basic.py 147 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
wxchan's avatar
wxchan committed
3
import ctypes
4
import json
5
import os
wxchan's avatar
wxchan committed
6
import warnings
7
from collections import OrderedDict
8
from copy import deepcopy
9
10
11
from functools import wraps
from logging import Logger
from tempfile import NamedTemporaryFile
12
from typing import Any, Dict, List, Set, Union
wxchan's avatar
wxchan committed
13
14
15
16

import numpy as np
import scipy.sparse

17
from .compat import PANDAS_INSTALLED, concat, dt_DataTable, is_dtype_sparse, pd_DataFrame, pd_Series
wxchan's avatar
wxchan committed
18
19
from .libpath import find_lib_path

wxchan's avatar
wxchan committed
20

21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
class _DummyLogger:
    def info(self, msg):
        print(msg)

    def warning(self, msg):
        warnings.warn(msg, stacklevel=3)


_LOGGER = _DummyLogger()


def register_logger(logger):
    """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


def _normalize_native_string(func):
    """Join log messages from native library which come by chunks."""
    msg_normalized = []

    @wraps(func)
    def wrapper(msg):
        nonlocal msg_normalized
        if msg.strip() == '':
            msg = ''.join(msg_normalized)
            msg_normalized = []
            return func(msg)
        else:
            msg_normalized.append(msg)

    return wrapper


def _log_info(msg):
    _LOGGER.info(msg)


def _log_warning(msg):
    _LOGGER.warning(msg)


@_normalize_native_string
def _log_native(msg):
    _LOGGER.info(msg)


76
def _log_callback(msg):
77
    """Redirect logs from native library into Python."""
78
    _log_native(str(msg.decode('utf-8')))
79
80


wxchan's avatar
wxchan committed
81
def _load_lib():
82
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
83
84
    lib_path = find_lib_path()
    if len(lib_path) == 0:
85
        return None
wxchan's avatar
wxchan committed
86
87
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
88
89
90
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
    lib.callback = callback(_log_callback)
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
91
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
92
93
    return lib

wxchan's avatar
wxchan committed
94

wxchan's avatar
wxchan committed
95
96
_LIB = _load_lib()

wxchan's avatar
wxchan committed
97

98
99
100
NUMERIC_TYPES = (int, float, bool)


wxchan's avatar
wxchan committed
101
def _safe_call(ret):
102
103
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
104
105
106
    Parameters
    ----------
    ret : int
107
        The return value from C API calls.
wxchan's avatar
wxchan committed
108
109
    """
    if ret != 0:
110
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
111

wxchan's avatar
wxchan committed
112

wxchan's avatar
wxchan committed
113
def is_numeric(obj):
114
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
115
116
117
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
118
119
120
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
121
122
        return False

wxchan's avatar
wxchan committed
123

wxchan's avatar
wxchan committed
124
def is_numpy_1d_array(data):
125
    """Check whether data is a numpy 1-D array."""
126
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
127

wxchan's avatar
wxchan committed
128

129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
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


def cast_numpy_1d_array_to_dtype(array, dtype):
    """Cast numpy 1d array to given dtype."""
    if array.dtype == dtype:
        return array
    return array.astype(dtype=dtype, copy=False)


wxchan's avatar
wxchan committed
144
def is_1d_list(data):
145
146
    """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
147

wxchan's avatar
wxchan committed
148

149
def list_to_1d_numpy(data, dtype=np.float32, name='list'):
150
    """Convert data to numpy 1-D array."""
wxchan's avatar
wxchan committed
151
    if is_numpy_1d_array(data):
152
153
154
155
156
        return cast_numpy_1d_array_to_dtype(data, dtype)
    elif is_numpy_column_array(data):
        _log_warning('Converting column-vector to 1d array')
        array = data.ravel()
        return cast_numpy_1d_array_to_dtype(array, dtype)
wxchan's avatar
wxchan committed
157
158
    elif is_1d_list(data):
        return np.array(data, dtype=dtype, copy=False)
159
    elif isinstance(data, pd_Series):
160
161
        if _get_bad_pandas_dtypes([data.dtypes]):
            raise ValueError('Series.dtypes must be int, float or bool')
162
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
163
    else:
164
165
        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
166

wxchan's avatar
wxchan committed
167

wxchan's avatar
wxchan committed
168
def cfloat32_array_to_numpy(cptr, length):
169
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
170
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
171
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
172
    else:
173
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
174

Guolin Ke's avatar
Guolin Ke committed
175

Guolin Ke's avatar
Guolin Ke committed
176
def cfloat64_array_to_numpy(cptr, length):
177
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
178
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
179
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
Guolin Ke's avatar
Guolin Ke committed
180
181
182
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
183

wxchan's avatar
wxchan committed
184
def cint32_array_to_numpy(cptr, length):
185
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
186
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
187
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
wxchan's avatar
wxchan committed
188
    else:
189
190
191
192
193
194
        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)):
195
        return np.ctypeslib.as_array(cptr, shape=(length,)).copy()
196
197
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
198

wxchan's avatar
wxchan committed
199

wxchan's avatar
wxchan committed
200
def c_str(string):
201
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
202
203
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
204

wxchan's avatar
wxchan committed
205
def c_array(ctype, values):
206
    """Convert a Python array to C array."""
wxchan's avatar
wxchan committed
207
208
    return (ctype * len(values))(*values)

wxchan's avatar
wxchan committed
209

210
211
212
213
214
215
216
217
218
219
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
220
def param_dict_to_str(data):
221
    """Convert Python dictionary to string, which is passed to C API."""
222
    if data is None or not data:
wxchan's avatar
wxchan committed
223
224
225
        return ""
    pairs = []
    for key, val in data.items():
226
        if isinstance(val, (list, tuple, set)) or is_numpy_1d_array(val):
227
228
            def to_string(x):
                if isinstance(x, list):
229
                    return f"[{','.join(map(str, x))}]"
230
231
                else:
                    return str(x)
232
            pairs.append(f"{key}={','.join(map(to_string, val))}")
233
        elif isinstance(val, (str, NUMERIC_TYPES)) or is_numeric(val):
234
            pairs.append(f"{key}={val}")
235
        elif val is not None:
236
            raise TypeError(f'Unknown type of parameter:{key}, got:{type(val).__name__}')
wxchan's avatar
wxchan committed
237
    return ' '.join(pairs)
238

wxchan's avatar
wxchan committed
239

240
class _TempFile:
241
242
243
244
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
        return self
wxchan's avatar
wxchan committed
245

246
247
248
    def __exit__(self, exc_type, exc_val, exc_tb):
        if os.path.isfile(self.name):
            os.remove(self.name)
wxchan's avatar
wxchan committed
249

250
251
252
253
    def readlines(self):
        with open(self.name, "r+") as f:
            ret = f.readlines()
        return ret
wxchan's avatar
wxchan committed
254

255
256
    def writelines(self, lines):
        with open(self.name, "w+") as f:
257
            f.writelines(lines)
258

wxchan's avatar
wxchan committed
259

260
class LightGBMError(Exception):
261
262
    """Error thrown by LightGBM."""

263
264
265
    pass


266
267
268
269
270
271
272
273
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
274
275
276
    aliases = {"bin_construct_sample_cnt": {"bin_construct_sample_cnt",
                                            "subsample_for_bin"},
               "boosting": {"boosting",
277
278
279
280
281
282
                            "boosting_type",
                            "boost"},
               "categorical_feature": {"categorical_feature",
                                       "cat_feature",
                                       "categorical_column",
                                       "cat_column"},
283
284
               "data_random_seed": {"data_random_seed",
                                    "data_seed"},
285
286
287
288
               "early_stopping_round": {"early_stopping_round",
                                        "early_stopping_rounds",
                                        "early_stopping",
                                        "n_iter_no_change"},
289
290
291
               "enable_bundle": {"enable_bundle",
                                 "is_enable_bundle",
                                 "bundle"},
292
293
294
295
296
               "eval_at": {"eval_at",
                           "ndcg_eval_at",
                           "ndcg_at",
                           "map_eval_at",
                           "map_at"},
297
298
299
300
301
302
               "group_column": {"group_column",
                                "group",
                                "group_id",
                                "query_column",
                                "query",
                                "query_id"},
303
304
               "header": {"header",
                          "has_header"},
305
306
307
308
309
310
311
312
313
               "ignore_column": {"ignore_column",
                                 "ignore_feature",
                                 "blacklist"},
               "is_enable_sparse": {"is_enable_sparse",
                                    "is_sparse",
                                    "enable_sparse",
                                    "sparse"},
               "label_column": {"label_column",
                                "label"},
Nikita Titov's avatar
Nikita Titov committed
314
315
               "linear_tree": {"linear_tree",
                               "linear_trees"},
316
317
318
               "local_listen_port": {"local_listen_port",
                                     "local_port",
                                     "port"},
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
               "machines": {"machines",
                            "workers",
                            "nodes"},
               "metric": {"metric",
                          "metrics",
                          "metric_types"},
               "num_class": {"num_class",
                             "num_classes"},
               "num_iterations": {"num_iterations",
                                  "num_iteration",
                                  "n_iter",
                                  "num_tree",
                                  "num_trees",
                                  "num_round",
                                  "num_rounds",
                                  "num_boost_round",
                                  "n_estimators"},
336
337
               "num_machines": {"num_machines",
                                "num_machine"},
338
339
340
341
342
               "num_threads": {"num_threads",
                               "num_thread",
                               "nthread",
                               "nthreads",
                               "n_jobs"},
343
344
345
346
               "objective": {"objective",
                             "objective_type",
                             "app",
                             "application"},
347
348
               "pre_partition": {"pre_partition",
                                 "is_pre_partition"},
349
350
351
352
               "tree_learner": {"tree_learner",
                                "tree",
                                "tree_type",
                                "tree_learner_type"},
353
354
355
               "two_round": {"two_round",
                             "two_round_loading",
                             "use_two_round_loading"},
356
               "verbosity": {"verbosity",
357
358
359
                             "verbose"},
               "weight_column": {"weight_column",
                                 "weight"}}
360
361
362
363
364

    @classmethod
    def get(cls, *args):
        ret = set()
        for i in args:
365
            ret |= cls.aliases.get(i, {i})
366
367
368
        return ret


369
370
371
372
373
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
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


