("Logger should inherit logging.Logger class")
    global _LOGGER
    _LOGGER = logger


61
def _normalize_native_string(func: Callable[[str], None]) -> Callable[[str], None]:
62
    """Join log messages from native library which come by chunks."""
63
    msg_normalized: List[str] = []
64
65

    @wraps(func)
66
    def wrapper(msg: str) -> None:
67
68
69
70
71
72
73
74
75
76
77
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


78
def _log_info(msg: str) -> None:
79
80
81
    _LOGGER.info(msg)


82
def _log_warning(msg: str) -> None:
83
84
85
86
    _LOGGER.warning(msg)


@_normalize_native_string
87
def _log_native(msg: str) -> None:
88
89
90
    _LOGGER.info(msg)


91
def _log_callback(msg: bytes) -> None:
92
    """Redirect logs from native library into Python."""
93
    _log_native(str(msg.decode('utf-8')))
94
95


wxchan's avatar
wxchan committed
96
def _load_lib():
97
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
98
99
    lib_path = find_lib_path()
    if len(lib_path) == 0:
100
        return None
wxchan's avatar
wxchan committed
101
102
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
103
104
105
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
    lib.callback = callback(_log_callback)
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
106
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
107
108
    return lib

wxchan's avatar
wxchan committed
109

wxchan's avatar
wxchan committed
110
111
_LIB = _load_lib()

wxchan's avatar
wxchan committed
112

113
NUMERIC_TYPES = (int, float, bool)
114
_ArrayLike = Union[List, np.ndarray, pd_Series]
115
116


117
def _safe_call(ret: int) -> None:
118
119
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
120
121
122
    Parameters
    ----------
    ret : int
123
        The return value from C API calls.
wxchan's avatar
wxchan committed
124
125
    """
    if ret != 0:
126
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
127

wxchan's avatar
wxchan committed
128

wxchan's avatar
wxchan committed
129
def is_numeric(obj):
130
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
131
132
133
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
134
135
136
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
137
138
        return False

wxchan's avatar
wxchan committed
139

wxchan's avatar
wxchan committed
140
def is_numpy_1d_array(data):
141
    """Check whether data is a numpy 1-D array."""
142
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
143

wxchan's avatar
wxchan committed
144

145
146
147
148
149
150
151
152
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


153
154
def cast_numpy_array_to_dtype(array, dtype):
    """Cast numpy array to given dtype."""
155
156
157
158
159
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


wxchan's avatar
wxchan committed
160
def is_1d_list(data):
161
162
    """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
163

wxchan's avatar
wxchan committed
164

165
166
167
168
169
170
171
172
173
174
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)
    )


175
def list_to_1d_numpy(data, dtype=np.float32, name='list'):
176
    """Convert data to numpy 1-D array."""
wxchan's avatar
wxchan committed
177
    if is_numpy_1d_array(data):
178
        return cast_numpy_array_to_dtype(data, dtype)
179
180
181
    elif is_numpy_column_array(data):
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
182
        return cast_numpy_array_to_dtype(array, dtype)
wxchan's avatar
wxchan committed
183
184
    elif is_1d_list(data):
        return np.array(data, dtype=dtype, copy=False)
185
    elif isinstance(data, pd_Series):
186
        _check_for_bad_pandas_dtypes(data.to_frame().dtypes)
187
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
188
    else:
189
190
        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
191

wxchan's avatar
wxchan committed
192

193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
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):
219
        _check_for_bad_pandas_dtypes(data.dtypes)
220
221
222
223
224
        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
225
def cfloat32_array_to_numpy(cptr, length):
226
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
227
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
228
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
229
    else:
230
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
231

Guolin Ke's avatar
Guolin Ke committed
232

Guolin Ke's avatar
Guolin Ke committed
233
def cfloat64_array_to_numpy(cptr, length):
234
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
235
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
236
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
237
238
239
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
240

wxchan's avatar
wxchan committed
241
def cint32_array_to_numpy(cptr, length):
242
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
243
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
244
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
245
    else:
246
247
248
249
250
251
        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)):
252
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
253
254
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
255

wxchan's avatar
wxchan committed
256

wxchan's avatar
wxchan committed
257
def c_str(string):
258
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
259
260
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
261

wxchan's avatar
wxchan committed
262
def c_array(ctype, values):
263
    """Convert a Python array to C array."""
wxchan's avatar
wxchan committed
264
265
    return (ctype * len(values))(*values)

wxchan's avatar
wxchan committed
266

267
268
269
270
271
272
273
274
275
276
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
277
def param_dict_to_str(data):
278
    """Convert Python dictionary to string, which is passed to C API."""
279
    if data is None or not data:
wxchan's avatar
wxchan committed
280
281
282
        return ""
    pairs = []
    for key, val in data.items():
283
        if isinstance(val, (list, tuple, set)) or is_numpy_1d_array(val):
284
285
            def to_string(x):
                if isinstance(x, list):
286
                    return f"[{','.join(map(str, x))}]"
287
288
                else:
                    return str(x)
289
            pairs.append(f"{key}={','.join(map(to_string, val))}")
290
        elif isinstance(val, (str, Path, NUMERIC_TYPES)) or is_numeric(val):
291
            pairs.append(f"{key}={val}")
292
        elif val is not None:
293
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
294
    return ' '.join(pairs)
295

wxchan's avatar
wxchan committed
296

297
class _TempFile:
298
299
    """Proxy class to workaround errors on Windows."""

300
301
302
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
303
            self.path = Path(self.name)
304
        return self
wxchan's avatar
wxchan committed
305

306
    def __exit__(self, exc_type, exc_val, exc_tb):
307
308
        if self.path.is_file():
            self.path.unlink()
309

wxchan's avatar
wxchan committed
310

311
class LightGBMError(Exception):
312
313
    """Error thrown by LightGBM."""

314
315
316
    pass


317
318
319
320
321
322
323
324
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
    # 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
352
353

    @classmethod
354
355
356
    def get(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
357
358
        ret = set()
        for i in args:
359
            ret |= cls.aliases.get(i, {i})
360
361
        return ret

362
    @classmethod
363
364
365
    def get_by_alias(cls, *args) -> Set[str]:
        if cls.aliases is None:
            cls.aliases = cls._get_all_param_aliases()
366
367
368
369
370
371
372
373
        ret = set(args)
        for arg in args:
            for aliases in cls.aliases.values():
                if arg in aliases:
                    ret |= aliases
                    break
        return ret

374

375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
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)

    # find a value, and remove other aliases with .pop()
    # prefer the value of 'main_param_name' if it exists, otherwise search the aliases
    found_value = None
    if main_param_name in params.keys():
        found_value = params[main_param_name]

    for param in _ConfigAliases.get(main_param_name):
        val = params.pop(param, None)
        if found_value is None and val is not None:
            found_value = val

    if found_value is not None:
        params[main_param_name] = found_value
    else:
        params[main_param_name] = default_value

    return params


415
416
MAX_INT32 = (1 << 31) - 1

417
"""Macro definition of data type in C API of LightGBM"""
wxchan's avatar
wxchan committed
418
419
420
421
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
422

423
"""Matrix is row major in Python"""
wxchan's avatar
wxchan committed
424
425
C_API_IS_ROW_MAJOR = 1

426
"""Macro definition of prediction type in C API of LightGBM"""
wxchan's avatar
wxchan committed
427
428
429
C_API_PREDICT_NORMAL = 0
C_API_PREDICT_RAW_SCORE = 1
C_API_PREDICT_LEAF_INDEX = 2
430
C_API_PREDICT_CONTRIB = 3
wxchan's avatar
wxchan committed
431

432
433
434
435
"""Macro definition of sparse matrix type"""
C_API_MATRIX_TYPE_CSR = 0
C_API_MATRIX_TYPE_CSC = 1

436
437
438
439
"""Macro definition of feature importance type"""
C_API_FEATURE_IMPORTANCE_SPLIT = 0
C_API_FEATURE_IMPORTANCE_GAIN = 1

440
"""Data type of data field"""
wxchan's avatar
wxchan committed
441
442
FIELD_TYPE_MAPPER = {"label": C_API_DTYPE_FLOAT32,
                     "weight": C_API_DTYPE_FLOAT32,
Guolin Ke's avatar
Guolin Ke committed
443
                     "init_score": C_API_DTYPE_FLOAT64,
444
                     "group": C_API_DTYPE_INT32}
wxchan's avatar
wxchan committed
445

446
447
448
449
"""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
450

451
def convert_from_sliced_object(data):
452
    """Fix the memory of multi-dimensional sliced object."""
453
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
454
        if not data.flags.c_contiguous:
455
456
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
457
458
459
460
            return np.copy(data)
    return data


wxchan's avatar
wxchan committed
461
def c_float_array(data):
462
    """Get pointer of float numpy array / list."""
wxchan's avatar
wxchan committed
463
464
465
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
466
467
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
468
469
470
471
472
473
474
        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:
475
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
476
    else:
477
        raise TypeError(f"Unknown type({type(data).__name__})")
478
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
479

wxchan's avatar
wxchan committed
480

wxchan's avatar
wxchan committed
481
def c_int_array(data):
482
    """Get pointer of int numpy array / list."""
wxchan's avatar
wxchan committed
483
484
485
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
486
487
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
488
489
490
491
492
493
494
        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:
495
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
496
    else:
497
        raise TypeError(f"Unknown type({type(data).__name__})")
498
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
499

wxchan's avatar
wxchan committed
500

501
def _check_for_bad_pandas_dtypes(pandas_dtypes_series):
502
503
504
505
506
507
508
509
    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))
        )

