-
Notifications
You must be signed in to change notification settings - Fork 177
/
flowtron.py
961 lines (811 loc) · 41 KB
/
flowtron.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
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
###############################################################################
#
# Copyright (c) 2020, NVIDIA CORPORATION. All rights reserved.
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
###############################################################################
from typing import Optional
import numpy as np
import torch
from torch import nn, Tensor
from torch.nn import functional as F
def get_gate_mask_from_lengths(lengths):
"""Constructs binary mask from a 1D torch tensor of input lengths
Args:
lengths (torch.tensor): 1D tensor
Returns:
mask (torch.tensor): num_sequences x max_length x 1 binary tensor
"""
max_len = torch.max(lengths).item()
ids = torch.arange(0, max_len, out=torch.cuda.LongTensor(max_len))
mask = (ids < lengths.unsqueeze(1)).bool()
return mask
def get_mask_from_lengths(lengths):
"""Constructs binary mask from a 1D torch tensor of input lengths
Args:
lengths (torch.tensor): 1D tensor
Returns:
mask (torch.tensor): num_sequences x max_length x 1 binary tensor
"""
max_len = torch.max(lengths).item()
ids = torch.arange(0, max_len, out=torch.cuda.LongTensor(max_len))
mask = (ids < lengths.unsqueeze(1)).bool()
return mask
def masked_instance_norm(input: Tensor, mask: Tensor, running_mean: Optional[Tensor], running_var: Optional[Tensor],
weight: Optional[Tensor], bias: Optional[Tensor], use_input_stats: bool,
momentum: float, eps: float = 1e-5) -> Tensor:
r"""Applies Masked Instance Normalization for each channel in each data sample in a batch.
See :class:`~MaskedInstanceNorm1d` for details.
"""
if not use_input_stats and (running_mean is None or running_var is None):
raise ValueError('Expected running_mean and running_var to be not None when use_input_stats=False')
shape = input.shape
b, c = shape[:2]
num_dims = len(shape[2:])
_dims = tuple(range(-num_dims, 0))
_slice = (...,) + (None,) * num_dims
running_mean_ = running_mean[None, :].repeat(b, 1) if running_mean is not None else None
running_var_ = running_var[None, :].repeat(b, 1) if running_mean is not None else None
if use_input_stats:
lengths = mask.sum(_dims)
mean = (input * mask).sum(_dims) / lengths # (N, C)
var = (((input - mean[_slice]) * mask) ** 2).sum(_dims) / lengths # (N, C)
if running_mean is not None:
running_mean_.mul_(1 - momentum).add_(momentum * mean.detach())
running_mean.copy_(running_mean_.view(b, c).mean(0, keepdim=False))
if running_var is not None:
running_var_.mul_(1 - momentum).add_(momentum * var.detach())
running_var.copy_(running_var_.view(b, c).mean(0, keepdim=False))
else:
mean, var = running_mean_.view(b, c), running_var_.view(b, c)
out = (input - mean[_slice]) / torch.sqrt(var[_slice] + eps) # (N, C, ...)
if weight is not None and bias is not None:
out = out * weight[None, :][_slice] + bias[None, :][_slice]
return out
class MaskedInstanceNorm1d(nn.InstanceNorm1d):
r"""Applies Instance Normalization over a masked 3D input
(a mini-batch of 1D inputs with additional channel dimension)..
See documentation of :class:`~torch.nn.InstanceNorm1d` for details.