409
410
MAX_INT32 = (1 << 31) - 1

411
"""Macro definition of data type in C API of LightGBM"""
wxchan's avatar
wxchan committed
412
413
414
415
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
416

417
"""Matrix is row major in Python"""
wxchan's avatar
wxchan committed
418
419
C_API_IS_ROW_MAJOR = 1

420
"""Macro definition of prediction type in C API of LightGBM"""
wxchan's avatar
wxchan committed
421
422
423
C_API_PREDICT_NORMAL = 0
C_API_PREDICT_RAW_SCORE = 1
C_API_PREDICT_LEAF_INDEX = 2
424
C_API_PREDICT_CONTRIB = 3
wxchan's avatar
wxchan committed
425

426
427
428
429
"""Macro definition of sparse matrix type"""
C_API_MATRIX_TYPE_CSR = 0
C_API_MATRIX_TYPE_CSC = 1

430
431
432
433
"""Macro definition of feature importance type"""
C_API_FEATURE_IMPORTANCE_SPLIT = 0
C_API_FEATURE_IMPORTANCE_GAIN = 1

434
"""Data type of data field"""
wxchan's avatar
wxchan committed
435
436
FIELD_TYPE_MAPPER = {"label": C_API_DTYPE_FLOAT32,
                     "weight": C_API_DTYPE_FLOAT32,
Guolin Ke's avatar
Guolin Ke committed
437
                     "init_score": C_API_DTYPE_FLOAT64,
438
                     "group": C_API_DTYPE_INT32}
wxchan's avatar
wxchan committed
439

440
441
442
443
"""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
444

445
def convert_from_sliced_object(data):
446
    """Fix the memory of multi-dimensional sliced object."""
447
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
448
        if not data.flags.c_contiguous:
449
450
            _log_warning("Usage of np.ndarray subset (sliced data) is not recommended "
                         "due to it will double the peak memory cost in LightGBM.")
451
452
453
454
            return np.copy(data)
    return data


wxchan's avatar
wxchan committed
455
def c_float_array(data):
456
    """Get pointer of float numpy array / list."""
wxchan's avatar
wxchan committed
457
458
459
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
460
461
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
462
463
464
465
466
467
468
        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:
469
            raise TypeError(f"Expected np.float32 or np.float64, met type({data.dtype})")
wxchan's avatar
wxchan committed
470
    else:
471
        raise TypeError(f"Unknown type({type(data).__name__})")
472
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
473

wxchan's avatar
wxchan committed
474

wxchan's avatar
wxchan committed
475
def c_int_array(data):
476
    """Get pointer of int numpy array / list."""
wxchan's avatar
wxchan committed
477
478
479
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
480
481
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
482
483
484
485
486
487
488
        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:
489
            raise TypeError(f"Expected np.int32 or np.int64, met type({data.dtype})")
wxchan's avatar
wxchan committed
490
    else:
491
        raise TypeError(f"Unknown type({type(data).__name__})")
492
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
493

wxchan's avatar
wxchan committed
494

495
496
497
498
499
500
501
502
503
504
505
def _get_bad_pandas_dtypes(dtypes):
    pandas_dtype_mapper = {'int8': 'int', 'int16': 'int', 'int32': 'int',
                           'int64': 'int', 'uint8': 'int', 'uint16': 'int',
                           'uint32': 'int', 'uint64': 'int', 'bool': 'int',
                           'float16': 'float', 'float32': 'float', 'float64': 'float'}
    bad_indices = [i for i, dtype in enumerate(dtypes) if (dtype.name not in pandas_dtype_mapper
                                                           and (not is_dtype_sparse(dtype)
                                                                or dtype.subtype.name not in pandas_dtype_mapper))]
    return bad_indices


506
def _data_from_pandas(data, feature_name, categorical_feature, pandas_categorical):
507
    if isinstance(data, pd_DataFrame):
508
509
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
510
511
        if feature_name == 'auto' or feature_name is None:
            data = data.rename(columns=str)
512
513
        cat_cols = list(data.select_dtypes(include=['category']).columns)
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
514
515
516
517
518
        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.')
519
            for col, category in zip(cat_cols, pandas_categorical):
520
521
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
522
        if len(cat_cols):  # cat_cols is list
523
            data = data.copy()  # not alter origin DataFrame
524
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
525
526
527
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
528
            if categorical_feature == 'auto':  # use cat cols from DataFrame
529
                categorical_feature = cat_cols_not_ordered
530
531
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
532
533
        if feature_name == 'auto':
            feature_name = list(data.columns)
534
535
        bad_indices = _get_bad_pandas_dtypes(data.dtypes)
        if bad_indices:
536
            bad_index_cols_str = ', '.join(data.columns[bad_indices])
537
            raise ValueError("DataFrame.dtypes for data must be int, float or bool.\n"
538
                             "Did not expect the data types in the following fields: "
539
                             f"{bad_index_cols_str}")
540
541
542
        data = data.values
        if data.dtype != np.float32 and data.dtype != np.float64:
            data = data.astype(np.float32)
543
544
545
546
547
548
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
549
550
551


def _label_from_pandas(label):
552
    if isinstance(label, pd_DataFrame):
553
554
        if len(label.columns) > 1:
            raise ValueError('DataFrame for label cannot have multiple columns')
555
        if _get_bad_pandas_dtypes(label.dtypes):
556
            raise ValueError('DataFrame.dtypes for label must be int, float or bool')
557
        label = np.ravel(label.values.astype(np.float32, copy=False))
558
559
560
    return label


561
def _dump_pandas_categorical(pandas_categorical, file_name=None):
562
563
    categorical_json = json.dumps(pandas_categorical, default=json_default_with_numpy)
    pandas_str = f'\npandas_categorical:{categorical_json}\n'
564
565
566
567
568
569
570
    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):
571
572
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
573
    if file_name is not None:
574
575
576
577
578
579
580
581
582
583
        max_offset = -os.path.getsize(file_name)
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
                f.seek(offset, os.SEEK_END)
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
584
        last_line = lines[-1].decode('utf-8').strip()
585
        if not last_line.startswith(pandas_key):
586
            last_line = lines[-2].decode('utf-8').strip()
587
    elif model_str is not None:
588
589
590
591
592
593
        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
594
595


596
class _InnerPredictor:
597
598
599
600
601
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
602
603
604
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
605
    """
606

607
    def __init__(self, model_file=None, booster_handle=None, pred_parameter=None):
608
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
609
610
611

        Parameters
        ----------
612
        model_file : string or None, optional (default=None)
wxchan's avatar
wxchan committed
613
            Path to the model file.
614
615
616
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
617
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
618
619
620
621
622
        """
        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
623
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
624
625
626
627
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
                c_str(model_file),
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
628
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
629
630
631
632
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
633
            self.num_total_iteration = out_num_iterations.value
634
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
635
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
636
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
637
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
638
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
639
640
641
642
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
643
            self.num_total_iteration = self.current_iteration()
644
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
645
        else:
646
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
647

648
649
        pred_parameter = {} if pred_parameter is None else pred_parameter
        self.pred_parameter = param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
650

wxchan's avatar
wxchan committed
651
    def __del__(self):
652
653
654
655
656
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
657

658
659
660
661
662
    def __getstate__(self):
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

663
    def predict(self, data, start_iteration=0, num_iteration=-1,
664
                raw_score=False, pred_leaf=False, pred_contrib=False, data_has_header=False,
wxchan's avatar
wxchan committed
665
                is_reshape=True):
666
        """Predict logic.
wxchan's avatar
wxchan committed
667
668
669

        Parameters
        ----------
670
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
671
672
            Data source for prediction.
            When data type is string, it represents the path of txt file.
673
674
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
675
676
677
678
679
680
681
682
683
684
685
686
687
        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
688
689
690

        Returns
        -------
691
        result : numpy array, scipy.sparse or list of scipy.sparse
692
            Prediction result.
693
            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
694
        """
wxchan's avatar
wxchan committed
695
        if isinstance(data, Dataset):
696
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
697
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
wxchan's avatar
wxchan committed
698
699
700
701
702
        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
703
704
        if pred_contrib:
            predict_type = C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
705
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
706

707
        if isinstance(data, str):
708
            with _TempFile() as f:
wxchan's avatar
wxchan committed
709
710
711
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
                    c_str(data),
Guolin Ke's avatar
Guolin Ke committed
712
713
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
714
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
715
                    ctypes.c_int(num_iteration),
716
                    c_str(self.pred_parameter),
wxchan's avatar
wxchan committed
717
718
                    c_str(f.name)))
                lines = f.readlines()
719
720
                nrow = len(lines)
                preds = [float(token) for line in lines for token in line.split('\t')]
Guolin Ke's avatar
Guolin Ke committed
721
                preds = np.array(preds, dtype=np.float64, copy=False)
wxchan's avatar
wxchan committed
722
        elif isinstance(data, scipy.sparse.csr_matrix):
723
            preds, nrow = self.__pred_for_csr(data, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
724
        elif isinstance(data, scipy.sparse.csc_matrix):
725
            preds, nrow = self.__pred_for_csc(data, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
726
        elif isinstance(data, np.ndarray):
727
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
728
729
730
        elif isinstance(data, list):
            try:
                data = np.array(data)
731
            except BaseException:
732
                raise ValueError('Cannot convert data list to numpy array.')
733
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
734
        elif isinstance(data, dt_DataTable):
735
            preds, nrow = self.__pred_for_np2d(data.to_numpy(), start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
736
737
        else:
            try:
738
                _log_warning('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
739
                csr = scipy.sparse.csr_matrix(data)
740
            except BaseException:
741
                raise TypeError(f'Cannot predict data for type {type(data).__name__}')
742
            preds, nrow = self.__pred_for_csr(csr, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
743
744
        if pred_leaf:
            preds = preds.astype(np.int32)
745
746
        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
747
            if preds.size % nrow == 0:
748
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
749
            else:
750
                raise ValueError(f'Length of predict result ({preds.size}) cannot be divide nrow ({nrow})')
wxchan's avatar
wxchan committed
751
752
        return preds

753
    def __get_num_preds(self, start_iteration, num_iteration, nrow, predict_type):
754
        """Get size of prediction result."""
755
756
        if nrow > MAX_INT32:
            raise LightGBMError('LightGBM cannot perform prediction for data'
757
                                f'with number of rows greater than MAX_INT32 ({MAX_INT32}).\n'
758
                                'You can split your data into chunks'
759
                                'and then concatenate predictions for them')
Guolin Ke's avatar
Guolin Ke committed
760
761
762
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
763
764
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
765
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
766
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
767
768
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
769

