ru-irina-0p51m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "ru_RU-flores-20260908", |
| 4 | "language": "ru_RU", |
| 5 | "voice": "ru_RU-irina-medium", |
| 6 | "created_from": "tools/export_roota_self_contained_package.py", |
| 7 | "sample_rate": 22050, |
| 8 | "hop_length": 256, |
| 9 | "inference": { |
| 10 | "duration_length_scale": 1.0, |
| 11 | "duration_policy": "round(exp(log_duration) * duration_length_scale)" |
| 12 | }, |
| 13 | "weights_file": "weights.fp16.bin", |
| 14 | "weights_size_bytes": 1024662, |
| 15 | "weights_sha256": "082c75f3f740e0ea624a79479503fe4395342068fe702a7edab14741b1f0b31c", |
| 16 | "total_parameters": 512331, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/ru_RU-tiny/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "36eef553c3a654725a44edd28c28871ce5c7d2e6d40c0c41d3a48012b9d9a3f5", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 151, |
| 24 | "hidden": 32, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 1019, |
| 28 | "max_duration": 80, |
| 29 | "target_duration_scale": 1.0, |
| 30 | "pause_preserve_scale": 0.0, |
| 31 | "pause_preserve_window": 0, |
| 32 | "pause_token_ids": [ |
| 33 | 4, |
| 34 | 8, |
| 35 | 10, |
| 36 | 11, |
| 37 | 12, |
| 38 | 13 |
| 39 | ], |
| 40 | "long_preserve_threshold": 0.0, |
| 41 | "long_preserve_scale": 0.0, |
| 42 | "duration_vocab_size_forced": 0, |
| 43 | "duration_oov_id": 59 |
| 44 | }, |
| 45 | "parameters": 36932, |
| 46 | "tensors": [ |
| 47 | { |
| 48 | "name": "blocks.0.net.0.bias", |
| 49 | "shape": [ |
| 50 | 32 |
| 51 | ], |
| 52 | "dtype": "float16", |
| 53 | "offset_bytes": 0, |
| 54 | "nbytes": 64, |
| 55 | "sha256": "f355d22765b835a2c6e580b1e19753439a4a6281a214b76ff2f9c7b729fb8a51" |
| 56 | }, |
| 57 | { |
| 58 | "name": "blocks.0.net.0.weight", |
| 59 | "shape": [ |
| 60 | 32, |
| 61 | 32, |
| 62 | 5 |
| 63 | ], |
| 64 | "dtype": "float16", |
| 65 | "offset_bytes": 64, |
| 66 | "nbytes": 10240, |
| 67 | "sha256": "5606e71be8cdc32cbd7be6ca447f10c4e061ac1f7fbab0689c2ea3f5214de955" |
| 68 | }, |
| 69 | { |
| 70 | "name": "blocks.0.net.2.bias", |
| 71 | "shape": [ |
| 72 | 32 |
| 73 | ], |
| 74 | "dtype": "float16", |
| 75 | "offset_bytes": 10304, |
| 76 | "nbytes": 64, |
| 77 | "sha256": "9ea2e6949c9d9864a3fee1941e4ee585a19c245f345e1dbf02e0d1f740ea5214" |
| 78 | }, |
| 79 | { |
| 80 | "name": "blocks.0.net.2.weight", |
| 81 | "shape": [ |
| 82 | 32, |
| 83 | 32, |
| 84 | 5 |
| 85 | ], |
| 86 | "dtype": "float16", |
| 87 | "offset_bytes": 10368, |
| 88 | "nbytes": 10240, |
| 89 | "sha256": "b598545eec11ca086cb3b895f5330f982ba51400b0735a9737c19af63676e910" |
| 90 | }, |
| 91 | { |
| 92 | "name": "blocks.0.scale", |
| 93 | "shape": [ |
| 94 | 1 |
| 95 | ], |
| 96 | "dtype": "float16", |
| 97 | "offset_bytes": 20608, |
| 98 | "nbytes": 2, |
| 99 | "sha256": "8b940be7fb78aaa6b6567dd7a3987996947460df1c668e698eb92ca77e425349" |
| 100 | }, |
| 101 | { |
| 102 | "name": "blocks.1.net.0.bias", |
| 103 | "shape": [ |
| 104 | 32 |
| 105 | ], |
| 106 | "dtype": "float16", |
| 107 | "offset_bytes": 20610, |
| 108 | "nbytes": 64, |
