-
Notifications
You must be signed in to change notification settings - Fork 31
/
hparams.py
125 lines (106 loc) · 3.61 KB
/
hparams.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
import tensorflow as tf
from text.symbols import symbols
def create_hparams(hparams_string=None, verbose=False):
"""Create model hyperparameters. Parse nondefault from given string."""
hparams = tf.contrib.training.HParams(
################################
# Experiment Parameters #
################################
epochs=50000,
iters_per_checkpoint=500,
seed=1234,
dynamic_loss_scaling=True,
fp16_run=False,
distributed_run=False,
dist_backend="nccl",
dist_url="tcp://localhost:54321",
cudnn_enabled=True,
cudnn_benchmark=False,
ignore_layers=['speaker_embedding.weight'],
################################
# Data Parameters #
################################
training_files='filelists/merge_korean_pron_train.txt',
validation_files='filelists/merge_korean_pron_valid.txt',
text_cleaners=['korean_cleaners'],
p_arpabet=1.0,
cmudict_path="data/cmu_dictionary",
################################
# Audio Parameters #
################################
max_wav_value=1.0,
sampling_rate=22050,
filter_length=1024,
hop_length=256,
win_length=1024,
n_mel_channels=80,
mel_fmin=0.0,
mel_fmax=8000.0,
f0_min=80,
f0_max=880,
harm_thresh=0.25,
################################
# Model Parameters #
################################
n_symbols=len(symbols),
symbols_embedding_dim=512,
# Encoder parameters
encoder_kernel_size=5,
encoder_n_convolutions=3,
encoder_embedding_dim=512,
# Decoder parameters
n_frames_per_step=1, # currently only 1 is supported
decoder_rnn_dim=1024,
prenet_dim=256,
prenet_f0_n_layers=1,
prenet_f0_dim=1,
prenet_f0_kernel_size=1,
prenet_rms_dim=0,
prenet_rms_kernel_size=1,
max_decoder_steps=1000,
gate_threshold=0.5,
p_attention_dropout=0.1,
p_decoder_dropout=0.1,
p_teacher_forcing=1.0,
# Attention parameters
attention_rnn_dim=1024,
attention_dim=128,
# Location Layer parameters
attention_location_n_filters=32,
attention_location_kernel_size=31,
# Mel-post processing network parameters
postnet_embedding_dim=512,
postnet_kernel_size=5,
postnet_n_convolutions=5,
# Speaker embedding
n_speakers=161,
speaker_embedding_dim=128,
# Reference encoder
with_gst=True,
ref_enc_filters=[32, 32, 64, 64, 128, 128],
ref_enc_size=[3, 3],
ref_enc_strides=[2, 2],
ref_enc_pad=[1, 1],
ref_enc_gru_size=128,
# Style Token Layer
token_embedding_size=256,
token_num=10,
num_heads=8,
################################
# Optimization Hyperparameters #
################################
use_saved_learning_rate=True,
learning_rate=1e-3,
learning_rate_min=1e-5,
learning_rate_anneal=50000,
weight_decay=1e-6,
grad_clip_thresh=1.0,
batch_size=8,
mask_padding=True, # set model's padded outputs to padded values
)
if hparams_string:
tf.compat.v1.logging.info('Parsing command line hparams: %s', hparams_string)
hparams.parse(hparams_string)
if verbose:
tf.compat.v1.logging.info('Final parsed hparams: %s', hparams.values())
return hparams