770
    def __pred_for_np2d(self, mat, start_iteration, num_iteration, predict_type):
771
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
772
        if len(mat.shape) != 2:
773
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
774

775
        def inner_predict(mat, start_iteration, num_iteration, predict_type, preds=None):
776
777
            if mat.dtype == np.float32 or mat.dtype == np.float64:
                data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
778
            else:  # change non-float data to float data, need to copy
779
780
                data = np.array(mat.reshape(mat.size), dtype=np.float32)
            ptr_data, type_ptr_data, _ = c_float_array(data)
781
            n_preds = self.__get_num_preds(start_iteration, num_iteration, mat.shape[0], predict_type)
782
            if preds is None:
783
                preds = np.empty(n_preds, dtype=np.float64)
784
785
786
787
788
789
790
            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),
791
792
                ctypes.c_int32(mat.shape[0]),
                ctypes.c_int32(mat.shape[1]),
793
794
                ctypes.c_int(C_API_IS_ROW_MAJOR),
                ctypes.c_int(predict_type),
795
                ctypes.c_int(start_iteration),
796
797
798
799
800
801
802
803
804
805
806
807
                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
808
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
809
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
810
            preds = np.empty(sum(n_preds), dtype=np.float64)
811
812
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
813
                # avoid memory consumption by arrays concatenation operations
814
                inner_predict(chunk, start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
815
            return preds, nrow
wxchan's avatar
wxchan committed
816
        else:
817
            return inner_predict(mat, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
818

819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
    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

865
    def __pred_for_csr(self, csr, start_iteration, num_iteration, predict_type):
866
        """Predict for a CSR data."""
867
        def inner_predict(csr, start_iteration, num_iteration, predict_type, preds=None):
868
            nrow = len(csr.indptr) - 1
869
            n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
870
            if preds is None:
871
                preds = np.empty(n_preds, dtype=np.float64)
872
873
874
875
876
877
878
            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)

879
            assert csr.shape[1] <= MAX_INT32
880
            csr_indices = csr.indices.astype(np.int32, copy=False)
881

882
883
884
            _safe_call(_LIB.LGBM_BoosterPredictForCSR(
                self.handle,
                ptr_indptr,
885
                ctypes.c_int(type_ptr_indptr),
886
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
887
888
889
890
891
892
                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),
893
                ctypes.c_int(start_iteration),
894
895
896
897
898
899
900
                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
901

902
        def inner_predict_sparse(csr, start_iteration, num_iteration, predict_type):
903
904
905
906
907
908
909
910
911
912
913
914
915
            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)()
916
            out_shape = np.empty(2, dtype=np.int64)
917
918
919
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
920
                ctypes.c_int(type_ptr_indptr),
921
922
923
924
925
926
927
                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),
928
                ctypes.c_int(start_iteration),
929
930
931
932
933
934
935
936
937
938
939
940
941
                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:
942
            return inner_predict_sparse(csr, start_iteration, num_iteration, predict_type)
943
944
945
946
        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
947
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
948
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
949
            preds = np.empty(sum(n_preds), dtype=np.float64)
950
951
            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:])):
952
                # avoid memory consumption by arrays concatenation operations
953
                inner_predict(csr[start_idx:end_idx], start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
954
955
            return preds, nrow
        else:
956
            return inner_predict(csr, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
957

958
    def __pred_for_csc(self, csc, start_iteration, num_iteration, predict_type):
959
        """Predict for a CSC data."""
960
        def inner_predict_sparse(csc, start_iteration, num_iteration, predict_type):
961
962
963
964
965
966
967
968
969
970
971
972
973
            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)()
974
            out_shape = np.empty(2, dtype=np.int64)
975
976
977
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
978
                ctypes.c_int(type_ptr_indptr),
979
980
981
982
983
984
985
                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),
986
                ctypes.c_int(start_iteration),
987
988
989
990
991
992
993
994
995
996
997
998
                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
999
        nrow = csc.shape[0]
1000
        if nrow > MAX_INT32:
1001
            return self.__pred_for_csr(csc.tocsr(), start_iteration, num_iteration, predict_type)
1002
        if predict_type == C_API_PREDICT_CONTRIB:
1003
1004
            return inner_predict_sparse(csc, start_iteration, num_iteration, predict_type)
        n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
1005
        preds = np.empty(n_preds, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
1006
1007
        out_num_preds = ctypes.c_int64(0)

1008
1009
        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
1010

1011
        assert csc.shape[0] <= MAX_INT32
1012
        csc_indices = csc.indices.astype(np.int32, copy=False)
1013

Guolin Ke's avatar
Guolin Ke committed
1014
1015
1016
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
1017
            ctypes.c_int(type_ptr_indptr),
1018
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1019
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1020
1021
1022
1023
1024
            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),
1025
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
1026
            ctypes.c_int(num_iteration),
1027
            c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
1028
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
1029
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
1030
        if n_preds != out_num_preds.value:
1031
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
1032
1033
        return preds, nrow

1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
    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
1048

1049
class Dataset:
wxchan's avatar
wxchan committed
1050
    """Dataset in LightGBM."""
1051

1052
    def __init__(self, data, label=None, reference=None,
1053
                 weight=None, group=None, init_score=None, silent=False,
1054
                 feature_name='auto', categorical_feature='auto', params=None,
wxchan's avatar
wxchan committed
1055
                 free_raw_data=True):
1056
        """Initialize Dataset.
1057

wxchan's avatar
wxchan committed
1058
1059
        Parameters
        ----------
1060
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse or list of numpy arrays
wxchan's avatar
wxchan committed
1061
            Data source of Dataset.
1062
            If string, it represents the path to txt file.
1063
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1064
1065
1066
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
1067
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
1068
            Weight for each instance.
1069
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1070
1071
1072
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1073
1074
            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.
1075
        init_score : list, numpy 1-D array, pandas Series or None, optional (default=None)
1076
            Init score for Dataset.
1077
1078
1079
1080
1081
1082
1083
1084
1085
        silent : bool, optional (default=False)
            Whether to print messages during construction.
        feature_name : list of strings or 'auto', optional (default="auto")
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
        categorical_feature : list of strings or int, or 'auto', optional (default="auto")
            Categorical features.
            If list of int, interpreted as indices.
            If list of strings, interpreted as feature names (need to specify ``feature_name`` as well).
1086
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
1087
            All values in categorical features should be less than int32 max value (2147483647).
1088
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
1089
            All negative values in categorical features will be treated as missing values.
1090
            The output cannot be monotonically constrained with respect to a categorical feature.
Nikita Titov's avatar
Nikita Titov committed
1091
        params : dict or None, optional (default=None)
1092
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
1093
        free_raw_data : bool, optional (default=True)
1094
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
1095
        """
wxchan's avatar
wxchan committed
1096
1097
1098
1099
1100
1101
        self.handle = None
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
1102
        self.init_score = init_score
wxchan's avatar
wxchan committed
1103
1104
        self.silent = silent
        self.feature_name = feature_name
1105
        self.categorical_feature = categorical_feature
1106
        self.params = deepcopy(params)
wxchan's avatar
wxchan committed
1107
1108
        self.free_raw_data = free_raw_data
        self.used_indices = None
1109
        self.need_slice = True
wxchan's avatar
wxchan committed
1110
        self._predictor = None
1111
        self.pandas_categorical = None
1112
        self.params_back_up = None
1113
1114
        self.feature_penalty = None
        self.monotone_constraints = None
1115
        self.version = 0
wxchan's avatar
wxchan committed
1116
1117

    def __del__(self):
1118
1119
1120
1121
        try:
            self._free_handle()
        except AttributeError:
            pass
1122

1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
    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",
1144
                                                "linear_tree",
1145
1146
1147
1148
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
Nikita Titov's avatar
Nikita Titov committed
1149
                                                "precise_float_parser",
1150
1151
1152
1153
1154
1155
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}

1156
    def _free_handle(self):
1157
        if self.handle is not None:
1158
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1159
            self.handle = None
Guolin Ke's avatar
Guolin Ke committed
1160
1161
1162
        self.need_slice = True
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1163
        return self
wxchan's avatar
wxchan committed
1164

Guolin Ke's avatar
Guolin Ke committed
1165
1166
    def _set_init_score_by_predictor(self, predictor, data, used_indices=None):
        data_has_header = False
1167
        if isinstance(data, str):
Guolin Ke's avatar
Guolin Ke committed
1168
            # check data has header or not
1169
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1170
        num_data = self.num_data()
1171
1172
1173
1174
1175
1176
1177
        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
1178
                if isinstance(data, str):
1179
                    sub_init_score = np.empty(num_data * predictor.num_class, dtype=np.float32)
1180
                    assert num_data == len(used_indices)
1181
1182
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1183
1184
1185
1186
                            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
1187
                new_init_score = np.empty(init_score.size, dtype=np.float32)
1188
1189
                for i in range(num_data):
                    for j in range(predictor.num_class):
1190
1191
1192
1193
1194
1195
                        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:
            init_score = np.zeros(self.init_score.shape, dtype=np.float32)
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1196
1197
        self.set_init_score(init_score)

1198
    def _lazy_init(self, data, label=None, reference=None,
1199
                   weight=None, group=None, init_score=None, predictor=None,
wxchan's avatar
wxchan committed
1200
                   silent=False, feature_name='auto',
1201
                   categorical_feature='auto', params=None):
wxchan's avatar
wxchan committed
1202
1203
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1204
            return self
Guolin Ke's avatar
Guolin Ke committed
1205
1206
1207
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1208
1209
1210
1211
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
wxchan's avatar
wxchan committed
1212
        label = _label_from_pandas(label)
Guolin Ke's avatar
Guolin Ke committed
1213

