Update ReFlow-VAE-SVC
Browse files
ReFlow-VAE-SVC/config.yaml
ADDED
|
@@ -0,0 +1,185 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data:
|
| 2 |
+
block_size: 512
|
| 3 |
+
duration: 1
|
| 4 |
+
encoder: contentvec768l12
|
| 5 |
+
encoder_ckpt: pretrain/contentvec/checkpoint_best_legacy_500.pt
|
| 6 |
+
encoder_hop_size: 320
|
| 7 |
+
encoder_out_channels: 768
|
| 8 |
+
encoder_sample_rate: 16000
|
| 9 |
+
extensions:
|
| 10 |
+
- wav
|
| 11 |
+
- flac
|
| 12 |
+
f0_extractor: rmvpe
|
| 13 |
+
f0_max: 1200
|
| 14 |
+
f0_min: 40
|
| 15 |
+
sampling_rate: 44100
|
| 16 |
+
train_path: data/train
|
| 17 |
+
valid_path: data/val
|
| 18 |
+
device: cuda
|
| 19 |
+
env:
|
| 20 |
+
expdir: exp/reflowvae-wavenet
|
| 21 |
+
gpu_id: 0
|
| 22 |
+
infer:
|
| 23 |
+
infer_step: 50
|
| 24 |
+
method: euler
|
| 25 |
+
model:
|
| 26 |
+
back_bone: wavenet
|
| 27 |
+
n_chans: 768
|
| 28 |
+
n_hidden: 256
|
| 29 |
+
n_layers: 30
|
| 30 |
+
n_spk: 133
|
| 31 |
+
type: RectifiedFlow_VAE
|
| 32 |
+
use_pitch_aug: false
|
| 33 |
+
spk_dict:
|
| 34 |
+
10KuroBousuku-CN: 0
|
| 35 |
+
11Barber-CN: 1
|
| 36 |
+
12ChuMing-CN: 2
|
| 37 |
+
13ChangGe-CN: 3
|
| 38 |
+
14QueHe-CN: 4
|
| 39 |
+
15QiXuan-CN: 5
|
| 40 |
+
16NamineRitsu-CN: 6
|
| 41 |
+
17LiYuan-CN: 7
|
| 42 |
+
18WenLi-CN: 8
|
| 43 |
+
19RanHuhu-CN: 9
|
| 44 |
+
1Crimson-CN: 10
|
| 45 |
+
20HuoLian-CN: 11
|
| 46 |
+
21XiaoYe-CN: 12
|
| 47 |
+
22QiLi-CN: 13
|
| 48 |
+
23QiFu-CN: 14
|
| 49 |
+
24XiaoMo-CN: 15
|
| 50 |
+
25YunHao-CN: 16
|
| 51 |
+
26ChuCi-CN: 17
|
| 52 |
+
27Tarara-CN: 18
|
| 53 |
+
28LuanMing-CN: 19
|
| 54 |
+
29HiragiYuki-CN: 20
|
| 55 |
+
2Orange-CN: 21
|
| 56 |
+
30TuYuan-CN: 22
|
| 57 |
+
3Tangerine-CN: 23
|
| 58 |
+
4Growl-CN: 24
|
| 59 |
+
5Steel-CN: 25
|
| 60 |
+
6Ghost-CN: 26
|
| 61 |
+
7David-CN: 27
|
| 62 |
+
8Lien-CN: 28
|
| 63 |
+
9Bianca-CN: 29
|
| 64 |
+
OpenSinger_ManRaw_0: 30
|
| 65 |
+
OpenSinger_ManRaw_1: 31
|
| 66 |
+
OpenSinger_ManRaw_10: 32
|
| 67 |
+
OpenSinger_ManRaw_11: 33
|
| 68 |
+
OpenSinger_ManRaw_12: 34
|
| 69 |
+
OpenSinger_ManRaw_13: 35
|
| 70 |
+
OpenSinger_ManRaw_14: 36
|
| 71 |
+
OpenSinger_ManRaw_15: 37