510
511
512
513
514
515
516
517
    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)}')
518
519


520
def _data_from_pandas(data, feature_name, categorical_feature, pandas_categorical):
521
    if isinstance(data, pd_DataFrame):
522
523
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
524
525
        if feature_name == 'auto' or feature_name is None:
            data = data.rename(columns=str)
526
        cat_cols = [col for col, dtype in zip(data.columns, data.dtypes) if isinstance(dtype, pd_CategoricalDtype)]
527
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
528
529
530
531
532
        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.')
533
            for col, category in zip(cat_cols, pandas_categorical):
534
535
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
536
        if len(cat_cols):  # cat_cols is list
537
            data = data.copy()  # not alter origin DataFrame
538
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
539
540
541
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
542
            if categorical_feature == 'auto':  # use cat cols from DataFrame
543
                categorical_feature = cat_cols_not_ordered
544
545
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
546
547
        if feature_name == 'auto':
            feature_name = list(data.columns)
548
        _check_for_bad_pandas_dtypes(data.dtypes)
549
550
551
552
        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
553
554
555
556
557
558
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
559
560
561


def _label_from_pandas(label):
562
    if isinstance(label, pd_DataFrame):
563
564
        if len(label.columns) > 1:
            raise ValueError('DataFrame for label cannot have multiple columns')
565
        _check_for_bad_pandas_dtypes(label.dtypes)
566
        label = np.ravel(label.values.astype(np.float32, copy=False))
567
568
569
    return label


570
def _dump_pandas_categorical(pandas_categorical, file_name=None):
571
572
    categorical_json = json.dumps(pandas_categorical, default=json_default_with_numpy)
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
573
574
575
576
577
578
579
    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):
580
581
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
582
    if file_name is not None:
583
        max_offset = -getsize(file_name)
584
585
586
587
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
588
                f.seek(offset, SEEK_END)
589
590
591
592
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
593
        last_line = lines[-1].decode('utf-8').strip()
594
        if not last_line.startswith(pandas_key):
595
            last_line = lines[-2].decode('utf-8').strip()
596
    elif model_str is not None:
597
598
599
600
601
602
        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
603
604


605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
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**.

625
626
    .. versionadded:: 3.3.0

627
628
629
630
631
632
633
634
635
    Attributes
    ----------
    batch_size : int
        Default size of a batch.
    """

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

    @abc.abstractmethod
636
    def __getitem__(self, idx: Union[int, slice, List[int]]) -> np.ndarray:
637
638
639
640
641
642
643
        """Return data for given row index.

        A basic implementation should look like this:

        .. code-block:: python

            if isinstance(idx, numbers.Integral):
644
                return self._get_one_line(idx)
645
            elif isinstance(idx, slice):
646
647
                return np.stack([self._get_one_line(i) for i in range(idx.start, idx.stop)])
            elif isinstance(idx, list):
648
                # Only required if using ``Dataset.subset()``.
649
                return np.array([self._get_one_line(i) for i in idx])
650
            else:
651
                raise TypeError(f"Sequence index must be integer, slice or list, got {type(idx).__name__}")
652
653
654

        Parameters
        ----------
655
        idx : int, slice[int], list[int]
656
657
658
659
            Item index.

        Returns
        -------
660
        result : numpy 1-D array or numpy 2-D array
661
            1-D array if idx is int, 2-D array if idx is slice or list.
662
663
664
665
666
667
668
669
670
        """
        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__()")


671
class _InnerPredictor:
672
673
674
675
676
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
677
678
679
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
680
    """
681

682
    def __init__(self, model_file=None, booster_handle=None, pred_parameter=None):
683
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
684
685
686

        Parameters
        ----------
687
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
688
            Path to the model file.
689
690
691
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
692
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
693
694
695
696
697
        """
        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
698
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
699
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
700
                c_str(str(model_file)),
wxchan's avatar
wxchan committed
701
702
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
703
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
704
705
706
707
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
708
            self.num_total_iteration = out_num_iterations.value
709
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
710
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
711
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
712
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
713
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
714
715
716
717
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
718
            self.num_total_iteration = self.current_iteration()
719
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
720
        else:
721
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
722

723
724
        pred_parameter = {} if pred_parameter is None else pred_parameter
        self.pred_parameter = param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
725

wxchan's avatar
wxchan committed
726
    def __del__(self):
727
728
729
730
731
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
732

733
734
735
736
737
    def __getstate__(self):
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

738
    def predict(self, data, start_iteration=0, num_iteration=-1,
739
                raw_score=False, pred_leaf=False, pred_contrib=False, data_has_header=False,
wxchan's avatar
wxchan committed
740
                is_reshape=True):
741
        """Predict logic.
wxchan's avatar
wxchan committed
742
743
744

        Parameters
        ----------
745
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
746
            Data source for prediction.
747
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
748
749
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
750
751
752
753
754
755
756
757
758
759
760
761
762
        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.
        is_reshape : bool, optional (default=True)
            Whether to reshape to (nrow, ncol).
wxchan's avatar
wxchan committed
763
764
765

        Returns
        -------
766
        result : numpy array, scipy.sparse or list of scipy.sparse
767
            Prediction result.
768
            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
769
        """
wxchan's avatar
wxchan committed
770
        if isinstance(data, Dataset):
771
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
772
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
wxchan's avatar
wxchan committed
773
774
775
776
777
        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
778
779
        if pred_contrib:
            predict_type = C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
780
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
781

782
        if isinstance(data, (str, Path)):
783
            with _TempFile() as f:
wxchan's avatar
wxchan committed
784
785
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
786
                    c_str(str(data)),
Guolin Ke's avatar
Guolin Ke committed
787
788
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
789
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
790
                    ctypes.c_int(num_iteration),
791
                    c_str(self.pred_parameter),
wxchan's avatar
wxchan committed
792
                    c_str(f.name)))
793
794
                preds = np.loadtxt(f.name, dtype=np.float64)
                nrow = preds.shape[0]
wxchan's avatar
wxchan committed
795
        elif isinstance(data, scipy.sparse.csr_matrix):
796
            preds, nrow = self.__pred_for_csr(data, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
797
        elif isinstance(data, scipy.sparse.csc_matrix):
798
            preds, nrow = self.__pred_for_csc(data, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
799
        elif isinstance(data, np.ndarray):
800
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
801
802
803
        elif isinstance(data, list):
            try:
                data = np.array(data)
804
            except BaseException:
805
                raise ValueError('Cannot convert data list to numpy array.')
806
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
807
        elif isinstance(data, dt_DataTable):
808
            preds, nrow = self.__pred_for_np2d(data.to_numpy(), start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
809
810
        else:
            try:
811
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
812
                csr = scipy.sparse.csr_matrix(data)
813
            except BaseException:
814
                raise TypeError(f'Cannot predict data for type {type(data).__name__}')
815
            preds, nrow = self.__pred_for_csr(csr, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
816
817
        if pred_leaf:
            preds = preds.astype(np.int32)
818
819
        is_sparse = scipy.sparse.issparse(preds) or isinstance(preds, list)
        if is_reshape and not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
820
            if preds.size % nrow == 0:
821
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
822
            else:
823
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
824
825
        return preds

826
    def __get_num_preds(self, start_iteration, num_iteration, nrow, predict_type):
827
        """Get size of prediction result."""
828
        if nrow > MAX_INT32:
829
            raise LightGBMError('LightGBM cannot perform prediction for data '
830
                                f'with number of rows greater than MAX_INT32 ({MAX_INT32}).\n'
831
                                'You can split your data into chunks '
832
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
833
834
835
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
836
837
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
838
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
839
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
840
841
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
842

843
    def __pred_for_np2d(self, mat, start_iteration, num_iteration, predict_type):
844
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
845
        if len(mat.shape) != 2:
846
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
847

848
        def inner_predict(mat, start_iteration, num_iteration, predict_type, preds=None):
849
850
            if mat.dtype == np.float32 or mat.dtype == np.float64:
                data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
851
            else:  # change non-float data to float data, need to copy
852
853
                data = np.array(mat.reshape(mat.size), dtype=np.float32)
            ptr_data, type_ptr_data, _ = c_float_array(data)
854
            n_preds = self.__get_num_preds(start_iteration, num_iteration, mat.shape[0], predict_type)
855
            if preds is None:
856
                preds = np.empty(n_preds, dtype=np.float64)
857
858
859
860
861
862
863
            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),
864
865
                ctypes.c_int32(mat.shape[0]),
                ctypes.c_int32(mat.shape[1]),
866
867
                ctypes.c_int(C_API_IS_ROW_MAJOR),
                ctypes.c_int(predict_type),
868
                ctypes.c_int(start_iteration),
869
870
871
872
873
874
875
876
877
878
879
880
                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
881
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
882
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
883
            preds = np.empty(sum(n_preds), dtype=np.float64)
884
885
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
886
                # avoid memory consumption by arrays concatenation operations
887
                inner_predict(chunk, start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
888
            return preds, nrow
wxchan's avatar
wxchan committed
889
        else:
890
            return inner_predict(mat, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
891

892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
    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

938
    def __pred_for_csr(self, csr, start_iteration, num_iteration, predict_type):
939
        """Predict for a CSR data."""
940
        def inner_predict(csr, start_iteration, num_iteration, predict_type, preds=None):
941
            nrow = len(csr.indptr) - 1
942
            n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
943
            if preds is None:
944
                preds = np.empty(n_preds, dtype=np.float64)
945
946
947
948
949
950
951
            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)

