-
Notifications
You must be signed in to change notification settings - Fork 1.7k
Expand file tree
/
Copy pathdata_utils.py
More file actions
1262 lines (1051 loc) · 35.4 KB
/
data_utils.py
File metadata and controls
1262 lines (1051 loc) · 35.4 KB
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
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
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
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
"""This library gathers utilities for data io operation.
Authors
* Mirco Ravanelli 2020
* Aku Rouhe 2020
* Samuele Cornell 2020
* Adel Moumen 2024
* Pierre Champion 2023
"""
import collections.abc
import csv
import gzip
import math
import os
import pathlib
import re
import shutil
import urllib.request
from numbers import Number
import torch
import tqdm
import speechbrain as sb
def undo_padding(batch, lengths):
"""Produces Python lists given a batch of sentences with
their corresponding relative lengths.
Arguments
---------
batch : torch.Tensor
Batch of sentences gathered in a batch.
lengths : torch.Tensor
Relative length of each sentence in the batch.
Returns
-------
as_list : list
A python list of the corresponding input tensor.
Example
-------
>>> batch = torch.rand([4, 100])
>>> lengths = torch.tensor([0.5, 0.6, 0.7, 1.0])
>>> snt_list = undo_padding(batch, lengths)
>>> len(snt_list)
4
"""
batch_max_len = batch.shape[1]
as_list = []
for seq, seq_length in zip(batch, lengths):
actual_size = int(torch.round(seq_length * batch_max_len))
seq_true = seq.narrow(0, 0, actual_size)
as_list.append(seq_true.tolist())
return as_list
def get_all_files(
dirName, match_and=None, match_or=None, exclude_and=None, exclude_or=None
):
"""Returns a list of files found within a folder.
Different options can be used to restrict the search to some specific
patterns.
Arguments
---------
dirName : str
The directory to search.
match_and : list
A list that contains patterns to match. The file is
returned if it matches all the entries in `match_and`.
match_or : list
A list that contains patterns to match. The file is
returned if it matches one or more of the entries in `match_or`.
exclude_and : list
A list that contains patterns to match. The file is
returned if it matches none of the entries in `exclude_and`.
exclude_or : list
A list that contains pattern to match. The file is
returned if it fails to match one of the entries in `exclude_or`.
Returns
-------
allFiles : list
The list of files matching the patterns.
Example
-------
>>> get_all_files("tests/samples/RIRs", match_and=["3.wav"])
['tests/samples/RIRs/rir3.wav']
"""
# Match/exclude variable initialization
match_and_entry = True
match_or_entry = True
exclude_or_entry = False
exclude_and_entry = False
# Create a list of file and sub directories
listOfFile = os.listdir(dirName)
allFiles = list()
# Iterate over all the entries
for entry in listOfFile:
# Create full path
fullPath = os.path.join(dirName, entry)
# If entry is a directory then get the list of files in this directory
if os.path.isdir(fullPath):
allFiles = allFiles + get_all_files(
fullPath,
match_and=match_and,
match_or=match_or,
exclude_and=exclude_and,
exclude_or=exclude_or,
)
else:
# Check match_and case
if match_and is not None:
match_and_entry = False
match_found = 0
for ele in match_and:
if ele in fullPath:
match_found = match_found + 1
if match_found == len(match_and):
match_and_entry = True
# Check match_or case
if match_or is not None:
match_or_entry = False
for ele in match_or:
if ele in fullPath:
match_or_entry = True
break
# Check exclude_and case
if exclude_and is not None:
match_found = 0
for ele in exclude_and:
if ele in fullPath:
match_found = match_found + 1
if match_found == len(exclude_and):
exclude_and_entry = True
# Check exclude_or case
if exclude_or is not None:
exclude_or_entry = False
for ele in exclude_or:
if ele in fullPath:
exclude_or_entry = True
break
# If needed, append the current file to the output list
if (
match_and_entry
and match_or_entry
and not (exclude_and_entry)
and not (exclude_or_entry)
):
allFiles.append(fullPath)
return allFiles
def get_list_from_csv(csvfile, field, delimiter=",", skipinitialspace=True):
"""Gets a list from the selected field of the input csv file.
Arguments
---------
csvfile: path
Path to the csv file.
field: str
Field of the csv file used to create the list.
delimiter: str
Delimiter of the csv file.
skipinitialspace: bool
Set it to true to skip initial spaces in the entries.
