-
Notifications
You must be signed in to change notification settings - Fork 17
/
dataset.py
418 lines (369 loc) · 18.4 KB
/
dataset.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
import numpy as np
import os
# os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
# os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import model
from tqdm import tqdm
import tensorflow as tf
# physical_devices = tf.config.list_physical_devices('GPU')
# try:
# tf.config.experimental.set_memory_growth(physical_devices[0], True)
# except:
# # Invalid device or cannot modify virtual devices once initialized.
# pass
import numpy as np
def sample_first_timestamps(y, n, window_length, seed):
'''
# TODO: No matter what seed is, fix labeled timestamps w.r.t 'n' to reduce effect of initial labels.
Args:
x: input time series
y: timestamp labels
n: number of timestamps for each class
Returns:
mask vector for timestamp-wise classification. If mask=1, label exists and back propagation occurs at the timestamp.
'''
if n > 0:
num_class = len(np.unique(y))
mask = np.zeros_like(y)
y_short = y[window_length // 2:-window_length // 2]
np.random.seed(
seed=0) # fix sampled timestamp for each dataset. change 0 to seed when randomizing initial labels.
for i in range(num_class):
indice = np.where(y_short == i)[0]
# print(indice,y,y_short)
if len(indice) < n:
print(f"number of timestamp label for class {i} is less than the number of required labels {n}")
spld_indice = np.random.choice(indice, size=n, replace=False)
spld_indice += window_length // 2
mask[spld_indice] = 1
np.random.seed()
else:
mask = np.ones_like(y)
print("full labels for each class are used")
return mask
def sample_true_timestamp_label_from_ratio(y, p, num_class):
'''
Return timestamp label array that p*len(y) timestamp labels are randomly changed to different labels.
Args:
y: true timestamp labels
p: ratio of true timestamp labels
num_class: the number of class
Returns:
Randomized y with ratio of p.
'''
if p == 1:
return y
NUM_SPL_TS = int(len(y) * p)
np.random.seed(0)
SPL_TS_TO_BE_CHGD = np.random.choice(list(range(len(y))), size=len(y)-NUM_SPL_TS, replace=False)
np.random.seed()
y[SPL_TS_TO_BE_CHGD] = np.random.choice(list(range(num_class)), size = len(SPL_TS_TO_BE_CHGD), replace=True)
return y
def get_dataset(DATA):
file_path = "datasets"
X_long = np.load(os.path.join(file_path, DATA + "_X_long.npy"))
y_long = np.load(os.path.join(file_path, DATA + "_y_long.npy"))
y_seg_long = np.load(os.path.join(file_path, DATA + "_y_seg_long.npy"))
file_boundaries = np.load(os.path.join(file_path, DATA + "_file_boundaries.npy"))
print(f"{DATA} loaded from preprocessed files from {file_path}")
print(X_long.shape, y_long.shape, y_seg_long.shape)
return X_long, y_long, y_seg_long, file_boundaries
def train_test_generator(X, y, y_seg, mask, file_boundaries, seed, K):
assert (seed <= K - 1)
test_data_start = len(X) // K * seed
if seed == K - 1:
test_data_end = len(X)
else:
test_data_end = len(X) // K * (seed + 1)
X_long_train = np.concatenate([X[:test_data_start], X[test_data_end:]])
y_long_train = np.concatenate([y[:test_data_start], y[test_data_end:]])
y_seg_long_train = np.concatenate([y_seg[:test_data_start], y_seg[test_data_end:]])
mask_long_train = np.concatenate([mask[:test_data_start], mask[test_data_end:]])
file_boundaries_train = np.concatenate(
[file_boundaries[:test_data_start], file_boundaries[test_data_end:]])
X_long_test = X[test_data_start:test_data_end]
y_long_test = y[test_data_start:test_data_end]
y_seg_long_test = y_seg[test_data_start:test_data_end]
mask_long_test = mask[test_data_start:test_data_end] # TODO: labeled_or_not index check needed
file_boundaries_test = file_boundaries[test_data_start:test_data_end]
print(f"NUM_CLASS: {len(np.unique(y))}")
print(X_long_train.shape, y_long_train.shape, y_seg_long_train.shape, mask_long_train.shape,
file_boundaries_train.shape)
print(X_long_test.shape, y_long_test.shape, y_seg_long_test.shape, mask_long_test.shape,
file_boundaries_test.shape)
return X_long_train, y_long_train, y_seg_long_train, mask_long_train, file_boundaries_train, X_long_test, y_long_test, y_seg_long_test, mask_long_test, file_boundaries_test
def masked_timeseries_dataset(
data,
targets,
mask,
sequence_length,
sequence_stride=1,
sampling_rate=1,
batch_size=128,
shuffle=False,
seed=None,
start_index=None,
end_index=None,
center_timestamps=[]):
if start_index:
if start_index < 0:
raise ValueError(f'`start_index` must be 0 or greater. Received: '
f'start_index={start_index}')
if start_index >= len(data):
raise ValueError(f'`start_index` must be lower than the length of the '