| 109 | "sha256": "34353c8364a630edc5788611b538f4fc7e2cee4ec6de8311bb3fffc37eac6543" |
| 110 | }, |
| 111 | { |
| 112 | "name": "blocks.1.net.0.weight", |
| 113 | "shape": [ |
| 114 | 32, |
| 115 | 32, |
| 116 | 5 |
| 117 | ], |
| 118 | "dtype": "float16", |
| 119 | "offset_bytes": 20674, |
| 120 | "nbytes": 10240, |
| 121 | "sha256": "fc03474103115d0c770c3c625a3002a460eba2f2b27078a76dda018c2ac14774" |
| 122 | }, |
| 123 | { |
| 124 | "name": "blocks.1.net.2.bias", |
| 125 | "shape": [ |
| 126 | 32 |
| 127 | ], |
| 128 | "dtype": "float16", |
| 129 | "offset_bytes": 30914, |
| 130 | "nbytes": 64, |
| 131 | "sha256": "f1300005392a5d41ee7554800eab428bf606a01386f064334c400afc6eeb146e" |
| 132 | }, |
| 133 | { |
| 134 | "name": "blocks.1.net.2.weight", |
| 135 | "shape": [ |
| 136 | 32, |
| 137 | 32, |
| 138 | 5 |
| 139 | ], |
| 140 | "dtype": "float16", |
| 141 | "offset_bytes": 30978, |
| 142 | "nbytes": 10240, |
| 143 | "sha256": "16b76ab083dfb175776983b5a1bbda1852755e39a000999e4536eddc44382086" |
| 144 | }, |
| 145 | { |
| 146 | "name": "blocks.1.scale", |
| 147 | "shape": [ |
| 148 | 1 |
| 149 | ], |
| 150 | "dtype": "float16", |
| 151 | "offset_bytes": 41218, |
| 152 | "nbytes": 2, |
| 153 | "sha256": "349c41201b62db851192665c504b350ff98c6b45fb62a8a2161f78b6534d8de9" |
| 154 | }, |
| 155 | { |
| 156 | "name": "blocks.2.net.0.bias", |
| 157 | "shape": [ |
| 158 | 32 |
| 159 | ], |
| 160 | "dtype": "float16", |
| 161 | "offset_bytes": 41220, |
| 162 | "nbytes": 64, |
| 163 | "sha256": "f5d82f97677f4fdb6ac4909b260145438ff8932d67274dea3fbfe808871c484b" |
| 164 | }, |
| 165 | { |
| 166 | "name": "blocks.2.net.0.weight", |
| 167 | "shape": [ |
| 168 | 32, |
| 169 | 32, |
| 170 | 5 |
| 171 | ], |
| 172 | "dtype": "float16", |
| 173 | "offset_bytes": 41284, |
| 174 | "nbytes": 10240, |
| 175 | "sha256": "9c98ff5febf489da0f172a3c598a032428d28728ad7c9d98902d59d8ad7ef736" |
| 176 | }, |
| 177 | { |
| 178 | "name": "blocks.2.net.2.bias", |
| 179 | "shape": [ |
| 180 | 32 |
| 181 | ], |
| 182 | "dtype": "float16", |
| 183 | "offset_bytes": 51524, |
| 184 | "nbytes": 64, |
| 185 | "sha256": "9c200de8e50b9f8900fe62de86e12cd68a68b2be35cbe0edfb6eb17ada13a6d6" |
| 186 | }, |
| 187 | { |
| 188 | "name": "blocks.2.net.2.weight", |
| 189 | "shape": [ |
| 190 | 32, |
| 191 | 32, |
| 192 | 5 |
| 193 | ], |
| 194 | "dtype": "float16", |
| 195 | "offset_bytes": 51588, |
| 196 | "nbytes": 10240, |
| 197 | "sha256": "5f237920d732bc384a2a9602178a2c50240bca476e4654921098707510acd946" |
| 198 | }, |
| 199 | { |
| 200 | "name": "blocks.2.scale", |
| 201 | "shape": [ |
| 202 | 1 |
| 203 | ], |
| 204 | "dtype": "float16", |
| 205 | "offset_bytes": 61828, |
| 206 | "nbytes": 2, |
| 207 | "sha256": "f80dd9a9fee29fe1d95104fd516c7774aa2c2f3ad96260ad947f7c9793ff0a99" |
| 208 | }, |
| 209 | { |
| 210 | "name": "embedding.weight", |
| 211 | "shape": [ |
| 212 | 151, |
| 213 | 32 |
| 214 | ], |
| 215 | "dtype": "float16", |
| 216 | "offset_bytes": 61830, |
| 217 | "nbytes": 9664, |
| 218 | "sha256": "fde1d15454369653f02e570cebf94eef599e65509a2954401cd3514b83600f3c" |