Shape:
- Input: :math:`(N, C, L)`
- Mask: :math:`(N, 1, L)`
- Output: :math:`(N, C, L)` (same shape as input)
"""
def __init__(self, num_features: int, eps: float = 1e-5, momentum: float = 0.1,
affine: bool = False, track_running_stats: bool = False) -> None:
super(MaskedInstanceNorm1d, self).__init__(
num_features, eps, momentum, affine, track_running_stats)
def forward(self, input: Tensor, mask: Tensor = None) -> Tensor:
self._check_input_dim(input)
if mask is not None:
self._check_input_dim(mask)
if mask is None:
return F.instance_norm(
input, self.running_mean, self.running_var, self.weight, self.bias,
self.training or not self.track_running_stats, self.momentum, self.eps
)
else:
return masked_instance_norm(
input, mask, self.running_mean, self.running_var, self.weight, self.bias,
self.training or not self.track_running_stats, self.momentum, self.eps
)
class AttentionConditioningLayer(nn.Module):
"""Adapted from the LocationLayer in
https://github.com/NVIDIA/tacotron2/blob/master/model.py
1D Conv model over a concatenation of the previous attention and the
accumulated attention values """
def __init__(self, input_dim=2, attention_n_filters=32,
attention_kernel_sizes=[5, 3], attention_dim=640):
super(AttentionConditioningLayer, self).__init__()
self.location_conv_hidden = ConvNorm(
input_dim, attention_n_filters,
kernel_size=attention_kernel_sizes[0], padding=None, bias=True,
stride=1, dilation=1, w_init_gain='relu')
self.location_conv_out = ConvNorm(
attention_n_filters, attention_dim,
kernel_size=attention_kernel_sizes[1], padding=None, bias=True,
stride=1, dilation=1, w_init_gain='sigmoid')
self.conv_layers = nn.Sequential(self.location_conv_hidden,
nn.ReLU(),
self.location_conv_out,
nn.Sigmoid())
def forward(self, attention_weights_cat):
return self.conv_layers(attention_weights_cat)
class AttentionCTCLoss(torch.nn.Module):
def __init__(self, blank_logprob=-1):
super(AttentionCTCLoss, self).__init__()
self.log_softmax = torch.nn.LogSoftmax(dim=3)
self.blank_logprob = blank_logprob
self.CTCLoss = nn.CTCLoss(zero_infinity=True)
def forward(self, attn, in_lens, out_lens, attn_logprob):
assert attn_logprob is not None
key_lens = in_lens
query_lens = out_lens
attn_logprob_padded = F.pad(input=attn_logprob,
pad=(1, 0, 0, 0, 0, 0, 0, 0),
value=self.blank_logprob)
cost_total = 0.0
for bid in range(attn_logprob.shape[0]):
target_seq = torch.arange(1, key_lens[bid]+1).unsqueeze(0)
curr_logprob = attn_logprob_padded[bid].permute(1, 0, 2)[
:query_lens[bid],
:,
:key_lens[bid]+1]
curr_logprob = self.log_softmax(curr_logprob[None])[0]
ctc_cost = self.CTCLoss(curr_logprob, target_seq,
input_lengths=query_lens[bid:bid+1],
target_lengths=key_lens[bid:bid+1])
cost_total += ctc_cost
cost = cost_total/attn_logprob.shape[0]
return cost
class FlowtronLoss(torch.nn.Module):
def __init__(self, sigma=1.0, gm_loss=False, gate_loss=True,
use_ctc_loss=False, ctc_loss_weight=0.0,
blank_logprob=-1):
super(FlowtronLoss, self).__init__()
self.sigma = sigma
self.gate_criterion = nn.BCEWithLogitsLoss(reduction='none')
self.gm_loss = gm_loss
self.gate_loss = gate_loss
self.use_ctc_loss = use_ctc_loss
self.ctc_loss_weight = ctc_loss_weight
self.blank_logprob = blank_logprob
self.attention_loss = AttentionCTCLoss(
blank_logprob=self.blank_logprob)
def forward(self, model_output, gate_target,
in_lengths, out_lengths, is_validation=False):
z, log_s_list, gate_pred, attn_list, attn_logprob_list, \
mean, log_var, prob = model_output
# create mask for outputs computed on padded data
mask = get_mask_from_lengths(out_lengths).transpose(0, 1)[..., None]
mask_inverse = ~mask
mask, mask_inverse = mask.float(), mask_inverse.float()
n_mel_dims = z.size(2)
n_elements = mask.sum()
for i, log_s in enumerate(log_s_list):
if i == 0:
log_s_total = torch.sum(log_s * mask)
else:
log_s_total = log_s_total + torch.sum(log_s * mask)
if self.gm_loss:
mask = mask[..., None] # T, B, 1, Dummy
z = z[..., None] # T, B, Mel, Dummy
mean = mean[None] # Dummy, Dummy or B, Mel, Components
log_var = log_var[None] # Dummy, Dummy or B, Mel, Components
prob = prob[None, :, None] # Dummy, B, Dummy, Components
_z = -(z - mean)**2 / (2 * torch.exp(log_var))
