-
Notifications
You must be signed in to change notification settings - Fork 0
/
loss.py
179 lines (152 loc) · 6.58 KB
/
loss.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
import torch
import torch.nn as nn
class FastSpeech2Loss(nn.Module):
""" FastSpeech2 Loss """
def __init__(self, preprocess_config, model_config):
super(FastSpeech2Loss, self).__init__()
self.pitch_feature_level = preprocess_config["preprocessing"]["pitch"][
"feature"
]
self.energy_feature_level = preprocess_config["preprocessing"]["energy"][
"feature"
]
self.mse_loss = nn.MSELoss()
self.mae_loss = nn.L1Loss()
def forward(self, inputs, predictions):
(
mel_targets,
_,
_,
pitch_targets,
energy_targets,
duration_targets,
) = inputs[6:]
(
mel_predictions,
postnet_mel_predictions,
pitch_predictions,
energy_predictions,
log_duration_predictions,
_,
src_masks,
mel_masks,
_,
_,
) = predictions
src_masks = ~src_masks
mel_masks = ~mel_masks
log_duration_targets = torch.log(duration_targets.float() + 1)
mel_targets = mel_targets[:, : mel_masks.shape[1], :]
mel_masks = mel_masks[:, :mel_masks.shape[1]]
log_duration_targets.requires_grad = False
pitch_targets.requires_grad = False
energy_targets.requires_grad = False
mel_targets.requires_grad = False
if self.pitch_feature_level == "phoneme_level":
pitch_predictions = pitch_predictions.masked_select(src_masks)
pitch_targets = pitch_targets.masked_select(src_masks)
elif self.pitch_feature_level == "frame_level":
pitch_predictions = pitch_predictions.masked_select(mel_masks)
pitch_targets = pitch_targets.masked_select(mel_masks)
if self.energy_feature_level == "phoneme_level":
energy_predictions = energy_predictions.masked_select(src_masks)
energy_targets = energy_targets.masked_select(src_masks)
if self.energy_feature_level == "frame_level":
energy_predictions = energy_predictions.masked_select(mel_masks)
energy_targets = energy_targets.masked_select(mel_masks)
log_duration_predictions = log_duration_predictions.masked_select(src_masks)
log_duration_targets = log_duration_targets.masked_select(src_masks)
mel_predictions = mel_predictions.masked_select(mel_masks.unsqueeze(-1))
postnet_mel_predictions = postnet_mel_predictions.masked_select(
mel_masks.unsqueeze(-1)
)
mel_targets = mel_targets.masked_select(mel_masks.unsqueeze(-1))
mel_loss = self.mae_loss(mel_predictions, mel_targets)
postnet_mel_loss = self.mae_loss(postnet_mel_predictions, mel_targets)
pitch_loss = self.mse_loss(pitch_predictions, pitch_targets)
energy_loss = self.mse_loss(energy_predictions, energy_targets)
duration_loss = self.mse_loss(log_duration_predictions, log_duration_targets)
total_loss = (
mel_loss + postnet_mel_loss + duration_loss + pitch_loss + energy_loss
)
return (
total_loss,
mel_loss,
postnet_mel_loss,
pitch_loss,
energy_loss,
duration_loss,
)
# import torch
# import torch.nn as nn
# class FastSpeech2Loss(nn.Module):
# """ FastSpeech2 Loss """
# def __init__(self, preprocess_config, model_config):
# super(FastSpeech2Loss, self).__init__()
# self.pitch_feature_level = preprocess_config["preprocessing"]["pitch"]["feature"]
# self.energy_feature_level = preprocess_config["preprocessing"]["energy"]["feature"]
# self.mse_loss = nn.MSELoss()
# self.mae_loss = nn.L1Loss()
# self.ce_loss = nn.CrossEntropyLoss()
# def forward(self, inputs, predictions):
# (
# mel_targets,
# _,
# pitch_targets,
# energy_targets,
# duration_targets,
# phoneme_targets,
# ) = inputs[6:]
# (
# mel_predictions,
# postnet_mel_predictions,
# pitch_predictions,
# energy_predictions,
# log_duration_predictions,
# phoneme_predictions,
# _,
# mel_masks,
# _,
# _,
# ) = predictions
# mel_masks = ~mel_masks
# log_duration_targets = torch.log(duration_targets.float() + 1)
# mel_targets = mel_targets[:, :mel_masks.shape[1], :]
# mel_masks = mel_masks[:, :mel_masks.shape[1]]
# log_duration_targets.requires_grad = False
# pitch_targets.requires_grad = False
# energy_targets.requires_grad = False
# mel_targets.requires_grad = False
# # Phoneme Reconstruction Loss
# phoneme_loss = self.ce_loss(phoneme_predictions.transpose(1, 2), phoneme_targets)
# # Mel Loss
# mel_predictions = mel_predictions.masked_select(mel_masks.unsqueeze(-1))
# mel_loss = self.mae_loss(mel_predictions, mel_targets)
# # Postnet Mel Loss
# postnet_mel_predictions = postnet_mel_predictions.masked_select(mel_masks.unsqueeze(-1))
# postnet_mel_loss = self.mae_loss(postnet_mel_predictions, mel_targets)
# # Pitch Loss
# if self.pitch_feature_level == "phoneme_level":
# pitch_predictions = pitch_predictions.masked_select(mel_masks)
# pitch_targets = pitch_targets.masked_select(mel_masks)
# pitch_loss = self.mse_loss(pitch_predictions, pitch_targets)
# # Energy Loss
# if self.energy_feature_level == "phoneme_level":
# energy_predictions = energy_predictions.masked_select(mel_masks)
# energy_targets = energy_targets.masked_select(mel_masks)
# energy_loss = self.mse_loss(energy_predictions, energy_targets)
# # Duration Loss
# log_duration_predictions = log_duration_predictions.masked_select(mel_masks)
# duration_loss = self.mse_loss(log_duration_predictions, log_duration_targets)
# total_loss = (
# mel_loss + postnet_mel_loss + pitch_loss + energy_loss + duration_loss + phoneme_loss
# )
# return (
# total_loss,
# mel_loss,
# postnet_mel_loss,
# pitch_loss,
# energy_loss,
# duration_loss,
# phoneme_loss,
# )