Hugging Face's logo

Spaces:
Nogizaka46
/
Nogizaka46-so
Running

App Files Community
darksakura commited on
Commit
da4aebf
•
1 Parent(s): a2aaaec

Upload 4 files

Browse files
Files changed (4) hide show
  1. N.zip +3 -0
  2. configs/44.json +149 -0
  3. configs/58v1.json +162 -0
  4. configs/58v2.json +162 -0
N.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b56e4467e1353d6475a17e6b4d67631103cd68df61108ddbe98ebb4962fbc1ec
3
+ size 447896729
configs/44.json ADDED
@@ -0,0 +1,149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 400,
4
+ "eval_interval": 1200,
5
+ "seed": 1234,
6
+ "epochs": 10000,
7
+ "learning_rate": 0.0001,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 6,
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": 50,
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": 45,
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
+ "AKIMOTO_MANATSU": 0,
104
+ "ENDO_SAKURA": 1,
105
+ "ETO_MISA": 2,
106
+ "HASHIMOTO_NANAMI": 3,
107
+ "HAYAKAWA_SEIRA": 4,
108
+ "HORI_MIONA": 5,
109
+ "HOSHINO_MINAMI": 6,
110
+ "ICHINOSE_MIKU": 7,
111
+ "IKEDA_TERESA": 8,
112
+ "IKUTA_ERIKA": 9,
113
+ "INOUE_NAGI": 10,
114
+ "INOUE_SAYURI": 11,
115
+ "IOKI_MAO": 12,
116
+ "IWAMOTO_RENKA": 13,
117
+ "KAKEHASHI_SAYAKA": 14,
118
+ "KAKI_HARUKA": 15,
119
+ "KANAGAWA_SAYA": 16,
120
+ "KUBO_SHIORI": 17,
121
+ "MATSUMURA_SAYURI": 18,
122
+ "NAKANISHI_ARUNO": 19,
123
+ "NAO_YUMIKI": 20,
124
+ "NISHINO_NANASE": 21,
125
+ "OGAWA_AYA": 22,
126
+ "OKUDA_IROHA": 23,
127
+ "OZONO_MOMOKO": 24,
128
+ "RIKA_SATO": 25,
129
+ "RUNA_HAYASHI": 26,
130
+ "SAITO_ASUKA": 27,
131
+ "SAKAGUCHI_TAMAMI": 28,
132
+ "SAKURAI_REIKA": 29,
133
+ "SATO_KAEDE": 30,
134
+ "SATO_YUURI": 31,
135
+ "SEIMIYA_REI": 32,
136
+ "SHIBATA_YUNA": 33,
137
+ "SHINUCHI_MAI": 34,
138
+ "SHIRAISHI_MAI": 35,
139
+ "SUGAWARA_SATSUKI": 36,
140
+ "TAKAYAMA_KAZUMI": 37,
141
+ "TAMURA_MAYU": 38,
142
+ "TERADA_RANZE": 39,
143
+ "TOMISATO_NAO": 40,
144
+ "TSUTSUI_AYAME": 41,
145
+ "UMEZAWA_MINAMI": 42,
146
+ "YAMASHITA_MIZUKI": 43,
147
+ "YODA_YUUKI": 44
148
+ }
149
+ }
configs/58v1.json ADDED
@@ -0,0 +1,162 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 500,
4
+ "eval_interval": 1000,
5
+ "seed": 1234,
6
+ "epochs": 10000,
7
+ "learning_rate": 0.0002,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 16,
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": 50,
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": 58,
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
+ "AKIMOTO_MANATSU": 0,
104
+ "ENDO_SAKURA": 1,
105
+ "ETO_MISA": 2,
106
+ "FUKAGAWA_MAI": 3,
107
+ "HARUKA_KUROMI": 4,
108
+ "HASHIMOTO_NANAMI": 5,
109
+ "HAYAKAWA_SEIRA": 6,
110
+ "HORI_MIONA": 7,
111
+ "HOSHINO_MINAMI": 8,
112
+ "ICHINOSE_MIKU": 9,
113
+ "IKEDA_TERESA": 10,
114
+ "IKUTA_ERIKA": 11,
115
+ "INOUE_NAGI": 12,
116
+ "INOUE_SAYURI": 13,
117
+ "IOKI_MAO": 14,
118
+ "ITO_RIRIA": 15,
119
+ "IWAMOTO_RENKA": 16,
120
+ "KAKEHASHI_SAYAKA": 17,
121
+ "KAKI_HARUKA": 18,
122
+ "KANAGAWA_SAYA": 19,