_zmax = _z.max(dim=3, keepdim=True)[0] # T, B, 80, Dummy
_z = prob * torch.exp(_z - _zmax) / torch.sqrt(torch.exp(log_var))
_z = _zmax + torch.log(torch.sum(_z, dim=3, keepdim=True))
nll = -torch.sum(mask * _z)
loss = nll - log_s_total
mask = mask[..., 0]
else:
z = z * mask
loss = torch.sum(z*z)/(2*self.sigma*self.sigma) - log_s_total
loss = loss / (n_elements * n_mel_dims)
gate_loss = torch.zeros(1, device=z.device)
if self.gate_loss > 0:
gate_pred = (gate_pred * mask)
gate_pred = gate_pred[..., 0].permute(1, 0)
gate_loss = self.gate_criterion(gate_pred, gate_target)
gate_loss = gate_loss.permute(1, 0) * mask[:, :, 0]
gate_loss = gate_loss.sum() / n_elements
loss_ctc = torch.zeros_like(gate_loss, device=z.device)
if self.use_ctc_loss:
for cur_flow_idx, flow_attn in enumerate(attn_list):
cur_attn_logprob = attn_logprob_list[cur_flow_idx]
# flip and send log probs for back step
if cur_flow_idx % 2 != 0:
if cur_attn_logprob is not None:
for k in range(cur_attn_logprob.size(0)):
cur_attn_logprob[k] = cur_attn_logprob[k].roll(
-out_lengths[k].item(),
dims=0)
cur_attn_logprob = torch.flip(cur_attn_logprob, (1, ))
cur_flow_ctc_loss = self.attention_loss(
flow_attn.unsqueeze(1),
in_lengths,
out_lengths,
attn_logprob=cur_attn_logprob.unsqueeze(1))
# flip the logprob back to be in backward direction
if cur_flow_idx % 2 != 0:
if cur_attn_logprob is not None:
cur_attn_logprob = torch.flip(cur_attn_logprob, (1, ))
for k in range(cur_attn_logprob.size(0)):
cur_attn_logprob[k] = cur_attn_logprob[k].roll(
out_lengths[k].item(),
dims=0)
loss_ctc += cur_flow_ctc_loss
# make CTC loss independent of number of flows by taking mean
loss_ctc = loss_ctc / float(len(attn_list))
return loss, gate_loss, loss_ctc
class LinearNorm(torch.nn.Module):
def __init__(self, in_dim, out_dim, bias=True, w_init_gain='linear'):
super(LinearNorm, self).__init__()
self.linear_layer = torch.nn.Linear(in_dim, out_dim, bias=bias)
torch.nn.init.xavier_uniform_(
self.linear_layer.weight,
gain=torch.nn.init.calculate_gain(w_init_gain))
def forward(self, x):
return self.linear_layer(x)
class ConvNorm(torch.nn.Module):
def __init__(self, in_channels, out_channels, kernel_size=1, stride=1,
padding=None, dilation=1, bias=True, w_init_gain='linear'):
super(ConvNorm, self).__init__()
if padding is None:
assert(kernel_size % 2 == 1)
padding = int(dilation * (kernel_size - 1) / 2)
self.conv = torch.nn.Conv1d(in_channels, out_channels,
kernel_size=kernel_size, stride=stride,
padding=padding, dilation=dilation,
bias=bias)
torch.nn.init.xavier_uniform_(
self.conv.weight, gain=torch.nn.init.calculate_gain(w_init_gain))
def forward(self, signal):
conv_signal = self.conv(signal)
return conv_signal
class GaussianMixture(torch.nn.Module):
def __init__(self, n_hidden, n_components, n_mel_channels, fixed_gaussian,
mean_scale):
super(GaussianMixture, self).__init__()
self.n_mel_channels = n_mel_channels
self.n_components = n_components
self.fixed_gaussian = fixed_gaussian
self.mean_scale = mean_scale
# TODO: fuse into one dense n_components * 3
self.prob_layer = LinearNorm(n_hidden, n_components)
if not fixed_gaussian:
self.mean_layer = LinearNorm(
n_hidden, n_mel_channels * n_components)
self.log_var_layer = LinearNorm(
n_hidden, n_mel_channels * n_components)
else:
mean = self.generate_mean(n_mel_channels, n_components, mean_scale)
log_var = self.generate_log_var(n_mel_channels, n_components)
self.register_buffer('mean', mean.float())
self.register_buffer('log_var', log_var.float())
def generate_mean(self, n_dimensions, n_components, scale=3):
means = torch.eye(n_dimensions).float()
ids = np.random.choice(
range(n_dimensions), n_components, replace=False)
means = means[ids] * scale
means = means.transpose(0, 1)
means = means[None]
return means
def generate_log_var(self, n_dimensions, n_components):
log_var = torch.zeros(1, n_dimensions, n_components).float()
return log_var
def generate_prob(self):
return torch.ones(1, 1).float()
def forward(self, outputs, bs):
prob = torch.softmax(self.prob_layer(outputs), dim=1)
if not self.fixed_gaussian:
mean = self.mean_layer(outputs).view(
bs, self.n_mel_channels, self.n_components)
