-
Notifications
You must be signed in to change notification settings - Fork 7
/
data_convertor.py
568 lines (481 loc) · 21.2 KB
/
data_convertor.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
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
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
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
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
from functools import partial
from inspect import signature
import pandas as pd
import torch
import torch.nn as nn
from fedot.core.data.data import InputData, OutputData
from fedot.core.repository.dataset_types import DataTypesEnum
from fedot.core.repository.tasks import Task, TaskTypesEnum
from pymonad.list import ListMonad
from sklearn.preprocessing import LabelEncoder
from fedot_ind.api.utils.data import check_multivariate_data
from fedot_ind.core.architecture.settings.computational import backend_methods as np
from fedot_ind.core.architecture.settings.computational import default_device
from fedot_ind.core.repository.constanst_repository import MATRIX, MULTI_ARRAY
class CustomDatasetTS:
def __init__(self, ts):
self.x = torch.from_numpy(DataConverter(
data=ts.features).convert_to_torch_format()).float()
self.y = torch.from_numpy(DataConverter(
data=ts.target).convert_to_torch_format()).float()
def __getitem__(self, index):
pass
def __len__(self):
pass
class CustomDatasetCLF:
def __init__(self, ts):
self.x = torch.from_numpy(ts.features).to(default_device()).float()
if ts.task.task_type == 'classification':
label_1 = max(ts.class_labels)
label_0 = min(ts.class_labels)
self.classes = ts.num_classes
if self.classes == 2 and label_1 != 1:
ts.target[ts.target == label_0] = 0
ts.target[ts.target == label_1] = 1
elif self.classes == 2 and label_0 != 0:
ts.target[ts.target == label_0] = 0
ts.target[ts.target == label_1] = 1
elif self.classes > 2 and label_0 == 1:
ts.target = ts.target - 1
if type(min(ts.target)[0]) is np.str_:
self.label_encoder = LabelEncoder()
ts.target = self.label_encoder.fit_transform(ts.target)
else:
self.label_encoder = None
try:
self.y = torch.nn.functional.one_hot(torch.from_numpy(ts.target).long(),
num_classes=self.classes).to(default_device()).squeeze(1)
except Exception:
self.y = torch.nn.functional.one_hot(torch.from_numpy(
ts.target).long()).to(default_device()).squeeze(1)
self.classes = self.y.shape[1]
else:
self.y = torch.from_numpy(ts.target).to(default_device()).float()
self.classes = 1
self.label_encoder = None
self.n_samples = ts.features.shape[0]
self.supplementary_data = ts.supplementary_data
def __getitem__(self, index):
return self.x[index], self.y[index]
def __len__(self):
return self.n_samples
class FedotConverter:
def __init__(self, data):
self.input_data = self.convert_to_input_data(data)
def convert_to_input_data(self, data):
if isinstance(data, InputData):
return data
elif isinstance(data, OutputData):
return data
elif isinstance(data[0], (np.ndarray, pd.DataFrame)):
return self.__init_input_data(features=data[0], target=data[1])
elif isinstance(data, list):
return data[0]
else:
try:
return torch.tensor(data)
except:
print(f"Can't convert {type(data)} to InputData", Warning)
def __init_input_data(self, features: pd.DataFrame,
target: np.ndarray,
task: str = 'classification') -> InputData:
if type(features) is np.ndarray:
features = pd.DataFrame(features)
is_multivariate_data = check_multivariate_data(features)
task_dict = {'classification': Task(TaskTypesEnum.classification),
'regression': Task(TaskTypesEnum.regression)}
if is_multivariate_data:
input_data = InputData(idx=np.arange(len(features)),
features=np.array(
features.values.tolist()).astype(float),
target=target.astype(
float).reshape(-1, 1),
task=task_dict[task],
data_type=MULTI_ARRAY)
else:
input_data = InputData(idx=np.arange(len(features)),
features=features.values,
target=np.ravel(target).reshape(-1, 1),
task=task_dict[task],
data_type=MATRIX)
return input_data
def convert_to_output_data(self,
prediction,
predict_data,
output_data_type):
if type(prediction) is OutputData:
output_data = prediction
elif type(prediction) is list:
output_data = prediction[0]
target = NumpyConverter(
data=np.concatenate([p.target for p in prediction], axis=0)).convert_to_torch_format()