f'data. Received: start_index={start_index}, for data '
f'of length {len(data)}')
if end_index:
if start_index and end_index <= start_index:
raise ValueError(f'`end_index` must be higher than `start_index`. '
f'Received: start_index={start_index}, and '
f'end_index={end_index} ')
if end_index >= len(data):
raise ValueError(f'`end_index` must be lower than the length of the '
f'data. Received: end_index={end_index}, for data of '
f'length {len(data)}')
if end_index <= 0:
raise ValueError('`end_index` must be higher than 0. '
f'Received: end_index={end_index}')
# Validate strides
if sampling_rate <= 0:
raise ValueError(f'`sampling_rate` must be higher than 0. Received: '
f'sampling_rate={sampling_rate}')
if sampling_rate >= len(data):
raise ValueError(f'`sampling_rate` must be lower than the length of the '
f'data. Received: sampling_rate={sampling_rate}, for data '
f'of length {len(data)}')
if sequence_stride <= 0:
raise ValueError(f'`sequence_stride` must be higher than 0. Received: '
f'sequence_stride={sequence_stride}')
if sequence_stride >= len(data):
raise ValueError(f'`sequence_stride` must be lower than the length of the '
f'data. Received: sequence_stride={sequence_stride}, for '
f'data of length {len(data)}')
if start_index is None:
start_index = 0
if end_index is None:
end_index = len(data)
# Determine the lowest dtype to store start positions (to lower memory usage).
num_seqs = end_index - start_index - (sequence_length * sampling_rate) + 1
if targets is not None:
num_seqs = min(num_seqs, len(targets))
if num_seqs < 2147483647:
index_dtype = 'int32'
else:
index_dtype = 'int64'
# Generate start positions
# max_num_window = (end_index - start_index-sequence_length)//sequence_stride
# for i in range(max_num_window):
# if np.sum(mask[i*sequence_stride:i*sequence_stride+sequence_length])>0 and (
# i*sequence_stride-sequence_length//2)>0:
# start_positions.append(i*sequence_stride-sequence_length//2)
if len(center_timestamps)>0:
start_positions = np.array(center_timestamps).astype(index_dtype)-sequence_length//2
else:
start_positions = np.array(np.where(mask == 1)[0] - sequence_length // 2, dtype=index_dtype)
# print(len(start_positions), np.sort(start_positions))
start_positions = start_positions[start_positions >= 0]
# print(len(start_positions), np.sort(start_positions))
# start_positions = np.arange(0, num_seqs, sequence_stride, dtype=index_dtype)
if shuffle:
if seed is None:
seed = np.random.randint(1e6)
rng = np.random.RandomState(seed)
rng.shuffle(start_positions)
sequence_length = tf.cast(sequence_length, dtype=index_dtype)
sampling_rate = tf.cast(sampling_rate, dtype=index_dtype)
positions_ds = tf.data.Dataset.from_tensors(start_positions).repeat()
# print(start_positions, sequence_length, sampling_rate)
# For each initial window position, generates indices of the window elements
indices = tf.data.Dataset.zip(
(tf.data.Dataset.range(len(start_positions)), positions_ds)).map(
lambda i, positions: tf.range( # pylint: disable=g-long-lambda
positions[i],
positions[i] + sequence_length * sampling_rate, #TODO: add varying sequence length
sampling_rate),
num_parallel_calls=tf.data.AUTOTUNE)
dataset = sequences_from_indices(data, indices, start_index, end_index)
if targets is not None:
indices = tf.data.Dataset.zip(
(tf.data.Dataset.range(len(start_positions)), positions_ds)).map(
lambda i, positions: positions[i],
num_parallel_calls=tf.data.AUTOTUNE)
target_ds = sequences_from_indices(
targets, indices, start_index, end_index)
dataset = tf.data.Dataset.zip((dataset, target_ds))
dataset = dataset.prefetch(tf.data.AUTOTUNE)
if batch_size is not None:
if shuffle:
# Shuffle locally at each iteration
dataset = dataset.shuffle(buffer_size=batch_size * 8, seed=seed)
dataset = dataset.batch(batch_size)
else:
if shuffle:
dataset = dataset.shuffle(buffer_size=1024, seed=seed)
return dataset
def sequences_from_indices(array, indices_ds, start_index, end_index):
dataset = tf.data.Dataset.from_tensors(array[start_index : end_index])
dataset = tf.data.Dataset.zip((dataset.repeat(), indices_ds)).map(
lambda steps, inds: tf.gather(steps, inds), # pylint: disable=unnecessary-lambda
num_parallel_calls=tf.data.AUTOTUNE)
return dataset
def varying_context_timeseries_dataset(
data,
targets,
overlap_length,
sequence_length, # maximum window length = overlap_length + 2 * context_length
sequence_stride=1,
sampling_rate=1,
batch_size=128,
iterations=50000,
varying_context_per_batch=True,
shuffle=False,
seed=None,
start_index=None,
end_index=None):
'''
Make batched windows with different length for each batch.