Returns
-------
The list of files in the given field of a csv
"""
lst = []
with open(csvfile, newline="", encoding="utf-8") as csvf:
reader = csv.DictReader(
csvf, delimiter=delimiter, skipinitialspace=skipinitialspace
)
for row in reader:
lst.append(row[field])
return lst
def split_list(seq, num):
"""Returns a list of splits in the sequence.
Arguments
---------
seq : iterable
The input list, to be split.
num : int
The number of chunks to produce.
Returns
-------
A list of lists, length num and containing all elements of seq.
Example
-------
>>> split_list([1, 2, 3, 4, 5, 6, 7, 8, 9], 4)
[[1, 2], [3, 4], [5, 6], [7, 8, 9]]
"""
# Average length of the chunk
avg = len(seq) / float(num)
out = []
last = 0.0
# Creating the chunks
while last < len(seq):
out.append(seq[int(last) : int(last + avg)])
last += avg
return out
def recursive_items(dictionary):
"""Yield each (key, value) of a nested dictionary.
Arguments
---------
dictionary : dict
The nested dictionary to list.
Yields
------
`(key, value)` tuples from the dictionary.
Example
-------
>>> rec_dict = {"lev1": {"lev2": {"lev3": "current_val"}}}
>>> [item for item in recursive_items(rec_dict)]
[('lev3', 'current_val')]
"""
for key, value in dictionary.items():
if type(value) is dict:
yield from recursive_items(value)
else:
yield (key, value)
def recursive_update(d, u, must_match=False):
"""Similar function to `dict.update`, but for a nested `dict`.
From: https://stackoverflow.com/a/3233356
If you have to a nested mapping structure, for example:
{"a": 1, "b": {"c": 2}}
Say you want to update the above structure with:
{"b": {"d": 3}}
This function will produce:
{"a": 1, "b": {"c": 2, "d": 3}}
Instead of:
{"a": 1, "b": {"d": 3}}
Arguments
---------
d : dict
Mapping to be updated.
u : dict
Mapping to update with.
must_match : bool
Whether to throw an error if the key in `u` does not exist in `d`.
Example
-------
>>> d = {"a": 1, "b": {"c": 2}}
>>> recursive_update(d, {"b": {"d": 3}})
>>> d
{'a': 1, 'b': {'c': 2, 'd': 3}}
"""
# TODO: Consider cases where u has branch off k, but d does not.
# e.g. d = {"a":1}, u = {"a": {"b": 2 }}
for k, v in u.items():
if isinstance(v, collections.abc.Mapping) and k in d:
recursive_update(d.get(k, {}), v)
elif must_match and k not in d:
raise KeyError(
f"Override '{k}' not found in: {[key for key in d.keys()]}"
)
else:
d[k] = v
def download_file(
source,
dest,
unpack=False,
dest_unpack=None,
replace_existing=False,
write_permissions=False,
):
"""Downloads the file from the given source and saves it in the given
destination path.
Arguments
---------
source : path or url
Path of the source file. If the source is an URL, it downloads it from
the web.
dest : path
Destination path.
unpack : bool
If True, it unpacks the data in the dest folder.
The archive is preserved.
File formats supported for unpacking/decompression are:
- any format enumerated by `shutil.get_archive_formats()`, usually
including `.tar`, `.tar.gz`, `.zip`.
- plain `.gz` file (when not a `.tar` archive)
Note that you should ALWAYS trust an archive you are extracting, for
security reasons.
dest_unpack: path
Path where to store the unpacked dataset
replace_existing : bool
If True, replaces the existing files.
write_permissions: bool
When set to True, all the files in the dest_unpack directory will be granted write permissions.
This option is active only when unpack=True.
