Upload 2 files
Browse filesUploading files.
- config.yml +102 -0
- epochs_2nd_00020.pth +3 -0
config.yml
ADDED
|
@@ -0,0 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
ASR_config: modules/models/asr/config.yml,
|
| 3 |
+
ASR_path: modules/models/asr/epoch_00080.pth,
|
| 4 |
+
F0_path: modules/models/f0/bst.t7,
|
| 5 |
+
PLBERT_dir: modules/models/plbert/,
|
| 6 |
+
batch_size: 8,
|
| 7 |
+
data_params:
|
| 8 |
+
{
|
| 9 |
+
OOD_data: Data/OOD_texts.txt,
|
| 10 |
+
min_length: 50,
|
| 11 |
+
root_path: "",
|
| 12 |
+
train_data: Data/train_list.txt,
|
| 13 |
+
val_data: Data/val_list.txt,
|
| 14 |
+
},
|
| 15 |
+
device: cuda,
|
| 16 |
+
epochs_1st: 40,
|
| 17 |
+
epochs_2nd: 25,
|
| 18 |
+
first_stage_path: first_stage.pth,
|
| 19 |
+
load_only_params: false,
|
| 20 |
+
log_dir: Models/LibriTTS,
|
| 21 |
+
log_interval: 10,
|
| 22 |
+
loss_params:
|
| 23 |
+
{
|
| 24 |
+
TMA_epoch: 4,
|
| 25 |
+
diff_epoch: 0,
|
| 26 |
+
joint_epoch: 0,
|
| 27 |
+
lambda_F0: 1.0,
|
| 28 |
+
lambda_ce: 20.0,
|
| 29 |
+
lambda_diff: 1.0,
|
| 30 |
+
lambda_dur: 1.0,
|
| 31 |
+
lambda_gen: 1.0,
|
| 32 |
+
lambda_mel: 5.0,
|
| 33 |
+
lambda_mono: 1.0,
|
| 34 |
+
lambda_norm: 1.0,
|
| 35 |
+
lambda_s2s: 1.0,
|
| 36 |
+
lambda_slm: 1.0,
|
| 37 |
+
lambda_sty: 1.0,
|
| 38 |
+
},
|
| 39 |
+
max_len: 300,
|
| 40 |
+
model_params:
|
| 41 |
+
{
|
| 42 |
+
decoder:
|
| 43 |
+
{
|
| 44 |
+
resblock_dilation_sizes: [[1, 3, 5], [1, 3, 5], [1, 3, 5]],
|
| 45 |
+
resblock_kernel_sizes: [3, 7, 11],
|
| 46 |
+
type: hifigan,
|
| 47 |
+
upsample_initial_channel: 512,
|
| 48 |
+
upsample_kernel_sizes: [20, 10, 6, 4],
|
| 49 |
+
upsample_rates: [10, 5, 3, 2],
|
| 50 |
+
},
|
| 51 |
+
diffusion:
|
| 52 |
+
{
|
| 53 |
+
dist:
|
| 54 |
+
{
|
| 55 |
+
estimate_sigma_data: true,
|
| 56 |
+
mean: -3.0,
|
| 57 |
+
sigma_data: 0.19926648961191362,
|
| 58 |
+
std: 1.0,
|
| 59 |
+
},
|
| 60 |
+
embedding_mask_proba: 0.1,
|
| 61 |
+
transformer:
|
| 62 |
+
{ head_features: 64, multiplier: 2, num_heads: 8, num_layers: 3 },
|
| 63 |
+
},
|
| 64 |
+
dim_in: 64,
|
| 65 |
+
dropout: 0.2,
|
| 66 |
+
hidden_dim: 512,
|
| 67 |
+
max_conv_dim: 512,
|
| 68 |
+
max_dur: 50,
|
| 69 |
+
multispeaker: true,
|
| 70 |
+
n_layer: 3,
|
| 71 |
+
n_mels: 80,
|
| 72 |
+
n_token: 178,
|
| 73 |
+
slm:
|
| 74 |
+
{
|
| 75 |
+
hidden: 768,
|
| 76 |
+
initial_channel: 64,
|
| 77 |
+
model: microsoft/wavlm-base-plus,
|
| 78 |
+
nlayers: 13,
|
| 79 |
+
sr: 16000,
|
| 80 |
+
},
|
| 81 |
+
style_dim: 128,
|
| 82 |
+
},
|
| 83 |
+
optimizer_params: { bert_lr: 1.0e-05, ft_lr: 1.0e-05, lr: 0.0001 },
|
| 84 |
+
preprocess_params:
|
| 85 |
+
{
|
| 86 |
+
spect_params: { hop_length: 300, n_fft: 2048, win_length: 1200 },
|
| 87 |
+
sr: 24000,
|
| 88 |
+
},
|
| 89 |
+
pretrained_model: Models/LibriTTS/epoch_2nd_00002.pth,
|
| 90 |
+
save_freq: 1,
|
| 91 |
+
second_stage_load_pretrained: true,
|
| 92 |
+
slmadv_params:
|
| 93 |
+
{
|
| 94 |
+
batch_percentage: 0.5,
|
| 95 |
+
iter: 20,
|
| 96 |
+
max_len: 500,
|
| 97 |
+
min_len: 400,
|
| 98 |
+
scale: 0.01,
|
| 99 |
+
sig: 1.5,
|
| 100 |
+
thresh: 5,
|
| 101 |
+
},
|
| 102 |
+
}
|
epochs_2nd_00020.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1164ffe19a17449d2c722234cecaf2836b35a698fb8ffd42562d2663657dca0a
|
| 3 |
+
size 771390526
|