:return: Tensorflow dataset
'''
if start_index:
if start_index < 0:
raise ValueError(f'`start_index` must be 0 or greater. Received: '
f'start_index={start_index}')
if start_index >= len(data):
raise ValueError(f'`start_index` must be lower than the length of the '
f'data. Received: start_index={start_index}, for data '
f'of length {len(data)}')
if end_index:
if start_index and end_index <= start_index:
raise ValueError(f'`end_index` must be higher than `start_index`. '
f'Received: start_index={start_index}, and '
f'end_index={end_index} ')
if end_index >= len(data):
raise ValueError(f'`end_index` must be lower than the length of the '
f'data. Received: end_index={end_index}, for data of '
f'length {len(data)}')
if end_index <= 0:
raise ValueError('`end_index` must be higher than 0. '
f'Received: end_index={end_index}')
# Validate strides
if sampling_rate <= 0:
raise ValueError(f'`sampling_rate` must be higher than 0. Received: '
f'sampling_rate={sampling_rate}')
if sampling_rate >= len(data):
raise ValueError(f'`sampling_rate` must be lower than the length of the '
f'data. Received: sampling_rate={sampling_rate}, for data '
f'of length {len(data)}')
if sequence_stride <= 0:
raise ValueError(f'`sequence_stride` must be higher than 0. Received: '
f'sequence_stride={sequence_stride}')
if sequence_stride >= len(data):
raise ValueError(f'`sequence_stride` must be lower than the length of the '
f'data. Received: sequence_stride={sequence_stride}, for '
f'data of length {len(data)}')
if start_index is None:
start_index = 0
if end_index is None:
end_index = len(data)
# Determine the lowest dtype to store start positions (to lower memory usage).
num_seqs = end_index - start_index - (sequence_length * sampling_rate) + 1
if targets is not None:
num_seqs = min(num_seqs, len(targets))
if num_seqs < 2147483647:
index_dtype = 'int32'
else:
index_dtype = 'int64'
total_length = len(data)
center_timestamps = np.arange(sequence_length,total_length-sequence_length,overlap_length)
start_positions = np.array(center_timestamps).astype(index_dtype)-sequence_length//2
start_positions = start_positions[start_positions >= 0].tolist()
start_positions = start_positions*((iterations*batch_size)//len(start_positions))+start_positions[:(iterations*batch_size)%len(start_positions)]
if shuffle:
if seed is None:
seed = np.random.randint(1e6)
rng = np.random.RandomState(seed)
rng.shuffle(start_positions)
if varying_context_per_batch:
context_lengths_sampled = np.repeat(np.random.choice(list(range(2,sequence_length-overlap_length,2)),size=iterations,replace=True), batch_size)
else:
context_lengths_sampled = np.repeat([sequence_length-overlap_length]*iterations, batch_size)
context_lengths_sampled = tf.cast(context_lengths_sampled,dtype=index_dtype)
context_lengths = tf.data.Dataset.from_tensors(context_lengths_sampled).repeat()
sampling_rate = tf.cast(sampling_rate, dtype=index_dtype)
positions_ds = tf.data.Dataset.from_tensors(start_positions).repeat()
indices = tf.data.Dataset.zip(
(tf.data.Dataset.range(len(start_positions)), positions_ds, context_lengths)).map(
lambda i, positions, context_lengths: tf.range( # pylint: disable=g-long-lambda
positions[i],
positions[i] + (overlap_length+context_lengths[i]) * sampling_rate, #TODO: add varying sequence length
