CoCalc -- melgan.v1.yaml

GitHub Repository: TensorSpeech/TensorFlowTTS
Path: blob/master/examples/melgan/conf/melgan.v1.yaml
¹⁵⁵⁹ views
1

2
# This is the hyperparameter configuration file for MelGAN.
3
# Please make sure this is adjusted for the LJSpeech dataset. If you want to
4
# apply to the other dataset, you might need to carefully change some parameters.
5
# This configuration performs 4000k iters.
6

7
###########################################################
8
#                FEATURE EXTRACTION SETTING               #
9
###########################################################
10
sampling_rate: 22050     # Sampling rate of dataset.
11
hop_size: 256            # Hop size.
12
format: "npy"
13

14

15
###########################################################
16
#         GENERATOR NETWORK ARCHITECTURE SETTING          #
17
###########################################################
18
model_type: "melgan_generator"
19

20
melgan_generator_params:
21
    out_channels: 1               # Number of output channels.
22
    kernel_size: 7                # Kernel size of initial and final conv layers.
23
    filters: 512                  # Initial number of channels for conv layers.
24
    upsample_scales: [8, 8, 2, 2] # List of Upsampling scales.
25
    stack_kernel_size: 3          # Kernel size of dilated conv layers in residual stack.
26
    stacks: 3                     # Number of stacks in a single residual stack module.
27
    is_weight_norm: false         # Use weight-norm or not.
28

29
###########################################################
30
#       DISCRIMINATOR NETWORK ARCHITECTURE SETTING        #
31
###########################################################
32
melgan_discriminator_params:
33
    out_channels: 1                   # Number of output channels.
34
    scales: 3                         # Number of multi-scales.
35
    downsample_pooling: "AveragePooling1D"   # Pooling type for the input downsampling.
36
    downsample_pooling_params:        # Parameters of the above pooling function.
37
        pool_size: 4
38
        strides: 2
39
    kernel_sizes: [5, 3]              # List of kernel size.
40
    filters: 16                       # Number of channels of the initial conv layer.
41
    max_downsample_filters: 1024      # Maximum number of channels of downsampling layers.
42
    downsample_scales: [4, 4, 4, 4]   # List of downsampling scales.
43
    nonlinear_activation: "LeakyReLU" # Nonlinear activation function.
44
    nonlinear_activation_params:      # Parameters of nonlinear activation function.
45
        alpha: 0.2
46
    is_weight_norm: false             # Use weight-norm or not.
47

48
###########################################################
49
#               ADVERSARIAL LOSS SETTING                  #
50
###########################################################
51
lambda_feat_match: 10.0
52

53
###########################################################
54
#                  DATA LOADER SETTING                    #
55
###########################################################
56
batch_size: 16                 # Batch size for each GPU with assuming that gradient_accumulation_steps == 1.
57
batch_max_steps: 8192          # Length of each audio in batch for training. Make sure dividable by hop_size.
58
batch_max_steps_valid: 81920   # Length of each audio for validation. Make sure dividable by hope_size.
59
remove_short_samples: true     # Whether to remove samples the length of which are less than batch_max_steps.
60
allow_cache: true              # Whether to allow cache in dataset. If true, it requires cpu memory.
61
is_shuffle: true               # shuffle dataset after each epoch.
62

63
###########################################################
64
#             OPTIMIZER & SCHEDULER SETTING               #
65
###########################################################
66
generator_optimizer_params:
67
    lr: 0.0001             # Generator's learning rate.
68
    beta_1: 0.5
69
    beta_2: 0.9
70
    
71
discriminator_optimizer_params:
72
    lr: 0.0001            # Discriminator's learning rate.
73
    beta_1: 0.5
74
    beta_2: 0.9
75

76
gradient_accumulation_steps: 1
77
###########################################################
78
#                    INTERVAL SETTING                     #
79
###########################################################
80
train_max_steps: 4000000                # Number of training steps.
81
save_interval_steps: 3               # Interval steps to save checkpoint.
82
eval_interval_steps: 2              # Interval steps to evaluate the network.
83
log_interval_steps: 1                # Interval steps to record the training log.
84
discriminator_train_start_steps: 0     # step to start training discriminator.
85

86
###########################################################
87
#                     OTHER SETTING                       #
88
###########################################################
89
num_save_intermediate_results: 1  # Number of batch to be saved as intermediate results.
90

91
Product

Resources

Company
