|
general: |
|
stage: "ssl" |
|
corpus_type: "single" |
|
source_path: "./data/jsut_22k-low" |
|
aux_path: "./data/jsut_22k" |
|
preprocessed_path: "./preprocessed/jsut-low" |
|
output_path: "./output/melspec/jsut-low" |
|
test_wav_path: null |
|
feature_type: "melspec" |
|
hifigan_path: "./hifigan/hifigan_melspec_universal" |
|
power_norm: True |
|
use_gst: False |
|
|
|
preprocess: |
|
n_train: 4950 |
|
n_val: 25 |
|
n_test: 25 |
|
sampling_rate: 22050 |
|
frame_length: 1024 |
|
frame_shift: 256 |
|
fft_length: 1024 |
|
fmin: 0 |
|
fmax: 8000 |
|
n_mels: 80 |
|
comp_factor: 1.0 |
|
min_magnitude: 0.00001 |
|
bitrate: "16k" |
|
max_wav_value: 32768.0 |
|
segment_length: -1 |
|
|
|
train: |
|
batchsize: 1 |
|
epoch: 50 |
|
epoch_channel: 25 |
|
multi_gpu_mode: False |
|
num_workers: 4 |
|
learning_rate: 0.001 |
|
alpha: 0.1 |
|
beta: 0.1 |
|
augment: False |
|
grad_clip_thresh: 1.0 |
|
logger_step: 1000 |
|
load_pretrained: False |
|
pretrained_path: null |
|
fix_channel: False |
|
early_stopping: False |
|
multi_scale_loss: |
|
use_linear: True |
|
gamma: 1.0 |
|
feature_loss: |
|
type: "mae" |
|
|
|
dual: |
|
enable: True |
|
config_path: ./configs/test/melspec/dual.yaml |