predict = NumpyConverter(
data=np.concatenate([p.predict for p in prediction], axis=0)).convert_to_torch_format()
output_data = OutputData(idx=predict_data.idx,
features=predict_data.features,
predict=predict,
task=predict_data.task,
target=target,
data_type=output_data_type,
supplementary_data=predict_data.supplementary_data)
else:
output_data = OutputData(idx=predict_data.idx,
features=predict_data.features,
predict=prediction,
task=predict_data.task,
target=predict_data.target,
data_type=output_data_type,
supplementary_data=predict_data.supplementary_data)
return output_data
def unwrap_list_to_output(self):
data_type = self.input_data.data_type
predict_data_copy = self.input_data
return data_type, predict_data_copy
def convert_input_to_output(self):
return OutputData(idx=self.input_data.idx,
features=self.input_data.features,
task=self.input_data.task,
data_type=self.input_data.data_type,
target=self.input_data.target,
predict=self.input_data.features)
def convert_to_industrial_composing_format(self, mode):
if mode == 'one_dimensional':
new_features, new_target = [
array.reshape(array.shape[0], array.shape[1] * array.shape[2])
if array is not None and len(array.shape) > 2 else array
for array in [self.input_data.features, self.input_data.target]]
input_data = InputData(idx=self.input_data.idx,
features=new_features,
target=new_target,
task=self.input_data.task,
data_type=self.input_data.data_type,
supplementary_data=self.input_data.supplementary_data)
elif mode == 'channel_independent':
if len(self.input_data.features.shape) == 1:
self.input_data.features = self.input_data.features.reshape(1, -1)
flat_input = self.input_data.features.shape[0] == 1
feats = self.input_data.features if flat_input else self.input_data.features.swapaxes(1, 0)
input_data = [InputData(idx=self.input_data.idx,
features=features,
target=self.input_data.target,
task=self.input_data.task,
data_type=self.input_data.data_type,
supplementary_data=self.input_data.supplementary_data) for features in
feats]
elif mode == 'multi_dimensional':
features = NumpyConverter(
data=self.input_data.features).convert_to_torch_format()
input_data = InputData(idx=self.input_data.idx,
features=features,
target=self.input_data.target,
task=self.input_data.task,
data_type=DataTypesEnum.image,
supplementary_data=self.input_data.supplementary_data)
return input_data
class TensorConverter:
def __init__(self, data):
self.tensor_data = self.convert_to_tensor(data)
def convert_to_tensor(self, data):
if isinstance(data, torch.Tensor):
return data
elif isinstance(data, np.ndarray):
return torch.from_numpy(data)
elif isinstance(data, pd.DataFrame):
return torch.from_numpy(data.values)
elif isinstance(data, InputData):
return torch.from_numpy(data.features)
else:
print(f"Can't convert {type(data)} to torch.Tensor", Warning)
def convert_to_1d_tensor(self):
if self.tensor_data.ndim == 1:
return self.tensor_data
elif self.tensor_data.ndim == 3:
return self.tensor_data[0, 0]
if self.tensor_data.ndim == 2:
return self.tensor_data[0]
assert False, f'Please, review input dimensions {self.tensor_data.ndim}'
def convert_to_2d_tensor(self):
if self.tensor_data.ndim == 2:
return self.tensor_data
elif self.tensor_data.ndim == 1:
return self.tensor_data[None]
elif self.tensor_data.ndim == 3:
return self.tensor_data[0]
assert False, f'Please, review input dimensions {self.tensor_data.ndim}'
def convert_to_3d_tensor(self):
if self.tensor_data.ndim == 3:
return self.tensor_data
elif self.tensor_data.ndim == 1:
return self.tensor_data[None, None]
elif self.tensor_data.ndim == 2:
return self.tensor_data[:, None]
assert False, f'Please, review input dimensions {self.tensor_data.ndim}'
class NumpyConverter:
def __init__(self, data):
self.numpy_data = self.convert_to_array(data)
self.numpy_data = np.where(
np.isnan(self.numpy_data), 0, self.numpy_data)
self.numpy_data = np.where(
np.isinf(self.numpy_data), 0, self.numpy_data)
def convert_to_array(self, data):
if isinstance(data, np.ndarray):
return data
elif isinstance(data, torch.Tensor):
return data.detach().numpy()
elif isinstance(data, pd.DataFrame):
return data.values
elif isinstance(data, InputData):
return data.features
elif isinstance(data, CustomDatasetTS):