|
| 72 |
+
OpenSinger_ManRaw_16: 38
|
| 73 |
+
OpenSinger_ManRaw_17: 39
|
| 74 |
+
OpenSinger_ManRaw_18: 40
|
| 75 |
+
OpenSinger_ManRaw_19: 41
|
| 76 |
+
OpenSinger_ManRaw_2: 42
|
| 77 |
+
OpenSinger_ManRaw_20: 43
|
| 78 |
+
OpenSinger_ManRaw_21: 44
|
| 79 |
+
OpenSinger_ManRaw_22: 45
|
| 80 |
+
OpenSinger_ManRaw_23: 46
|
| 81 |
+
OpenSinger_ManRaw_24: 47
|
| 82 |
+
OpenSinger_ManRaw_25: 48
|
| 83 |
+
OpenSinger_ManRaw_26: 49
|
| 84 |
+
OpenSinger_ManRaw_27: 50
|
| 85 |
+
OpenSinger_ManRaw_3: 51
|
| 86 |
+
OpenSinger_ManRaw_4: 52
|
| 87 |
+
OpenSinger_ManRaw_5: 53
|
| 88 |
+
OpenSinger_ManRaw_6: 54
|
| 89 |
+
OpenSinger_ManRaw_7: 55
|
| 90 |
+
OpenSinger_ManRaw_8: 56
|
| 91 |
+
OpenSinger_ManRaw_9: 57
|
| 92 |
+
OpenSinger_WomanRaw_0: 58
|
| 93 |
+
OpenSinger_WomanRaw_1: 59
|
| 94 |
+
OpenSinger_WomanRaw_10: 60
|
| 95 |
+
OpenSinger_WomanRaw_11: 61
|
| 96 |
+
OpenSinger_WomanRaw_12: 62
|
| 97 |
+
OpenSinger_WomanRaw_13: 63
|
| 98 |
+
OpenSinger_WomanRaw_14: 64
|
| 99 |
+
OpenSinger_WomanRaw_15: 65
|
| 100 |
+
OpenSinger_WomanRaw_16: 66
|
| 101 |
+
OpenSinger_WomanRaw_17: 67
|
| 102 |
+
OpenSinger_WomanRaw_18: 68
|
| 103 |
+
OpenSinger_WomanRaw_19: 69
|
| 104 |
+
OpenSinger_WomanRaw_2: 70
|
| 105 |
+
OpenSinger_WomanRaw_20: 71
|
| 106 |
+
OpenSinger_WomanRaw_21: 72
|
| 107 |
+
OpenSinger_WomanRaw_22: 73
|
| 108 |
+
OpenSinger_WomanRaw_23: 74
|
| 109 |
+
OpenSinger_WomanRaw_24: 75
|
| 110 |
+
OpenSinger_WomanRaw_25: 76
|
| 111 |
+
OpenSinger_WomanRaw_26: 77
|
| 112 |
+
OpenSinger_WomanRaw_27: 78
|
| 113 |
+
OpenSinger_WomanRaw_28: 79
|
| 114 |
+
OpenSinger_WomanRaw_29: 80
|
| 115 |
+
OpenSinger_WomanRaw_3: 81
|
| 116 |
+
OpenSinger_WomanRaw_30: 82
|
| 117 |
+
OpenSinger_WomanRaw_31: 83
|
| 118 |
+
OpenSinger_WomanRaw_32: 84
|
| 119 |
+
OpenSinger_WomanRaw_33: 85
|
| 120 |
+
OpenSinger_WomanRaw_34: 86
|
| 121 |
+
OpenSinger_WomanRaw_35: 87
|
| 122 |
+
OpenSinger_WomanRaw_36: 88
|
| 123 |
+
OpenSinger_WomanRaw_37: 89
|
| 124 |
+
OpenSinger_WomanRaw_38: 90
|
| 125 |
+
OpenSinger_WomanRaw_39: 91
|
| 126 |
+
OpenSinger_WomanRaw_4: 92
|
| 127 |
+
OpenSinger_WomanRaw_40: 93
|
| 128 |
+
