Hyper_Parameters.yaml

Sound:
    N_FFT: 2048
    Mel_Dim: 80
    Frame_Length: 1024
    Frame_Shift: 256
    Sample_Rate: 22050
    Mel_F_Min: 0
    Mel_F_Max: 8000

Feature_Type: 'Mel' #'Spectrogram', 'Mel'

Tokens: 77
Notes: 128
Durations: 5000
Genres: 1
Singers: 1
Duration:
    Equality: false
    Consonant_Duration: 3   # This is only used when Equality is False.

Encoder:
    Size: 384
    ConvFFT:
        Stack: 6
        Head: 2
        Dropout_Rate: 0.1
        Conv:
            Stack: 2
            Kernel_Size: 5
        FFN:
            Kernel_Size: 17

Diffusion:
    Max_Step: 100
    Size: 256
    Kernel_Size: 5
    Stack: 20

Token_Path: './YAML/Token.yaml'
Spectrogram_Range_Info_Path: './YAML/Spectrogram_Range_Info.yaml'
Mel_Range_Info_Path: './YAML/Mel_Range_Info.yaml'
Log_F0_Info_Path: './YAML/Log_F0_Info.yaml'
Log_Energy_Info_Path: './YAML/Log_Energy_Info.yaml'
Singer_Info_Path: './YAML/Singer_Info.yaml'
Genre_Info_Path: './YAML/Genre_Info.yaml'