return data.x
elif isinstance(data, CustomDatasetCLF):
return data.x
else:
try:
return np.asarray(data)
except:
print(f"Can't convert {type(data)} to np.array", Warning)
def convert_to_1d_array(self):
if self.numpy_data.ndim == 1:
return self.numpy_data
elif self.numpy_data.ndim > 2:
return np.squeeze(self.numpy_data)
elif self.numpy_data.ndim == 2:
return self.numpy_data.flatten()
assert False, print(
f'Please, review input dimensions {self.numpy_data.ndim}')
def convert_to_2d_array(self):
if self.numpy_data.ndim == 2:
return self.numpy_data
elif self.numpy_data.ndim == 1:
return self.numpy_data.reshape(1, -1)
elif self.numpy_data.ndim == 3:
return self.numpy_data[0]
assert False, print(
f'Please, review input dimensions {self.numpy_data.ndim}')
def convert_to_3d_array(self):
if self.numpy_data.ndim == 3:
return self.numpy_data
elif self.numpy_data.ndim == 1:
return self.numpy_data[None, None]
elif self.numpy_data.ndim == 2:
return self.numpy_data[:, None]
assert False, print(
f'Please, review input dimensions {self.numpy_data.ndim}')
def convert_to_torch_format(self):
if self.numpy_data.ndim == 3:
return self.numpy_data
elif self.numpy_data.ndim == 1:
return self.numpy_data.reshape(self.numpy_data.shape[0],
1,
1)
elif self.numpy_data.ndim == 2 and self.numpy_data.shape[0] != 1:
# add 1 channel
return self.numpy_data.reshape(self.numpy_data.shape[0],
1,
self.numpy_data.shape[1])
elif self.numpy_data.ndim == 2 and self.numpy_data.shape[0] == 1:
# add transpose data to (features, channel = 1 , sample = 1)
return self.numpy_data.reshape(self.numpy_data.shape[1],
1,
self.numpy_data.shape[0])
elif self.numpy_data.ndim > 3:
return self.numpy_data.squeeze()
assert False, print(
f'Please, review input dimensions {self.numpy_data.ndim}')
def convert_to_ts_format(self):
if self.numpy_data.ndim > 1:
return self.numpy_data.squeeze()
else:
return self.numpy_data
class ConditionConverter:
def __init__(self, train_data, operation_implementation, mode):
self.train_data = train_data
self.operation_implementation = operation_implementation
self.mode = mode
@property
def have_transform_method(self):
return 'transform' in dir(self.operation_implementation)
@property
def have_fit_method(self):
return 'fit' in dir(self.operation_implementation)
@property
def have_predict_method(self):
return 'predict' in dir(self.operation_implementation)
@property
def have_predict_for_fit_method(self):
return 'predict_for_fit' in dir(self.operation_implementation)
@property
def is_one_dim_operation(self):
return self.mode == 'one_dimensional'
@property
def is_channel_independent_operation(self):
return self.mode == 'channel_independent'
@property
def is_multi_dimensional_operation(self):
return self.mode == 'multi_dimensional'
@property
def is_list_container(self):
return type(self.train_data) is list
@property
def is_operation_is_list_container(self):
return type(self.operation_implementation) is list
@property
def have_predict_atr(self):
return 'predict' in vars(self.operation_implementation[0]) if self.is_operation_is_list_container else False
@property
def is_fit_input_fedot(self):
return str(list(signature(self.operation_implementation.fit).parameters.keys())[0]) == 'input_data'
@property
def is_transform_input_fedot(self):
return str(list(signature(self.operation_implementation.transform).parameters.keys())[0]) == 'input_data'
@property
def is_predict_input_fedot(self):
return str(list(signature(self.operation_implementation.predict).parameters.keys())[0]) == 'input_data'
@property
def is_regression_of_forecasting_task(self):
return self.train_data.task.task_type.value in ['regression', 'ts_forecasting']
@property
def is_multi_output_target(self):
return isinstance(self.operation_implementation.classes_, list)
def output_mode_converter(self, output_mode, n_classes):
return self.operation_implementation.predict(self.train_data.features).reshape(-1, 1) if output_mode == 'labels' \
else self.probs_prediction_converter(output_mode, n_classes)
def probs_prediction_converter(self, output_mode, n_classes):
prediction = self.operation_implementation.predict_proba(
self.train_data.features)
if n_classes < 2:
raise ValueError(
'Data set contain only 1 target class. Please reformat your data.')