123
+ "KAWASAKI_SAKURA": 20,
124
+ "KITAGAWA_YURI": 21,
125
+ "KITANO_HINAKO": 22,
126
+ "KUBO_SHIORI": 23,
127
+ "MATSUMURA_SAYURI": 24,
128
+ "MIYU_MATSUO": 25,
129
+ "MUKAI_HAZUKI": 26,
130
+ "NAKAMOTO_HIMEKA": 27,
131
+ "NAKANISHI_ARUNO": 28,
132
+ "NAO_YUMIKI": 29,
133
+ "NISHINO_NANASE": 30,
134
+ "OGAWA_AYA": 31,
135
+ "OKAMOTO_HINA": 32,
136
+ "OKUDA_IROHA": 33,
137
+ "OZONO_MOMOKO": 34,
138
+ "RIKA_SATO": 35,
139
+ "RUNA_HAYASHI": 36,
140
+ "SAITO_ASUKA": 37,
141
+ "SAKAGUCHI_TAMAMI": 38,
142
+ "SAKURAI_REIKA": 39,
143
+ "SASAKI_KOTOKO": 40,
144
+ "SATO_KAEDE": 41,
145
+ "SATO_YUURI": 42,
146
+ "SEIMIYA_REI": 43,
147
+ "SHIBATA_YUNA": 44,
148
+ "SHINUCHI_MAI": 45,
149
+ "SHIRAISHI_MAI": 46,
150
+ "SUGAWARA_SATSUKI": 47,
151
+ "SUZUKI_AYANE": 48,
152
+ "TAKAYAMA_KAZUMI": 49,
153
+ "TAMURA_MAYU": 50,
154
+ "TERADA_RANZE": 51,
155
+ "TOMISATO_NAO": 52,
156
+ "TSUTSUI_AYAME": 53,
157
+ "UMEZAWA_MINAMI": 54,
158
+ "YAKUBO_MIO": 55,
159
+ "YAMASHITA_MIZUKI": 56,
160
+ "YODA_YUUKI": 57
161
+ }
162
+ }
configs/58v2.json ADDED
@@ -0,0 +1,162 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 500,
4
+ "eval_interval": 1000,
5
+ "seed": 1234,
6
+ "epochs": 10000,
7
+ "learning_rate": 0.0002,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 16,
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": 50,
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": 58,
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
+ "AKIMOTO_MANATSU": 0,
104
+ "ENDO_SAKURA": 1,
105
+ "ETO_MISA": 2,
106
+ "FUKAGAWA_MAI": 3,
107
+ "HARUKA_KUROMI": 4,
108
+ "HASHIMOTO_NANAMI": 5,
109
+ "HAYAKAWA_SEIRA": 6,
110
+ "HORI_MIONA": 7,
111
+ "HOSHINO_MINAMI": 8,
112
+ "ICHINOSE_MIKU": 9,
113
+ "IKEDA_TERESA": 10,
114
+ "IKUTA_ERIKA": 11,
115
+ "INOUE_NAGI": 12,
116
+ "INOUE_SAYURI": 13,
117
+ "IOKI_MAO": 14,
118
+ "ITO_RIRIA": 15,
119
+ "IWAMOTO_RENKA": 16,
120
+ "KAKEHASHI_SAYAKA": 17,
121
+ "KAKI_HARUKA": 18,
122
+ "KANAGAWA_SAYA": 19,
123
+ "KAWASAKI_SAKURA": 20,
124
+ "KITAGAWA_YURI": 21,
125
+ "KITANO_HINAKO": 22,
126
+ "KUBO_SHIORI": 23,
127
+ "MATSUMURA_SAYURI": 24,
128
+ "MIYU_MATSUO": 25,
129
+ "MUKAI_HAZUKI": 26,
130
+ "NAKAMOTO_HIMEKA": 27,
131
+ "NAKANISHI_ARUNO": 28,
132
+ "NAO_YUMIKI": 29,
133
+ "NISHINO_NANASE": 30,
134
+ "OGAWA_AYA": 31,
135
+ "OKAMOTO_HINA": 32,
136
+ "OKUDA_IROHA": 33,
137
+ "OZONO_MOMOKO": 34,
138
+ "RIKA_SATO": 35,
139
+ "RUNA_HAYASHI": 36,
140
+ "SAITO_ASUKA": 37,
141
+ "SAKAGUCHI_TAMAMI": 38,
142
+ "SAKURAI_REIKA": 39,
143
+ "SASAKI_KOTOKO": 40,
144
+ "SATO_KAEDE": 41,
145
+ "SATO_YUURI": 42,
146
+ "SEIMIYA_REI": 43,
147
+ "SHIBATA_YUNA": 44,
148
+ "SHINUCHI_MAI": 45,
149
+ "SHIRAISHI_MAI": 46,
150
+ "SUGAWARA_SATSUKI": 47,
151
+ "SUZUKI_AYANE": 48,
152
+ "TAKAYAMA_KAZUMI": 49,
153
+ "TAMURA_MAYU": 50,
154
+ "TERADA_RANZE": 51,
155
+ "TOMISATO_NAO": 52,
156
+ "TSUTSUI_AYAME": 53,
157
+ "UMEZAWA_MINAMI": 54,
158
+ "YAKUBO_MIO": 55,
159
+ "YAMASHITA_MIZUKI": 56,
160
+ "YODA_YUUKI": 57
161
+ }
162
+ }