952
            assert csr.shape[1] <= MAX_INT32
953
            csr_indices = csr.indices.astype(np.int32, copy=False)
954

955
956
957
            _safe_call(_LIB.LGBM_BoosterPredictForCSR(
                self.handle,
                ptr_indptr,
958
                ctypes.c_int(type_ptr_indptr),
959
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
960
961
962
963
964
965
                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),
966
                ctypes.c_int(start_iteration),
967
968
969
970
971
972
973
                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
974

975
        def inner_predict_sparse(csr, start_iteration, num_iteration, predict_type):
976
977
978
979
980
981
982
983
984
985
986
987
988
            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)()
989
            out_shape = np.empty(2, dtype=np.int64)
990
991
992
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
993
                ctypes.c_int(type_ptr_indptr),
994
995
996
997
998
999
1000
                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),
1001
                ctypes.c_int(start_iteration),
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
                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:
1015
            return inner_predict_sparse(csr, start_iteration, num_iteration, predict_type)
1016
1017
1018
1019
        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
1020
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
1021
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
1022
            preds = np.empty(sum(n_preds), dtype=np.float64)
1023
1024
            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:])):
1025
                # avoid memory consumption by arrays concatenation operations
1026
                inner_predict(csr[start_idx:end_idx], start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
1027
1028
            return preds, nrow
        else:
1029
            return inner_predict(csr, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
1030

1031
    def __pred_for_csc(self, csc, start_iteration, num_iteration, predict_type):
1032
        """Predict for a CSC data."""
1033
        def inner_predict_sparse(csc, start_iteration, num_iteration, predict_type):
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
            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)()
1047
            out_shape = np.empty(2, dtype=np.int64)
1048
1049
1050
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
1051
                ctypes.c_int(type_ptr_indptr),
1052
1053
1054
1055
1056
1057
1058
                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),
1059
                ctypes.c_int(start_iteration),
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
                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
1072
        nrow = csc.shape[0]
1073
        if nrow > MAX_INT32:
1074
            return self.__pred_for_csr(csc.tocsr(), start_iteration, num_iteration, predict_type)
1075
        if predict_type == C_API_PREDICT_CONTRIB:
1076
1077
            return inner_predict_sparse(csc, start_iteration, num_iteration, predict_type)
        n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
1078
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1079
1080
        out_num_preds = ctypes.c_int64(0)

1081
1082
        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
1083

1084
        assert csc.shape[0] <= MAX_INT32
1085
        csc_indices = csc.indices.astype(np.int32, copy=False)
1086

Guolin Ke's avatar
Guolin Ke committed
1087
1088
1089
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
1090
            ctypes.c_int(type_ptr_indptr),
1091
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1092
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1093
1094
1095
1096
1097
            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),
1098
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1099
            ctypes.c_int(num_iteration),
1100
            c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1101
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1102
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1103
        if n_preds != out_num_preds.value:
1104
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1105
1106
        return preds, nrow

1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
    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
1121

1122
class Dataset:
wxchan's avatar
wxchan committed
1123
    """Dataset in LightGBM."""
1124

1125
    def __init__(self, data, label=None, reference=None,
1126
                 weight=None, group=None, init_score=None,
1127
                 feature_name='auto', categorical_feature='auto', params=None,
wxchan's avatar
wxchan committed
1128
                 free_raw_data=True):
1129
        """Initialize Dataset.
1130

wxchan's avatar
wxchan committed
1131
1132
        Parameters
        ----------
1133
        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
1134
            Data source of Dataset.
1135
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1136
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1137
1138
1139
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1140
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1141
            Weight for each instance. Weights should be non-negative.
1142
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1143
1144
1145
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1146
1147
            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.
1148
        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)
1149
            Init score for Dataset.
1150
        feature_name : list of str, or 'auto', optional (default="auto")
1151
1152
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
1153
        categorical_feature : list of str or int, or 'auto', optional (default="auto")
1154
1155
            Categorical features.
            If list of int, interpreted as indices.
1156
            If list of str, interpreted as feature names (need to specify ``feature_name`` as well).
1157
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1158
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
1159
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1160
            All negative values in categorical features will be treated as missing values.
1161
            The output cannot be monotonically constrained with respect to a categorical feature.
1162
            Floating point numbers in categorical features will be rounded towards 0.
Nikita Titov's avatar
Nikita Titov committed
1163
        params : dict or None, optional (default=None)
1164
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1165
        free_raw_data : bool, optional (default=True)
1166
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
1167
        """
wxchan's avatar
wxchan committed
1168
1169
1170
1171
1172
1173
        self.handle = None
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1174
        self.init_score = init_score
wxchan's avatar
wxchan committed
1175
        self.feature_name = feature_name
1176
        self.categorical_feature = categorical_feature
1177
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1178
1179
        self.free_raw_data = free_raw_data
        self.used_indices = None
1180
        self.need_slice = True
wxchan's avatar
wxchan committed
1181
        self._predictor = None
1182
        self.pandas_categorical = None
1183
        self.params_back_up = None
1184
1185
        self.feature_penalty = None
        self.monotone_constraints = None
1186
        self.version = 0
1187
        self._start_row = 0  # Used when pushing rows one by one.
wxchan's avatar
wxchan committed
1188
1189

    def __del__(self):
1190
1191
1192
1193
        try:
            self._free_handle()
        except AttributeError:
            pass
1194

1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
    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),
        ))
1224
1225
        assert sample_cnt == actual_sample_cnt.value
        return indices
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260

    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
        ----------
1261
        sample_data : list of numpy array
1262
            Sample data for each column.
1263
        sample_indices : list of numpy array
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
1293
1294
1295
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
            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

1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
    def get_params(self):
        """Get the used parameters in the Dataset.

        Returns
        -------
        params : dict or None
            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",
1360
                                                "linear_tree",
1361
1362
1363
1364
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1365
                                                "precise_float_parser",
1366
1367
1368
1369
1370
1371
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}

1372
    def _free_handle(self):
1373
        if self.handle is not None:
1374
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1375
            self.handle = None
Guolin Ke's avatar
Guolin Ke committed
1376
1377
1378
        self.need_slice = True
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1379
        return self
wxchan's avatar
wxchan committed
1380

Guolin Ke's avatar
Guolin Ke committed
1381
1382
    def _set_init_score_by_predictor(self, predictor, data, used_indices=None):
        data_has_header = False
1383
        if isinstance(data, (str, Path)):
Guolin Ke's avatar
Guolin Ke committed
1384
            # check data has header or not
1385
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1386
        num_data = self.num_data()
1387
1388
1389
1390
1391
1392
1393
        if predictor is not None:
            init_score = predictor.predict(data,
                                           raw_score=True,
                                           data_has_header=data_has_header,
                                           is_reshape=False)
            if used_indices is not None:
                assert not self.need_slice
1394
                if isinstance(data, (str, Path)):
1395
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float64)
1396
                    assert num_data == len(used_indices)
1397
1398
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1399
1400
1401
1402
                            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
1403
                new_init_score = np.empty(init_score.size, dtype=np.float64)
1404
1405
                for i in range(num_data):
                    for j in range(predictor.num_class):
1406
1407
1408
                        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:
1409
            init_score = np.zeros(self.init_score.shape, dtype=np.float64)
1410
1411
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1412
1413
        self.set_init_score(init_score)

1414
    def _lazy_init(self, data, label=None, reference=None,
1415
                   weight=None, group=None, init_score=None, predictor=None,
1416
                   feature_name='auto', categorical_feature='auto', params=None):
wxchan's avatar
wxchan committed
1417
1418
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1419
            return self
Guolin Ke's avatar
Guolin Ke committed
1420
1421
1422
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1423
1424
1425
1426
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
wxchan's avatar
wxchan committed
1427
        label = _label_from_pandas(label)
Guolin Ke's avatar
Guolin Ke committed
1428

1429
        # process for args
wxchan's avatar
wxchan committed
1430
        params = {} if params is None else params
1431
1432
1433
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1434
        for key in params.keys():
1435
            if key in args_names:
1436
1437
                _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.')
1438
        # get categorical features
1439
1440
1441
1442
1443
1444
        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:
1445
                if isinstance(name, str) and name in feature_dict:
1446
                    categorical_indices.add(feature_dict[name])
1447
                elif isinstance(name, int):
1448
1449
                    categorical_indices.add(name)
                else:
1450
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1451
            if categorical_indices:
1452
1453
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1454
1455
1456
                        # 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.')
1457
                        params.pop(cat_alias, None)
1458
                params['categorical_column'] = sorted(categorical_indices)
1459

wxchan's avatar
wxchan committed
1460
        params_str = param_dict_to_str(params)
1461
        self.params = params
1462
        # process for reference dataset
wxchan's avatar
wxchan committed
1463
        ref_dataset = None
wxchan's avatar
wxchan committed
1464
        if isinstance(reference, Dataset):
1465
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1466
1467
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1468
        # start construct data
1469
        if isinstance(data, (str, Path)):
wxchan's avatar
wxchan committed
1470
1471
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
1472
                c_str(str(data)),
wxchan's avatar
wxchan committed
1473
1474
1475
1476
1477
                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
1478
1479
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1480
1481
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1482
1483
1484
1485
1486
1487
1488
1489
1490
        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)
1491
        elif isinstance(data, dt_DataTable):
1492
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1493
1494
1495
1496
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1497
            except BaseException:
1498
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}')
wxchan's avatar
wxchan committed
1499
1500
1501
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1502
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1503
1504
1505
1506
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1507
1508
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1509
                _log_warning("The init_score will be overridden by the prediction of init_model.")
Guolin Ke's avatar
Guolin Ke committed
1510
            self._set_init_score_by_predictor(predictor, data)
1511
1512
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1513
        elif predictor is not None:
1514
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1515
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1516
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1517