1214
        # process for args
wxchan's avatar
wxchan committed
1215
        params = {} if params is None else params
1216
1217
1218
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1219
1220
        for key, _ in params.items():
            if key in args_names:
1221
1222
                _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.')
1223
        # user can set verbose with params, it has higher priority
1224
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
1225
            params["verbose"] = -1
1226
        # get categorical features
1227
1228
1229
1230
1231
1232
        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:
1233
                if isinstance(name, str) and name in feature_dict:
1234
                    categorical_indices.add(feature_dict[name])
1235
                elif isinstance(name, int):
1236
1237
                    categorical_indices.add(name)
                else:
1238
                    raise TypeError(f"Wrong type({type(name).__name__}) or unknown name({name}) in categorical_feature")
1239
            if categorical_indices:
1240
1241
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
1242
                        _log_warning(f'{cat_alias} in param dict is overridden.')
1243
                        params.pop(cat_alias, None)
1244
                params['categorical_column'] = sorted(categorical_indices)
1245

wxchan's avatar
wxchan committed
1246
        params_str = param_dict_to_str(params)
1247
        self.params = params
1248
        # process for reference dataset
wxchan's avatar
wxchan committed
1249
        ref_dataset = None
wxchan's avatar
wxchan committed
1250
        if isinstance(reference, Dataset):
1251
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1252
1253
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1254
        # start construct data
1255
        if isinstance(data, str):
wxchan's avatar
wxchan committed
1256
1257
1258
1259
1260
1261
1262
1263
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
                c_str(data),
                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
1264
1265
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1266
1267
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1268
1269
        elif isinstance(data, list) and len(data) > 0 and all(isinstance(x, np.ndarray) for x in data):
            self.__init_from_list_np2d(data, params_str, ref_dataset)
1270
        elif isinstance(data, dt_DataTable):
1271
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1272
1273
1274
1275
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1276
            except BaseException:
1277
                raise TypeError(f'Cannot initialize Dataset from {type(data).__name__}')
wxchan's avatar
wxchan committed
1278
1279
1280
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1281
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1282
1283
1284
1285
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1286
1287
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
1288
                _log_warning("The init_score will be overridden by the prediction of init_model.")
Guolin Ke's avatar
Guolin Ke committed
1289
            self._set_init_score_by_predictor(predictor, data)
1290
1291
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1292
        elif predictor is not None:
1293
            raise TypeError(f'Wrong predictor type {type(predictor).__name__}')
Guolin Ke's avatar
Guolin Ke committed
1294
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1295
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1296
1297

    def __init_from_np2d(self, mat, params_str, ref_dataset):
1298
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1299
1300
1301
1302
1303
1304
        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)
1305
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1306
1307
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1308
        ptr_data, type_ptr_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1309
1310
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1311
            ctypes.c_int(type_ptr_data),
1312
1313
            ctypes.c_int32(mat.shape[0]),
            ctypes.c_int32(mat.shape[1]),
Guolin Ke's avatar
Guolin Ke committed
1314
            ctypes.c_int(C_API_IS_ROW_MAJOR),
wxchan's avatar
wxchan committed
1315
1316
1317
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1318
        return self
wxchan's avatar
wxchan committed
1319

1320
    def __init_from_list_np2d(self, mats, params_str, ref_dataset):
1321
        """Initialize data from a list of 2-D numpy matrices."""
1322
        ncol = mats[0].shape[1]
1323
        nrow = np.empty((len(mats),), np.int32)
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
        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)
1343
            else:  # change non-float data to float data, need to copy
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
                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(
1355
            ctypes.c_int32(len(mats)),
1356
1357
1358
            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)),
1359
            ctypes.c_int32(ncol),
1360
1361
1362
1363
            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
1364
        return self
1365

wxchan's avatar
wxchan committed
1366
    def __init_from_csr(self, csr, params_str, ref_dataset):
1367
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
1368
        if len(csr.indices) != len(csr.data):
1369
            raise ValueError(f'Length mismatch: {len(csr.indices)} vs {len(csr.data)}')
wxchan's avatar
wxchan committed
1370
1371
        self.handle = ctypes.c_void_p()

1372
1373
        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
1374

1375
        assert csr.shape[1] <= MAX_INT32
1376
        csr_indices = csr.indices.astype(np.int32, copy=False)
1377

wxchan's avatar
wxchan committed
1378
1379
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1380
            ctypes.c_int(type_ptr_indptr),
1381
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
1382
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1383
1384
1385
1386
            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
1387
1388
1389
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1390
        return self
wxchan's avatar
wxchan committed
1391

Guolin Ke's avatar
Guolin Ke committed
1392
    def __init_from_csc(self, csc, params_str, ref_dataset):
1393
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
1394
        if len(csc.indices) != len(csc.data):
1395
            raise ValueError(f'Length mismatch: {len(csc.indices)} vs {len(csc.data)}')
Guolin Ke's avatar
Guolin Ke committed
1396
1397
        self.handle = ctypes.c_void_p()

1398
1399
        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
1400

1401
        assert csc.shape[0] <= MAX_INT32
1402
        csc_indices = csc.indices.astype(np.int32, copy=False)
1403

Guolin Ke's avatar
Guolin Ke committed
1404
1405
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1406
            ctypes.c_int(type_ptr_indptr),
1407
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1408
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1409
1410
1411
1412
            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
1413
1414
1415
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1416
        return self
Guolin Ke's avatar
Guolin Ke committed
1417

wxchan's avatar
wxchan committed
1418
    def construct(self):
1419
1420
1421
1422
1423
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
1424
            Constructed Dataset object.
1425
        """
1426
        if self.handle is None:
wxchan's avatar
wxchan committed
1427
            if self.reference is not None:
1428
1429
                reference_params = self.reference.get_params()
                if self.get_params() != reference_params:
1430
                    _log_warning('Overriding the parameters from Reference Dataset.')
1431
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
1432
                if self.used_indices is None:
1433
                    # create valid
1434
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
1435
1436
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
1437
                                    silent=self.silent, feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
1438
                else:
1439
                    # construct subset
wxchan's avatar
wxchan committed
1440
                    used_indices = list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
1441
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
1442
                    if self.reference.group is not None:
1443
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
1444
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
1445
                                                  return_counts=True)
1446
                    self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1447
1448
                    params_str = param_dict_to_str(self.params)
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
1449
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
1450
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
1451
                        ctypes.c_int32(used_indices.shape[0]),
wxchan's avatar
wxchan committed
1452
1453
                        c_str(params_str),
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
1454
1455
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
1456
1457
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
1458
1459
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
1460
1461
1462
                    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
1463
            else:
1464
                # create train
1465
                self._lazy_init(self.data, label=self.label,
1466
1467
1468
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
                                silent=self.silent, feature_name=self.feature_name,
1469
                                categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
1470
1471
1472
            if self.free_raw_data:
                self.data = None
        return self
wxchan's avatar
wxchan committed
1473

wxchan's avatar
wxchan committed
1474
    def create_valid(self, data, label=None, weight=None, group=None,
1475
                     init_score=None, silent=False, params=None):
1476
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
1477
1478
1479

        Parameters
        ----------
1480
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse or list of numpy arrays
wxchan's avatar
wxchan committed
1481
            Data source of Dataset.
1482
            If string, it represents the path to txt file.
1483
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1484
1485
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
1486
            Weight for each instance.
1487
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1488
1489
1490
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1491
1492
            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.
1493
        init_score : list, numpy 1-D array, pandas Series or None, optional (default=None)
1494
            Init score for Dataset.
1495
1496
        silent : bool, optional (default=False)
            Whether to print messages during construction.
Nikita Titov's avatar
Nikita Titov committed
1497
        params : dict or None, optional (default=None)
1498
            Other parameters for validation Dataset.
1499
1500
1501

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1502
1503
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
1504
        """
1505
        ret = Dataset(data, label=label, reference=self,
1506
1507
                      weight=weight, group=group, init_score=init_score,
                      silent=silent, params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1508
        ret._predictor = self._predictor
1509
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
1510
        return ret
wxchan's avatar
wxchan committed
1511

wxchan's avatar
wxchan committed
1512
    def subset(self, used_indices, params=None):
1513
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
1514
1515
1516
1517

        Parameters
        ----------
        used_indices : list of int
1518
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
1519
        params : dict or None, optional (default=None)
1520
            These parameters will be passed to Dataset constructor.
1521
1522
1523
1524
1525

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
1526
        """
wxchan's avatar
wxchan committed
1527
1528
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
1529
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
1530
1531
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1532
        ret._predictor = self._predictor
1533
        ret.pandas_categorical = self.pandas_categorical
1534
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
1535
1536
1537
        return ret

    def save_binary(self, filename):
1538
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
1539

1540
1541
1542
1543
1544
        .. 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
1545
1546
1547
1548
        Parameters
        ----------
        filename : string
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
1549
1550
1551
1552
1553

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
1554
1555
1556
1557
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
            c_str(filename)))
Nikita Titov's avatar
Nikita Titov committed
1558
        return self
wxchan's avatar
wxchan committed
1559
1560

    def _update_params(self, params):
1561
1562
        if not params:
            return self
1563
        params = deepcopy(params)
1564
1565
1566
1567
1568

        def update():
            if not self.params:
                self.params = params
            else:
1569
                self.params_back_up = deepcopy(self.params)
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
                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:
1584
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
1585
        return self
wxchan's avatar
wxchan committed
1586

1587
    def _reverse_update_params(self):
1588
        if self.handle is None:
1589
            self.params = deepcopy(self.params_back_up)
1590
            self.params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
1591
        return self
1592

wxchan's avatar
wxchan committed
1593
    def set_field(self, field_name, data):
wxchan's avatar
wxchan committed
1594
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
1595
1596
1597

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
1598
        field_name : string
1599
            The field name of the information.
1600
        data : list, numpy 1-D array, pandas Series or None
1601
            The array of data to be set.
Nikita Titov's avatar
Nikita Titov committed
1602
1603
1604
1605
1606

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
1607
        """
1608
        if self.handle is None:
1609
            raise Exception(f"Cannot set {field_name} before construct dataset")
wxchan's avatar
wxchan committed
1610
        if data is None:
1611
            # set to None
wxchan's avatar
wxchan committed
1612
1613
1614
1615
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
                c_str(field_name),
                None,
Guolin Ke's avatar
Guolin Ke committed
1616
1617
                ctypes.c_int(0),
                ctypes.c_int(FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
1618
            return self
Guolin Ke's avatar
Guolin Ke committed
1619
1620
1621
1622
1623
        dtype = np.float32
        if field_name == 'group':
            dtype = np.int32
        elif field_name == 'init_score':
            dtype = np.float64
1624
        data = list_to_1d_numpy(data, dtype, name=field_name)
1625
1626
        if data.dtype == np.float32 or data.dtype == np.float64:
            ptr_data, type_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1627
        elif data.dtype == np.int32:
1628
            ptr_data, type_data, _ = c_int_array(data)
wxchan's avatar
wxchan committed
1629
        else:
1630
            raise TypeError(f"Expected np.float32/64 or np.int32, met type({data.dtype})")
wxchan's avatar
wxchan committed
1631
        if type_data != FIELD_TYPE_MAPPER[field_name]:
1632
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
1633
1634
1635
1636
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
            c_str(field_name),
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1637
1638
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
1639
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
1640
        return self
wxchan's avatar
wxchan committed
1641

wxchan's avatar
wxchan committed
1642
1643
    def get_field(self, field_name):
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
1644
1645
1646