"""
try:
# make sure all processing reached here before main process create dest_dir
sb.utils.distributed.ddp_barrier()
if sb.utils.distributed.if_main_process():
class DownloadProgressBar(tqdm.tqdm):
"""DownloadProgressBar class."""
def update_to(self, b=1, bsize=1, tsize=None):
"""Needed to support multigpu training."""
if tsize is not None:
self.total = tsize
self.update(b * bsize - self.n)
# Create the destination directory if it doesn't exist
dest_dir = pathlib.Path(dest).resolve().parent
dest_dir.mkdir(parents=True, exist_ok=True)
if "http" not in source:
shutil.copyfile(source, dest)
elif not os.path.isfile(dest) or (
os.path.isfile(dest) and replace_existing
):
print(f"Downloading {source} to {dest}")
with DownloadProgressBar(
unit="B",
unit_scale=True,
miniters=1,
desc=source.split("/")[-1],
) as t:
urllib.request.urlretrieve(
source, filename=dest, reporthook=t.update_to
)
else:
print(f"{dest} exists. Skipping download")
# Unpack if necessary
if unpack:
if dest_unpack is None:
dest_unpack = os.path.dirname(dest)
print(f"Extracting {dest} to {dest_unpack}")
if dest.endswith(".gz") and not dest.endswith(".tar.gz"):
# just a gzip'd file, but not an actual archive.
# merely uncompress it and remove the `.gz`.
with gzip.open(dest, "rb") as f_in:
with open(dest[:-3], "wb") as f_out:
shutil.copyfileobj(f_in, f_out)
else:
shutil.unpack_archive(dest, dest_unpack)
if write_permissions:
set_writing_permissions(dest_unpack)
finally:
sb.utils.distributed.ddp_barrier()
def set_writing_permissions(folder_path):
"""
This function sets user writing permissions to all the files in the given folder.
Arguments
---------
folder_path : folder
Folder whose files will be granted write permissions.
"""
for root, dirs, files in os.walk(folder_path):
for file_name in files:
file_path = os.path.join(root, file_name)
# Set writing permissions (mode 0o666) to the file
os.chmod(file_path, 0o666)
def pad_right_to(tensor, target_shape, mode="constant", value=0):
"""
This function takes a torch tensor of arbitrary shape and pads it to target
shape by appending values on the right.
Arguments
---------
tensor : torch.Tensor
Input tensor whose dimension we need to pad.
target_shape : (list, tuple)
Target shape we want for the target tensor its len must be equal to tensor.ndim
mode : str
Pad mode, please refer to torch.nn.functional.pad documentation.
value : float
Pad value, please refer to torch.nn.functional.pad documentation.
Returns
-------
tensor : torch.Tensor
Padded tensor.
valid_vals : list
List containing proportion for each dimension of original, non-padded values.
"""
assert len(target_shape) == tensor.ndim
pads = [] # this contains the abs length of the padding for each dimension.
valid_vals = [] # this contains the relative lengths for each dimension.
i = len(target_shape) - 1 # iterating over target_shape ndims
j = 0
while i >= 0:
assert target_shape[i] >= tensor.shape[i], (
"Target shape must be >= original shape for every dim"
)
pads.extend([0, target_shape[i] - tensor.shape[i]])
valid_vals.append(tensor.shape[j] / target_shape[j])
i -= 1
j += 1
tensor = torch.nn.functional.pad(tensor, pads, mode=mode, value=value)
return tensor, valid_vals
def batch_pad_right(tensors: list, mode="constant", value=0):
"""Given a list of torch tensors it batches them together by padding to the right
on each dimension in order to get same length for all.
Arguments
---------
tensors : list
List of tensor we wish to pad together.
mode : str
Padding mode see torch.nn.functional.pad documentation.
value : float
Padding value see torch.nn.functional.pad documentation.
Returns
-------
tensor : torch.Tensor
Padded tensor.
valid_vals : list
List containing proportion for each dimension of original, non-padded values.
"""
if not len(tensors):
raise IndexError("Tensors list must not be empty")
if len(tensors) == 1:
# if there is only one tensor in the batch we simply unsqueeze it.
return tensors[0].unsqueeze(0), torch.tensor([1.0])
if not (
all(
[tensors[i].ndim == tensors[0].ndim for i in range(1, len(tensors))]
)
):
raise IndexError("All tensors must have same number of dimensions")
# FIXME we limit the support here: we allow padding of only the first dimension
# need to remove this when feat extraction is updated to handle multichannel.
max_shape = []
for dim in range(tensors[0].ndim):
if dim != 0:
if not all(
[x.shape[dim] == tensors[0].shape[dim] for x in tensors[1:]]
):
raise OSError(
"Tensors should have same dimensions except for the first one"
)
max_shape.append(max([x.shape[dim] for x in tensors]))
batched = []
valid = []
for t in tensors:
# for each tensor we apply pad_right_to
padded, valid_percent = pad_right_to(
t, max_shape, mode=mode, value=value
)
batched.append(padded)
valid.append(valid_percent[0])
batched = torch.stack(batched)
return batched, torch.tensor(valid)
def split_by_whitespace(text):
"""A very basic functional version of str.split"""
return text.split()
def recursive_to(data, *args, **kwargs):
"""Moves data to device, or other type, and handles containers.