1518
1519
    @staticmethod
    def _yield_row_from_seqlist(seqs: List[Sequence], indices: Iterable[int]):
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
        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.
1545
        sampled = np.array([row for row in self._yield_row_from_seqlist(seqs, indices)])
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
        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
1590
    def __init_from_np2d(self, mat, params_str, ref_dataset):
1591
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1592
1593
1594
1595
1596
1597
        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)
1598
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1599
1600
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1601
        ptr_data, type_ptr_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1602
1603
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1604
            ctypes.c_int(type_ptr_data),
1605
1606
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
Guolin Ke's avatar
Guolin Ke committed
1607
            ctypes.c_int(C_API_IS_ROW_MAJOR),
wxchan's avatar
wxchan committed
1608
1609
1610
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1611
        return self
wxchan's avatar
wxchan committed
1612

1613
    def __init_from_list_np2d(self, mats, params_str, ref_dataset):
1614
        """Initialize data from a list of 2-D numpy matrices."""
1615
        ncol = mats[0].shape[1]
1616
        nrow = np.empty((len(mats),), np.int32)
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
        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)
1636
            else:  # change non-float data to float data, need to copy
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
                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(
1648
            ctypes.c_int32(len(mats)),
1649
1650
1651
            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)),
1652
            ctypes.c_int32(ncol),
1653
1654
1655
1656
            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
1657
        return self
1658

wxchan's avatar
wxchan committed
1659
    def __init_from_csr(self, csr, params_str, ref_dataset):
1660
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
1661
        if len(csr.indices) != len(csr.data):
1662
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
wxchan's avatar
wxchan committed
1663
1664
        self.handle = ctypes.c_void_p()

1665
1666
        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
1667

1668
        assert csr.shape[1] <= MAX_INT32
1669
        csr_indices = csr.indices.astype(np.int32, copy=False)
1670

wxchan's avatar
wxchan committed
1671
1672
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1673
            ctypes.c_int(type_ptr_indptr),
1674
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
1675
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1676
1677
1678
1679
            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
1680
1681
1682
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1683
        return self
wxchan's avatar
wxchan committed
1684

Guolin Ke's avatar
Guolin Ke committed
1685
    def __init_from_csc(self, csc, params_str, ref_dataset):
1686
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
1687
        if len(csc.indices) != len(csc.data):
1688
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
Guolin Ke's avatar
Guolin Ke committed
1689
1690
        self.handle = ctypes.c_void_p()

1691
1692
        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
1693

1694
        assert csc.shape[0] <= MAX_INT32
1695
        csc_indices = csc.indices.astype(np.int32, copy=False)
1696

Guolin Ke's avatar
Guolin Ke committed
1697
1698
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1699
            ctypes.c_int(type_ptr_indptr),
1700
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1701
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1702
1703
1704
1705
            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
1706
1707
1708
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1709
        return self
Guolin Ke's avatar
Guolin Ke committed
1710

1711
    @staticmethod
1712
1713
1714
1715
1716
1717
    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.
1718

1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
        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.
1729
1730
1731

        Returns
        -------
1732
1733
        compare_result : bool
          Returns whether two dictionaries with params are equal.
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
        """
        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
1749
    def construct(self):
1750
1751
1752
1753
1754
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
1755
            Constructed Dataset object.
1756
        """
1757
        if self.handle is None:
wxchan's avatar
wxchan committed
1758
            if self.reference is not None:
1759
                reference_params = self.reference.get_params()
1760
1761
                params = self.get_params()
                if params != reference_params:
1762
1763
1764
1765
1766
                    if not self._compare_params_for_warning(
                        params=params,
                        other_params=reference_params,
                        ignore_keys=_ConfigAliases.get("categorical_feature")
                    ):
1767
                        _log_warning('Overriding the parameters from Reference Dataset.')
1768
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
1769
                if self.used_indices is None:
1770
                    # create valid
1771
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
1772
1773
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
1774
                                    feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
1775
                else:
1776
                    # construct subset
wxchan's avatar
wxchan committed
1777
                    used_indices = list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
1778
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
1779
                    if self.reference.group is not None:
1780
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
1781
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
1782
                                                  return_counts=True)
1783
                    self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1784
1785
                    params_str = param_dict_to_str(self.params)
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
1786
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
1787
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
1788
                        ctypes.c_int32(used_indices.shape[0]),
wxchan's avatar
wxchan committed
1789
1790
                        c_str(params_str),
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
1791
1792
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
1793
1794
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
1795
1796
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
1797
1798
1799
                    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
1800
            else:
1801
                # create train
1802
                self._lazy_init(self.data, label=self.label,
1803
1804
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
1805
                                feature_name=self.feature_name, categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
1806
1807
1808
            if self.free_raw_data:
                self.data = None
        return self
wxchan's avatar
wxchan committed
1809

1810
    def create_valid(self, data, label=None, weight=None, group=None, init_score=None, params=None):
1811
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
1812
1813
1814

        Parameters
        ----------
1815
        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
1816
            Data source of Dataset.
1817
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM) or a LightGBM Dataset binary file.
1818
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1819
1820
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
1821
            Weight for each instance. Weights should be non-negative.
1822
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1823
1824
1825
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1826
1827
            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.
1828
        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)
1829
            Init score for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1830
        params : dict or None, optional (default=None)
1831
            Other parameters for validation Dataset.
1832
1833
1834

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1835
1836
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
1837
        """
1838
        ret = Dataset(data, label=label, reference=self,
1839
                      weight=weight, group=group, init_score=init_score,
1840
                      params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1841
        ret._predictor = self._predictor
1842
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
1843
        return ret
wxchan's avatar
wxchan committed
1844

wxchan's avatar
wxchan committed
1845
    def subset(self, used_indices, params=None):
1846
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
1847
1848
1849
1850

        Parameters
        ----------
        used_indices : list of int
1851
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
1852
        params : dict or None, optional (default=None)
1853
            These parameters will be passed to Dataset constructor.
1854
1855
1856
1857
1858

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
1859
        """
wxchan's avatar
wxchan committed
1860
1861
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
1862
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
1863
1864
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1865
        ret._predictor = self._predictor
1866
        ret.pandas_categorical = self.pandas_categorical
1867
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
1868
1869
1870
        return ret

    def save_binary(self, filename):
1871
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
1872

1873
1874
1875
1876
1877
        .. 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
1878
1879
        Parameters
        ----------
1880
        filename : str or pathlib.Path
wxchan's avatar
wxchan committed
1881
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
1882
1883
1884
1885
1886

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
1887
1888
1889
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
1890
            c_str(str(filename))))
Nikita Titov's avatar
Nikita Titov committed
1891
        return self
wxchan's avatar
wxchan committed
1892
1893

    def _update_params(self, params):
1894
1895
        if not params:
            return self
1896
        params = deepcopy(params)
1897
1898
1899
1900
1901

        def update():
            if not self.params:
                self.params = params
            else:
1902
                self.params_back_up = deepcopy(self.params)
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
                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:
1917
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
1918
        return self
wxchan's avatar
wxchan committed
1919

1920
    def _reverse_update_params(self):
1921
        if self.handle is None:
1922
            self.params = deepcopy(self.params_back_up)
1923
            self.params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
1924
        return self
1925

wxchan's avatar
wxchan committed
1926
    def set_field(self, field_name, data):
wxchan's avatar
wxchan committed
1927
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
1928
1929
1930

        Parameters
        ----------
1931
        field_name : str
1932
            The field name of the information.
1933
1934
        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
1935
1936
1937
1938
1939

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
1940
        """
1941
        if self.handle is None:
1942
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
1943
        if data is None:
1944
            # set to None
wxchan's avatar
wxchan committed
1945
1946
1947
1948
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
                c_str(field_name),
                None,
Guolin Ke's avatar
Guolin Ke committed
1949
1950
                ctypes.c_int(0),
                ctypes.c_int(FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
1951
            return self
1952
        if field_name == 'init_score':
Guolin Ke's avatar
Guolin Ke committed
1953
            dtype = np.float64
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
            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)

1968
1969
        if data.dtype == np.float32 or data.dtype == np.float64:
            ptr_data, type_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1970
        elif data.dtype == np.int32:
1971
            ptr_data, type_data, _ = c_int_array(data)
wxchan's avatar
wxchan committed
1972
        else:
1973
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
wxchan's avatar
wxchan committed
1974
        if type_data != FIELD_TYPE_MAPPER[field_name]:
1975
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
1976
1977
1978
1979
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
            c_str(field_name),
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1980
1981
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
1982
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
1983
        return self
wxchan's avatar
wxchan committed
1984

wxchan's avatar
wxchan committed
1985
1986
    def get_field(self, field_name):
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
1987
1988
1989

        Parameters
        ----------
1990
        field_name : str
1991
            The field name of the information.
wxchan's avatar
wxchan committed
1992
1993
1994