| 219 | }, |
| 220 | { |
| 221 | "name": "input_proj.bias", |
| 222 | "shape": [ |
| 223 | 32 |
| 224 | ], |
| 225 | "dtype": "float16", |
| 226 | "offset_bytes": 71494, |
| 227 | "nbytes": 64, |
| 228 | "sha256": "ebb6d885baabd6386204be84d71bf83f780110c9fa9ac4f63714c37b18d83bd6" |
| 229 | }, |
| 230 | { |
| 231 | "name": "input_proj.weight", |
| 232 | "shape": [ |
| 233 | 32, |
| 234 | 35, |
| 235 | 1 |
| 236 | ], |
| 237 | "dtype": "float16", |
| 238 | "offset_bytes": 71558, |
| 239 | "nbytes": 2240, |
| 240 | "sha256": "97887a749c0077c1dcb3e6781a04fb7fb7a1e5a059c0a3848e347aad2831a2ba" |
| 241 | }, |
| 242 | { |
| 243 | "name": "output.bias", |
| 244 | "shape": [ |
| 245 | 1 |
| 246 | ], |
| 247 | "dtype": "float16", |
| 248 | "offset_bytes": 73798, |
| 249 | "nbytes": 2, |
| 250 | "sha256": "efdd30d1dda5a1a329ee9d0bf61387ee7e254ec69477e17ce1c4370d28e2b2db" |
| 251 | }, |
| 252 | { |
| 253 | "name": "output.weight", |
| 254 | "shape": [ |
| 255 | 1, |
| 256 | 32, |
| 257 | 1 |
| 258 | ], |
| 259 | "dtype": "float16", |
| 260 | "offset_bytes": 73800, |
| 261 | "nbytes": 64, |
| 262 | "sha256": "49caa887cec0f62a10c71bffd3ec6fc1e98c1cfc5ea15b64c5cc4ce63249b429" |
| 263 | } |
| 264 | ] |
| 265 | }, |
| 266 | "acoustic": { |
| 267 | "checkpoint": "artifacts/voices/ru_RU-tiny/joint/latent-student.pt", |
| 268 | "checkpoint_sha256": "a61b804a6b03009f2e65c34e1aaeeccce1fab25417f4ccbed3200e69bf8f16e7", |
| 269 | "config": { |
| 270 | "architecture": "token_context", |
| 271 | "vocab_size": 151, |
| 272 | "hidden": 48, |
| 273 | "depth": 3, |
| 274 | "kernel_size": 5, |
| 275 | "out_channels": 192, |
| 276 | "token_depth": 3 |
| 277 | }, |
| 278 | "parameters": 160422, |
| 279 | "tensors": [ |
| 280 | { |
| 281 | "name": "embedding.weight", |
| 282 | "shape": [ |
| 283 | 151, |
| 284 | 48 |
| 285 | ], |
| 286 | "dtype": "float16", |
| 287 | "offset_bytes": 73864, |
| 288 | "nbytes": 14496, |
| 289 | "sha256": "2c407abcfedbf55c8a76e9c7cc93f3fc56b56dce0b73976a14d0cc683dde13fa" |
| 290 | }, |
| 291 | { |
| 292 | "name": "frame_blocks.0.net.0.bias", |
| 293 | "shape": [ |
| 294 | 48 |
| 295 | ], |
| 296 | "dtype": "float16", |
| 297 | "offset_bytes": 88360, |
| 298 | "nbytes": 96, |
| 299 | "sha256": "f44bfa64f345c67648118cd80cdb50edd7c8efa992797dbbda384e4f6f8b3898" |
| 300 | }, |
| 301 | { |
| 302 | "name": "frame_blocks.0.net.0.weight", |
| 303 | "shape": [ |
| 304 | 48, |
| 305 | 48, |
| 306 | 5 |
| 307 | ], |
| 308 | "dtype": "float16", |
| 309 | "offset_bytes": 88456, |
| 310 | "nbytes": 23040, |
| 311 | "sha256": "82c31c6fb48dc3e25bf5b870dd7f0f211638c25df6b2be57767a7234a6400973" |
| 312 | }, |
| 313 | { |
| 314 | "name": "frame_blocks.0.net.2.bias", |
| 315 | "shape": [ |
| 316 | 48 |
| 317 | ], |
| 318 | "dtype": "float16", |
| 319 | "offset_bytes": 111496, |
| 320 | "nbytes": 96, |
| 321 | "sha256": "5c694125b7d83927131fbaa7fd37db6c6df00e0828f99154c5b5f0ecf0689d40" |
| 322 | }, |
| 323 | { |
| 324 | "name": "frame_blocks.0.net.2.weight", |
| 325 | "shape": [ |
| 326 | 48, |
| 327 | 48, |
| 328 | 5 |
| 329 | ], |