Very similar to torch.utils.data._utils.pin_memory.pin_memory,
but applies .to() instead.
"""
if isinstance(data, torch.Tensor):
return data.to(*args, **kwargs)
elif isinstance(data, collections.abc.Mapping):
return {
k: recursive_to(sample, *args, **kwargs)
for k, sample in data.items()
}
elif isinstance(data, tuple) and hasattr(data, "_fields"): # namedtuple
return type(data)(
*(recursive_to(sample, *args, **kwargs) for sample in data)
)
elif isinstance(data, collections.abc.Sequence):
return [recursive_to(sample, *args, **kwargs) for sample in data]
elif hasattr(data, "to"):
return data.to(*args, **kwargs)
# What should be done with unknown data?
# For now, just return as they are
else:
return data
np_str_obj_array_pattern = re.compile(r"[SaUO]")
def mod_default_collate(batch):
"""Makes a tensor from list of batch values.
Note that this doesn't need to zip(*) values together
as PaddedBatch connects them already (by key).
Here the idea is not to error out.
This is modified from:
https://github.com/pytorch/pytorch/blob/c0deb231db76dbea8a9d326401417f7d1ce96ed5/torch/utils/data/_utils/collate.py#L42
"""
elem = batch[0]
elem_type = type(elem)
if isinstance(elem, torch.Tensor):
out = None
try:
if torch.utils.data.get_worker_info() is not None:
# If we're in a background process, concatenate directly into a
# shared memory tensor to avoid an extra copy
numel = sum([x.numel() for x in batch])
storage = elem.storage()._new_shared(numel)
out = elem.new(storage)
return torch.stack(batch, 0, out=out)
except RuntimeError: # Unequal size:
return batch
elif (
elem_type.__module__ == "numpy"
and elem_type.__name__ != "str_"
and elem_type.__name__ != "string_"
):
try:
if (
elem_type.__name__ == "ndarray"
or elem_type.__name__ == "memmap"
):
# array of string classes and object
if np_str_obj_array_pattern.search(elem.dtype.str) is not None:
return batch
return mod_default_collate([torch.as_tensor(b) for b in batch])
elif elem.shape == (): # scalars
return torch.as_tensor(batch)
except RuntimeError: # Unequal size
return batch
elif isinstance(elem, float):
return torch.tensor(batch, dtype=torch.float64)
elif isinstance(elem, int):
return torch.tensor(batch)
else:
return batch
def split_path(path):
"""Splits a path to source and filename
This also handles URLs and Huggingface hub paths, in addition to
regular paths.
Arguments
---------
path : str or FetchSource
Returns
-------
str
Source
str
Filename
"""
def split(src):
"""Core function to split path."""
if "/" in src:
return src.rsplit("/", maxsplit=1)
else:
