Upload 2 files
Browse files- checkpoint.pth +3 -0
- config.json +186 -0
checkpoint.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a74e9eadffff065c75eb6dfa040efa72cad23e72cfea70d39190bc174fb97093
|
| 3 |
+
size 207770124
|
config.json
ADDED
|
@@ -0,0 +1,186 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"train": {
|
| 3 |
+
"segment_size": 16384
|
| 4 |
+
},
|
| 5 |
+
"data": {
|
| 6 |
+
"sampling_rate": 44100,
|
| 7 |
+
"filter_length": 2048,
|
| 8 |
+
"hop_length": 512,
|
| 9 |
+
"add_blank": true,
|
| 10 |
+
"n_speakers": 256,
|
| 11 |
+
"spk2id": {
|
| 12 |
+
"ZH": 1
|
| 13 |
+
}
|
| 14 |
+
},
|
| 15 |
+
"model": {
|
| 16 |
+
"use_spk_conditioned_encoder": true,
|
| 17 |
+
"use_noise_scaled_mas": true,
|
| 18 |
+
"use_mel_posterior_encoder": false,
|
| 19 |
+
"use_duration_discriminator": true,
|
| 20 |
+
"inter_channels": 192,
|
| 21 |
+
"hidden_channels": 192,
|
| 22 |
+
"filter_channels": 768,
|
| 23 |
+
"n_heads": 2,
|
| 24 |
+
"n_layers": 6,
|
| 25 |
+
"n_layers_trans_flow": 3,
|
| 26 |
+
"kernel_size": 3,
|
| 27 |
+
"p_dropout": 0.1,
|
| 28 |
+
"resblock": "1",
|
| 29 |
+
"resblock_kernel_sizes": [
|
| 30 |
+
3,
|
| 31 |
+
7,
|
| 32 |
+
11
|
| 33 |
+
],
|
| 34 |
+
"resblock_dilation_sizes": [
|
| 35 |
+
[
|
| 36 |
+
1,
|
| 37 |
+
3,
|
| 38 |
+
5
|
| 39 |
+
],
|
| 40 |
+
[
|
| 41 |
+
1,
|
| 42 |
+
3,
|
| 43 |
+
5
|
| 44 |
+
],
|
| 45 |
+
[
|
| 46 |
+
1,
|
| 47 |
+
3,
|
| 48 |
+
5
|
| 49 |
+
]
|
| 50 |
+
],
|
| 51 |
+
"upsample_rates": [
|
| 52 |
+
8,
|
| 53 |
+
8,
|
| 54 |
+
2,
|
| 55 |
+
2,
|
| 56 |
+
2
|
| 57 |
+
],
|
| 58 |
+
"upsample_initial_channel": 512,
|
| 59 |
+
"upsample_kernel_sizes": [
|
| 60 |
+
16,
|
| 61 |
+
16,
|
| 62 |
+
8,
|
| 63 |
+
2,
|
| 64 |
+
2
|
| 65 |
+
],
|
| 66 |
+
"n_layers_q": 3,
|
| 67 |
+
"use_spectral_norm": false,
|
| 68 |
+
"gin_channels": 256
|
| 69 |
+
},
|
| 70 |
+
"symbols": [
|
| 71 |
+
"_",
|
| 72 |
+
"AA",
|
| 73 |
+
"E",
|
| 74 |
+
"EE",
|
| 75 |
+
"En",
|
| 76 |
+
"N",
|
| 77 |
+
"OO",
|
| 78 |
+
"V",
|
| 79 |
+
"a",
|
| 80 |
+
"a:",
|
| 81 |
+
"aa",
|
| 82 |
+
"ae",
|
| 83 |
+
"ah",
|
| 84 |
+
"ai",
|
| 85 |
+
"an",
|
| 86 |
+
"ang",
|
| 87 |
+
"ao",
|
| 88 |
+
"aw",
|
| 89 |
+
"ay",
|
| 90 |
+
"b",
|
| 91 |
+
"by",
|
| 92 |
+
"c",
|
| 93 |
+
"ch",
|
| 94 |
+
"d",
|
| 95 |
+
"dh",
|
| 96 |
+
"dy",
|
| 97 |
+
"e",
|
| 98 |
+
"e:",
|
| 99 |
+
"eh",
|
| 100 |
+
"ei",
|
| 101 |
+
"en",
|
| 102 |
+
"eng",
|
| 103 |
+
"er",
|
| 104 |
+
"ey",
|
| 105 |
+
"f",
|
| 106 |
+
"g",
|
| 107 |
+
"gy",
|
| 108 |
+
"h",
|
| 109 |
+
"hh",
|
| 110 |
+
"hy",
|
| 111 |
+
"i",
|
| 112 |
+
"i0",
|
| 113 |
+
"i:",
|
| 114 |
+
"ia",
|
| 115 |
+
"ian",
|
| 116 |
+
"iang",
|
| 117 |
+
"iao",
|
| 118 |
+
"ie",
|
| 119 |
+
"ih",
|
| 120 |
+
"in",
|
| 121 |
+
"ing",
|
| 122 |
+
"iong",
|
| 123 |
+
"ir",
|
| 124 |
+
"iu",
|
| 125 |
+
"iy",
|
| 126 |
+
"j",
|
| 127 |
+
"jh",
|
| 128 |
+
"k",
|
| 129 |
+
"ky",
|
| 130 |
+
"l",
|
| 131 |
+
"m",
|
| 132 |
+
"my",
|
| 133 |
+
"n",
|
| 134 |
+
"ng",
|
| 135 |
+
"ny",
|
| 136 |
+
"o",
|
| 137 |
+
"o:",
|
| 138 |
+
"ong",
|
| 139 |
+
"ou",
|
| 140 |
+
"ow",
|
| 141 |
+
"oy",
|
| 142 |
+
"p",
|
| 143 |
+
"py",
|
| 144 |
+
"q",
|
| 145 |
+
"r",
|
| 146 |
+
"ry",
|
| 147 |
+
"s",
|
| 148 |
+
"sh",
|
| 149 |
+
"t",
|
| 150 |
+
"th",
|
| 151 |
+
"ts",
|
| 152 |
+
"ty",
|
| 153 |
+
"u",
|
| 154 |
+
"u:",
|
| 155 |
+
"ua",
|
| 156 |
+
"uai",
|
| 157 |
+
"uan",
|
| 158 |
+
"uang",
|
| 159 |
+
"uh",
|
| 160 |
+
"ui",
|
| 161 |
+
"un",
|
| 162 |
+
"uo",
|
| 163 |
+
"uw",
|
| 164 |
+
"v",
|
| 165 |
+
"van",
|
| 166 |
+
"ve",
|
| 167 |
+
"vn",
|
| 168 |
+
"w",
|
| 169 |
+
"x",
|
| 170 |
+
"y",
|
| 171 |
+
"z",
|
| 172 |
+
"zh",
|
| 173 |
+
"zy",
|
| 174 |
+
"!",
|
| 175 |
+
"?",
|
| 176 |
+
"…",
|
| 177 |
+
",",
|
| 178 |
+
".",
|
| 179 |
+
"'",
|
| 180 |
+
"-",
|
| 181 |
+
"SP",
|
| 182 |
+
"UNK"
|
| 183 |
+
],
|
| 184 |
+
"num_tones": 11,
|
| 185 |
+
"num_languages": 4
|
| 186 |
+
}
|