log_var = self.log_var_layer(outputs).view(
bs, self.n_mel_channels, self.n_components)
else:
mean = self.mean
log_var = self.log_var
return mean, log_var, prob
class MelEncoder(nn.Module):
"""Encoder module:
- Three 1-d convolution banks
- Bidirectional LSTM
"""
def __init__(self, encoder_embedding_dim=512, encoder_kernel_size=3,
encoder_n_convolutions=2, norm_fn=MaskedInstanceNorm1d):
super(MelEncoder, self).__init__()
convolutions = []
for i in range(encoder_n_convolutions):
conv_layer = nn.Sequential(
ConvNorm(80 if i == 0 else encoder_embedding_dim,
encoder_embedding_dim,
kernel_size=encoder_kernel_size, stride=1,
padding=int((encoder_kernel_size - 1) / 2),
dilation=1, w_init_gain='relu'),
norm_fn(encoder_embedding_dim, affine=True))
convolutions.append(conv_layer)
self.convolutions = nn.ModuleList(convolutions)
self.lstm = nn.LSTM(encoder_embedding_dim,
int(encoder_embedding_dim / 2), 1,
bidirectional=True)
def run_padded_sequence(self, sorted_idx, unsort_idx,
lens, padded_data, recurrent_model):
"""Sorts input data by previded ordering (and un-ordering)
and runs the packed data through the recurrent model
Args:
sorted_idx (torch.tensor): 1D sorting index
unsort_idx (torch.tensor): 1D unsorting index (inverse of sorted_idx)
lens: lengths of input data (sorted in descending order)
padded_data (torch.tensor): input sequences (padded)
recurrent_model (nn.Module): recurrent model through which to run the data
Returns:
hidden_vectors (torch.tensor): outputs of the RNN, in the original, unsorted, ordering
"""
# sort the data by decreasing length using provided index
# we assume batch index is in dim=1
padded_data = padded_data[:, sorted_idx]
padded_data = nn.utils.rnn.pack_padded_sequence(padded_data, lens)
hidden_vectors = recurrent_model(padded_data)[0]
hidden_vectors, _ = nn.utils.rnn.pad_packed_sequence(hidden_vectors)
# unsort the results at dim=1 and return
hidden_vectors = hidden_vectors[:, unsort_idx]
return hidden_vectors
def forward(self, x, lens):
mask = get_mask_from_lengths(lens).unsqueeze(1) if x.size(0) > 1 else None
for conv, norm in self.convolutions:
if mask is not None:
x.masked_fill_(~mask, 0.) # zero out padded values before applying convolution
x = F.dropout(F.relu(norm(conv(x), mask=mask)), 0.5, self.training)
del mask
x = x.permute(2, 0, 1) # (N, C, L) -> (L, N, C)
self.lstm.flatten_parameters()
if lens is not None:
# collect decreasing length indices
lens, ids = torch.sort(lens, descending=True)
original_ids = [0] * lens.size(0)
for i in range(len(ids)):
original_ids[ids[i]] = i
x = self.run_padded_sequence(ids, original_ids, lens, x, self.lstm)
else:
x, _ = self.lstm(x)
# average pooling over time dimension
x = torch.mean(x, dim=0)
return x
def infer(self, x):
for conv in self.convolutions:
x = F.dropout(F.relu(conv(x)), 0.5, self.training)
x = x.transpose(1, 2)
self.lstm.flatten_parameters()
outputs, _ = self.lstm(x)
return outputs
class DenseLayer(nn.Module):
def __init__(self, in_dim=1024, sizes=[1024, 1024]):
super(DenseLayer, self).__init__()
in_sizes = [in_dim] + sizes[:-1]
self.layers = nn.ModuleList(
[LinearNorm(in_size, out_size, bias=True)
for (in_size, out_size) in zip(in_sizes, sizes)])
def forward(self, x):
for linear in self.layers:
x = torch.tanh(linear(x))
return x
class Encoder(nn.Module):
"""Encoder module:
- Three 1-d convolution banks
- Bidirectional LSTM
"""
def __init__(self, encoder_n_convolutions=3, encoder_embedding_dim=512,
encoder_kernel_size=5, norm_fn=nn.BatchNorm1d):
super(Encoder, self).__init__()
convolutions = []
for _ in range(encoder_n_convolutions):
conv_layer = nn.Sequential(
ConvNorm(encoder_embedding_dim,
encoder_embedding_dim,
kernel_size=encoder_kernel_size, stride=1,
padding=int((encoder_kernel_size - 1) / 2),
dilation=1, w_init_gain='relu'),
norm_fn(encoder_embedding_dim, affine=True))
convolutions.append(conv_layer)
self.convolutions = nn.ModuleList(convolutions)
self.lstm = nn.LSTM(encoder_embedding_dim,
int(encoder_embedding_dim / 2), 1,
batch_first=True, bidirectional=True)
def forward(self, x, in_lens):
"""
Args:
x (torch.tensor): N x C x L padded input of text embeddings
in_lens (torch.tensor): 1D tensor of sequence lengths