# Interpret as path to file in current directory.
return "./", src
if isinstance(path, sb.utils.fetching.FetchSource):
fetch_from, fetch_path = path
source, filename = split(fetch_path)
return sb.utils.fetching.FetchSource(fetch_from, source), filename
else:
return split(path)
def scalarize(value):
"""Converts a namedtuple or dictionary containing tensors
to their scalar value
Arguments
---------
value: dict or namedtuple
a dictionary or named tuple of tensors
Returns
-------
result: dict
a result dictionary
"""
if hasattr(value, "_asdict"):
value_dict = value._asdict()
else:
value_dict = value
return {key: item_value.item() for key, item_value in value_dict.items()}
def unsqueeze_as(x, target):
"""Reshape the tensor to be of a shape compatible with the target
tensor, only valid if x.dim() <= y.dim()
Arguments
---------
x: torch.Tensor
the original tensor
target: torch.Tensor
the tensor whose shape
Returns
-------
result: torch.Tensor
a view of tensor x reshaped to a shape compatible with y
"""
return x.view(x.shape + (1,) * (target.dim() - x.dim()))
def pad_divisible(tensor, length=None, factor=2, len_dim=1, pad_value=0):
"""Adds extra padding to the specified dimension of a tensor to make
it divisible by the specified factor. This is useful when passing
variable-length sequences to downsampling UNets or other similar
architectures in which inputs are expected to be divisible by the
downsampling factor
Arguments
---------
tensor: torch.Tensor
the tensor to be padded, of arbitrary dimension
length: torch.Tensor
a 1-D tensor of relative lengths
factor: int
the divisibility factor
len_dim: int
the index of the dimension used as the length
pad_value: int
the value with which outputs will be padded
Returns
-------
tensor_padded: torch.Tensor
the tensor, with additional padding if required
length: torch.Tensor
the adjusted length tensor, if provided
Example
-------
>>> x = torch.tensor([[1, 2, 3, 4], [5, 6, 0, 0]])
>>> lens = torch.tensor([1.0, 0.5])
>>> x_pad, lens_pad = pad_divisible(x, length=lens, factor=5)
>>> x_pad
tensor([[1, 2, 3, 4, 0],
[5, 6, 0, 0, 0]])
>>> lens_pad
tensor([0.8000, 0.4000])
"""
time_dim = tensor.size(len_dim)
desired_time_dim = time_dim
gap = time_dim % factor
if gap > 0:
desired_time_dim += factor - gap
new_shape = list(tensor.shape)
new_shape[len_dim] = desired_time_dim
tensor_padded, _ = pad_right_to(tensor, new_shape, value=pad_value)
# Adjust lengths to the new dimension, post-padding
if length is not None:
length = length * (time_dim / desired_time_dim)
return tensor_padded, length
def trim_to_shape(tensor, shape):
"""Trims the specified tensor to match the specified shape
Arguments
---------
tensor: torch.Tensor
a tensor
shape: enumerable
the desired shape
Returns
-------
tensor: torch.Tensor
the trimmed tensor
"""
for dim, size in enumerate(shape):
tensor = tensor.narrow(dim, 0, size)
return tensor
def trim_as(tensor, other):
"""Trims the specified tensor to match the shape of another
tensor (at most)
Arguments
---------
tensor: torch.Tensor:
a tensor
other: torch.Tensor
the tensor whose shape to match
Returns
-------
tensor: torch.Tensor
the trimmed tensor
"""
return trim_to_shape(tensor, other.shape)
def match_shape(tensor, other):
"""A swiss-army-knife helper function to match the shape of a tensor to
match that of another tensor - useful for masks, etc.
Arguments
---------
tensor: torch.Tensor:
a tensor
other: torch.Tensor
the tensor whose shape to match
Returns
-------
tensor: torch.Tensor
the tensor with matching shape
"""
result = unsqueeze_as(tensor, other)
result = result.expand_as(other)
result = trim_as(result, other)
return result
def batch_shuffle(items, batch_size):
"""Shuffles batches of fixed size within a sequence
Arguments
---------
items: sequence
a tensor or an indexable sequence, such as a list
batch_size: int
the batch size
Returns
-------
items: sequence
the original items. If a tensor was passed, a tensor
will be returned. Otherwise, it will return a list
"""
batch_count = math.floor(len(items) / batch_size)
batches = torch.randperm(batch_count)
batch_idx = (
batches.unsqueeze(-1).expand(batch_count, batch_size) * batch_size
)
batch_offset = torch.arange(batch_size).unsqueeze(0)
batch_idx += batch_offset
tail = torch.arange(batch_count * batch_size, len(items))
batch_idx = torch.concat((batch_idx.flatten(), tail))
if torch.is_tensor(items):
result = items[batch_idx]
else:
result = [items[idx] for idx in batch_idx]
return result
def concat_padded_features(
feats, lens, dim=1, feats_slice_start=None, feats_slice_end=None
):
"""Concatenates multiple padded feature tensors into a single
padded tensor in a vectorized manner without including the
padding in the final tensor, adding padding only at the end.
The function supports optional relative sicing of the tensors.
One possible use case is to concatenate batches of spectrograms
or audio.