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
1647
        field_name : string
1648
            The field name of the information.
wxchan's avatar
wxchan committed
1649
1650
1651

        Returns
        -------
1652
1653
        info : numpy array
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1654
        """
1655
        if self.handle is None:
1656
            raise Exception(f"Cannot get {field_name} before construct Dataset")
1657
1658
        tmp_out_len = ctypes.c_int(0)
        out_type = ctypes.c_int(0)
wxchan's avatar
wxchan committed
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
        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:
            return cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
        elif out_type.value == C_API_DTYPE_FLOAT32:
            return cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
Guolin Ke's avatar
Guolin Ke committed
1674
1675
        elif out_type.value == C_API_DTYPE_FLOAT64:
            return cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
1676
        else:
wxchan's avatar
wxchan committed
1677
            raise TypeError("Unknown type")
Guolin Ke's avatar
Guolin Ke committed
1678

1679
    def set_categorical_feature(self, categorical_feature):
1680
        """Set categorical features.
1681
1682
1683

        Parameters
        ----------
1684
1685
        categorical_feature : list of int or strings
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
1686
1687
1688
1689
1690

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
1691
1692
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
1693
            return self
1694
        if self.data is not None:
1695
1696
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
1697
                return self._free_handle()
1698
            elif categorical_feature == 'auto':
1699
                _log_warning('Using categorical_feature in Dataset.')
Nikita Titov's avatar
Nikita Titov committed
1700
                return self
1701
            else:
1702
                _log_warning('categorical_feature in Dataset is overridden.\n'
1703
                             f'New categorical_feature is {sorted(list(categorical_feature))}')
1704
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
1705
                return self._free_handle()
1706
        else:
1707
1708
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
1709

Guolin Ke's avatar
Guolin Ke committed
1710
    def _set_predictor(self, predictor):
1711
1712
1713
1714
        """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
1715
        """
1716
        if predictor is self._predictor and (predictor is None or predictor.current_iteration() == self._predictor.current_iteration()):
Nikita Titov's avatar
Nikita Titov committed
1717
            return self
1718
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
1719
            self._predictor = predictor
1720
1721
1722
1723
1724
1725
        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
1726
        else:
1727
1728
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
1729
        return self
Guolin Ke's avatar
Guolin Ke committed
1730
1731

    def set_reference(self, reference):
1732
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
1733
1734
1735
1736

        Parameters
        ----------
        reference : Dataset
1737
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
1738
1739
1740
1741
1742

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
1743
        """
1744
1745
1746
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
1747
1748
        # we're done if self and reference share a common upstrem reference
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
1749
            return self
Guolin Ke's avatar
Guolin Ke committed
1750
1751
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
1752
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
1753
        else:
1754
1755
            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
1756
1757

    def set_feature_name(self, feature_name):
1758
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
1759
1760
1761

        Parameters
        ----------
1762
1763
        feature_name : list of strings
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
1764
1765
1766
1767
1768

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
1769
        """
1770
1771
        if feature_name != 'auto':
            self.feature_name = feature_name
1772
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
1773
            if len(feature_name) != self.num_feature():
1774
                raise ValueError(f"Length of feature_name({len(feature_name)}) and num_feature({self.num_feature()}) don't match")
1775
            c_feature_name = [c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
1776
1777
1778
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
                c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
1779
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
1780
        return self
Guolin Ke's avatar
Guolin Ke committed
1781
1782

    def set_label(self, label):
1783
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
1784
1785
1786

        Parameters
        ----------
1787
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
1788
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
1789
1790
1791
1792
1793

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
1794
1795
        """
        self.label = label
1796
        if self.handle is not None:
1797
            label = list_to_1d_numpy(_label_from_pandas(label), name='label')
wxchan's avatar
wxchan committed
1798
            self.set_field('label', label)
1799
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1800
        return self
Guolin Ke's avatar
Guolin Ke committed
1801
1802

    def set_weight(self, weight):
1803
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
1804
1805
1806

        Parameters
        ----------
1807
        weight : list, numpy 1-D array, pandas Series or None
1808
            Weight to be set for each data point.
Nikita Titov's avatar
Nikita Titov committed
1809
1810
1811
1812
1813

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
1814
        """
1815
1816
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
1817
        self.weight = weight
1818
        if self.handle is not None and weight is not None:
wxchan's avatar
wxchan committed
1819
1820
            weight = list_to_1d_numpy(weight, name='weight')
            self.set_field('weight', weight)
1821
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1822
        return self
Guolin Ke's avatar
Guolin Ke committed
1823
1824

    def set_init_score(self, init_score):
1825
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
1826
1827
1828

        Parameters
        ----------
1829
        init_score : list, numpy 1-D array, pandas Series or None
1830
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
1831
1832
1833
1834
1835

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
1836
1837
        """
        self.init_score = init_score
1838
        if self.handle is not None and init_score is not None:
Guolin Ke's avatar
Guolin Ke committed
1839
            init_score = list_to_1d_numpy(init_score, np.float64, name='init_score')
wxchan's avatar
wxchan committed
1840
            self.set_field('init_score', init_score)
1841
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1842
        return self
Guolin Ke's avatar
Guolin Ke committed
1843
1844

    def set_group(self, group):
1845
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
1846
1847
1848

        Parameters
        ----------
1849
        group : list, numpy 1-D array, pandas Series or None
1850
1851
1852
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1853
1854
            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
1855
1856
1857
1858
1859

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
1860
1861
        """
        self.group = group
1862
        if self.handle is not None and group is not None:
wxchan's avatar
wxchan committed
1863
1864
            group = list_to_1d_numpy(group, np.int32, name='group')
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
1865
        return self
Guolin Ke's avatar
Guolin Ke committed
1866

1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
    def get_feature_name(self):
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
        feature_names : list
            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)
1881
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
1882
1883
1884
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
1885
            ctypes.c_int(num_feature),
1886
            ctypes.byref(tmp_out_len),
1887
            ctypes.c_size_t(reserved_string_buffer_size),
1888
1889
1890
1891
            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")
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
        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))
1904
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
1905

Guolin Ke's avatar
Guolin Ke committed
1906
    def get_label(self):
1907
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1908
1909
1910

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1911
        label : numpy array or None
1912
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1913
        """
1914
        if self.label is None:
wxchan's avatar
wxchan committed
1915
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
1916
1917
1918
        return self.label

    def get_weight(self):
1919
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1920
1921
1922

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1923
        weight : numpy array or None
1924
            Weight for each data point from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1925
        """
1926
        if self.weight is None:
wxchan's avatar
wxchan committed
1927
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
1928
1929
1930
        return self.weight

    def get_init_score(self):
1931
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1932
1933
1934

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1935
        init_score : numpy array or None
1936
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
1937
        """
1938
        if self.init_score is None:
wxchan's avatar
wxchan committed
1939
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
1940
1941
        return self.init_score

1942
1943
1944
1945
1946
    def get_data(self):
        """Get the raw data of the Dataset.

        Returns
        -------
1947
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, list of numpy arrays or None
1948
1949
1950
1951
            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
1952
1953
1954
1955
1956
        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, :]
1957
                elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
1958
                    self.data = self.data.iloc[self.used_indices].copy()
1959
                elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
1960
1961
                    self.data = self.data[self.used_indices, :]
                else:
1962
1963
                    _log_warning(f"Cannot subset {type(self.data).__name__} type of raw data.\n"
                                 "Returning original raw data")
1964
            self.need_slice = False
Guolin Ke's avatar
Guolin Ke committed
1965
1966
1967
        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.")
1968
1969
        return self.data

Guolin Ke's avatar
Guolin Ke committed
1970
    def get_group(self):
1971
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1972
1973
1974

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1975
        group : numpy array or None
1976
1977
1978
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
1979
1980
            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
1981
        """
1982
        if self.group is None:
wxchan's avatar
wxchan committed
1983
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
1984
1985
            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
1986
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
1987
1988
1989
        return self.group

    def num_data(self):
1990
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1991
1992
1993

        Returns
        -------
1994
1995
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1996
        """
1997
        if self.handle is not None:
1998
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
1999
2000
2001
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2002
        else:
2003
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2004
2005

    def num_feature(self):
2006
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2007
2008
2009

        Returns
        -------
2010
2011
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
2012
        """
2013
        if self.handle is not None:
2014
            ret = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2015
2016
2017
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
2018
        else:
2019
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
2020

2021
    def get_ref_chain(self, ref_limit=100):
2022
2023
2024
2025
2026
        """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.
2027
2028
2029
2030
2031

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
2032
2033
2034

        Returns
        -------
2035
2036
2037
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
2038
        head = self
2039
        ref_chain = set()
2040
2041
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
2042
                ref_chain.add(head)
2043
2044
2045
2046
2047
2048
                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
2049
        return ref_chain
2050

2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
    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
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
        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()))
2079
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2080
                    self.data = np.hstack((self.data, other.data.values))
2081
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2082
2083
2084
2085
2086
2087
2088
                    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)
2089
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2090
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
2091
                elif isinstance(other.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2092
2093
2094
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
2095
            elif isinstance(self.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2096
2097
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
2098
2099
                                        "without pandas installed. "
                                        "Install pandas and restart your session.")
Guolin Ke's avatar
Guolin Ke committed
2100
                if isinstance(other.data, np.ndarray):
2101
                    self.data = concat((self.data, pd_DataFrame(other.data)),
Guolin Ke's avatar
Guolin Ke committed
2102
2103
                                       axis=1, ignore_index=True)
                elif scipy.sparse.issparse(other.data):
2104
                    self.data = concat((self.data, pd_DataFrame(other.data.toarray())),
Guolin Ke's avatar
Guolin Ke committed
2105
                                       axis=1, ignore_index=True)
2106
                elif isinstance(other.data, pd_DataFrame):
Guolin Ke's avatar
Guolin Ke committed
2107
2108
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
2109
2110
                elif isinstance(other.data, dt_DataTable):
                    self.data = concat((self.data, pd_DataFrame(other.data.to_numpy())),
Guolin Ke's avatar
Guolin Ke committed
2111
2112
2113
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
2114
            elif isinstance(self.data, dt_DataTable):
Guolin Ke's avatar
Guolin Ke committed
2115
                if isinstance(other.data, np.ndarray):
2116
                    self.data = dt_DataTable(np.hstack((self.data.to_numpy(), other.data)))
Guolin Ke's avatar
Guolin Ke committed
2117
                elif scipy.sparse.issparse(other.data):
2118
2119
2120
2121
2122
                    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
2123
2124
2125
2126
2127
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
2128
2129
            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
2130
2131
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
2132
            _log_warning(err_msg)
Guolin Ke's avatar
Guolin Ke committed
2133
        self.feature_name = self.get_feature_name()
2134
2135
        _log_warning("Reseting categorical features.\n"
                     "You can set new categorical features via ``set_categorical_feature`` method")
Guolin Ke's avatar
Guolin Ke committed
2136
2137
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2138
2139
        return self

2140
    def _dump_text(self, filename):
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
        """Save Dataset to a text file.