| 330 | "dtype": "float16", |
| 331 | "offset_bytes": 111592, |
| 332 | "nbytes": 23040, |
| 333 | "sha256": "b15aebdcc5a9e4287369eef746ecc2ede1a6082c7317535a40a009518ea12981" |
| 334 | }, |
| 335 | { |
| 336 | "name": "frame_blocks.0.scale", |
| 337 | "shape": [ |
| 338 | 1 |
| 339 | ], |
| 340 | "dtype": "float16", |
| 341 | "offset_bytes": 134632, |
| 342 | "nbytes": 2, |
| 343 | "sha256": "7251f996505ca03a2014baa9b01c485f137725f513f520f09610cb1248324aa5" |
| 344 | }, |
| 345 | { |
| 346 | "name": "frame_blocks.1.net.0.bias", |
| 347 | "shape": [ |
| 348 | 48 |
| 349 | ], |
| 350 | "dtype": "float16", |
| 351 | "offset_bytes": 134634, |
| 352 | "nbytes": 96, |
| 353 | "sha256": "21ef8a2ee6659b7a21bd4c3837b32c03f3b5dfecdda3919527fa1653810fac26" |
| 354 | }, |
| 355 | { |
| 356 | "name": "frame_blocks.1.net.0.weight", |
| 357 | "shape": [ |
| 358 | 48, |
| 359 | 48, |
| 360 | 5 |
| 361 | ], |
| 362 | "dtype": "float16", |
| 363 | "offset_bytes": 134730, |
| 364 | "nbytes": 23040, |
| 365 | "sha256": "df1ed59bf52dc9fbd012ba791400e65617196e75335315e99845b97bc886c2ca" |
| 366 | }, |
| 367 | { |
| 368 | "name": "frame_blocks.1.net.2.bias", |
| 369 | "shape": [ |
| 370 | 48 |
| 371 | ], |
| 372 | "dtype": "float16", |
| 373 | "offset_bytes": 157770, |
| 374 | "nbytes": 96, |
| 375 | "sha256": "c0824288effc40f3452bc2fcf6c22feb0cbebc0d7c33ed701a722c1b297f9f22" |
| 376 | }, |
| 377 | { |
| 378 | "name": "frame_blocks.1.net.2.weight", |
| 379 | "shape": [ |
| 380 | 48, |
| 381 | 48, |
| 382 | 5 |
| 383 | ], |
| 384 | "dtype": "float16", |
| 385 | "offset_bytes": 157866, |
| 386 | "nbytes": 23040, |
| 387 | "sha256": "87a849d741ad9303db7541f9d1f8c27c082fd1ecc3d416d1b2ed5a46638fba0f" |
| 388 | }, |
| 389 | { |
| 390 | "name": "frame_blocks.1.scale", |
| 391 | "shape": [ |
| 392 | 1 |
| 393 | ], |
| 394 | "dtype": "float16", |
| 395 | "offset_bytes": 180906, |
| 396 | "nbytes": 2, |
| 397 | "sha256": "12ff27f82eb1fec60296d1e0176f4e158fbae7748cc092e5e145404c6e8473c5" |
| 398 | }, |
| 399 | { |
| 400 | "name": "frame_blocks.2.net.0.bias", |
| 401 | "shape": [ |
| 402 | 48 |
| 403 | ], |
| 404 | "dtype": "float16", |
| 405 | "offset_bytes": 180908, |
| 406 | "nbytes": 96, |
| 407 | "sha256": "57fa96afcd035ab87fd9e2c3724e98dcf4fc0840a5411f14dca9ee339ac0b98f" |
| 408 | }, |
| 409 | { |
| 410 | "name": "frame_blocks.2.net.0.weight", |
| 411 | "shape": [ |
| 412 | 48, |
| 413 | 48, |
| 414 | 5 |
| 415 | ], |
| 416 | "dtype": "float16", |
| 417 | "offset_bytes": 181004, |
| 418 | "nbytes": 23040, |
| 419 | "sha256": "f63b422c8a9c09393c614f7044485fe99ef5c3a00de1ae65075bdc8b1467699f" |
| 420 | }, |
| 421 | { |
| 422 | "name": "frame_blocks.2.net.2.bias", |
| 423 | "shape": [ |
| 424 | 48 |
| 425 | ], |
| 426 | "dtype": "float16", |
| 427 | "offset_bytes": 204044, |
| 428 | "nbytes": 96, |
| 429 | "sha256": "34644233629a0ffb1dfe711230499cdf698db711b902995d779e9a4c718e6133" |
| 430 | }, |
| 431 | { |
| 432 | "name": "frame_blocks.2.net.2.weight", |
| 433 | "shape": [ |
| 434 | 48, |
| 435 | 48, |
| 436 | 5 |
| 437 | ], |