        Returns
        -------
1995
        info : numpy array or None
1996
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1997
        """
1998
        if self.handle is None:
1999
            raise Exception(f"Cannot get {field_name} before construct Dataset")
2000
2001
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
        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:
2014
            arr = cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
wxchan's avatar
wxchan committed
2015
        elif out_type.value == C_API_DTYPE_FLOAT32:
2016
            arr = cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
Guolin Ke's avatar
Guolin Ke committed
2017
        elif out_type.value == C_API_DTYPE_FLOAT64:
2018
            arr = cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
2019
        else:
wxchan's avatar
wxchan committed
2020
            raise TypeError("Unknown type")
2021
2022
2023
2024
2025
2026
        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
2027

2028
    def set_categorical_feature(self, categorical_feature):
2029
        """Set categorical features.
2030
2031
2032

        Parameters
        ----------
2033
        categorical_feature : list of int or str
2034
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
2035
2036
2037
2038
2039

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
2040
2041
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
2042
            return self
2043
        if self.data is not None:
2044
2045
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2046
                return self._free_handle()
2047
            elif categorical_feature == 'auto':
Nikita Titov's avatar
Nikita Titov committed
2048
                return self
2049
            else:
2050
2051
2052
                if self.categorical_feature != 'auto':
                    _log_warning('categorical_feature in Dataset is overridden.\n'
                                 f'New categorical_feature is {sorted(list(categorical_feature))}')
2053
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
2054
                return self._free_handle()
2055
        else:
2056
2057
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2058

Guolin Ke's avatar
Guolin Ke committed
2059
    def _set_predictor(self, predictor):
2060
2061
2062
2063
        """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
2064
        """
2065
        if predictor is self._predictor and (predictor is None or predictor.current_iteration() == self._predictor.current_iteration()):
Nikita Titov's avatar
Nikita Titov committed
2066
            return self
2067
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
2068
            self._predictor = predictor
2069
2070
2071
2072
2073
2074
        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
2075
        else:
2076
2077
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
2078
        return self
Guolin Ke's avatar
Guolin Ke committed
2079
2080

    def set_reference(self, reference):
2081
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
2082
2083
2084
2085

        Parameters
        ----------
        reference : Dataset
2086
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
2087
2088
2089
2090
2091

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
2092
        """
2093
2094
2095
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
2096
        # we're done if self and reference share a common upstream reference
2097
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
2098
            return self
Guolin Ke's avatar
Guolin Ke committed
2099
2100
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
2101
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
2102
        else:
2103
2104
            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
2105
2106

    def set_feature_name(self, feature_name):
2107
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
2108
2109
2110

        Parameters
        ----------
2111
        feature_name : list of str
2112
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
2113
2114
2115
2116
2117

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
2118
        """
2119
2120
        if feature_name != 'auto':
            self.feature_name = feature_name
2121
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
2122
            if len(feature_name) != self.num_feature():
2123
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
2124
            c_feature_name = [c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
2125
2126
2127
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
                c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
2128
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
2129
        return self
Guolin Ke's avatar
Guolin Ke committed
2130
2131

    def set_label(self, label):
2132
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
2133
2134
2135

        Parameters
        ----------
2136
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
2137
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
2138
2139
2140
2141
2142

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
2143
2144
        """
        self.label = label
2145
        if self.handle is not None:
2146
            label = list_to_1d_numpy(_label_from_pandas(label), name='label')
wxchan's avatar
wxchan committed
2147
            self.set_field('label', label)
2148
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2149
        return self
Guolin Ke's avatar
Guolin Ke committed
2150
2151

    def set_weight(self, weight):
2152
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
2153
2154
2155

        Parameters
        ----------
2156
        weight : list, numpy 1-D array, pandas Series or None
2157
            Weight to be set for each data point. Weights should be non-negative.
Nikita Titov's avatar
Nikita Titov committed
2158
2159
2160
2161
2162

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
2163
        """
2164
2165
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
2166
        self.weight = weight
2167
        if self.handle is not None and weight is not None:
wxchan's avatar
wxchan committed
2168
2169
            weight = list_to_1d_numpy(weight, name='weight')
            self.set_field('weight', weight)
2170
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2171
        return self
Guolin Ke's avatar
Guolin Ke committed
2172
2173

    def set_init_score(self, init_score):
2174
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
2175
2176
2177

        Parameters
        ----------
2178
        init_score : list, list of lists (for multi-class task), numpy array, pandas Series, pandas DataFrame (for multi-class task), or None
2179
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
2180
2181
2182
2183
2184

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
2185
2186
        """
        self.init_score = init_score
2187
        if self.handle is not None and init_score is not None:
wxchan's avatar
wxchan committed
2188
            self.set_field('init_score', init_score)
2189
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
2190
        return self
Guolin Ke's avatar
Guolin Ke committed
2191
2192

    def set_group(self, group):
2193
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
2194
2195
2196

        Parameters
        ----------
2197
        group : list, numpy 1-D array, pandas Series or None
2198
2199
2200
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2201
2202
            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
2203
2204
2205
2206
2207

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
2208
2209
        """
        self.group = group
2210
        if self.handle is not None and group is not None:
wxchan's avatar
wxchan committed
2211
2212
            group = list_to_1d_numpy(group, np.int32, name='group')
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
2213
        return self
Guolin Ke's avatar
Guolin Ke committed
2214

2215
2216
2217
2218
2219
    def get_feature_name(self):
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
2220
        feature_names : list of str
2221
2222
2223
2224
2225
2226
2227
2228
            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)
2229
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
2230
2231
2232
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
2233
            ctypes.c_int(num_feature),
2234
            ctypes.byref(tmp_out_len),
2235
            ctypes.c_size_t(reserved_string_buffer_size),
2236
2237
2238
2239
            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")
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
        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))
2252
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
2253

Guolin Ke's avatar
Guolin Ke committed
2254
    def get_label(self):
2255
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2256
2257
2258

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2259
        label : numpy array or None
2260
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2261
        """
2262
        if self.label is None:
wxchan's avatar
wxchan committed
2263
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
2264
2265
2266
        return self.label

    def get_weight(self):
2267
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2268
2269
2270

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2271
        weight : numpy array or None
2272
            Weight for each data point from the Dataset. Weights should be non-negative.
Guolin Ke's avatar
Guolin Ke committed
2273
        """
2274
        if self.weight is None:
wxchan's avatar
wxchan committed
2275
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
2276
2277
2278
        return self.weight

    def get_init_score(self):
2279
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2280
2281
2282

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2283
        init_score : numpy array or None
2284
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
2285
        """
2286
        if self.init_score is None:
wxchan's avatar
wxchan committed
2287
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
2288
2289
        return self.init_score

2290
2291
2292
2293
2294
    def get_data(self):
        """Get the raw data of the Dataset.

        Returns
        -------
2295
        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
2296
2297
2298
2299
            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
2300
2301
2302
2303
2304
        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, :]
2305
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2306
                    self.data = self.data.iloc[self.used_indices].copy()
2307
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2308
                    self.data = self.data[self.used_indices, :]
2309
2310
2311
2312
                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
2313
                else:
2314
2315
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
2316
            self.need_slice = False
Guolin Ke's avatar
Guolin Ke committed
2317
2318
2319
        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.")
2320
2321
        return self.data

Guolin Ke's avatar
Guolin Ke committed
2322
    def get_group(self):
2323
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2324
2325
2326

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2327
        group : numpy array or None
2328
2329
2330
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
2331
2332
            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
2333
        """
2334
        if self.group is None:
wxchan's avatar
wxchan committed
2335
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
2336
2337
            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
2338
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
2339
2340
2341
        return self.group

    def num_data(self):
2342
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2343
2344
2345

        Returns
        -------
2346
2347
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2348
        """
2349
        if self.handle is not None:
2350
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2351
2352
2353
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2354
        else:
2355
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2356
2357

    def num_feature(self):
2358
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2359
2360
2361

        Returns
        -------
2362
2363
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2364
        """
2365
        if self.handle is not None:
2366
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2367
2368
2369
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2370
        else:
2371
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2372

2373
    def get_ref_chain(self, ref_limit=100):
2374
2375
2376
2377
2378
        """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.
2379
2380
2381
2382
2383

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2384
2385
2386

        Returns
        -------
2387
2388
2389
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2390
        head = self
2391
        ref_chain = set()
2392
2393
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2394
                ref_chain.add(head)
2395
2396
2397
2398
2399
2400
                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
2401
        return ref_chain
2402

2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
    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
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
        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()))
2431
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2432
                    self.data = np.hstack((self.data, other.data.values))
2433
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2434
2435
2436
2437
2438
2439
2440
                    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)
2441
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2442
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
2443
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2444
2445
2446
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
2447
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2448
2449
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
2450
2451
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
2452
                if isinstance(other.data, np.ndarray):
2453
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
2454
2455
                                       axis=1, ignore_index=True)
                elif scipy.sparse.issparse(other.data):
2456
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
2457
                                       axis=1, ignore_index=True)
2458
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2459
2460
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
2461
2462
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
2463
2464
2465
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
2466
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2467
                if isinstance(other.data, np.ndarray):
2468
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
Guolin Ke's avatar
Guolin Ke committed
2469
                elif scipy.sparse.issparse(other.data):
2470
2471
2472
2473
2474
                    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
2475
2476
2477
2478
2479
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
2480
2481
            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
2482
2483
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
2484
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
2485
        self.feature_name = self.get_feature_name()
2486
2487
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
2488
2489
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2490
2491
        return self

2492
    def _dump_text(self, filename):
2493
2494
2495
2496
2497
2498
        """Save Dataset to a text file.