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

        Parameters
        ----------
        filename : string
            Name of the output file.

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

wxchan's avatar
wxchan committed
2160

2161
class Booster:
2162
    """Booster in LightGBM."""
2163

2164
    def __init__(self, params=None, train_set=None, model_file=None, model_str=None, silent=False):
2165
        """Initialize the Booster.
wxchan's avatar
wxchan committed
2166
2167
2168

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2169
        params : dict or None, optional (default=None)
2170
2171
2172
2173
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
        model_file : string or None, optional (default=None)
wxchan's avatar
wxchan committed
2174
            Path to the model file.
2175
2176
        model_str : string or None, optional (default=None)
            Model will be loaded from this string.
2177
2178
        silent : bool, optional (default=False)
            Whether to print messages during construction.
wxchan's avatar
wxchan committed
2179
        """
2180
        self.handle = None
2181
        self.network = False
wxchan's avatar
wxchan committed
2182
        self.__need_reload_eval_info = True
2183
        self._train_data_name = "training"
wxchan's avatar
wxchan committed
2184
        self.__attr = {}
2185
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
2186
        self.best_iteration = -1
wxchan's avatar
wxchan committed
2187
        self.best_score = {}
2188
        params = {} if params is None else deepcopy(params)
2189
        # user can set verbose with params, it has higher priority
2190
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
2191
            params["verbose"] = -1
wxchan's avatar
wxchan committed
2192
        if train_set is not None:
2193
            # Training task
wxchan's avatar
wxchan committed
2194
            if not isinstance(train_set, Dataset):
2195
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
            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"]
                )
2230
            # construct booster object
2231
2232
2233
2234
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
            params_str = param_dict_to_str(params)
2235
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2236
            _safe_call(_LIB.LGBM_BoosterCreate(
2237
                train_set.handle,
wxchan's avatar
wxchan committed
2238
2239
                c_str(params_str),
                ctypes.byref(self.handle)))
2240
            # save reference to data
wxchan's avatar
wxchan committed
2241
2242
2243
2244
            self.train_set = train_set
            self.valid_sets = []
            self.name_valid_sets = []
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
2245
2246
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
2247
2248
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
2249
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
2250
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2251
2252
2253
2254
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2255
            # buffer for inner predict
wxchan's avatar
wxchan committed
2256
2257
2258
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
2259
            self.pandas_categorical = train_set.pandas_categorical
2260
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
2261
        elif model_file is not None:
2262
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
2263
            out_num_iterations = ctypes.c_int(0)
2264
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2265
2266
2267
2268
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
                c_str(model_file),
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2269
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2270
2271
2272
2273
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2274
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
2275
2276
        elif model_str is not None:
            self.model_from_string(model_str, not silent)
wxchan's avatar
wxchan committed
2277
        else:
2278
2279
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
2280
        self.params = params
wxchan's avatar
wxchan committed
2281
2282

    def __del__(self):
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
        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
2293

wxchan's avatar
wxchan committed
2294
2295
2296
2297
    def __copy__(self):
        return self.__deepcopy__(None)

    def __deepcopy__(self, _):
2298
        model_str = self.model_to_string(num_iteration=-1)
2299
        booster = Booster(model_str=model_str)
2300
        return booster
wxchan's avatar
wxchan committed
2301
2302
2303
2304
2305
2306
2307

    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:
2308
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
2309
2310
2311
        return this

    def __setstate__(self, state):
2312
2313
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
2314
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2315
            out_num_iterations = ctypes.c_int(0)
2316
2317
2318
2319
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
                c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
2320
2321
2322
            state['handle'] = handle
        self.__dict__.update(state)

wxchan's avatar
wxchan committed
2323
    def free_dataset(self):
Nikita Titov's avatar
Nikita Titov committed
2324
2325
2326
2327
2328
2329
2330
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
2331
2332
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
2333
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
2334
        return self
wxchan's avatar
wxchan committed
2335

2336
2337
2338
    def _free_buffer(self):
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
2339
        return self
2340

2341
2342
2343
2344
2345
2346
2347
    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":
2348
2349
2350
2351
        """Set the network configuration.

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2352
        machines : list, set or string
2353
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
2354
        local_listen_port : int, optional (default=12400)
2355
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
2356
        listen_time_out : int, optional (default=120)
2357
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
2358
        num_machines : int, optional (default=1)
2359
            The number of machines for distributed learning application.
Nikita Titov's avatar
Nikita Titov committed
2360
2361
2362
2363
2364

        Returns
        -------
        self : Booster
            Booster with set network.
2365
        """
2366
2367
        if isinstance(machines, (list, set)):
            machines = ','.join(machines)
2368
2369
2370
2371
2372
        _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
2373
        return self
2374
2375

    def free_network(self):
Nikita Titov's avatar
Nikita Titov committed
2376
2377
2378
2379
2380
2381
2382
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
2383
2384
        _safe_call(_LIB.LGBM_NetworkFree())
        self.network = False
Nikita Titov's avatar
Nikita Titov committed
2385
        return self
2386

2387
2388
2389
    def trees_to_dataframe(self):
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
        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.
            - ``node_index`` : string, unique identifier for a node.
            - ``left_child`` : string, ``node_index`` of the child node to the left of a split. ``None`` for leaf nodes.
            - ``right_child`` : string, ``node_index`` of the child node to the right of a split. ``None`` for leaf nodes.
            - ``parent_index`` : string, ``node_index`` of this node's parent. ``None`` for the root node.
            - ``split_feature`` : string, name of the feature used for splitting. ``None`` for leaf nodes.
            - ``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.
2401
2402
2403
            - ``decision_type`` : string, logical operator describing how to compare a value to ``threshold``.
              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.
2404
2405
2406
2407
2408
2409
            - ``missing_direction`` : string, split direction that missing values should go to. ``None`` for leaf nodes.
            - ``missing_type`` : string, describes what types of values are treated as missing.
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
            - ``weight`` : float64 or int64, sum of hessian (second-order derivative of objective), summed over observations that fall in this node.
            - ``count`` : int64, number of records in the training data that fall into this node.

2410
2411
2412
2413
2414
2415
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
2416
2417
            raise LightGBMError('This method cannot be run without pandas installed. '
                                'You must install pandas and restart your session to use this method.')
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428

        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):
2429
                tree_num = f'{tree_index}-' if tree_index is not None else ''
2430
2431
2432
                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
2433
2434
                node_num = tree.get('split_index' if is_split else 'leaf_index', 0)
                return f"{tree_num}{node_type}{node_num}"
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446

            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):
2447
                return set(tree.keys()) == {'leaf_value'}
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520

            # 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))

2521
        return pd_DataFrame(model_list, columns=model_list[0].keys())
2522

wxchan's avatar
wxchan committed
2523
    def set_train_data_name(self, name):
2524
2525
2526
2527
        """Set the name to the training Dataset.

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2528
2529
2530
2531
2532
2533
2534
        name : string
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
2535
        """
2536
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
2537
        return self
wxchan's avatar
wxchan committed
2538
2539

    def add_valid(self, data, name):
2540
        """Add validation data.
wxchan's avatar
wxchan committed
2541
2542
2543
2544

        Parameters
        ----------
        data : Dataset
2545
2546
2547
            Validation data.
        name : string
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
2548
2549
2550
2551
2552

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
2553
        """
Guolin Ke's avatar
Guolin Ke committed
2554
        if not isinstance(data, Dataset):
2555
            raise TypeError(f'Validation data should be Dataset instance, met {type(data).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2556
        if data._predictor is not self.__init_predictor:
2557
2558
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2559
2560
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
2561
            data.construct().handle))
wxchan's avatar
wxchan committed
2562
2563
2564
2565
2566
        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
2567
        return self
wxchan's avatar
wxchan committed
2568
2569

    def reset_parameter(self, params):
2570
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
2571
2572
2573
2574

        Parameters
        ----------
        params : dict
2575
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
2576
2577
2578
2579
2580

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
2581
2582
2583
2584
2585
2586
        """
        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
2587
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
2588
        return self
wxchan's avatar
wxchan committed
2589
2590

    def update(self, train_set=None, fobj=None):
Nikita Titov's avatar
Nikita Titov committed
2591
        """Update Booster for one iteration.
2592

wxchan's avatar
wxchan committed
2593
2594
        Parameters
        ----------
2595
2596
2597
2598
        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
2599
            Customized objective function.
2600
2601
2602
2603
2604
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

                preds : list or numpy 1-D array
                    The predicted values.
2605
2606
                    Predicted values are returned before any transformation,
                    e.g. they are raw margin instead of probability of positive class for binary task.
2607
2608
2609
                train_data : Dataset
                    The training dataset.
                grad : list or numpy 1-D array
2610
2611
                    The value of the first order derivative (gradient) of the loss
                    with respect to the elements of preds for each sample point.
2612
                hess : list or numpy 1-D array
2613
2614
                    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
2615

2616
2617
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is score[j * num_data + i]
2618
2619
            and you should group grad and hess in this way as well.

