{ "train": { "log_interval": 100, "eval_interval": 1000, "seed": 42, "epochs": 100, "learning_rate": 0.0001, "betas": [ 0.8, 0.99 ], "eps": 1e-09, "batch_size": 16, "bf16_run": false, "lr_decay": 0.99995, "segment_size": 16384, "init_lr_ratio": 1, "warmup_epochs": 1, "c_mel": 45, "c_kl": 1.0, "c_commit": 100, "skip_optimizer": true, "freeze_ZH_bert": false, "freeze_JP_bert": false, "freeze_EN_bert": false, "freeze_YUE_bert": false, "freeze_emo": false }, "data": { "training_files": "/notebooks/bert-vits2/dataset/commonvoice16_1_bertvits2/train.list", "validation_files": "/notebooks/bert-vits2/dataset/commonvoice16_1_bertvits2/val.list", "max_wav_value": 32768.0, "sampling_rate": 44100, "filter_length": 2048, "hop_length": 512, "win_length": 2048, "n_mel_channels": 128, "mel_fmin": 0.0, "mel_fmax": null, "add_blank": true, "n_speakers": 887, "cleaned_text": true, "spk2id": { "SPK1887": 0, "SPK1844": 1, "SPK2377": 2, "SPK393": 3, "SPK861": 4, "SPK1979": 5, "SPK2004": 6, "SPK664": 7, "SPK1156": 8, "SPK199": 9, "SPK2280": 10, "SPK472": 11, "SPK1158": 12, "SPK963": 13, "SPK2162": 14, "SPK1044": 15, "SPK1922": 16, "SPK2480": 17, "SPK2424": 18, "SPK2547": 19, "SPK2473": 20, "SPK1359": 21, "SPK835": 22, "SPK1912": 23, "SPK77": 24, "SPK324": 25, "SPK1435": 26, "SPK2698": 27, "SPK313": 28, "SPK483": 29, "SPK1148": 30, "SPK1257": 31, "SPK471": 32, "SPK1293": 33, "SPK83": 34, "SPK730": 35, "SPK2615": 36, "SPK1967": 37, "SPK2124": 38, "SPK2561": 39, "SPK534": 40, "SPK540": 41, "SPK573": 42, "SPK1760": 43, "SPK470": 44, "SPK999": 45, "SPK805": 46, "SPK1899": 47, "SPK957": 48, "SPK687": 49, "SPK782": 50, "SPK828": 51, "SPK656": 52, "SPK512": 53, "SPK453": 54, "SPK1988": 55, "SPK929": 56, "SPK1079": 57, "SPK1026": 58, "SPK1401": 59, "SPK987": 60, "SPK1521": 61, "SPK2057": 62, "SPK1892": 63, "SPK514": 64, "SPK2432": 65, "SPK1296": 66, "SPK2376": 67, "SPK2667": 68, "SPK768": 69, "SPK1451": 70, "SPK912": 71, "SPK173": 72, "SPK1943": 73, "SPK2085": 74, "SPK2575": 75, "SPK1335": 76, "SPK244": 77, "SPK561": 78, "SPK24": 79, "SPK721": 80, "SPK2516": 81, "SPK1366": 82, "SPK1213": 83, "SPK1642": 84, "SPK1841": 85, "SPK2390": 86, "SPK964": 87, "SPK1015": 88, "SPK493": 89, "SPK1053": 90, "SPK49": 91, "SPK1152": 92, "SPK1885": 93, "SPK1095": 94, "SPK400": 95, "SPK1267": 96, "SPK528": 97, "SPK1858": 98, "SPK744": 99, "SPK221": 100, "SPK502": 101, "SPK2614": 102, "SPK2173": 103, "SPK2445": 104, "SPK1243": 105, "SPK421": 106, "SPK2620": 107, "SPK578": 108, "SPK2433": 109, "SPK1610": 110, "SPK1810": 111, "SPK1666": 112, "SPK855": 113, "SPK2041": 114, "SPK2138": 115, "SPK1544": 116, "SPK738": 