sampling_rate),
num_parallel_calls=tf.data.AUTOTUNE)
dataset = sequences_from_indices(data, indices, start_index, end_index)
if targets is not None:
indices = tf.data.Dataset.zip(
(tf.data.Dataset.range(len(start_positions)), positions_ds)).map(
lambda i, positions: positions[i],
num_parallel_calls=tf.data.AUTOTUNE)
target_ds = sequences_from_indices(
targets, indices, start_index, end_index)
dataset = tf.data.Dataset.zip((dataset, target_ds))
dataset = dataset.prefetch(tf.data.AUTOTUNE)
if batch_size is not None:
if shuffle:
# Shuffle locally at each iteration
dataset = dataset.shuffle(buffer_size=batch_size * 8, seed=seed)
dataset = dataset.batch(batch_size)
else:
if shuffle:
dataset = dataset.shuffle(buffer_size=1024, seed=seed)
return dataset
if __name__=="__main__":
@tf.function
def train_cls(model, x, y, lossMask, lambd=0.15):
with tf.GradientTape() as tape:
outputs = model.call_classifier(x, training=True)
lossMask = tf.cast(lossMask, dtype=tf.float32)
# cls_loss = model.cls_loss(y, outputs[0])
# masked_cls_loss = tf.math.reduce_sum(tf.math.multiply(cls_loss,lossMask))
# mean_masked_cls_loss = tf.math.divide(masked_cls_loss,tf.math.reduce_sum(tf.cast(lossMask!=0,tf.float32)))
# loss = mean_masked_cls_loss + lambd*model.seg_loss([],outputs[0])
# print(f"{cls_loss}\n{masked_cls_loss}\n{mean_masked_cls_loss}")
loss = tf.math.divide(tf.math.reduce_sum(tf.math.multiply(model.cls_loss(y, outputs[0]),lossMask)),tf.math.reduce_sum(tf.cast(lossMask!=0,tf.float32))) + lambd*model.seg_loss([],outputs[0])
for i in range(len(model.tcn_stage)-1):
loss += tf.math.divide(tf.math.reduce_sum(tf.math.multiply(model.cls_loss(y, outputs[i+1]),lossMask)), tf.math.reduce_sum(tf.cast(lossMask!=0,tf.float32))) + lambd*model.seg_loss([],outputs[i+1])
gradients = tape.gradient(loss, model.trainable_variables)
model.optimizer.apply_gradients(zip(gradients, model.trainable_variables))
return loss
DATA = "50salads"
file_path = "datasets"
X_long = np.load(os.path.join(file_path, DATA + "_X_long.npy"))
y_long = np.load(os.path.join(file_path, DATA + "_y_long.npy"))
y_seg_long = np.load(os.path.join(file_path, DATA + "_y_seg_long.npy"))
file_boundaries = np.load(os.path.join(file_path, DATA + "_file_boundaries.npy"))
# mask = np.random.choice([0,1],size=len(y_long))
mask = np.ones_like(y_long)
mask = np.reshape(mask,(len(y_long),1))
y_long = np.reshape(y_long,(len(y_long),1))
X_mask_y = np.concatenate((X_long,mask,y_long),axis=1)
NUM_CLASS = len(np.unique(y_long))
dim = X_long.shape[1]
print(X_mask_y.shape)
X_mask_y_dataset = masked_timeseries_dataset(
data=X_mask_y,
targets=None,
mask=mask,
sequence_length=2048,
sequence_stride=2048,
shuffle=True,
batch_size=1, )
for example_inputs in X_mask_y_dataset.take(1):
print(f'Inputs shape (batch, time, features): {example_inputs.shape}')
print(X_mask_y_dataset, len(X_mask_y_dataset))
models = model.MSTCN(NUM_CLASS, lr=0.005, num_stage=1, num_filters=64) # models include contrastive layer and classificaiton layer
models(np.zeros((1, 2048, dim)))
for i in range(50):
for X_mask_y_batch in tqdm(X_mask_y_dataset, leave=False):
X_batch = X_mask_y_batch[:,:,:-2]
mask_batch = X_mask_y_batch[:,:,-2]
y_batch = X_mask_y_batch[:,:,-1]
loss = train_cls(models, X_batch, y_batch, mask_batch)