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

        Parameters
        ----------
2499
        filename : str or pathlib.Path
2500
2501
2502
2503
2504
2505
2506
2507
2508
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
            self.construct().handle,
2509
            c_str(str(filename))))
2510
2511
        return self

wxchan's avatar
wxchan committed
2512

2513
class Booster:
2514
    """Booster in LightGBM."""
2515

2516
    def __init__(self, params=None, train_set=None, model_file=None, model_str=None):
2517
        """Initialize the Booster.
wxchan's avatar
wxchan committed
2518
2519
2520

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2521
        params : dict or None, optional (default=None)
2522
2523
2524
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
2525
        model_file : str, pathlib.Path or None, optional (default=None)
wxchan's avatar
wxchan committed
2526
            Path to the model file.
2527
        model_str : str or None, optional (default=None)
2528
            Model will be loaded from this string.
wxchan's avatar
wxchan committed
2529
        """
2530
        self.handle = None
2531
        self.network = False
wxchan's avatar
wxchan committed
2532
        self.__need_reload_eval_info = True
2533
        self._train_data_name = "training"
wxchan's avatar
wxchan committed
2534
        self.__attr = {}
2535
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
2536
        self.best_iteration = -1
wxchan's avatar
wxchan committed
2537
        self.best_score = {}
2538
        params = {} if params is None else deepcopy(params)
wxchan's avatar
wxchan committed
2539
        if train_set is not None:
2540
            # Training task
wxchan's avatar
wxchan committed
2541
            if not isinstance(train_set, Dataset):
2542
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
            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"]
                )
2577
            # construct booster object
2578
2579
2580
2581
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
            params_str = param_dict_to_str(params)
2582
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2583
            _safe_call(_LIB.LGBM_BoosterCreate(
2584
                train_set.handle,
wxchan's avatar
wxchan committed
2585
2586
                c_str(params_str),
                ctypes.byref(self.handle)))
2587
            # save reference to data
wxchan's avatar
wxchan committed
2588
2589
2590
2591
            self.train_set = train_set
            self.valid_sets = []
            self.name_valid_sets = []
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
2592
2593
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
2594
2595
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
2596
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
2597
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2598
2599
2600
2601
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2602
            # buffer for inner predict
wxchan's avatar
wxchan committed
2603
2604
2605
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
2606
            self.pandas_categorical = train_set.pandas_categorical
2607
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
2608
        elif model_file is not None:
2609
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
2610
            out_num_iterations = ctypes.c_int(0)
2611
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2612
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
2613
                c_str(str(model_file)),
wxchan's avatar
wxchan committed
2614
2615
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2616
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2617
2618
2619
2620
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2621
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
2622
        elif model_str is not None:
2623
            self.model_from_string(model_str)
wxchan's avatar
wxchan committed
2624
        else:
2625
2626
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
2627
        self.params = params
wxchan's avatar
wxchan committed
2628
2629

    def __del__(self):
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
        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
2640

wxchan's avatar
wxchan committed
2641
2642
2643
2644
    def __copy__(self):
        return self.__deepcopy__(None)

    def __deepcopy__(self, _):
2645
        model_str = self.model_to_string(num_iteration=-1)
2646
        booster = Booster(model_str=model_str)
2647
        return booster
wxchan's avatar
wxchan committed
2648
2649
2650
2651
2652
2653
2654

    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:
2655
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
2656
2657
2658
        return this

    def __setstate__(self, state):
2659
2660
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
2661
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2662
            out_num_iterations = ctypes.c_int(0)
2663
2664
2665
2666
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
                c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
2667
2668
2669
            state['handle'] = handle
        self.__dict__.update(state)

wxchan's avatar
wxchan committed
2670
    def free_dataset(self):
Nikita Titov's avatar
Nikita Titov committed
2671
2672
2673
2674
2675
2676
2677
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
2678
2679
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
2680
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
2681
        return self
wxchan's avatar
wxchan committed
2682

2683
2684
2685
    def _free_buffer(self):
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
2686
        return self
2687

2688
2689
2690
2691
2692
2693
2694
    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":
2695
2696
2697
2698
        """Set the network configuration.

        Parameters
        ----------
2699
        machines : list, set or str
2700
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
2701
        local_listen_port : int, optional (default=12400)
2702
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
2703
        listen_time_out : int, optional (default=120)
2704
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
2705
        num_machines : int, optional (default=1)
2706
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
2707
2708
2709
2710
2711

        Returns
        -------
        self : Booster
            Booster with set network.
2712
        """
2713
2714
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
2715
2716
2717
2718
2719
        _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
2720
        return self
2721
2722

    def free_network(self):
Nikita Titov's avatar
Nikita Titov committed
2723
2724
2725
2726
2727
2728
2729
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
2730
2731
        _safe_call(_LIB.LGBM_NetworkFree())
        self.network = False
Nikita Titov's avatar
Nikita Titov committed
2732
        return self
2733

2734
2735
2736
    def trees_to_dataframe(self):
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

2737
2738
2739
2740
        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.
2741
2742
2743
2744
2745
            - ``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.
2746
2747
            - ``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.
2748
            - ``decision_type`` : str, logical operator describing how to compare a value to ``threshold``.
2749
2750
              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.
2751
2752
            - ``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.
2753
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
2754
            - ``weight`` : float64 or int64, sum of Hessian (second-order derivative of objective), summed over observations that fall in this node.
2755
2756
            - ``count`` : int64, number of records in the training data that fall into this node.

2757
2758
2759
2760
2761
2762
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
2763
2764
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775

        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):
2776
                tree_num = f'{tree_index}-' if tree_index is not None else ''
2777
2778
2779
                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
2780
2781
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793

            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):
2794
                return set(tree.keys()) == {'leaf_value'}
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867

            # 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))

2868
        return pd_DataFrame(model_list, columns=model_list[0].keys())
2869

wxchan's avatar
wxchan committed
2870
    def set_train_data_name(self, name):
2871
2872
2873
2874
        """Set the name to the training Dataset.

        Parameters
        ----------
2875
        name : str
Nikita Titov's avatar
Nikita Titov committed
2876
2877
2878
2879
2880
2881
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
2882
        """
2883
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
2884
        return self
wxchan's avatar
wxchan committed
2885
2886

    def add_valid(self, data, name):
2887
        """Add validation data.
wxchan's avatar
wxchan committed
2888
2889
2890
2891

        Parameters
        ----------
        data : Dataset
2892
            Validation data.
2893
        name : str
2894
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
2895
2896
2897
2898
2899

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
2900
        """
Guolin Ke's avatar
Guolin Ke committed
2901
        if not isinstance(data, Dataset):
2902
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2903
        if data._predictor is not self.__init_predictor:
2904
2905
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2906
2907
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
2908
            data.construct().handle))
wxchan's avatar
wxchan committed
2909
2910
2911
2912
2913
        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
2914
        return self
wxchan's avatar
wxchan committed
2915
2916

    def reset_parameter(self, params):
2917
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
2918
2919
2920
2921

        Parameters
        ----------
        params : dict
2922
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
2923
2924
2925
2926
2927

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
2928
2929
2930
2931
2932
2933
        """
        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
2934
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
2935
        return self
wxchan's avatar
wxchan committed
2936
2937

    def update(self, train_set=None, fobj=None):
Nikita Titov's avatar
Nikita Titov committed
2938
        """Update Booster for one iteration.
2939

wxchan's avatar
wxchan committed
2940
2941
        Parameters
        ----------
2942
2943
2944
2945
        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
2946
            Customized objective function.
2947
2948
2949
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

2950
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
2951
                    The predicted values.
2952
2953
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
2954
2955
                train_data : Dataset
                    The training dataset.
2956
                grad : numpy 1-D array or numpy 2-D array (for multi-class task)
2957
2958
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
2959
                hess : numpy 1-D array or numpy 2-D array (for multi-class task)
2960
2961
                    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
2962

2963
            For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes],
2964
            and grad and hess should be returned in the same format.
2965

wxchan's avatar
wxchan committed
2966
2967
        Returns
        -------
2968
2969
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
2970
        """
2971
        # need reset training data
2972
2973
2974
2975
2976
2977
        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
2978
            if not isinstance(train_set, Dataset):
2979
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2980
            if train_set._predictor is not self.__init_predictor:
2981
2982
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2983
2984
2985
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
2986
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
2987
            self.__inner_predict_buffer[0] = None
2988
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
2989
2990
        is_finished = ctypes.c_int(0)
        if fobj is None:
2991
            if self.__set_objective_to_none:
2992
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
2993
2994
2995
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
2996
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
2997
2998
            return is_finished.value == 1
        else:
2999
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
3000
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
3001
3002
3003
3004
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

    def __boost(self, grad, hess):
3005
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
3006

Nikita Titov's avatar
Nikita Titov committed
3007
3008
        .. note::

3009
3010
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
3011
            For multi-class task, score are numpy 2-D array of shape = [n_samples, n_classes],
3012
            and grad and hess should be returned in the same format.
3013

wxchan's avatar
wxchan committed
3014
3015
        Parameters
        ----------
3016
        grad : numpy 1-D array or numpy 2-D array (for multi-class task)
3017
3018
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
3019
        hess : numpy 1-D array or numpy 2-D array (for multi-class task)
3020
3021
            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
3022
3023
3024

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3025
3026
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
3027
        """
3028
3029
3030
        if self.__num_class > 1:
            grad = grad.ravel(order='F')
            hess = hess.ravel(order='F')
3031
3032
        grad = list_to_1d_numpy(grad, name='gradient')
        hess = list_to_1d_numpy(hess, name='hessian')
3033
3034
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
3035
        if len(grad) != len(hess):
3036
3037
            raise ValueError(f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) don't match")
        num_train_data = self.train_set.num_data()
3038
        if len(grad) != num_train_data * self.__num_class:
3039
3040
3041
            raise ValueError(
                f"Lengths of gradient ({len(grad)}) and Hessian ({len(hess)}) "
                f"don't match training data length ({num_train_data}) * "
3042
                f"number of models per one iteration ({self.__num_class})"
3043
            )
wxchan's avatar
wxchan committed
3044
3045
3046
3047
3048
3049
        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)))
3050
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
3051
3052
3053
        return is_finished.value == 1

    def rollback_one_iter(self):
Nikita Titov's avatar
Nikita Titov committed
3054
3055
3056
3057
3058
3059
3060
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
3061
3062
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
3063
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
3064
        return self
wxchan's avatar
wxchan committed
3065
3066

    def current_iteration(self):
3067
3068
3069
3070
3071
3072
3073
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
3074
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3075
3076
3077
3078
3079
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
    def num_model_per_iteration(self):
        """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

    def num_trees(self):
        """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

3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
    def upper_bound(self):
        """Get upper bound value of a model.