117, "SPK759": 118, "SPK916": 119, "SPK156": 120, "SPK2139": 121, "SPK78": 122, "SPK896": 123, "SPK1185": 124, "SPK746": 125, "SPK1455": 126, "SPK2423": 127, "SPK458": 128, "SPK359": 129, "SPK2245": 130, "SPK2203": 131, "SPK1248": 132, "SPK1565": 133, "SPK207": 134, "SPK760": 135, "SPK1805": 136, "SPK2072": 137, "SPK2034": 138, "SPK2259": 139, "SPK2438": 140, "SPK1098": 141, "SPK680": 142, "SPK2548": 143, "SPK2178": 144, "SPK1699": 145, "SPK1204": 146, "SPK1939": 147, "SPK2523": 148, "SPK1983": 149, "SPK2530": 150, "SPK373": 151, "SPK283": 152, "SPK2502": 153, "SPK1745": 154, "SPK385": 155, "SPK1615": 156, "SPK785": 157, "SPK1794": 158, "SPK587": 159, "SPK801": 160, "SPK1896": 161, "SPK2130": 162, "SPK494": 163, "SPK2492": 164, "SPK2682": 165, "SPK2461": 166, "SPK1683": 167, "SPK2172": 168, "SPK650": 169, "SPK1127": 170, "SPK2189": 171, "SPK486": 172, "SPK291": 173, "SPK1363": 174, "SPK1970": 175, "SPK2094": 176, "SPK288": 177, "SPK1740": 178, "SPK1283": 179, "SPK1978": 180, "SPK442": 181, "SPK249": 182, "SPK482": 183, "SPK465": 184, "SPK204": 185, "SPK1300": 186, "SPK2521": 187, "SPK2474": 188, "SPK193": 189, "SPK2595": 190, "SPK174": 191, "SPK1295": 192, "SPK2291": 193, "SPK1282": 194, "SPK124": 195, "SPK2537": 196, "SPK1934": 197, "SPK645": 198, "SPK2672": 199, "SPK2412": 200, "SPK287": 201, "SPK334": 202, "SPK1902": 203, "SPK1801": 204, "SPK2621": 205, "SPK570": 206, "SPK501": 207, "SPK604": 208, "SPK2479": 209, "SPK2657": 210, "SPK823": 211, "SPK1324": 212, "SPK2482": 213, "SPK93": 214, "SPK1010": 215, "SPK1588": 216, "SPK950": 217, "SPK1897": 218, "SPK988": 219, "SPK2182": 220, "SPK717": 221, "SPK1469": 222, "SPK2338": 223, "SPK2400": 224, "SPK2261": 225, "SPK1360": 226, "SPK1933": 227, "SPK1963": 228, "SPK1478": 229, "SPK2417": 230, "SPK409": 231, "SPK989": 232, "SPK961": 233, "SPK420": 234, "SPK1520": 235, "SPK2311": 236, "SPK913": 237, "SPK1445": 238, "SPK492": 239, "SPK2633": 240, "SPK1516": 241, "SPK2418": 242, "SPK1743": 243, "SPK1163": 244, "SPK713": 245, "SPK2588": 246, "SPK2120": 247, "SPK524": 248, "SPK2453": 249, "SPK1725": 250, "SPK2360": 251, "SPK2344": 252, "SPK1355": 253, "SPK1677": 254, "SPK1223": 255, "SPK1233": 256, "SPK2688": 257, "SPK2230": 258, "SPK2421": 259, "SPK2416": 260, "SPK1173": 261, "SPK1070": 262, "SPK969": 263, "SPK2022": 264, "SPK2680": 265, "SPK945": 266, "SPK1539": 267, "SPK86": 268, "SPK836": 269, "SPK2602": 270, "SPK2101": 271, "SPK1161": 272, "SPK1946": 273, "SPK1691": 274, "SPK463": 275, "SPK2109": 276, "SPK1214": 277, "SPK1454": 278, "SPK590": 279, "SPK934": 280, "SPK1001": 281, "SPK2358": 282, "SPK1262": 283, "SPK694": 