Arguments
---------
feats: list
a list of padded tensors
lens: list
a list of length tensors
dim: int
The dimension on which to perform concatenation
feats_slice_start: list
offsets, relative to the beginning of the sequence, for each
of the tensors being concatenated. This is useful if only
a subsequence of some slices is included
feats_slice_end: list
offsets, relative to the end of the sequence, for each
of the tensors being concatenated. This is useful if only
a subsequence of some slices is included
Returns
-------
out: torch.Tensor
a concatenated tensor
"""
first_item = feats[0]
item_lengths = torch.tensor([item.size(dim) for item in feats]).to(
first_item.device
)
lens = torch.concat([len_rel.unsqueeze(0) for len_rel in lens])
lens_abs = (lens * item_lengths.unsqueeze(-1)).int()
feats_slice_start = _offset_to_tensor(feats_slice_start, lens_abs)
feats_slice_end = _offset_to_tensor(feats_slice_end, lens_abs)
out_start, out_end = _lens_to_boundaries(
lens_abs, feats_slice_start, feats_slice_end, cumulative=True
)
in_start, in_end = _lens_to_boundaries(
lens_abs, feats_slice_start, feats_slice_end, cumulative=False
)
total_length = out_end.max().int().item()
out_shape = list(first_item.shape)
out_shape[dim] = total_length
out = torch.zeros(out_shape).to(first_item.device)
for item, item_in_start, item_in_end, item_out_start, item_out_end in zip(
feats, in_start, in_end, out_start, out_end
):
in_mask = _boundaries_to_mask(item, item_in_start, item_in_end, dim)
out_mask = _boundaries_to_mask(out, item_out_start, item_out_end, dim)
out[out_mask] = item[in_mask]
out_lens = out_end[-1, :].float() / total_length
return out, out_lens
def _offset_to_tensor(offset, lengths):
"""Converts a variety of offset representations to a component x batch tensor,
used by concat_padded_features. offset can be a tensor, a list of tensors (where
each element is a tensor of relative offsets similar to lengths), a list of floats
(in which case all batch elements are presumed to have the same offset)
Arguments
---------
offset: list|Tensor
a list or tensor of offsets
lengths: torch.Tensor
a length tensor
Returns
-------
result: torch.Tensor
a tensor of offsets
"""
if offset is None:
result = None
elif torch.is_tensor(offset):
result = offset
elif isinstance(offset, Number):
result = torch.ones_like(lengths) * offset
elif isinstance(offset, list):
if isinstance(offset[0], Number):
result = torch.tensor(offset).unsqueeze(-1).to(lengths.device)
else:
result = torch.concat([item.unsqueeze(0) for item in offset])
else:
raise ValueError(
"The offset must be a number, a tensor or a list of tensors"
)
return result
def _lens_to_boundaries(
lengths, slice_start=None, slice_end=None, cumulative=True
):
"""Converts a tensor of lengths to a tensor of start and end
boundaries, used for concat_padded_features
Arguments
---------
lengths: torch.Tensor
a (component x batch) tensor of absolute lengths
slice_start: torch.Tensor
a (component x batch) tensor of relative start offsets
slice_end: torch.Tensor
a (component x batch) tensor of relative end offsets
cumulative: True
if true, the start of a given component is assumed to
be at the end of the previous component.
if false, all components start at the beginning of the
length dimension
Returns
-------
start: torch.Tensor
the starting boundary
end: torch.Tensor
the ending boundary
"""
batch_size = lengths.size(-1)
batch_padding = torch.zeros((1, batch_size)).int().to(lengths.device)
if slice_start is None:
start_offset = torch.tensor(0).to(lengths.device)
else:
start_offset = (lengths * slice_start).floor().int()
if slice_end is None:
end_offset = torch.tensor(0).to(lengths.device)
else:
end_offset = (lengths * slice_end).floor().int()
if cumulative:
effective_lengths = lengths - start_offset - end_offset
effective_lengths_zpad = torch.concat(
[batch_padding, effective_lengths], dim=0
)
start = effective_lengths_zpad.cumsum(dim=0)[:-1, :]
else:
start = torch.zeros(*lengths.shape).to(lengths.device)
start += start_offset
end = start + lengths - end_offset
return start, end
def _boundaries_to_mask(target, start, end, len_dim=1):
"""For a given features tensor and tensors of start and end indexes,
computes the corresponding Boolean mask
Arguments
---------
target: torch.Tensor
the target tensor
start: torch.Tensor