        Returns
        -------
        upper_bound : double
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

    def lower_bound(self):
        """Get lower bound value of a model.

        Returns
        -------
        lower_bound : double
            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
3136
    def eval(self, data, name, feval=None):
3137
        """Evaluate for data.
wxchan's avatar
wxchan committed
3138
3139
3140

        Parameters
        ----------
3141
3142
        data : Dataset
            Data for the evaluating.
3143
        name : str
3144
            Name of the data.
3145
        feval : callable, list of callable, or None, optional (default=None)
3146
            Customized evaluation function.
3147
            Each evaluation function should accept two parameters: preds, eval_data,
3148
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3149

3150
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3151
                    The predicted values.
3152
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3153
3154
                    If ``fobj`` is specified, predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task in this case.
3155
                eval_data : Dataset
3156
                    A ``Dataset`` to evaluate.
3157
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3158
                    The name of evaluation function (without whitespace).
3159
3160
3161
3162
3163
                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
3164
3165
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3166
        result : list
3167
            List with evaluation results.
wxchan's avatar
wxchan committed
3168
        """
Guolin Ke's avatar
Guolin Ke committed
3169
3170
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
3171
3172
3173
3174
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
3175
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
3176
3177
3178
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
3179
        # need to push new valid data
wxchan's avatar
wxchan committed
3180
3181
3182
3183
3184
3185
3186
        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):
3187
        """Evaluate for training data.
wxchan's avatar
wxchan committed
3188
3189
3190

        Parameters
        ----------
3191
        feval : callable, list of callable, or None, optional (default=None)
3192
            Customized evaluation function.
3193
            Each evaluation function should accept two parameters: preds, eval_data,
3194
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3195

3196
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3197
                    The predicted values.
3198
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3199
3200
                    If ``fobj`` is specified, predicted values are returned before any transformation,
                    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
3201
                eval_data : Dataset
3202
                    The training dataset.
3203
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3204
                    The name of evaluation function (without whitespace).
3205
3206
3207
3208
3209
                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
3210
3211
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3212
        result : list
3213
            List with evaluation results.
wxchan's avatar
wxchan committed
3214
        """
3215
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
3216
3217

    def eval_valid(self, feval=None):
3218
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
3219
3220
3221

        Parameters
        ----------
3222
        feval : callable, list of callable, or None, optional (default=None)
3223
            Customized evaluation function.
3224
            Each evaluation function should accept two parameters: preds, eval_data,
3225
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
3226

3227
                preds : numpy 1-D array or numpy 2-D array (for multi-class task)
3228
                    The predicted values.
3229
                    For multi-class task, preds are numpy 2-D array of shape = [n_samples, n_classes].
3230
3231
                    If ``fobj`` is specified, predicted values are returned before any transformation,
                    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
3232
                eval_data : Dataset
3233
                    The validation dataset.
3234
                eval_name : str
Andrew Ziem's avatar
Andrew Ziem committed
3235
                    The name of evaluation function (without whitespace).
3236
3237
3238
3239
3240
                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
3241
3242
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3243
        result : list
3244
            List with evaluation results.
wxchan's avatar
wxchan committed
3245
        """
3246
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
3247
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
3248

3249
    def save_model(self, filename, num_iteration=None, start_iteration=0, importance_type='split'):
3250
        """Save Booster to file.
wxchan's avatar
wxchan committed
3251
3252
3253

        Parameters
        ----------
3254
        filename : str or pathlib.Path
3255
            Filename to save Booster.
3256
3257
3258
3259
        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
3260
        start_iteration : int, optional (default=0)
3261
            Start index of the iteration that should be saved.
3262
        importance_type : str, optional (default="split")
3263
3264
3265
            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
3266
3267
3268
3269
3270

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
3271
        """
3272
        if num_iteration is None:
3273
            num_iteration = self.best_iteration
3274
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3275
3276
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
3277
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3278
            ctypes.c_int(num_iteration),
3279
            ctypes.c_int(importance_type_int),
3280
            c_str(str(filename))))
3281
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
3282
        return self
wxchan's avatar
wxchan committed
3283

3284
    def shuffle_models(self, start_iteration=0, end_iteration=-1):
3285
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
3286

3287
3288
3289
        Parameters
        ----------
        start_iteration : int, optional (default=0)
3290
            The first iteration that will be shuffled.
3291
3292
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
3293
            If <= 0, means the last available iteration.
3294

Nikita Titov's avatar
Nikita Titov committed
3295
3296
3297
3298
        Returns
        -------
        self : Booster
            Booster with shuffled models.
3299
        """
3300
3301
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
3302
3303
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
3304
        return self
3305

3306
    def model_from_string(self, model_str):
3307
3308
3309
3310
        """Load Booster from a string.

        Parameters
        ----------
3311
        model_str : str
3312
3313
3314
3315
            Model will be loaded from this string.

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3316
        self : Booster
3317
3318
            Loaded Booster object.
        """
3319
3320
3321
3322
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
        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
3333
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
3334
3335
        return self

3336
    def model_to_string(self, num_iteration=None, start_iteration=0, importance_type='split'):
3337
        """Save Booster to string.
3338

3339
3340
3341
3342
3343
3344
        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
3345
        start_iteration : int, optional (default=0)
3346
            Start index of the iteration that should be saved.
3347
        importance_type : str, optional (default="split")
3348
3349
3350
            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.
3351
3352
3353

        Returns
        -------
3354
        str_repr : str
3355
3356
            String representation of Booster.
        """
3357
        if num_iteration is None:
3358
            num_iteration = self.best_iteration
3359
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3360
        buffer_len = 1 << 20
3361
        tmp_out_len = ctypes.c_int64(0)
3362
3363
3364
3365
        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,
3366
            ctypes.c_int(start_iteration),
3367
            ctypes.c_int(num_iteration),
3368
            ctypes.c_int(importance_type_int),
3369
            ctypes.c_int64(buffer_len),
3370
3371
3372
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
3373
        # if buffer length is not long enough, re-allocate a buffer
3374
3375
3376
3377
3378
        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,
3379
                ctypes.c_int(start_iteration),
3380
                ctypes.c_int(num_iteration),
3381
                ctypes.c_int(importance_type_int),
3382
                ctypes.c_int64(actual_len),
3383
3384
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
3385
        ret = string_buffer.value.decode('utf-8')
3386
3387
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
3388

3389
    def dump_model(self, num_iteration=None, start_iteration=0, importance_type='split', object_hook=None):
Nikita Titov's avatar
Nikita Titov committed
3390
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
3391

3392
3393
        Parameters
        ----------
3394
3395
3396
3397
        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
3398
        start_iteration : int, optional (default=0)
3399
            Start index of the iteration that should be dumped.
3400
        importance_type : str, optional (default="split")
3401
3402
3403
            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.
3404
3405
3406
3407
3408
3409
3410
3411
3412
        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.
3413

wxchan's avatar
wxchan committed
3414
3415
        Returns
        -------
3416
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
3417
            JSON format of Booster.
wxchan's avatar
wxchan committed
3418
        """
3419
        if num_iteration is None:
3420
            num_iteration = self.best_iteration
3421
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3422
        buffer_len = 1 << 20
3423
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
3424
3425
3426
3427
        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,
3428
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3429
            ctypes.c_int(num_iteration),
3430
            ctypes.c_int(importance_type_int),
3431
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
3432
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3433
            ptr_string_buffer))
wxchan's avatar
wxchan committed
3434
        actual_len = tmp_out_len.value
3435
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
3436
3437
3438
3439
3440
        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,
3441
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3442
                ctypes.c_int(num_iteration),
3443
                ctypes.c_int(importance_type_int),
3444
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
3445
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3446
                ptr_string_buffer))
3447
        ret = json.loads(string_buffer.value.decode('utf-8'), object_hook=object_hook)
3448
3449
3450
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
                                                          default=json_default_with_numpy))
        return ret
wxchan's avatar
wxchan committed
3451

3452
    def predict(self, data, start_iteration=0, num_iteration=None,
3453
                raw_score=False, pred_leaf=False, pred_contrib=False,
3454
                data_has_header=False, is_reshape=True, **kwargs):
3455
        """Make a prediction.
wxchan's avatar
wxchan committed
3456
3457
3458

        Parameters
        ----------
3459
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
3460
            Data source for prediction.
3461
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3462
        start_iteration : int, optional (default=0)
3463
            Start index of the iteration to predict.
3464
            If <= 0, starts from the first iteration.
3465
        num_iteration : int or None, optional (default=None)
3466
3467
3468
3469
            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).
3470
3471
3472
3473
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
3474
3475
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
3476

Nikita Titov's avatar
Nikita Titov committed
3477
3478
3479
3480
3481
3482
3483
            .. 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.
3484

3485
3486
        data_has_header : bool, optional (default=False)
            Whether the data has header.
3487
            Used only if data is str.
3488
3489
        is_reshape : bool, optional (default=True)
            If True, result is reshaped to [nrow, ncol].
3490
3491
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
3492
3493
3494

        Returns
        -------
3495
        result : numpy array, scipy.sparse or list of scipy.sparse
3496
            Prediction result.
3497
            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
3498
        """
3499
        predictor = self._to_predictor(deepcopy(kwargs))
3500
        if num_iteration is None:
3501
            if start_iteration <= 0:
3502
3503
3504
3505
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
        return predictor.predict(data, start_iteration, num_iteration,
3506
3507
                                 raw_score, pred_leaf, pred_contrib,
                                 data_has_header, is_reshape)
wxchan's avatar
wxchan committed
3508

3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
    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
3524
3525
3526
3527
        """Refit the existing Booster by new data.