284, "SPK762": 285, "SPK1433": 286, "SPK433": 287, "SPK210": 288, "SPK323": 289, "SPK2364": 290, "SPK1125": 291, "SPK1479": 292, "SPK677": 293, "SPK2105": 294, "SPK1923": 295, "SPK1021": 296, "SPK711": 297, "SPK2246": 298, "SPK1782": 299, "SPK2213": 300, "SPK18": 301, "SPK1944": 302, "SPK1689": 303, "SPK2326": 304, "SPK301": 305, "SPK1168": 306, "SPK2401": 307, "SPK2483": 308, "SPK413": 309, "SPK464": 310, "SPK1890": 311, "SPK2168": 312, "SPK1863": 313, "SPK1264": 314, "SPK2662": 315, "SPK2198": 316, "SPK1973": 317, "SPK415": 318, "SPK361": 319, "SPK2249": 320, "SPK1093": 321, "SPK414": 322, "SPK136": 323, "SPK2159": 324, "SPK1050": 325, "SPK60": 326, "SPK517": 327, "SPK894": 328, "SPK2484": 329, "SPK2434": 330, "SPK1066": 331, "SPK1814": 332, "SPK985": 333, "SPK1103": 334, "SPK1102": 335, "SPK613": 336, "SPK1986": 337, "SPK1828": 338, "SPK722": 339, "SPK1517": 340, "SPK1618": 341, "SPK2195": 342, "SPK1675": 343, "SPK338": 344, "SPK1477": 345, "SPK2251": 346, "SPK542": 347, "SPK771": 348, "SPK706": 349, "SPK2076": 350, "SPK725": 351, "SPK2493": 352, "SPK1190": 353, "SPK427": 354, "SPK2609": 355, "SPK2070": 356, "SPK145": 357, "SPK737": 358, "SPK2229": 359, "SPK1481": 360, "SPK263": 361, "SPK635": 362, "SPK2104": 363, "SPK1708": 364, "SPK1853": 365, "SPK1279": 366, "SPK2149": 367, "SPK2301": 368, "SPK1274": 369, "SPK35": 370, "SPK557": 371, "SPK169": 372, "SPK1658": 373, "SPK690": 374, "SPK2244": 375, "SPK16": 376, "SPK2111": 377, "SPK459": 378, "SPK742": 379, "SPK2035": 380, "SPK2089": 381, "SPK31": 382, "SPK1926": 383, "SPK714": 384, "SPK2325": 385, "SPK2605": 386, "SPK1217": 387, "SPK811": 388, "SPK1351": 389, "SPK456": 390, "SPK200": 391, "SPK2191": 392, "SPK1855": 393, "SPK41": 394, "SPK1072": 395, "SPK2074": 396, "SPK2354": 397, "SPK953": 398, "SPK1976": 399, "SPK376": 400, "SPK1399": 401, "SPK2368": 402, "SPK1470": 403, "SPK102": 404, "SPK1848": 405, "SPK1358": 406, "SPK830": 407, "SPK1661": 408, "SPK1060": 409, "SPK1690": 410, "SPK123": 411, "SPK2304": 412, "SPK2696": 413, "SPK1654": 414, "SPK522": 415, "SPK976": 416, "SPK755": 417, "SPK2699": 418, "SPK2129": 419, "SPK702": 420, "SPK2061": 421, "SPK1905": 422, "SPK351": 423, "SPK119": 424, "SPK2611": 425, "SPK2233": 426, "SPK1386": 427, "SPK1506": 428, "SPK2180": 429, "SPK2507": 430, "SPK2312": 431, "SPK2285": 432, "SPK2023": 433, "SPK990": 434, "SPK1752": 435, "SPK1002": 436, "SPK789": 437, "SPK1714": 438, "SPK84": 439, "SPK2046": 440, "SPK2459": 441, "SPK2520": 442, "SPK1815": 443, "SPK917": 444, "SPK7": 445, "SPK1391": 446, "SPK2123": 447, "SPK2392": 448, "SPK396": 449, "SPK1354": 450, "SPK2415": 