"""
mask = get_mask_from_lengths(in_lens).unsqueeze(1) if x.size(0) > 1 else None
for conv, norm in self.convolutions:
if mask is not None:
x.masked_fill_(~mask, 0.) # zero out padded values before applying convolution
x = F.dropout(F.relu(norm(conv(x), mask=mask)), 0.5, self.training)
del mask
x = x.transpose(1, 2)
x = nn.utils.rnn.pack_padded_sequence(x, in_lens.cpu(), batch_first=True)
self.lstm.flatten_parameters()
outputs, _ = self.lstm(x)
outputs, _ = nn.utils.rnn.pad_packed_sequence(
outputs, batch_first=True)
return outputs
def infer(self, x):
for conv in self.convolutions:
x = F.dropout(F.relu(conv(x)), 0.5, self.training)
x = x.transpose(1, 2)
self.lstm.flatten_parameters()
outputs, _ = self.lstm(x)
return outputs
class Attention(torch.nn.Module):
def __init__(self, n_mel_channels=80, n_speaker_dim=128,
n_text_channels=512, n_att_channels=128, temperature=1.0):
super(Attention, self).__init__()
self.temperature = temperature
self.softmax = torch.nn.Softmax(dim=2)
self.query = LinearNorm(n_mel_channels,
n_att_channels, bias=False, w_init_gain='tanh')
self.key = LinearNorm(n_text_channels+n_speaker_dim,
n_att_channels, bias=False, w_init_gain='tanh')
self.value = LinearNorm(n_text_channels+n_speaker_dim,
n_att_channels, bias=False,
w_init_gain='tanh')
self.v = LinearNorm(n_att_channels, 1, bias=False, w_init_gain='tanh')
self.score_mask_value = -float("inf")
def compute_attention_posterior(self, attn, attn_prior, mask=None,
eps=1e-20):
attn_prior = torch.log(attn_prior.float() + eps)
attn = torch.log(attn.float() + eps)
attn_posterior = attn + attn_prior
attn_logprob = attn_posterior.clone()
if mask is not None:
attn_posterior.data.masked_fill_(
mask.transpose(1, 2), self.score_mask_value)
attn_posterior = self.softmax(attn_posterior)
return attn_posterior, attn_logprob
def forward(self, queries, keys, values, mask=None, attn=None,
attn_prior=None):
"""
returns:
attention weights batch x mel_seq_len x text_seq_len
attention_context batch x featdim x mel_seq_len
sums to 1 over text_seq_len(keys)
"""
if attn is None:
keys = self.key(keys).transpose(0, 1)
values = self.value(values) if hasattr(self, 'value') else values
values = values.transpose(0, 1)
queries = self.query(queries).transpose(0, 1)
attn = self.v(torch.tanh((queries[:, :, None] + keys[:, None])))
attn = attn[..., 0] / self.temperature
if mask is not None:
attn.data.masked_fill_(mask.transpose(1, 2),
self.score_mask_value)
attn = self.softmax(attn)
if attn_prior is not None:
attn, attn_logprob = self.compute_attention_posterior(
attn, attn_prior, mask)
else:
attn_logprob = torch.log(attn.float() + 1e-8)
else:
attn_logprob = None
values = self.value(values)
values = values.transpose(0, 1)
output = torch.bmm(attn, values)
output = output.transpose(1, 2)
return output, attn, attn_logprob
class AR_Back_Step(torch.nn.Module):
def __init__(self, n_mel_channels, n_speaker_dim, n_text_dim,
n_in_channels, n_hidden, n_attn_channels, n_lstm_layers,
add_gate, use_cumm_attention):
super(AR_Back_Step, self).__init__()
self.ar_step = AR_Step(n_mel_channels, n_speaker_dim, n_text_dim,
n_mel_channels+n_speaker_dim, n_hidden,
n_attn_channels, n_lstm_layers, add_gate,
use_cumm_attention)
def forward(self, mel, text, mask, out_lens, attn_prior=None):
mel = torch.flip(mel, (0, ))
if attn_prior is not None:
attn_prior = torch.flip(attn_prior, (1, )) # (B, M, T)
# backwards flow, send padded zeros back to end
for k in range(mel.size(1)):
mel[:, k] = mel[:, k].roll(out_lens[k].item(), dims=0)
if attn_prior is not None:
attn_prior[k] = attn_prior[k].roll(out_lens[k].item(), dims=0)
mel, log_s, gates, attn_out, attention_logprobs = self.ar_step(
mel, text, mask, out_lens, attn_prior)
# move padded zeros back to beginning
for k in range(mel.size(1)):
mel[:, k] = mel[:, k].roll(-out_lens[k].item(), dims=0)
if attn_prior is not None:
attn_prior[k] = attn_prior[k].roll(-out_lens[k].item(), dims=0)
if attn_prior is not None:
attn_prior = torch.flip(attn_prior, (1, ))
return (torch.flip(mel, (0, )), log_s, gates,
attn_out, attention_logprobs)
def infer(self, residual, text, attns, attn_prior=None):