| 438 | "dtype": "float16", |
| 439 | "offset_bytes": 204140, |
| 440 | "nbytes": 23040, |
| 441 | "sha256": "3881f18fedb7a91d2d9f9c39d9bfc682bec6a18bbc2966e09eac7dd072c85de9" |
| 442 | }, |
| 443 | { |
| 444 | "name": "frame_blocks.2.scale", |
| 445 | "shape": [ |
| 446 | 1 |
| 447 | ], |
| 448 | "dtype": "float16", |
| 449 | "offset_bytes": 227180, |
| 450 | "nbytes": 2, |
| 451 | "sha256": "99ff31f7289a579a15ee200f4fa8e003f73e24d557e98e34e00f326e3f09cd70" |
| 452 | }, |
| 453 | { |
| 454 | "name": "frame_input_proj.bias", |
| 455 | "shape": [ |
| 456 | 48 |
| 457 | ], |
| 458 | "dtype": "float16", |
| 459 | "offset_bytes": 227182, |
| 460 | "nbytes": 96, |
| 461 | "sha256": "fd68e91d1dbe43d1f9b3016ba755f40da5f0769096f5f3abf6b1c63fdad667d3" |
| 462 | }, |
| 463 | { |
| 464 | "name": "frame_input_proj.weight", |
| 465 | "shape": [ |
| 466 | 48, |
| 467 | 51, |
| 468 | 1 |
| 469 | ], |
| 470 | "dtype": "float16", |
| 471 | "offset_bytes": 227278, |
| 472 | "nbytes": 4896, |
| 473 | "sha256": "98a7721adf8e1633be3547037aafc5ddbb8fa0d31ea0fc1aff69d739b13b242b" |
| 474 | }, |
| 475 | { |
| 476 | "name": "output.bias", |
| 477 | "shape": [ |
| 478 | 192 |
| 479 | ], |
| 480 | "dtype": "float16", |
| 481 | "offset_bytes": 232174, |
| 482 | "nbytes": 384, |
| 483 | "sha256": "15778bbc9b7aeb84a50fc350ac406c1192e9e172d58a828ff23232aa4da5fa0a" |
| 484 | }, |
| 485 | { |
| 486 | "name": "output.weight", |
| 487 | "shape": [ |
| 488 | 192, |
| 489 | 48, |
| 490 | 1 |
| 491 | ], |
| 492 | "dtype": "float16", |
| 493 | "offset_bytes": 232558, |
| 494 | "nbytes": 18432, |
| 495 | "sha256": "0073fd4bdeb6ddc84537c314637f588971765da6fea0936653aea453008b747b" |
| 496 | }, |
| 497 | { |
| 498 | "name": "token_blocks.0.net.0.bias", |
| 499 | "shape": [ |
| 500 | 48 |
| 501 | ], |
| 502 | "dtype": "float16", |
| 503 | "offset_bytes": 250990, |
| 504 | "nbytes": 96, |
| 505 | "sha256": "33b4a572bcb20b7d309e90653857e3ce725f31b8ee1ec9fb05b99713f3249aac" |
| 506 | }, |
| 507 | { |
| 508 | "name": "token_blocks.0.net.0.weight", |
| 509 | "shape": [ |
| 510 | 48, |
| 511 | 48, |
| 512 | 5 |
| 513 | ], |
| 514 | "dtype": "float16", |
| 515 | "offset_bytes": 251086, |
| 516 | "nbytes": 23040, |
| 517 | "sha256": "6c7e4033897cb8f750265d1d125f8c680b97cb4d989680ed6cbf141f573018dc" |
| 518 | }, |
| 519 | { |
| 520 | "name": "token_blocks.0.net.2.bias", |
| 521 | "shape": [ |
| 522 | 48 |
| 523 | ], |
| 524 | "dtype": "float16", |
| 525 | "offset_bytes": 274126, |
| 526 | "nbytes": 96, |
| 527 | "sha256": "cd3cf4d4f4b37b732a35eb816f0e550e707761b3ac14a39d10a9c813b9d2cc62" |
| 528 | }, |
| 529 | { |
| 530 | "name": "token_blocks.0.net.2.weight", |
| 531 | "shape": [ |
| 532 | 48, |
| 533 | 48, |
| 534 | 5 |
| 535 | ], |
| 536 | "dtype": "float16", |
| 537 | "offset_bytes": 274222, |
| 538 | "nbytes": 23040, |
| 539 | "sha256": "abec3a84704de428c87caa353b12a4128dc284ab9ce55585de98e765323d6d3e" |
| 540 | }, |
| 541 | { |
| 542 | "name": "token_blocks.0.scale", |
| 543 | "shape": [ |
| 544 | 1 |
| 545 | ], |
| 546 | "dtype": "float16", |
| 547 | "offset_bytes": 297262, |