451, "SPK1122": 452, "SPK317": 453, "SPK58": 454, "SPK1292": 455, "SPK319": 456, "SPK129": 457, "SPK2572": 458, "SPK975": 459, "SPK1261": 460, "SPK2084": 461, "SPK2529": 462, "SPK1437": 463, "SPK549": 464, "SPK1191": 465, "SPK1080": 466, "SPK1131": 467, "SPK686": 468, "SPK2152": 469, "SPK224": 470, "SPK1600": 471, "SPK435": 472, "SPK1875": 473, "SPK784": 474, "SPK2408": 475, "SPK1024": 476, "SPK2221": 477, "SPK1031": 478, "SPK863": 479, "SPK986": 480, "SPK219": 481, "SPK390": 482, "SPK68": 483, "SPK2573": 484, "SPK1442": 485, "SPK1346": 486, "SPK2190": 487, "SPK2458": 488, "SPK2524": 489, "SPK1818": 490, "SPK1862": 491, "SPK1620": 492, "SPK2468": 493, "SPK716": 494, "SPK1347": 495, "SPK1203": 496, "SPK235": 497, "SPK484": 498, "SPK149": 499, "SPK551": 500, "SPK2235": 501, "SPK215": 502, "SPK676": 503, "SPK897": 504, "SPK684": 505, "SPK1700": 506, "SPK419": 507, "SPK2078": 508, "SPK996": 509, "SPK150": 510, "SPK1023": 511, "SPK2581": 512, "SPK1883": 513, "SPK52": 514, "SPK1188": 515, "SPK2471": 516, "SPK705": 517, "SPK1999": 518, "SPK1170": 519, "SPK1164": 520, "SPK1850": 521, "SPK154": 522, "SPK1571": 523, "SPK265": 524, "SPK1851": 525, "SPK2268": 526, "SPK2475": 527, "SPK1311": 528, "SPK1652": 529, "SPK1075": 530, "SPK852": 531, "SPK2263": 532, "SPK135": 533, "SPK346": 534, "SPK428": 535, "SPK1108": 536, "SPK2324": 537, "SPK2150": 538, "SPK2491": 539, "SPK1463": 540, "SPK2503": 541, "SPK1096": 542, "SPK2486": 543, "SPK1318": 544, "SPK813": 545, "SPK1888": 546, "SPK233": 547, "SPK2289": 548, "SPK1417": 549, "SPK1290": 550, "SPK1299": 551, "SPK2513": 552, "SPK2334": 553, "SPK440": 554, "SPK1393": 555, "SPK1396": 556, "SPK959": 557, "SPK581": 558, "SPK1062": 559, "SPK247": 560, "SPK593": 561, "SPK840": 562, "SPK1564": 563, "SPK504": 564, "SPK356": 565, "SPK1276": 566, "SPK477": 567, "SPK1310": 568, "SPK2098": 569, "SPK2685": 570, "SPK1750": 571, "SPK2627": 572, "SPK392": 573, "SPK2040": 574, "SPK2654": 575, "SPK341": 576, "SPK1519": 577, "SPK1420": 578, "SPK2514": 579, "SPK446": 580, "SPK2025": 581, "SPK1550": 582, "SPK1294": 583, "SPK2148": 584, "SPK838": 585, "SPK369": 586, "SPK2560": 587, "SPK152": 588, "SPK1877": 589, "SPK1114": 590, "SPK1032": 591, "SPK365": 592, "SPK397": 593, "SPK1298": 594, "SPK2666": 595, "SPK1372": 596, "SPK1138": 597, "SPK2331": 598, "SPK1": 599, "SPK796": 600, "SPK867": 601, "SPK2277": 602, "SPK315": 603, "SPK2016": 604, "SPK45": 605, "SPK925": 606, "SPK1870": 607, "SPK2252": 608, "SPK65": 609, "SPK278": 610, "SPK2651": 611, "SPK1370": 612, "SPK1781": 613, "SPK209": 614, "SPK1104": 615, "SPK337": 616, "SPK1429": 617, "SPK256": 618, "SPK2613": 