# only need to flip, no need for padding since bs=1
if attn_prior is not None:
# (B, M, T)
attn_prior = torch.flip(attn_prior, (1, ))
residual, attention_weights = self.ar_step.infer(
torch.flip(residual, (0, )), text, attns, attn_prior=attn_prior)
if attn_prior is not None:
attn_prior = torch.flip(attn_prior, (1, ))
residual = torch.flip(residual, (0, ))
return residual, attention_weights
class AR_Step(torch.nn.Module):
def __init__(self, n_mel_channels, n_speaker_dim, n_text_channels,
n_in_channels, n_hidden, n_attn_channels, n_lstm_layers,
add_gate, use_cumm_attention):
super(AR_Step, self).__init__()
self.use_cumm_attention = use_cumm_attention
self.conv = torch.nn.Conv1d(n_hidden, 2*n_mel_channels, 1)
self.conv.weight.data = 0.0*self.conv.weight.data
self.conv.bias.data = 0.0*self.conv.bias.data
self.lstm = torch.nn.LSTM(n_hidden+n_attn_channels, n_hidden, n_lstm_layers)
self.attention_lstm = torch.nn.LSTM(n_mel_channels, n_hidden)
self.attention_layer = Attention(n_hidden, n_speaker_dim,
n_text_channels, n_attn_channels)
if self.use_cumm_attention:
self.attn_cond_layer = AttentionConditioningLayer(
input_dim=2, attention_n_filters=32,
attention_kernel_sizes=[5, 3],
attention_dim=n_text_channels + n_speaker_dim)
self.dense_layer = DenseLayer(in_dim=n_hidden,
sizes=[n_hidden, n_hidden])
if add_gate:
self.gate_threshold = 0.5
self.gate_layer = LinearNorm(
n_hidden+n_attn_channels, 1, bias=True,
w_init_gain='sigmoid')
def run_padded_sequence(self, sorted_idx, unsort_idx, lens, padded_data,
recurrent_model):
"""Sorts input data by previded ordering (and un-ordering) and runs the
packed data through the recurrent model
Args:
sorted_idx (torch.tensor): 1D sorting index
unsort_idx (torch.tensor): 1D unsorting index (inverse of sorted_idx)
lens: lengths of input data (sorted in descending order)
padded_data (torch.tensor): input sequences (padded)
recurrent_model (nn.Module): recurrent model to run data through
Returns:
hidden_vectors (torch.tensor): outputs of the RNN, in the original,
unsorted, ordering
"""
# sort the data by decreasing length using provided index
# we assume batch index is in dim=1
padded_data = padded_data[:, sorted_idx]
padded_data = nn.utils.rnn.pack_padded_sequence(padded_data, lens.cpu())
hidden_vectors = recurrent_model(padded_data)[0]
hidden_vectors, _ = nn.utils.rnn.pad_packed_sequence(hidden_vectors)
# unsort the results at dim=1 and return
hidden_vectors = hidden_vectors[:, unsort_idx]
return hidden_vectors
def run_cumm_attn_sequence(self, attn_lstm_outputs, text, mask,
attn_prior=None):
seq_len, bsize, text_feat_dim = text.shape
# strangely, appending to a list is faster than pre-allocation
attention_context_all = []
attention_weights_all = []
attention_logprobs_all = []
attn_cumm_tensor = text[:, :, 0:1].permute(1, 2, 0)*0
attention_weights = attn_cumm_tensor*0
for i in range(attn_lstm_outputs.shape[0]):
attn_cat = torch.cat((attn_cumm_tensor, attention_weights), 1)
attn_cond_vector = self.attn_cond_layer(attn_cat).permute(2, 0, 1)
output = attn_lstm_outputs[i:i+1:, :]
(attention_context, attention_weights,
attention_logprobs) = self.attention_layer(
output, text*attn_cond_vector, text, mask=mask,
attn_prior=attn_prior)
attention_context_all += [attention_context]
attention_weights_all += [attention_weights]
attention_logprobs_all += [attention_logprobs]
attn_cumm_tensor = attn_cumm_tensor + attention_weights
attention_context_all = torch.cat(attention_context_all, 2)
attention_weights_all = torch.cat(attention_weights_all, 1)
attention_logprobs_all = torch.cat(attention_logprobs_all, 1)
return {'attention_context': attention_context_all,
'attention_weights': attention_weights_all,
'attention_logprobs': attention_logprobs_all}
def forward(self, mel, text, mask, out_lens, attn_prior=None):
dummy = torch.FloatTensor(1, mel.size(1), mel.size(2)).zero_()
dummy = dummy.type(mel.type())
# seq_len x batch x dim
mel0 = torch.cat([dummy, mel[:-1, :, :]], 0)
if out_lens is not None:
# collect decreasing length indices
lens, ids = torch.sort(out_lens, descending=True)
original_ids = [0] * lens.size(0)
for i in range(len(ids)):
original_ids[ids[i]] = i
# mel_seq_len x batch x hidden_dim
attention_hidden = self.run_padded_sequence(