elif n_classes == 2 and output_mode != 'full_probs':
if self.is_multi_output_target:
prediction = np.stack([pred[:, 1]
for pred in prediction]).T
else:
prediction = prediction[:, 1]
return prediction
class DataConverter(TensorConverter, NumpyConverter):
def __init__(self, data):
super().__init__(data)
self.data = data
self.numpy_data = self.convert_to_array(data)
@property
def is_nparray(self):
return isinstance(self.data, np.ndarray)
@property
def is_tensor(self):
return isinstance(self.data, torch.Tensor)
@property
def is_zarr(self):
return hasattr(self.data, 'oindex')
@property
def is_dask(self):
return hasattr(self.data, 'compute')
@property
def is_memmap(self):
return isinstance(self.data, np.memmap)
@property
def is_slice(self):
return isinstance(self.data, slice)
@property
def is_tuple(self):
return isinstance(self.data, tuple)
@property
def is_none(self):
return self.data is None
@property
def is_fedot_data(self):
return isinstance(self.data, InputData)
@property
def is_exist(self):
return self.data is not None
def convert_to_data_type(self):
if isinstance(self.data, torch.Tensor):
self.data = self.data.to(dtype=torch.Tensor)
elif isinstance(self.data, np.ndarray):
self.data = self.data.astype(np.ndarray)
def convert_to_list(self):
if isinstance(self.data, list):
return self.data
elif isinstance(self.data, (np.ndarray, torch.Tensor)):
return self.data.tolist()
else:
try:
return list(self.data)
except:
print(f'passed object needs to be of type L, list, np.ndarray or torch.Tensor but is {type(self.data)}',
Warning)
def convert_data_to_1d(self):
if self.data.ndim == 1:
return self.data
if isinstance(self.data, np.ndarray):
return self.convert_to_1d_array()
if isinstance(self.data, torch.Tensor):
return self.convert_to_1d_tensor()
def convert_data_to_2d(self):
if self.data.ndim == 2:
return self.data
if isinstance(self.data, np.ndarray):
return self.convert_to_2d_array()
if isinstance(self.data, torch.Tensor):
return self.convert_to_2d_tensor()
def convert_data_to_3d(self):
if self.data.ndim == 3:
return self.data
if isinstance(self.data, (np.ndarray, pd.self.dataFrame)):
return self.convert_to_3d_array()
if isinstance(self.data, torch.Tensor):
return self.convert_to_3d_tensor()
def convert_to_monad_data(self):
if self.is_fedot_data:
features = np.array(ListMonad(*self.data.features.tolist()).value)
else:
features = np.array(ListMonad(*self.data.tolist()).value)
if len(features.shape) == 2 and features.shape[1] == 1:
features = features.reshape(1, -1)
elif len(features.shape) == 3 and features.shape[1] == 1:
features = features.squeeze()
return features
def convert_to_eigen_basis(self):
if self.is_fedot_data:
features = self.data.features
else:
features = np.array(ListMonad(*self.data.values.tolist()).value)
features = np.array([series[~np.isnan(series)]
for series in features])
return features
class NeuralNetworkConverter:
def __init__(self, layer):
self.layer = layer
@property
def is_layer(self, *args):
def _is_layer(cond=args):
return isinstance(self.layer, cond)
return partial(_is_layer, cond=args)
@property
def is_linear(self):
return isinstance(self.layer, nn.Linear)
@property
def is_batch_norm(self):
types = (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)
return isinstance(self.layer, types)
@property
def is_convolutional_linear(self):
types = (nn.Conv1d, nn.Conv2d, nn.Conv3d, nn.Linear)
return isinstance(self.layer, types)
@property
def is_affine(self):
return self.has_bias or self.has_weight
@property
def is_convolutional(self):
types = (nn.Conv1d, nn.Conv2d, nn.Conv3d)
return isinstance(self.layer, types)
@property
def has_bias(self):
return hasattr(self.layer, 'bias') and self.layer.bias is not None
@property
def has_weight(self):
return hasattr(self.layer, 'weight')
@property
def has_weight_or_bias(self):
return any((self.has_weight, self.has_bias))