-
-
Notifications
You must be signed in to change notification settings - Fork 341
/
multi_band_melgan.v2.yaml
150 lines (140 loc) · 8.21 KB
/
multi_band_melgan.v2.yaml
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
# This is the hyperparameter configuration file for MelGAN.
# Please make sure this is adjusted for the JSUT dataset. If you want to
# apply to the other dataset, you might need to carefully change some parameters.
# This configuration requires ~ 8GB memory and will finish within 4 days on Titan V.
# This configuration is based on full-band MelGAN but the hop size and sampling
# rate is different from the paper (16kHz vs 24kHz). The number of iteraions
# is now shown in the paper so currently we train 1M iterations (not sure enough
# to converge). The optimizer setting is based on @dathudeptrai advice.
# https://github.com/kan-bayashi/ParallelWaveGAN/issues/143#issuecomment-632539906
###########################################################
# FEATURE EXTRACTION SETTING #
###########################################################
sampling_rate: 24000 # Sampling rate.
fft_size: 2048 # FFT size.
hop_size: 300 # Hop size.
win_length: 1200 # Window length.
# If set to null, it will be the same as fft_size.
window: "hann" # Window function.
num_mels: 80 # Number of mel basis.
fmin: 80 # Minimum freq in mel basis calculation.
fmax: 7600 # Maximum frequency in mel basis calculation.
global_gain_scale: 1.0 # Will be multiplied to all of waveform.
trim_silence: false # Whether to trim the start and end of silence.
trim_threshold_in_db: 60 # Need to tune carefully if the recording is not good.
trim_frame_size: 2048 # Frame size in trimming.
trim_hop_size: 512 # Hop size in trimming.
format: "hdf5" # Feature file format. "npy" or "hdf5" is supported.
###########################################################
# GENERATOR NETWORK ARCHITECTURE SETTING #
###########################################################
generator_type: "MelGANGenerator" # Generator type.
generator_params:
in_channels: 80 # Number of input channels.
out_channels: 4 # Number of output channels.
kernel_size: 7 # Kernel size of initial and final conv layers.
channels: 384 # Initial number of channels for conv layers.
upsample_scales: [5, 5, 3] # List of Upsampling scales.
stack_kernel_size: 3 # Kernel size of dilated conv layers in residual stack.
stacks: 4 # Number of stacks in a single residual stack module.
use_weight_norm: True # Whether to use weight normalization.
use_causal_conv: False # Whether to use causal convolution.
###########################################################
# DISCRIMINATOR NETWORK ARCHITECTURE SETTING #
###########################################################
discriminator_type: "MelGANMultiScaleDiscriminator" # Discriminator type.
discriminator_params:
in_channels: 1 # Number of input channels.
out_channels: 1 # Number of output channels.
scales: 3 # Number of multi-scales.
downsample_pooling: "AvgPool1d" # Pooling type for the input downsampling.
downsample_pooling_params: # Parameters of the above pooling function.
kernel_size: 4
stride: 2
padding: 1
count_include_pad: False
kernel_sizes: [5, 3] # List of kernel size.
channels: 16 # Number of channels of the initial conv layer.
max_downsample_channels: 512 # Maximum number of channels of downsampling layers.
downsample_scales: [4, 4, 4] # List of downsampling scales.
nonlinear_activation: "LeakyReLU" # Nonlinear activation function.
nonlinear_activation_params: # Parameters of nonlinear activation function.
negative_slope: 0.2
use_weight_norm: True # Whether to use weight norm.
###########################################################
# STFT LOSS SETTING #
###########################################################
stft_loss_params:
fft_sizes: [1024, 2048, 512] # List of FFT size for STFT-based loss.
hop_sizes: [120, 240, 50] # List of hop size for STFT-based loss
win_lengths: [600, 1200, 240] # List of window length for STFT-based loss.
window: "hann_window" # Window function for STFT-based loss
use_subband_stft_loss: true
subband_stft_loss_params:
fft_sizes: [384, 683, 171] # List of FFT size for STFT-based loss.
hop_sizes: [30, 60, 10] # List of hop size for STFT-based loss
win_lengths: [150, 300, 60] # List of window length for STFT-based loss.
window: "hann_window" # Window function for STFT-based loss
###########################################################
# ADVERSARIAL LOSS SETTING #
###########################################################
use_feat_match_loss: false # Whether to use feature matching loss.
lambda_adv: 2.5 # Loss balancing coefficient for adversarial loss.
###########################################################
# DATA LOADER SETTING #
###########################################################
batch_size: 64 # Batch size.
batch_max_steps: 16200 # Length of each audio in batch. Make sure dividable by hop_size.
pin_memory: true # Whether to pin memory in Pytorch DataLoader.
num_workers: 4 # Number of workers in Pytorch DataLoader.
remove_short_samples: true # Whether to remove samples the length of which are less than batch_max_steps.
allow_cache: true # Whether to allow cache in dataset. If true, it requires cpu memory.
###########################################################
# OPTIMIZER & SCHEDULER SETTING #
###########################################################
generator_optimizer_type: "Adam" # Generator's optimizer type.
generator_optimizer_params:
lr: 1.0e-3 # Generator's learning rate.
eps: 1.0e-7 # Generator's epsilon.
weight_decay: 0.0 # Generator's weight decay coefficient.
amsgrad: true
generator_grad_norm: -1 # Generator's gradient norm.
generator_scheduler_type: "MultiStepLR" # Generator's scheduler type.
generator_scheduler_params:
gamma: 0.5 # Generator's scheduler gamma.
milestones: # At each milestone, lr will be multiplied by gamma.
- 100000
- 200000
- 300000
- 400000
- 500000
- 600000
discriminator_optimizer_type: "Adam" # Discriminator's optimizer type.
discriminator_optimizer_params:
lr: 1.0e-3 # Discriminator's learning rate.
eps: 1.0e-7 # Discriminator's epsilon.
weight_decay: 0.0 # Discriminator's weight decay coefficient.
amsgrad: true
discriminator_grad_norm: -1 # Discriminator's gradient norm.
discriminator_scheduler_type: "MultiStepLR" # Discriminator's scheduler type.
discriminator_scheduler_params:
gamma: 0.5 # Discriminator's scheduler gamma.
milestones: # At each milestone, lr will be multiplied by gamma.
- 100000
- 200000
- 300000
- 400000
- 500000
- 600000
###########################################################
# INTERVAL SETTING #
###########################################################
discriminator_train_start_steps: 200000 # Number of steps to start to train discriminator.
train_max_steps: 1000000 # Number of training steps.
save_interval_steps: 50000 # Interval steps to save checkpoint.
eval_interval_steps: 1000 # Interval steps to evaluate the network.
log_interval_steps: 1000 # Interval steps to record the training log.
###########################################################
# OTHER SETTING #
###########################################################
num_save_intermediate_results: 4 # Number of results to be saved as intermediate results.