ids, original_ids, lens, mel0, self.attention_lstm)
else:
attention_hidden = self.attention_lstm(mel0)[0]
if hasattr(self, 'use_cumm_attention') and self.use_cumm_attention:
cumm_attn_output_dict = self.run_cumm_attn_sequence(
attention_hidden, text, mask)
attention_context = cumm_attn_output_dict['attention_context']
attention_weights = cumm_attn_output_dict['attention_weights']
attention_logprobs = cumm_attn_output_dict['attention_logprobs']
else:
(attention_context, attention_weights,
attention_logprobs) = self.attention_layer(
attention_hidden, text, text, mask=mask, attn_prior=attn_prior)
attention_context = attention_context.permute(2, 0, 1)
decoder_input = torch.cat((attention_hidden, attention_context), -1)
gates = None
if hasattr(self, 'gate_layer'):
# compute gates before packing
gates = self.gate_layer(decoder_input)
if out_lens is not None:
# reorder, run padded sequence and undo reordering
lstm_hidden = self.run_padded_sequence(
ids, original_ids, lens, decoder_input, self.lstm)
else:
lstm_hidden = self.lstm(decoder_input)[0]
lstm_hidden = self.dense_layer(lstm_hidden).permute(1, 2, 0)
decoder_output = self.conv(lstm_hidden).permute(2, 0, 1)
log_s = decoder_output[:, :, :mel.size(2)]
b = decoder_output[:, :, mel.size(2):]
mel = torch.exp(log_s) * mel + b
return mel, log_s, gates, attention_weights, attention_logprobs
def infer(self, residual, text, attns, attn_prior=None):
attn_cond_vector = 1.0
if hasattr(self, 'use_cumm_attention') and self.use_cumm_attention:
attn_cumm_tensor = text[:, :, 0:1].permute(1, 2, 0)*0
attention_weight = attn_cumm_tensor*0
attention_weights = []
total_output = [] # seems 10FPS faster than pre-allocation
output = None
attn = None
dummy = torch.cuda.FloatTensor(
1, residual.size(1), residual.size(2)).zero_()
for i in range(0, residual.size(0)):
if i == 0:
attention_hidden, (h, c) = self.attention_lstm(dummy)
else:
attention_hidden, (h, c) = self.attention_lstm(output, (h, c))
if hasattr(self, 'use_cumm_attention') and self.use_cumm_attention:
attn_cat = torch.cat((attn_cumm_tensor, attention_weight), 1)
attn_cond_vector = self.attn_cond_layer(attn_cat).permute(2, 0, 1)
attn = None if attns is None else attns[i][None, None]
attn_prior_i = None if attn_prior is None else attn_prior[:, i][None]
(attention_context, attention_weight,
attention_logprob) = self.attention_layer(
attention_hidden, text * attn_cond_vector, text, attn=attn,
attn_prior=attn_prior_i)
if hasattr(self, 'use_cumm_attention') and self.use_cumm_attention:
attn_cumm_tensor = attn_cumm_tensor + attention_weight
attention_weights.append(attention_weight)
attention_context = attention_context.permute(2, 0, 1)
decoder_input = torch.cat((
attention_hidden, attention_context), -1)
if i == 0:
lstm_hidden, (h1, c1) = self.lstm(decoder_input)
else:
lstm_hidden, (h1, c1) = self.lstm(decoder_input, (h1, c1))
lstm_hidden = self.dense_layer(lstm_hidden).permute(1, 2, 0)
decoder_output = self.conv(lstm_hidden).permute(2, 0, 1)
log_s = decoder_output[:, :, :decoder_output.size(2)//2]
b = decoder_output[:, :, decoder_output.size(2)//2:]
output = (residual[i, :, :] - b)/torch.exp(log_s)
total_output.append(output)
if (hasattr(self, 'gate_layer') and
torch.sigmoid(self.gate_layer(decoder_input)) > self.gate_threshold):
print("Hitting gate limit")
break
total_output = torch.cat(total_output, 0)
return total_output, attention_weights
class Flowtron(torch.nn.Module):
def __init__(self, n_speakers, n_speaker_dim, n_text, n_text_dim, n_flows,
n_mel_channels, n_hidden, n_attn_channels, n_lstm_layers,
use_gate_layer, mel_encoder_n_hidden, n_components,
fixed_gaussian, mean_scale, dummy_speaker_embedding,
use_cumm_attention):
super(Flowtron, self).__init__()
norm_fn = MaskedInstanceNorm1d
self.speaker_embedding = torch.nn.Embedding(n_speakers, n_speaker_dim)
self.embedding = torch.nn.Embedding(n_text, n_text_dim)
self.flows = torch.nn.ModuleList()
self.encoder = Encoder(norm_fn=norm_fn, encoder_embedding_dim=n_text_dim)
self.dummy_speaker_embedding = dummy_speaker_embedding
if n_components > 1:
self.mel_encoder = MelEncoder(mel_encoder_n_hidden, norm_fn=norm_fn)
