Upload 8 files
Browse files- HoshinoAi(new)/Hoshino2.json +105 -0
- HoshinoAi(new)/Hoshino2_G_54000.pth +3 -0
- HoshinoAi(new)/Hoshino2_diffusion.yaml +51 -0
- HoshinoAi(new)/Hoshino2_feature_and_index.pkl +3 -0
- HoshinoAi(new)/diffusion/Hoshino2_model_10000.pt +3 -0
- HoshinoAi(new)/diffusion/Hoshino2_model_20000.pt +3 -0
- HoshinoAi(new)/diffusion/Hoshino2_model_30000.pt +3 -0
- HoshinoAi(new)/diffusion/Hoshino2_model_52000.pt +3 -0
HoshinoAi(new)/Hoshino2.json
ADDED
@@ -0,0 +1,105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"train": {
|
3 |
+
"log_interval": 100,
|
4 |
+
"eval_interval": 1000,
|
5 |
+
"seed": 1234,
|
6 |
+
"epochs": 20000,
|
7 |
+
"learning_rate": 0.0002,
|
8 |
+
"betas": [
|
9 |
+
0.8,
|
10 |
+
0.99
|
11 |
+
],
|
12 |
+
"eps": 1e-09,
|
13 |
+
"batch_size": 24,
|
14 |
+
"fp16_run": false,
|
15 |
+
"half_type": "fp16",
|
16 |
+
"lr_decay": 0.999875,
|
17 |
+
"segment_size": 10240,
|
18 |
+
"init_lr_ratio": 1,
|
19 |
+
"warmup_epochs": 0,
|
20 |
+
"c_mel": 45,
|
21 |
+
"c_kl": 1.0,
|
22 |
+
"use_sr": true,
|
23 |
+
"max_speclen": 512,
|
24 |
+
"port": "8001",
|
25 |
+
"keep_ckpts": 10,
|
26 |
+
"all_in_mem": false,
|
27 |
+
"vol_aug": true
|
28 |
+
},
|
29 |
+
"data": {
|
30 |
+
"training_files": "filelists/train.txt",
|
31 |
+
"validation_files": "filelists/val.txt",
|
32 |
+
"max_wav_value": 32768.0,
|
33 |
+
"sampling_rate": 44100,
|
34 |
+
"filter_length": 2048,
|
35 |
+
"hop_length": 512,
|
36 |
+
"win_length": 2048,
|
37 |
+
"n_mel_channels": 80,
|
38 |
+
"mel_fmin": 0.0,
|
39 |
+
"mel_fmax": 22050,
|
40 |
+
"unit_interpolate_mode": "nearest"
|
41 |
+
},
|
42 |
+
"model": {
|
43 |
+
"inter_channels": 192,
|
44 |
+
"hidden_channels": 192,
|
45 |
+
"filter_channels": 768,
|
46 |
+
"n_heads": 2,
|
47 |
+
"n_layers": 6,
|
48 |
+
"kernel_size": 3,
|
49 |
+
"p_dropout": 0.1,
|
50 |
+
"resblock": "1",
|
51 |
+
"resblock_kernel_sizes": [
|
52 |
+
3,
|
53 |
+
7,
|
54 |
+
11
|
55 |
+
],
|
56 |
+
"resblock_dilation_sizes": [
|
57 |
+
[
|
58 |
+
1,
|
59 |
+
3,
|
60 |
+
5
|
61 |
+
],
|
62 |
+
[
|
63 |
+
1,
|
64 |
+
3,
|
65 |
+
5
|
66 |
+
],
|
67 |
+
[
|
68 |
+
1,
|
69 |
+
3,
|
70 |
+
5
|
71 |
+
]
|
72 |
+
],
|
73 |
+
"upsample_rates": [
|
74 |
+
8,
|
75 |
+
8,
|
76 |
+
2,
|
77 |
+
2,
|
78 |
+
2
|
79 |
+
],
|
80 |
+
"upsample_initial_channel": 512,
|
81 |
+
"upsample_kernel_sizes": [
|
82 |
+
16,
|
83 |
+
16,
|
84 |
+
4,
|
85 |
+
4,
|
86 |
+
4
|
87 |
+
],
|
88 |
+
"n_layers_q": 3,
|
89 |
+
"n_flow_layer": 4,
|
90 |
+
"use_spectral_norm": false,
|
91 |
+
"gin_channels": 768,
|
92 |
+
"ssl_dim": 768,
|
93 |
+
"n_speakers": 1,
|
94 |
+
"vocoder_name": "nsf-hifigan",
|
95 |
+
"speech_encoder": "vec768l12",
|
96 |
+
"speaker_embedding": false,
|
97 |
+