wxchan's avatar
wxchan committed
2620
2621
        Returns
        -------
2622
2623
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
2624
        """
2625
        # need reset training data
2626
2627
2628
2629
2630
2631
        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
2632
            if not isinstance(train_set, Dataset):
2633
                raise TypeError(f'Training data should be Dataset instance, met {type(train_set).__name__}')
Guolin Ke's avatar
Guolin Ke committed
2634
            if train_set._predictor is not self.__init_predictor:
2635
2636
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2637
2638
2639
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
2640
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
2641
            self.__inner_predict_buffer[0] = None
2642
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
2643
2644
        is_finished = ctypes.c_int(0)
        if fobj is None:
2645
            if self.__set_objective_to_none:
2646
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
2647
2648
2649
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
2650
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
2651
2652
            return is_finished.value == 1
        else:
2653
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
2654
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
2655
2656
2657
2658
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

    def __boost(self, grad, hess):
2659
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
2660

Nikita Titov's avatar
Nikita Titov committed
2661
2662
        .. note::

2663
2664
            Score is returned before any transformation,
            e.g. it is raw margin instead of probability of positive class for binary task.
Nikita Titov's avatar
Nikita Titov committed
2665
2666
2667
            For multi-class task, the score is group by class_id first, then group by row_id.
            If you want to get i-th row score in j-th class, the access way is score[j * num_data + i]
            and you should group grad and hess in this way as well.
2668

wxchan's avatar
wxchan committed
2669
2670
        Parameters
        ----------
2671
        grad : list or numpy 1-D array
2672
2673
            The value of the first order derivative (gradient) of the loss
            with respect to the elements of score for each sample point.
2674
        hess : list or numpy 1-D array
2675
2676
            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
2677
2678
2679

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2680
2681
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
2682
        """
2683
2684
        grad = list_to_1d_numpy(grad, name='gradient')
        hess = list_to_1d_numpy(hess, name='hessian')
2685
2686
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
2687
        if len(grad) != len(hess):
2688
            raise ValueError(f"Lengths of gradient({len(grad)}) and hessian({len(hess)}) don't match")
wxchan's avatar
wxchan committed
2689
2690
2691
2692
2693
2694
        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)))
2695
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
2696
2697
2698
        return is_finished.value == 1

    def rollback_one_iter(self):
Nikita Titov's avatar
Nikita Titov committed
2699
2700
2701
2702
2703
2704
2705
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
2706
2707
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
2708
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
2709
        return self
wxchan's avatar
wxchan committed
2710
2711

    def current_iteration(self):
2712
2713
2714
2715
2716
2717
2718
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
2719
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2720
2721
2722
2723
2724
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
    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

2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
    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
2781
    def eval(self, data, name, feval=None):
2782
        """Evaluate for data.
wxchan's avatar
wxchan committed
2783
2784
2785

        Parameters
        ----------
2786
2787
2788
2789
2790
        data : Dataset
            Data for the evaluating.
        name : string
            Name of the data.
        feval : callable or None, optional (default=None)
2791
            Customized evaluation function.
2792
            Should accept two parameters: preds, eval_data,
2793
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2794
2795
2796

                preds : list or numpy 1-D array
                    The predicted values.
2797
2798
                    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.
2799
2800
2801
                eval_data : Dataset
                    The evaluation dataset.
                eval_name : string
Andrew Ziem's avatar
Andrew Ziem committed
2802
                    The name of evaluation function (without whitespace).
2803
2804
2805
2806
2807
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2808
2809
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
2810

wxchan's avatar
wxchan committed
2811
2812
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2813
        result : list
2814
            List with evaluation results.
wxchan's avatar
wxchan committed
2815
        """
Guolin Ke's avatar
Guolin Ke committed
2816
2817
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
2818
2819
2820
2821
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
2822
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
2823
2824
2825
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
2826
        # need to push new valid data
wxchan's avatar
wxchan committed
2827
2828
2829
2830
2831
2832
2833
        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):
2834
        """Evaluate for training data.
wxchan's avatar
wxchan committed
2835
2836
2837

        Parameters
        ----------
2838
        feval : callable or None, optional (default=None)
2839
            Customized evaluation function.
2840
2841
            Should accept two parameters: preds, train_data,
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2842
2843
2844

                preds : list or numpy 1-D array
                    The predicted values.
2845
2846
                    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.
2847
2848
2849
                train_data : Dataset
                    The training dataset.
                eval_name : string
Andrew Ziem's avatar
Andrew Ziem committed
2850
                    The name of evaluation function (without whitespace).
2851
2852
2853
2854
2855
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2856
2857
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
wxchan's avatar
wxchan committed
2858
2859
2860

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2861
        result : list
2862
            List with evaluation results.
wxchan's avatar
wxchan committed
2863
        """
2864
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
2865
2866

    def eval_valid(self, feval=None):
2867
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
2868
2869
2870

        Parameters
        ----------
2871
        feval : callable or None, optional (default=None)
2872
            Customized evaluation function.
2873
            Should accept two parameters: preds, valid_data,
2874
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2875
2876
2877

                preds : list or numpy 1-D array
                    The predicted values.
2878
2879
                    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.
2880
2881
2882
                valid_data : Dataset
                    The validation dataset.
                eval_name : string
Andrew Ziem's avatar
Andrew Ziem committed
2883
                    The name of evaluation function (without whitespace).
2884
2885
2886
2887
2888
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2889
2890
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
wxchan's avatar
wxchan committed
2891
2892
2893

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2894
        result : list
2895
            List with evaluation results.
wxchan's avatar
wxchan committed
2896
        """
2897
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
2898
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
2899

2900
    def save_model(self, filename, num_iteration=None, start_iteration=0, importance_type='split'):
2901
        """Save Booster to file.
wxchan's avatar
wxchan committed
2902
2903
2904

        Parameters
        ----------
2905
2906
        filename : string
            Filename to save Booster.
2907
2908
2909
2910
        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
2911
        start_iteration : int, optional (default=0)
2912
            Start index of the iteration that should be saved.
2913
2914
2915
2916
        importance_type : string, optional (default="split")
            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
2917
2918
2919
2920
2921

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
2922
        """
2923
        if num_iteration is None:
2924
            num_iteration = self.best_iteration
2925
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
2926
2927
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
2928
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
2929
            ctypes.c_int(num_iteration),
2930
            ctypes.c_int(importance_type_int),
wxchan's avatar
wxchan committed
2931
            c_str(filename)))
2932
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
2933
        return self
wxchan's avatar
wxchan committed
2934

2935
    def shuffle_models(self, start_iteration=0, end_iteration=-1):
2936
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
2937

2938
2939
2940
        Parameters
        ----------
        start_iteration : int, optional (default=0)
2941
            The first iteration that will be shuffled.
2942
2943
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
2944
            If <= 0, means the last available iteration.
2945

Nikita Titov's avatar
Nikita Titov committed
2946
2947
2948
2949
        Returns
        -------
        self : Booster
            Booster with shuffled models.
2950
        """
2951
2952
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
2953
2954
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
2955
        return self
2956
2957
2958
2959
2960
2961

    def model_from_string(self, model_str, verbose=True):
        """Load Booster from a string.

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2962
        model_str : string
2963
            Model will be loaded from this string.
Nikita Titov's avatar
Nikita Titov committed
2964
2965
        verbose : bool, optional (default=True)
            Whether to print messages while loading model.
2966
2967
2968

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2969
        self : Booster
2970
2971
            Loaded Booster object.
        """
2972
2973
2974
2975
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
2976
2977
2978
2979
2980
2981
2982
2983
2984
        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)))
2985
        if verbose:
2986
            _log_info(f'Finished loading model, total used {int(out_num_iterations.value)} iterations')
2987
        self.__num_class = out_num_class.value
2988
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
2989
2990
        return self

2991
    def model_to_string(self, num_iteration=None, start_iteration=0, importance_type='split'):
2992
        """Save Booster to string.
2993

2994
2995
2996
2997
2998
2999
        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
3000
        start_iteration : int, optional (default=0)
3001
            Start index of the iteration that should be saved.
3002
3003
3004
3005
        importance_type : string, optional (default="split")
            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.
3006
3007
3008

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
3009
        str_repr : string
3010
3011
            String representation of Booster.
        """
3012
        if num_iteration is None:
3013
            num_iteration = self.best_iteration
3014
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3015
        buffer_len = 1 << 20
3016
        tmp_out_len = ctypes.c_int64(0)
3017
3018
3019
3020
        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,
3021
            ctypes.c_int(start_iteration),
3022
            ctypes.c_int(num_iteration),
3023
            ctypes.c_int(importance_type_int),
3024
            ctypes.c_int64(buffer_len),
3025
3026
3027
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
3028
        # if buffer length is not long enough, re-allocate a buffer
3029
3030
3031
3032
3033
        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,
3034
                ctypes.c_int(start_iteration),
3035
                ctypes.c_int(num_iteration),
3036
                ctypes.c_int(importance_type_int),
3037
                ctypes.c_int64(actual_len),
3038
3039
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
3040
        ret = string_buffer.value.decode('utf-8')
3041
3042
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
3043

3044
    def dump_model(self, num_iteration=None, start_iteration=0, importance_type='split'):
Nikita Titov's avatar
Nikita Titov committed
3045
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
3046

3047
3048
        Parameters
        ----------
3049
3050
3051
3052
        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
3053
        start_iteration : int, optional (default=0)
3054
            Start index of the iteration that should be dumped.
3055
3056
3057
3058
        importance_type : string, optional (default="split")
            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.
3059

wxchan's avatar
wxchan committed
3060
3061
        Returns
        -------
3062
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
3063
            JSON format of Booster.
wxchan's avatar
wxchan committed
3064
        """
3065
        if num_iteration is None:
3066
            num_iteration = self.best_iteration
3067
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
3068
        buffer_len = 1 << 20
3069
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
3070
3071
3072
3073
        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,
3074
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3075
            ctypes.c_int(num_iteration),
3076
            ctypes.c_int(importance_type_int),
3077
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
3078
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3079
            ptr_string_buffer))
wxchan's avatar
wxchan committed
3080
        actual_len = tmp_out_len.value
3081
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
3082
3083
3084
3085
3086
        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,
3087
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
3088
                ctypes.c_int(num_iteration),
3089
                ctypes.c_int(importance_type_int),
3090
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
3091
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3092
                ptr_string_buffer))
3093
        ret = json.loads(string_buffer.value.decode('utf-8'))
3094
3095
3096
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
                                                          default=json_default_with_numpy))
        return ret
wxchan's avatar
wxchan committed
3097

