Upload 2 files
Browse files- Multi0/config.yml +112 -0
- Multi0/epoch_2nd_23_9ab0.pth +3 -0
Multi0/config.yml
ADDED
@@ -0,0 +1,112 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
ASR_config: Utils/ASR/config.yml
|
2 |
+
ASR_path: Utils/ASR/epoch_00080.pth
|
3 |
+
F0_path: Utils/JDC/bst.t7
|
4 |
+
PLBERT_dir: Utils/PLBERT/
|
5 |
+
batch_size: 6
|
6 |
+
data_params:
|
7 |
+
OOD_data: Data/OOD_texts.txt
|
8 |
+
min_length: 50
|
9 |
+
root_path: /root/StyleTTS2/Omni1_data
|
10 |
+
train_data: Data/train_list.txt
|
11 |
+
val_data: Data/val_list.txt
|
12 |
+
device: cuda
|
13 |
+
epochs_1st: 200
|
14 |
+
epochs_2nd: 60
|
15 |
+
first_stage_path: first_stage.pth
|
16 |
+
log_dir: Models/Omni1
|
17 |
+
log_interval: 10
|
18 |
+
loss_params:
|
19 |
+
TMA_epoch: 50
|
20 |
+
diff_epoch: 14
|
21 |
+
joint_epoch: 19
|
22 |
+
lambda_F0: 1.0
|
23 |
+
lambda_ce: 20.0
|
24 |
+
lambda_diff: 1.0
|
25 |
+
lambda_dur: 1.0
|
26 |
+
lambda_gen: 1.0
|
27 |
+
lambda_mel: 5.0
|
28 |
+
lambda_mono: 1.0
|
29 |
+
lambda_norm: 1.0
|
30 |
+
lambda_s2s: 1.0
|
31 |
+
lambda_slm: 1.0
|
32 |
+
lambda_sty: 1.0
|
33 |
+
max_len: 400
|
34 |
+
model_params:
|
35 |
+
decoder:
|
36 |
+
gen_istft_hop_size: 5
|
37 |
+
gen_istft_n_fft: 20
|
38 |
+
resblock_dilation_sizes:
|
39 |
+
- - 1
|
40 |
+
- 3
|
41 |
+
- 5
|
42 |
+
- - 1
|
43 |
+
- 3
|
44 |
+
- 5
|
45 |
+
- - 1
|
46 |
+
- 3
|
47 |
+
- 5
|
48 |
+
resblock_kernel_sizes:
|
49 |
+
- 3
|
50 |
+
- 7
|
51 |
+
- 11
|
52 |
+
type: istftnet
|
53 |
+
upsample_initial_channel: 512
|
54 |
+
upsample_kernel_sizes:
|
55 |
+
- 20
|
56 |
+
- 12
|
57 |
+
upsample_rates:
|
58 |
+
- 10
|
59 |
+
- 6
|
60 |
+
diffusion:
|
61 |
+
dist:
|
62 |
+
estimate_sigma_data: true
|
63 |
+
mean: -3.0
|
64 |
+
sigma_data: 0.3141927569675583
|
65 |
+
std: 1.0
|
66 |
+
embedding_mask_proba: 0.1
|
67 |
+
transformer:
|
68 |
+
head_features: 64
|
69 |
+
multiplier: 2
|
70 |
+
num_heads: 8
|
71 |
+
num_layers: 3
|
72 |
+
dim_in: 64
|
73 |
+
dropout: 0.2
|
74 |
+
hidden_dim: 512
|
75 |
+
max_conv_dim: 512
|
76 |
+
max_dur: 50
|
77 |
+
multispeaker: true
|
78 |
+
n_layer: 3
|
79 |
+
n_mels: 80
|
80 |
+
n_token: 178
|
81 |
+
slm:
|
82 |
+
hidden: 768
|
83 |
+
initial_channel: 64
|
84 |
+
model: microsoft/wavlm-base-plus
|
85 |
+
nlayers: 13
|
86 |
+
sr: 16000
|
87 |
+
style_dim: 128
|
88 |
+
optimizer_params:
|
89 |
+
bert_lr: 1.0e-05
|
90 |
+
ft_lr: 1.0e-05
|
91 |
+
lr: 0.0001
|
92 |
+
preprocess_params:
|
93 |
+
spect_params:
|
94 |
+
hop_length: 300
|
95 |
+
n_fft: 2048
|
96 |
+
win_length: 1200
|
97 |
+
sr: 24000
|
98 |
+
pretrained_model: /root/StyleTTS2/Models/Omni1/epoch_2nd_pretrained.pth
|
99 |
+
resume: true
|
100 |
+
save_freq: 1
|
101 |
+
saver_freq_steps: 150
|
102 |
+
saver_max_ckpts: 5
|
103 |
+
saver_mode: ITER
|
104 |
+
second_stage_load_pretrained: true
|
105 |
+
slmadv_params:
|
106 |
+
batch_percentage: 0.5
|
107 |
+
iter: 10
|
108 |
+
max_len: 400
|
109 |
+
min_len: 160
|
110 |
+
scale: 0.01
|
111 |
+
sig: 1.5
|
112 |
+
thresh: 5
|
Multi0/epoch_2nd_23_9ab0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fbe9be5c4c2df12b5ddb65cce7e45849d3ed674db1fcb89eb7f1bafc65f05ade
|
3 |
+
size 2132412506
|