| 548 | "nbytes": 2, |
| 549 | "sha256": "41d5ad5d80850bbff97bc4ce1510db47b597210a8bb7b52bf3789ffb209d06d2" |
| 550 | }, |
| 551 | { |
| 552 | "name": "token_blocks.1.net.0.bias", |
| 553 | "shape": [ |
| 554 | 48 |
| 555 | ], |
| 556 | "dtype": "float16", |
| 557 | "offset_bytes": 297264, |
| 558 | "nbytes": 96, |
| 559 | "sha256": "b3dc7234fe511922e905e6f5dc06956349fdbf075174d188113767ce2a5849ee" |
| 560 | }, |
| 561 | { |
| 562 | "name": "token_blocks.1.net.0.weight", |
| 563 | "shape": [ |
| 564 | 48, |
| 565 | 48, |
| 566 | 5 |
| 567 | ], |
| 568 | "dtype": "float16", |
| 569 | "offset_bytes": 297360, |
| 570 | "nbytes": 23040, |
| 571 | "sha256": "7527769033be332886ebb1b6d643cc0c88bd2d16797a4c9a25238a7f597cffd8" |
| 572 | }, |
| 573 | { |
| 574 | "name": "token_blocks.1.net.2.bias", |
| 575 | "shape": [ |
| 576 | 48 |
| 577 | ], |
| 578 | "dtype": "float16", |
| 579 | "offset_bytes": 320400, |
| 580 | "nbytes": 96, |
| 581 | "sha256": "5e543f34f0978ead47e1c1b2cc64e58c427e51689294500dcbb8ac4368018220" |
| 582 | }, |
| 583 | { |
| 584 | "name": "token_blocks.1.net.2.weight", |
| 585 | "shape": [ |
| 586 | 48, |
| 587 | 48, |
| 588 | 5 |
| 589 | ], |
| 590 | "dtype": "float16", |
| 591 | "offset_bytes": 320496, |
| 592 | "nbytes": 23040, |
| 593 | "sha256": "08eece9b482f0a80bfde76a458bd97419acff52c8c206adf39e017a14487fd63" |
| 594 | }, |
| 595 | { |
| 596 | "name": "token_blocks.1.scale", |
| 597 | "shape": [ |
| 598 | 1 |
| 599 | ], |
| 600 | "dtype": "float16", |
| 601 | "offset_bytes": 343536, |
| 602 | "nbytes": 2, |
| 603 | "sha256": "f21b0ef89d8ee5daf6ba6cd441652a647229941e85b1537b5c35ede6b4d519a8" |
| 604 | }, |
| 605 | { |
| 606 | "name": "token_blocks.2.net.0.bias", |
| 607 | "shape": [ |
| 608 | 48 |
| 609 | ], |
| 610 | "dtype": "float16", |
| 611 | "offset_bytes": 343538, |
| 612 | "nbytes": 96, |
| 613 | "sha256": "13f214a46554b21ff52f043a0ea36903d7e336f31c6e50c35df93976b00968af" |
| 614 | }, |
| 615 | { |
| 616 | "name": "token_blocks.2.net.0.weight", |
| 617 | "shape": [ |
| 618 | 48, |
| 619 | 48, |
| 620 | 5 |
| 621 | ], |
| 622 | "dtype": "float16", |
| 623 | "offset_bytes": 343634, |
| 624 | "nbytes": 23040, |
| 625 | "sha256": "438a0977bd36a4bad8d0ab06c315c0d38e3f85795bf1ee0646c4a5eb9fdf6518" |
| 626 | }, |
| 627 | { |
| 628 | "name": "token_blocks.2.net.2.bias", |
| 629 | "shape": [ |
| 630 | 48 |
| 631 | ], |
| 632 | "dtype": "float16", |
| 633 | "offset_bytes": 366674, |
| 634 | "nbytes": 96, |
| 635 | "sha256": "b682c7aeedc5dfb7702d366fb0e5a134ee81b77ba565b9c97f1a8a5f81280cd1" |
| 636 | }, |
| 637 | { |
| 638 | "name": "token_blocks.2.net.2.weight", |
| 639 | "shape": [ |
| 640 | 48, |
| 641 | 48, |
| 642 | 5 |
| 643 | ], |
| 644 | "dtype": "float16", |
| 645 | "offset_bytes": 366770, |
| 646 | "nbytes": 23040, |
| 647 | "sha256": "c48bc3ae5210058f460fe7c4a9717c9cd0a62e62adb0e08d16621cf47ebde13a" |
| 648 | }, |
| 649 | { |
| 650 | "name": "token_blocks.2.scale", |
| 651 | "shape": [ |
| 652 | 1 |
| 653 | ], |
| 654 | "dtype": "float16", |
| 655 | "offset_bytes": 389810, |
| 656 | "nbytes": 2, |