        Parameters
        ----------
3528
        data : str, pathlib.Path, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
Guolin Ke's avatar
Guolin Ke committed
3529
            Data source for refit.
3530
            If str or pathlib.Path, it represents the path to a text file (CSV, TSV, or LibSVM).
3531
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
3532
3533
            Label for refit.
        decay_rate : float, optional (default=0.9)
3534
3535
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
3536
3537
3538
        reference : Dataset or None, optional (default=None)
            Reference for ``data``.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
3539
            Weight for each ``data`` instance. Weights should be non-negative.
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
        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.
3556
            All values in categorical features will be cast to int32 and thus should be less than int32 max value (2147483647).
3557
3558
3559
            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.
3560
            Floating point numbers in categorical features will be rounded towards 0.
3561
3562
3563
3564
        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``.
3565
3566
        **kwargs
            Other parameters for refit.
3567
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
3568
3569
3570
3571
3572
3573

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
3574
3575
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
3576
3577
        if dataset_params is None:
            dataset_params = {}
3578
        predictor = self._to_predictor(deepcopy(kwargs))
3579
        leaf_preds = predictor.predict(data, -1, pred_leaf=True)
3580
        nrow, ncol = leaf_preds.shape
3581
        out_is_linear = ctypes.c_int(0)
3582
3583
3584
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
3585
3586
3587
3588
3589
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
3590
        new_params["linear_tree"] = bool(out_is_linear.value)
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
        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,
        )
3604
        new_params['refit_decay_rate'] = decay_rate
3605
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
3606
3607
3608
3609
3610
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
3611
        ptr_data, _, _ = c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
3612
3613
3614
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
3615
3616
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
3617
3618
        new_booster.network = self.network
        new_booster.__attr = self.__attr.copy()
Guolin Ke's avatar
Guolin Ke committed
3619
3620
        return new_booster

3621
    def get_leaf_output(self, tree_id, leaf_id):
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
        """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.
        """
3636
3637
3638
3639
3640
3641
3642
3643
        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

3644
    def _to_predictor(self, pred_parameter=None):
3645
        """Convert to predictor."""
3646
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
3647
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
3648
3649
        return predictor

3650
    def num_feature(self):
3651
3652
3653
3654
3655
3656
3657
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
3658
3659
3660
3661
3662
3663
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
            self.handle,
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

wxchan's avatar
wxchan committed
3664
    def feature_name(self):
3665
        """Get names of features.
wxchan's avatar
wxchan committed
3666
3667
3668

        Returns
        -------
3669
        result : list of str
3670
            List with names of features.
wxchan's avatar
wxchan committed
3671
        """
3672
        num_feature = self.num_feature()
3673
        # Get name of features
wxchan's avatar
wxchan committed
3674
        tmp_out_len = ctypes.c_int(0)
3675
3676
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
3677
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
3678
3679
3680
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
3681
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
3682
            ctypes.byref(tmp_out_len),
3683
            ctypes.c_size_t(reserved_string_buffer_size),
3684
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3685
3686
3687
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
        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))
3700
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
3701

3702
    def feature_importance(self, importance_type='split', iteration=None):
3703
        """Get feature importances.
3704

3705
3706
        Parameters
        ----------
3707
        importance_type : str, optional (default="split")
3708
3709
3710
            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.
3711
3712
3713
3714
        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).
3715

3716
3717
        Returns
        -------
3718
3719
        result : numpy array
            Array with feature importances.
3720
        """
3721
3722
        if iteration is None:
            iteration = self.best_iteration
3723
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3724
        result = np.empty(self.num_feature(), dtype=np.float64)
3725
3726
3727
3728
3729
3730
        _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))))
        if importance_type_int == 0:
3731
            return result.astype(np.int32)
3732
3733
        else:
            return result
3734

3735
3736
3737
3738
3739
    def get_split_value_histogram(self, feature, bins=None, xgboost_style=False):
        """Get split value histogram for the specified feature.

        Parameters
        ----------
3740
        feature : int or str
3741
3742
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
3743
            If str, interpreted as name.
3744

Nikita Titov's avatar
Nikita Titov committed
3745
3746
3747
            .. warning::

                Categorical features are not supported.
3748

3749
        bins : int, str or None, optional (default=None)
3750
3751
3752
            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.
3753
            If str, it should be one from the list of the supported values by ``numpy.histogram()`` function.
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
        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
3771
                if feature_names is not None and isinstance(feature, str):
3772
3773
3774
3775
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
3776
                    if isinstance(root['threshold'], str):
3777
3778
3779
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
                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'])

3790
        if bins is None or isinstance(bins, int) and xgboost_style:
3791
3792
3793
3794
3795
3796
3797
            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:
3798
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
3799
3800
3801
3802
3803
            else:
                return ret
        else:
            return hist, bin_edges

wxchan's avatar
wxchan committed
3804
    def __inner_eval(self, data_name, data_idx, feval=None):
3805
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
3806
        if data_idx >= self.__num_dataset:
3807
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3808
3809
3810
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
3811
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
3812
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3813
3814
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3815
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3816
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3817
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
3818
            if tmp_out_len.value != self.__num_inner_eval:
3819
                raise ValueError("Wrong length of eval results")
3820
            for i in range(self.__num_inner_eval):
3821
3822
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
3823
3824
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
3825
3826
3827
3828
3829
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
3830
3831
3832
3833
3834
3835
3836
3837
3838
            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
3839
3840
3841
3842
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

    def __inner_predict(self, data_idx):
3843
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
3844
        if data_idx >= self.__num_dataset:
3845
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3846
3847
3848
3849
3850
        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
3851
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
3852
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
3853
3854
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
3855
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
3856
3857
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3858
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3859
3860
3861
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
3862
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
3863
            self.__is_predicted_cur_iter[data_idx] = True
3864
3865
3866
3867
3868
        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
3869
3870

    def __get_eval_info(self):
3871
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
3872
3873
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
3874
            out_num_eval = ctypes.c_int(0)
3875
            # Get num of inner evals
wxchan's avatar
wxchan committed
3876
3877
3878
3879
3880
            _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:
3881
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
3882
                tmp_out_len = ctypes.c_int(0)
3883
3884
3885
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
3886
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
3887
                ]
wxchan's avatar
wxchan committed
3888
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
3889
3890
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
3891
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
3892
                    ctypes.byref(tmp_out_len),
3893
                    ctypes.c_size_t(reserved_string_buffer_size),
3894
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3895
3896
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
3897
                    raise ValueError("Length of eval names doesn't equal with num_evals")
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
                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
                ]
3918

wxchan's avatar
wxchan committed
3919
    def attr(self, key):
3920
        """Get attribute string from the Booster.
wxchan's avatar
wxchan committed
3921
3922
3923

        Parameters
        ----------
3924
        key : str
3925
            The name of the attribute.
wxchan's avatar
wxchan committed
3926
3927
3928

        Returns
        -------
3929
        value : str or None
3930
            The attribute value.
Nikita Titov's avatar
Nikita Titov committed
3931
            Returns None if attribute does not exist.
wxchan's avatar
wxchan committed
3932
        """
3933
        return self.__attr.get(key, None)
wxchan's avatar
wxchan committed
3934
3935

    def set_attr(self, **kwargs):
3936
        """Set attributes to the Booster.
wxchan's avatar
wxchan committed
3937
3938
3939
3940

        Parameters
        ----------
        **kwargs
3941
3942
            The attributes to set.
            Setting a value to None deletes an attribute.
Nikita Titov's avatar
Nikita Titov committed
3943
3944
3945
3946

        Returns
        -------
        self : Booster
3947
            Booster with set attributes.
wxchan's avatar
wxchan committed
3948
3949
3950
        """
        for key, value in kwargs.items():
            if value is not None:
3951
                if not isinstance(value, str):
Nikita Titov's avatar
Nikita Titov committed
3952
                    raise ValueError("Only string values are accepted")
wxchan's avatar
wxchan committed
3953
3954
3955
                self.__attr[key] = value
            else:
                self.__attr.pop(key, None)
Nikita Titov's avatar
Nikita Titov committed
3956
        return self