self.gaussian_mixture = GaussianMixture(mel_encoder_n_hidden,
n_components,
n_mel_channels,
fixed_gaussian, mean_scale)
for i in range(n_flows):
add_gate = True if (i == (n_flows-1) and use_gate_layer) else False
if i % 2 == 0:
self.flows.append(AR_Step(n_mel_channels, n_speaker_dim,
n_text_dim,
n_mel_channels+n_speaker_dim,
n_hidden, n_attn_channels,
n_lstm_layers, add_gate,
use_cumm_attention))
else:
self.flows.append(AR_Back_Step(n_mel_channels, n_speaker_dim,
n_text_dim,
n_mel_channels+n_speaker_dim,
n_hidden, n_attn_channels,
n_lstm_layers, add_gate,
use_cumm_attention))
def forward(self, mel, speaker_ids, text, in_lens, out_lens,
attn_prior=None):
speaker_ids = speaker_ids*0 if self.dummy_speaker_embedding else speaker_ids
speaker_vecs = self.speaker_embedding(speaker_ids)
text = self.embedding(text).transpose(1, 2)
text = self.encoder(text, in_lens)
mean, log_var, prob = None, None, None
if hasattr(self, 'gaussian_mixture'):
mel_embedding = self.mel_encoder(mel, out_lens)
mean, log_var, prob = self.gaussian_mixture(
mel_embedding, mel_embedding.size(0))
text = text.transpose(0, 1)
mel = mel.permute(2, 0, 1)
encoder_outputs = torch.cat(
[text, speaker_vecs.expand(text.size(0), -1, -1)], 2)
log_s_list = []
attns_list = []
attns_logprob_list = []
mask = ~get_mask_from_lengths(in_lens)[..., None]
for i, flow in enumerate(self.flows):
mel, log_s, gate, attn_out, attn_logprob_out = flow(
mel, encoder_outputs, mask, out_lens, attn_prior)
log_s_list.append(log_s)
attns_list.append(attn_out)
attns_logprob_list.append(attn_logprob_out)
return (mel, log_s_list, gate, attns_list,
attns_logprob_list, mean, log_var, prob)
def infer(self, residual, speaker_ids, text, temperature=1.0,
gate_threshold=0.5, attns=None, attn_prior=None):
"""Inference function. Inverse of the forward pass
Args:
residual: 1 x 80 x N_residual tensor of sampled z values
speaker_ids: 1 x 1 tensor of integral speaker ids (should be a single value)
text (torch.int64): 1 x N_text tensor holding text-token ids
Returns:
residual: input residual after flow transformation. Technically the mel spectrogram values
attention_weights: attention weights predicted by each flow step for mel-text alignment
"""
speaker_ids = speaker_ids*0 if self.dummy_speaker_embedding else speaker_ids
speaker_vecs = self.speaker_embedding(speaker_ids)
text = self.embedding(text).transpose(1, 2)
text = self.encoder.infer(text)
text = text.transpose(0, 1)
encoder_outputs = torch.cat(
[text, speaker_vecs.expand(text.size(0), -1, -1)], 2)
residual = residual.permute(2, 0, 1)
attention_weights = []
for i, flow in enumerate(reversed(self.flows)):
attn = None if attns is None else reversed(attns)[i]
self.set_temperature_and_gate(flow, temperature, gate_threshold)
residual, attention_weight = flow.infer(
residual, encoder_outputs, attn, attn_prior=attn_prior)
attention_weights.append(attention_weight)
return residual.permute(1, 2, 0), attention_weights
def test_invertibility(self, residual, speaker_ids, text, temperature=1.0,
gate_threshold=0.5, attns=None):
"""Model invertibility check. Call this the same way you would call self.infer()
Args:
residual: 1 x 80 x N_residual tensor of sampled z values
speaker_ids: 1 x 1 tensor of integral speaker ids (should be a single value)
text (torch.int64): 1 x N_text tensor holding text-token ids
Returns:
error: should be in the order of 1e-5 or less, or there may be an invertibility bug
"""
mel, attn_weights = self.infer(residual, speaker_ids, text)
in_lens = torch.LongTensor([text.shape[1]]).cuda()
residual_recon, log_s_list, gate, _, _, _, _ = self.forward(mel,
speaker_ids, text,
in_lens, None)
residual_permuted = residual.permute(2, 0, 1)
if len(self.flows) % 2 == 0:
residual_permuted = torch.flip(residual_permuted, (0,))
residual_recon = torch.flip(residual_recon, (0,))
error = (residual_recon - residual_permuted[0:residual_recon.shape[0]]).abs().mean()
return error
@staticmethod
def set_temperature_and_gate(flow, temperature, gate_threshold):
flow = flow.ar_step if hasattr(flow, "ar_step") else flow
flow.attention_layer.temperature = temperature
if hasattr(flow, 'gate_layer'):
flow.gate_threshold = gate_threshold