| 657 | "sha256": "ba8500497f8609171da5819bac45acf711441cf1942ee52bf8f4d49341e16d39" |
| 658 | }, |
| 659 | { |
| 660 | "name": "token_input_proj.bias", |
| 661 | "shape": [ |
| 662 | 48 |
| 663 | ], |
| 664 | "dtype": "float16", |
| 665 | "offset_bytes": 389812, |
| 666 | "nbytes": 96, |
| 667 | "sha256": "16a4e42a51f4e23107a9ff1f2eb70c2033ffe266f22bb43b9dc3dde9dba06553" |
| 668 | }, |
| 669 | { |
| 670 | "name": "token_input_proj.weight", |
| 671 | "shape": [ |
| 672 | 48, |
| 673 | 50, |
| 674 | 1 |
| 675 | ], |
| 676 | "dtype": "float16", |
| 677 | "offset_bytes": 389908, |
| 678 | "nbytes": 4800, |
| 679 | "sha256": "f56d7595a803b1ef108dc82bd2e3e17dfb30617f878d12eba6ca6e1b466e8e14" |
| 680 | } |
| 681 | ] |
| 682 | }, |
| 683 | "decoder": { |
| 684 | "checkpoint": "artifacts/voices/ru_RU-tiny/joint/decoder-student.pt", |
| 685 | "checkpoint_sha256": "e9a3d937c0abe69d23da7cf87884282cd29814fa92c785c705611ca05083b9bf", |
| 686 | "config": { |
| 687 | "in_channels": 192, |
| 688 | "source_in_channels": 192, |
| 689 | "channels": [ |
| 690 | 96, |
| 691 | 48, |
| 692 | 24, |
| 693 | 12 |
| 694 | ], |
| 695 | "res_layers": 1, |
| 696 | "variant": "piperlite", |
| 697 | "lrc_code_dim": 40, |
| 698 | "lrc_encoder_hidden": 64, |
| 699 | "rank_ratio": 0.5, |
| 700 | "activation": "leaky_relu", |
| 701 | "stage_affine": false, |
| 702 | "factorized_pre_rank": 0, |
| 703 | "piper_res_factor_rank_ratio": 0.0, |
| 704 | "res_bank_scale_mode": "kept", |
| 705 | "stage0_branches": [ |
| 706 | 0, |
| 707 | 1, |
| 708 | 2 |
| 709 | ], |
| 710 | "stage1_branches": [ |
| 711 | 0, |
| 712 | 1, |
| 713 | 2 |
| 714 | ], |
| 715 | "stage2_branches": [ |
| 716 | 0, |
| 717 | 1, |
| 718 | 2 |
| 719 | ], |
| 720 | "stage3_branches": [ |
| 721 | 0, |
| 722 | 1, |
| 723 | 2 |
| 724 | ], |
| 725 | "fsd_dim": 72, |
| 726 | "fsd_blocks": 5, |
| 727 | "fsd_film_rank": 12, |
| 728 | "fsd_head_rank": 48, |
| 729 | "wavehax_channels": 32, |
| 730 | "wavehax_blocks": 8, |
| 731 | "wavehax_mult_channels": 2, |
| 732 | "wavehax_kernel_freq": 7, |
| 733 | "wavehax_kernel_time": 7, |
| 734 | "wavehax_f0_channel": 80, |
| 735 | "wavehax_voiced_channel": 81, |
| 736 | "wavehax_sample_rate": 24000, |
| 737 | "wavehax_prior_power": 0.1, |
| 738 | "wavehax_prior_noise": 0.01, |
| 739 | "whx_channels": 32, |
| 740 | "whx_hidden": 64, |
| 741 | "whx_blocks": 8, |
| 742 | "whx_n_fft": 512, |
| 743 | "whx_f0_channel": 80, |
| 744 | "whx_voiced_channel": 81, |
| 745 | "whx_sample_rate": 24000, |
| 746 | "hift_width": 224, |
| 747 | "hift_gen_channels": 192, |
| 748 | "hift_style_dim": 64, |
| 749 | "hift_body_blocks": 3, |
| 750 | "hift_f0_channel": 80, |
| 751 | "hift_voiced_channel": 81, |
| 752 | "hift_sample_rate": 24000, |
| 753 | "stage_projection_bottlenecks": [], |
| 754 | "teacher_init_checkpoint": null, |
| 755 | "teacher_init_method": "first", |
| 756 | "teacher_init": null, |
| 757 | "init_decoder_checkpoint": "artifacts/voices/ru_RU-tiny/decoder/decoder-student.pt", |
| 758 | "decoder_init": { |
| 759 | "checkpoint": "artifacts/voices/ru_RU-tiny/decoder/decoder-student.pt", |