619, "SPK1286": 620, "SPK1194": 621, "SPK2343": 622, "SPK398": 623, "SPK1956": 624, "SPK914": 625, "SPK844": 626, "SPK1043": 627, "SPK1336": 628, "SPK795": 629, "SPK2467": 630, "SPK1484": 631, "SPK971": 632, "SPK2294": 633, "SPK112": 634, "SPK2541": 635, "SPK401": 636, "SPK688": 637, "SPK1201": 638, "SPK2444": 639, "SPK2533": 640, "SPK1731": 641, "SPK1438": 642, "SPK1728": 643, "SPK610": 644, "SPK1827": 645, "SPK155": 646, "SPK1047": 647, "SCC_SPK1": 648, "SCC_SPK2": 649, "ZH_SPK14": 650, "ZH_SPK39": 651, "ZH_SPK12": 652, "ZH_SPK4": 653, "ZH_SPK65": 654, "ZH_SPK35": 655, "ZH_SPK28": 656, "ZH_SPK78": 657, "ZH_SPK41": 658, "ZH_SPK66": 659, "ZH_SPK13": 660, "ZH_SPK21": 661, "ZH_SPK33": 662, "ZH_SPK11": 663, "ZH_SPK30": 664, "ZH_SPK26": 665, "ZH_SPK72": 666, "ZH_SPK60": 667, "ZH_SPK22": 668, "ZH_SPK68": 669, "ZH_SPK63": 670, "ZH_SPK59": 671, "ZH_SPK1": 672, "ZH_SPK94": 673, "ZH_SPK19": 674, "ZH_SPK55": 675, "ZH_SPK56": 676, "ZH_SPK75": 677, "ZH_SPK17": 678, "ZH_SPK70": 679, "ZH_SPK88": 680, "ZH_SPK46": 681, "ZH_SPK40": 682, "ZH_SPK62": 683, "ZH_SPK48": 684, "ZH_SPK85": 685, "ZH_SPK31": 686, "ZH_SPK57": 687, "ZH_SPK42": 688, "ZH_SPK91": 689, "ZH_SPK29": 690, "ZH_SPK73": 691, "ZH_SPK64": 692, "ZH_SPK23": 693, "ZH_SPK16": 694, "ZH_SPK61": 695, "ZH_SPK76": 696, "ZH_SPK37": 697, "ZH_SPK8": 698, "ZH_SPK83": 699, "ZH_SPK34": 700, "ZH_SPK77": 701, "ZH_SPK9": 702, "ZH_SPK53": 703, "ZH_SPK44": 704, "ZH_SPK71": 705, "ZH_SPK84": 706, "ZH_SPK67": 707, "ZH_SPK49": 708, "ZH_SPK50": 709, "ZH_SPK15": 710, "ZH_SPK79": 711, "ZH_SPK90": 712, "ZH_SPK10": 713, "ZH_SPK95": 714, "ZH_SPK24": 715, "ZH_SPK89": 716, "ZH_SPK36": 717, "ZH_SPK51": 718, "ZH_SPK43": 719, "ZH_SPK45": 720, "ZH_SPK27": 721, "ZH_SPK69": 722, "ZH_SPK5": 723, "ZH_SPK74": 724, "ZH_SPK80": 725, "ZH_SPK0": 726, "ZH_SPK93": 727, "ZH_SPK92": 728, "ZH_SPK47": 729, "ZH_SPK7": 730, "ZH_SPK87": 731, "ZH_SPK32": 732, "ZH_SPK2": 733, "ZH_SPK86": 734, "ZH_SPK6": 735, "ZH_SPK3": 736, "ZH_SPK18": 737, "ZH_SPK81": 738, "ZH_SPK82": 739, "ZH_SPK38": 740, "EN_SPK6": 741, "EN_SPK25": 742, "EN_SPK98": 743, "EN_SPK85": 744, "EN_SPK96": 745, "EN_SPK46": 746, "EN_SPK9": 747, "EN_SPK15": 748, "EN_SPK42": 749, "EN_SPK56": 750, "EN_SPK28": 751, "EN_SPK32": 752, "EN_SPK37": 753, "EN_SPK81": 754, "EN_SPK52": 755, "EN_SPK66": 756, "EN_SPK39": 757, "EN_SPK63": 758, "EN_SPK61": 759, "EN_SPK47": 760, "EN_SPK12": 761, "EN_SPK69": 762, "EN_SPK76": 763, "EN_SPK13": 764, "EN_SPK70": 765, "EN_SPK87": 766, "EN_SPK38": 767, "EN_SPK68": 768, "EN_SPK99": 769, "EN_SPK34": 770, "EN_SPK73": 771, "EN_SPK5": 772, "EN_SPK14": 773, "EN_SPK33": 774, "EN_SPK75": 