3098
    def predict(self, data, start_iteration=0, num_iteration=None,
3099
                raw_score=False, pred_leaf=False, pred_contrib=False,
3100
                data_has_header=False, is_reshape=True, **kwargs):
3101
        """Make a prediction.
wxchan's avatar
wxchan committed
3102
3103
3104

        Parameters
        ----------
3105
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
3106
3107
            Data source for prediction.
            If string, it represents the path to txt file.
3108
        start_iteration : int, optional (default=0)
3109
            Start index of the iteration to predict.
3110
            If <= 0, starts from the first iteration.
3111
        num_iteration : int or None, optional (default=None)
3112
3113
3114
3115
            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).
3116
3117
3118
3119
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
3120
3121
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
3122

Nikita Titov's avatar
Nikita Titov committed
3123
3124
3125
3126
3127
3128
3129
            .. 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.
3130

3131
3132
3133
3134
3135
        data_has_header : bool, optional (default=False)
            Whether the data has header.
            Used only if data is string.
        is_reshape : bool, optional (default=True)
            If True, result is reshaped to [nrow, ncol].
3136
3137
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
3138
3139
3140

        Returns
        -------
3141
        result : numpy array, scipy.sparse or list of scipy.sparse
3142
            Prediction result.
3143
            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
3144
        """
3145
        predictor = self._to_predictor(deepcopy(kwargs))
3146
        if num_iteration is None:
3147
            if start_iteration <= 0:
3148
3149
3150
3151
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
        return predictor.predict(data, start_iteration, num_iteration,
3152
3153
                                 raw_score, pred_leaf, pred_contrib,
                                 data_has_header, is_reshape)
wxchan's avatar
wxchan committed
3154

3155
    def refit(self, data, label, decay_rate=0.9, **kwargs):
Guolin Ke's avatar
Guolin Ke committed
3156
3157
3158
3159
        """Refit the existing Booster by new data.

        Parameters
        ----------
3160
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
Guolin Ke's avatar
Guolin Ke committed
3161
3162
            Data source for refit.
            If string, it represents the path to txt file.
3163
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
3164
3165
            Label for refit.
        decay_rate : float, optional (default=0.9)
3166
3167
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
3168
3169
        **kwargs
            Other parameters for refit.
3170
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
3171
3172
3173
3174
3175
3176

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
3177
3178
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
3179
        predictor = self._to_predictor(deepcopy(kwargs))
3180
        leaf_preds = predictor.predict(data, -1, pred_leaf=True)
3181
        nrow, ncol = leaf_preds.shape
3182
3183
3184
3185
        out_is_linear = ctypes.c_bool(False)
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
Nikita Titov's avatar
Nikita Titov committed
3186
3187
3188
3189
3190
        new_params = _choose_param_value(
            main_param_name="linear_tree",
            params=self.params,
            default_value=None
        )
3191
3192
        new_params["linear_tree"] = out_is_linear.value
        train_set = Dataset(data, label, silent=True, params=new_params)
3193
        new_params['refit_decay_rate'] = decay_rate
3194
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
3195
3196
3197
3198
3199
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
3200
        ptr_data, _, _ = c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
3201
3202
3203
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
3204
3205
            ctypes.c_int32(nrow),
            ctypes.c_int32(ncol)))
3206
3207
        new_booster.network = self.network
        new_booster.__attr = self.__attr.copy()
Guolin Ke's avatar
Guolin Ke committed
3208
3209
        return new_booster

3210
    def get_leaf_output(self, tree_id, leaf_id):
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
        """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.
        """
3225
3226
3227
3228
3229
3230
3231
3232
        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

3233
    def _to_predictor(self, pred_parameter=None):
3234
        """Convert to predictor."""
3235
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
3236
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
3237
3238
        return predictor

3239
    def num_feature(self):
3240
3241
3242
3243
3244
3245
3246
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
3247
3248
3249
3250
3251
3252
        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
3253
    def feature_name(self):
3254
        """Get names of features.
wxchan's avatar
wxchan committed
3255
3256
3257

        Returns
        -------
3258
3259
        result : list
            List with names of features.
wxchan's avatar
wxchan committed
3260
        """
3261
        num_feature = self.num_feature()
3262
        # Get name of features
wxchan's avatar
wxchan committed
3263
        tmp_out_len = ctypes.c_int(0)
3264
3265
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
3266
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(num_feature)]
wxchan's avatar
wxchan committed
3267
3268
3269
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
3270
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
3271
            ctypes.byref(tmp_out_len),
3272
            ctypes.c_size_t(reserved_string_buffer_size),
3273
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3274
3275
3276
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
        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))
3289
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
3290

3291
    def feature_importance(self, importance_type='split', iteration=None):
3292
        """Get feature importances.
3293

3294
3295
        Parameters
        ----------
3296
3297
3298
3299
        importance_type : string, optional (default="split")
            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.
3300
3301
3302
3303
        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).
3304

3305
3306
        Returns
        -------
3307
3308
        result : numpy array
            Array with feature importances.
3309
        """
3310
3311
        if iteration is None:
            iteration = self.best_iteration
3312
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
3313
        result = np.empty(self.num_feature(), dtype=np.float64)
3314
3315
3316
3317
3318
3319
        _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:
3320
            return result.astype(np.int32)
3321
3322
        else:
            return result
3323

3324
3325
3326
3327
3328
3329
3330
3331
3332
    def get_split_value_histogram(self, feature, bins=None, xgboost_style=False):
        """Get split value histogram for the specified feature.

        Parameters
        ----------
        feature : int or string
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
            If string, interpreted as name.
3333

Nikita Titov's avatar
Nikita Titov committed
3334
3335
3336
            .. warning::

                Categorical features are not supported.
3337

3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
        bins : int, string or None, optional (default=None)
            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.
            If string, it should be one from the list of the supported values by ``numpy.histogram()`` function.
        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
3360
                if feature_names is not None and isinstance(feature, str):
3361
3362
3363
3364
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
3365
                    if isinstance(root['threshold'], str):
3366
3367
3368
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
                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'])

3379
        if bins is None or isinstance(bins, int) and xgboost_style:
3380
3381
3382
3383
3384
3385
3386
            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:
3387
                return pd_DataFrame(ret, columns=['SplitValue', 'Count'])
3388
3389
3390
3391
3392
            else:
                return ret
        else:
            return hist, bin_edges

wxchan's avatar
wxchan committed
3393
    def __inner_eval(self, data_name, data_idx, feval=None):
3394
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
3395
        if data_idx >= self.__num_dataset:
3396
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3397
3398
3399
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
3400
            result = np.empty(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
3401
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3402
3403
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3404
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3405
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3406
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
3407
            if tmp_out_len.value != self.__num_inner_eval:
3408
                raise ValueError("Wrong length of eval results")
3409
            for i in range(self.__num_inner_eval):
3410
3411
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
3412
3413
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
3414
3415
3416
3417
3418
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
3419
3420
3421
3422
3423
3424
3425
3426
3427
            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
3428
3429
3430
3431
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

    def __inner_predict(self, data_idx):
3432
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
3433
        if data_idx >= self.__num_dataset:
3434
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3435
3436
3437
3438
3439
        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
3440
            self.__inner_predict_buffer[data_idx] = np.empty(n_preds, dtype=np.float64)
3441
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
3442
3443
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
3444
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
3445
3446
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3447
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3448
3449
3450
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
3451
                raise ValueError(f"Wrong length of predict results for data {data_idx}")
wxchan's avatar
wxchan committed
3452
3453
3454
3455
            self.__is_predicted_cur_iter[data_idx] = True
        return self.__inner_predict_buffer[data_idx]

    def __get_eval_info(self):
3456
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
3457
3458
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
3459
            out_num_eval = ctypes.c_int(0)
3460
            # Get num of inner evals
wxchan's avatar
wxchan committed
3461
3462
3463
3464
3465
            _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:
3466
                # Get name of eval metrics
Guolin Ke's avatar
Guolin Ke committed
3467
                tmp_out_len = ctypes.c_int(0)
3468
3469
3470
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
3471
                    ctypes.create_string_buffer(reserved_string_buffer_size) for _ in range(self.__num_inner_eval)
3472
                ]
wxchan's avatar
wxchan committed
3473
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
3474
3475
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
3476
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
3477
                    ctypes.byref(tmp_out_len),
3478
                    ctypes.c_size_t(reserved_string_buffer_size),
3479
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3480
3481
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
3482
                    raise ValueError("Length of eval names doesn't equal with num_evals")
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
                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
                ]
3503

wxchan's avatar
wxchan committed
3504
    def attr(self, key):
3505
        """Get attribute string from the Booster.
wxchan's avatar
wxchan committed
3506
3507
3508

        Parameters
        ----------
3509
3510
        key : string
            The name of the attribute.
wxchan's avatar
wxchan committed
3511
3512
3513

        Returns
        -------
3514
3515
        value : string or None
            The attribute value.
Nikita Titov's avatar
Nikita Titov committed
3516
            Returns None if attribute does not exist.
wxchan's avatar
wxchan committed
3517
        """
3518
        return self.__attr.get(key, None)
wxchan's avatar
wxchan committed
3519
3520

    def set_attr(self, **kwargs):
3521
        """Set attributes to the Booster.
wxchan's avatar
wxchan committed
3522
3523
3524
3525

        Parameters
        ----------
        **kwargs
3526
3527
            The attributes to set.
            Setting a value to None deletes an attribute.
Nikita Titov's avatar
Nikita Titov committed
3528
3529
3530
3531

        Returns
        -------
        self : Booster
3532
            Booster with set attributes.
wxchan's avatar
wxchan committed
3533
3534
3535
        """
        for key, value in kwargs.items():
            if value is not None:
3536
                if not isinstance(value, str):
Nikita Titov's avatar
Nikita Titov committed
3537
                    raise ValueError("Only string values are accepted")
wxchan's avatar
wxchan committed
3538
3539
3540
                self.__attr[key] = value
            else:
                self.__attr.pop(key, None)
Nikita Titov's avatar
Nikita Titov committed
3541
        return self