"vol_embedding": true,
|
98 |
+
"use_depthwise_conv": false,
|
99 |
+
"flow_share_parameter": false,
|
100 |
+
"use_automatic_f0_prediction": true
|
101 |
+
},
|
102 |
+
"spk": {
|
103 |
+
"HoshinoV2": 0
|
104 |
+
}
|
105 |
+
}
|
HoshinoAi(new)/Hoshino2_G_54000.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8dd8e4c21586564ab48897480b0e9349b0af21c4b3358ac1e6a72f7cb252b859
|
3 |
+
size 627912149
|
HoshinoAi(new)/Hoshino2_diffusion.yaml
ADDED
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
data:
|
2 |
+
block_size: 512
|
3 |
+
cnhubertsoft_gate: 10
|
4 |
+
duration: 2
|
5 |
+
encoder: vec768l12
|
6 |
+
encoder_hop_size: 320
|
7 |
+
encoder_out_channels: 768
|
8 |
+
encoder_sample_rate: 16000
|
9 |
+
extensions:
|
10 |
+
- wav
|
11 |
+
sampling_rate: 44100
|
12 |
+
training_files: filelists/train.txt
|
13 |
+
unit_interpolate_mode: nearest
|
14 |
+
validation_files: filelists/val.txt
|
15 |
+
device: cuda
|
16 |
+
env:
|
17 |
+
expdir: logs/44k/diffusion
|
18 |
+
gpu_id: 0
|
19 |
+
infer:
|
20 |
+
method: dpm-solver++
|
21 |
+
speedup: 10
|
22 |
+
model:
|
23 |
+
k_step_max: 0
|
24 |
+
n_chans: 512
|
25 |
+
n_hidden: 256
|
26 |
+
n_layers: 20
|
27 |
+
n_spk: 1
|
28 |
+
timesteps: 1000
|
29 |
+
type: Diffusion
|
30 |
+
use_pitch_aug: true
|
31 |
+
spk:
|
32 |
+
HoshinoV2: 0
|
33 |
+
train:
|
34 |
+
amp_dtype: fp32
|
35 |
+
batch_size: 48
|
36 |
+
cache_all_data: true
|
37 |
+
cache_device: cpu
|
38 |
+
cache_fp16: true
|
39 |
+
decay_step: 100000
|
40 |
+
epochs: 100000
|
41 |
+
gamma: 0.5
|
42 |
+
interval_force_save: 5000
|
43 |
+
interval_log: 10
|
44 |
+
interval_val: 2000
|
45 |
+
lr: 0.0001
|
46 |
+
num_workers: 4
|
47 |
+
save_opt: false
|
48 |
+
weight_decay: 0
|
49 |
+
vocoder:
|
50 |
+
ckpt: pretrain/nsf_hifigan/model
|
51 |
+
type: nsf-hifigan
|
HoshinoAi(new)/Hoshino2_feature_and_index.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:55e1d4f29db85dd07ba0cdc85e6868e3a82d246a159438b59b3d00fdb1e1155a
|
3 |
+
size 111138940
|
HoshinoAi(new)/diffusion/Hoshino2_model_10000.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e15ec606fae46d6ce5853c48274f5056bd81ec4416ebc899c8a89b5ec6618eef
|
3 |
+
size 220893767
|
HoshinoAi(new)/diffusion/Hoshino2_model_20000.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:394046e5f466685234f03417a26f1e0a49d7bbeafb8db307fe9ca934654d6a6c
|
3 |
+
size 220893767
|
HoshinoAi(new)/diffusion/Hoshino2_model_30000.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:363ed8385370ca0a3061c8fe3e96003c3dd54179ca7fe13f2c393d84f525ff7a
|
3 |
+
size 220893767
|
HoshinoAi(new)/diffusion/Hoshino2_model_52000.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:239589f3b5a4a1e192e42b0863dcf786ac646751853635813dbd69384f047234
|
3 |
+
size 220893767
|