775, "EN_SPK19": 776, "EN_SPK8": 777, "EN_SPK4": 778, "EN_SPK59": 779, "EN_SPK89": 780, "EN_SPK72": 781, "EN_SPK23": 782, "EN_SPK30": 783, "EN_SPK95": 784, "EN_SPK94": 785, "EN_SPK78": 786, "EN_SPK88": 787, "EN_SPK64": 788, "EN_SPK92": 789, "EN_SPK84": 790, "EN_SPK58": 791, "EN_SPK16": 792, "EN_SPK51": 793, "EN_SPK48": 794, "EN_SPK44": 795, "EN_SPK31": 796, "EN_SPK17": 797, "EN_SPK55": 798, "EN_SPK49": 799, "EN_SPK71": 800, "EN_SPK54": 801, "EN_SPK36": 802, "EN_SPK40": 803, "EN_SPK74": 804, "EN_SPK24": 805, "EN_SPK57": 806, "EN_SPK10": 807, "EN_SPK18": 808, "EN_SPK11": 809, "EN_SPK20": 810, "EN_SPK2": 811, "EN_SPK45": 812, "EN_SPK65": 813, "JP_SPK70": 814, "JP_SPK1": 815, "JP_SPK73": 816, "JP_SPK42": 817, "JP_SPK62": 818, "JP_SPK44": 819, "JP_SPK43": 820, "JP_SPK69": 821, "JP_SPK12": 822, "JP_SPK65": 823, "JP_SPK49": 824, "JP_SPK78": 825, "JP_SPK81": 826, "JP_SPK26": 827, "JP_SPK29": 828, "JP_SPK77": 829, "JP_SPK56": 830, "JP_SPK59": 831, "JP_SPK6": 832, "JP_SPK88": 833, "JP_SPK89": 834, "JP_SPK52": 835, "JP_SPK8": 836, "JP_SPK61": 837, "JP_SPK19": 838, "JP_SPK83": 839, "JP_SPK82": 840, "JP_SPK46": 841, "JP_SPK3": 842, "JP_SPK92": 843, "JP_SPK63": 844, "JP_SPK34": 845, "JP_SPK10": 846, "JP_SPK35": 847, "JP_SPK9": 848, "JP_SPK39": 849, "JP_SPK87": 850, "JP_SPK79": 851, "JP_SPK86": 852, "JP_SPK24": 853, "JP_SPK58": 854, "JP_SPK71": 855, "JP_SPK45": 856, "JP_SPK72": 857, "JP_SPK7": 858, "JP_SPK51": 859, "JP_SPK68": 860, "JP_SPK18": 861, "JP_SPK64": 862, "JP_SPK93": 863, "JP_SPK41": 864, "JP_SPK96": 865, "JP_SPK66": 866, "JP_SPK38": 867, "JP_SPK31": 868, "JP_SPK21": 869, "JP_SPK75": 870, "JP_SPK27": 871, "JP_SPK32": 872, "JP_SPK22": 873, "JP_SPK85": 874, "JP_SPK23": 875, "JP_SPK33": 876, "JP_SPK80": 877, "JP_SPK25": 878, "JP_SPK74": 879, "JP_SPK36": 880, "JP_SPK17": 881, "JP_SPK0": 882, "JP_SPK5": 883, "JP_SPK37": 884, "JP_SPK60": 885, "JP_SPK91": 886 } }, "model": { "use_spk_conditioned_encoder": true, "use_noise_scaled_mas": true, "use_mel_posterior_encoder": false, "use_duration_discriminator": true, "inter_channels": 192, "hidden_channels": 192, "filter_channels": 768, "n_heads": 2, "n_layers": 6, "kernel_size": 3, "p_dropout": 0.1, "resblock": "1", "resblock_kernel_sizes": [ 3, 7, 11 ], "resblock_dilation_sizes": [ [ 1, 3, 5 ], [ 1, 3, 5 ], [ 1, 3, 5 ] ], "upsample_rates": [ 8, 8, 2, 2, 2 ], "upsample_initial_channel": 512, "upsample_kernel_sizes": [ 16, 16, 8, 2, 2 ], "n_layers_q": 3, "use_spectral_norm": false, "gin_channels": 512, "slm": { "model": "./slm/wavlm-base-plus", "sr": 16000, "hidden": 768, "nlayers": 13, "initial_channel": 64 } }, "version": "2.3" }