OpenSinger_WomanRaw_41: 94
|
| 129 |
+
OpenSinger_WomanRaw_42: 95
|
| 130 |
+
OpenSinger_WomanRaw_43: 96
|
| 131 |
+
OpenSinger_WomanRaw_44: 97
|
| 132 |
+
OpenSinger_WomanRaw_45: 98
|
| 133 |
+
OpenSinger_WomanRaw_46: 99
|
| 134 |
+
OpenSinger_WomanRaw_47: 100
|
| 135 |
+
OpenSinger_WomanRaw_5: 101
|
| 136 |
+
OpenSinger_WomanRaw_6: 102
|
| 137 |
+
OpenSinger_WomanRaw_7: 103
|
| 138 |
+
OpenSinger_WomanRaw_8: 104
|
| 139 |
+
OpenSinger_WomanRaw_9: 105
|
| 140 |
+
Opencpop: 106
|
| 141 |
+
Popcs: 107
|
| 142 |
+
m4singer_Alto-1: 108
|
| 143 |
+
m4singer_Alto-2: 109
|
| 144 |
+
m4singer_Alto-3: 110
|
| 145 |
+
m4singer_Alto-4: 111
|
| 146 |
+
m4singer_Alto-5: 112
|
| 147 |
+
m4singer_Alto-6: 113
|
| 148 |
+
m4singer_Alto-7: 114
|
| 149 |
+
m4singer_Bass-1: 115
|
| 150 |
+
m4singer_Bass-2: 116
|
| 151 |
+
m4singer_Bass-3: 117
|
| 152 |
+
m4singer_Soprano-1: 118
|
| 153 |
+
m4singer_Soprano-2: 119
|
| 154 |
+
m4singer_Soprano-3: 120
|
| 155 |
+
m4singer_Tenor-1: 121
|
| 156 |
+
m4singer_Tenor-2: 122
|
| 157 |
+
m4singer_Tenor-3: 123
|
| 158 |
+
m4singer_Tenor-4: 124
|
| 159 |
+
m4singer_Tenor-5: 125
|
| 160 |
+
m4singer_Tenor-6: 126
|
| 161 |
+
m4singer_Tenor-7: 127
|
| 162 |
+
"\u6771\u5317\u304D\u308A\u305F\u3093": 128
|
| 163 |
+
"\u6771\u5317\u30A4\u30BF\u30B3": 129
|
| 164 |
+
"\u6D3E\u8499_CHS": 130
|
| 165 |
+
"\u6D3E\u8499_EN": 131
|
| 166 |
+
"\u6D3E\u8499_JA": 132
|
| 167 |
+
train:
|
| 168 |
+
amp_dtype: fp16
|
| 169 |
+
batch_size: 200
|
| 170 |
+
cache_all_data: false
|
| 171 |
+
cache_device: cpu
|
| 172 |
+
cache_fp16: true
|
| 173 |
+
decay_step: 200000
|
| 174 |
+
epochs: 100000
|
| 175 |
+
gamma: 0.5
|
| 176 |
+
interval_force_save: 10000
|
| 177 |
+
interval_log: 100
|
| 178 |
+
interval_val: 5000
|
| 179 |
+
lr: 0.00015
|
| 180 |
+
num_workers: 4
|
| 181 |
+
save_opt: false
|
| 182 |
+
weight_decay: 0
|
| 183 |
+
vocoder:
|
| 184 |
+
ckpt: pretrain/nsf_hifigan/model
|
| 185 |
+
type: nsf-hifigan
|
ReFlow-VAE-SVC/reflowvae-wavenet_325000.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:11b531e0cce1e903e339035586591784a0ebfe87d0eec6f645f8c3307b6cea10
|
| 3 |
+
size 707522569
|