| 760 | "decoder_parameters": 314977, |
| 761 | "matched_config": { |
| 762 | "in_channels": 192, |
| 763 | "channels": [ |
| 764 | 96, |
| 765 | 48, |
| 766 | 24, |
| 767 | 12 |
| 768 | ], |
| 769 | "res_layers": 1, |
| 770 | "variant": "piperlite", |
| 771 | "rank_ratio": 0.5, |
| 772 | "activation": "leaky_relu", |
| 773 | "stage_affine": false, |
| 774 | "factorized_pre_rank": 0, |
| 775 | "piper_res_factor_rank_ratio": 0.0, |
| 776 | "res_bank_scale_mode": "kept", |
| 777 | "stage0_branches": [ |
| 778 | 0, |
| 779 | 1, |
| 780 | 2 |
| 781 | ], |
| 782 | "stage1_branches": [ |
| 783 | 0, |
| 784 | 1, |
| 785 | 2 |
| 786 | ], |
| 787 | "stage2_branches": [ |
| 788 | 0, |
| 789 | 1, |
| 790 | 2 |
| 791 | ], |
| 792 | "stage3_branches": [ |
| 793 | 0, |
| 794 | 1, |
| 795 | 2 |
| 796 | ], |
| 797 | "fsd_dim": 72, |
| 798 | "fsd_blocks": 5, |
| 799 | "fsd_film_rank": 12, |
| 800 | "fsd_head_rank": 48, |
| 801 | "wavehax_channels": 32, |
| 802 | "wavehax_blocks": 8, |
| 803 | "wavehax_mult_channels": 2, |
| 804 | "wavehax_kernel_freq": 7, |
| 805 | "wavehax_kernel_time": 7, |
| 806 | "wavehax_f0_channel": 80, |
| 807 | "wavehax_voiced_channel": 81, |
| 808 | "wavehax_sample_rate": 24000, |
| 809 | "wavehax_prior_power": 0.1, |
| 810 | "wavehax_prior_noise": 0.01, |
| 811 | "whx_channels": 32, |
| 812 | "whx_hidden": 64, |
| 813 | "whx_blocks": 8, |
| 814 | "whx_n_fft": 512, |
| 815 | "whx_f0_channel": 80, |
| 816 | "whx_voiced_channel": 81, |
| 817 | "whx_sample_rate": 24000, |
| 818 | "hift_width": 224, |
| 819 | "hift_gen_channels": 192, |
| 820 | "hift_style_dim": 64, |
| 821 | "hift_body_blocks": 3, |
| 822 | "hift_f0_channel": 80, |
| 823 | "hift_voiced_channel": 81, |
| 824 | "hift_sample_rate": 24000, |
| 825 | "stage_projection_bottlenecks": [], |
| 826 | "post_filter_channels": 0, |
| 827 | "post_filter_layers": 0, |
| 828 | "post_filter_kernel": 9, |
| 829 | "post_filter_scale": 0.25, |
| 830 | "pre_tanh_repair_channels": 0, |
| 831 | "pre_tanh_repair_layers": 0, |
| 832 | "pre_tanh_repair_kernel": 7, |
| 833 | "pre_tanh_repair_scale": 0.15, |
| 834 | "istft_n_fft": 512, |
| 835 | "istft_tail_head_rank": 20 |
| 836 | }, |
| 837 | "allowed_post_filter_mismatches": [], |
| 838 | "allowed_pre_tanh_repair_mismatches": [], |
| 839 | "allowed_activation_mismatches": [], |
| 840 | "missing_keys": [], |
| 841 | "unexpected_keys": [] |
| 842 | }, |
| 843 | "post_filter_channels": 0, |
| 844 | "post_filter_layers": 0, |
| 845 | "post_filter_kernel": 9, |
| 846 | "post_filter_scale": 0.25, |
| 847 | "pre_tanh_repair_channels": 0, |
| 848 | "pre_tanh_repair_layers": 0, |
| 849 | "pre_tanh_repair_kernel": 7, |
| 850 | "pre_tanh_repair_scale": 0.15, |
| 851 | "istft_n_fft": 512, |
| 852 | "istft_tail_head_rank": 20, |
| 853 | "ap_amplitude_weight": 0.0, |
| 854 | "ap_phase_weight": 0.0, |
| 855 | "ap_complex_weight": 0.0, |
| 856 | "spectral_head_init": "default", |
| 857 | "spectral_head_init_scale": 0.001, |
| 858 | "ap_amp_init_bias": -6.0, |
| 859 | "ap_phase_real_init_bias": 1.0, |
| 860 | "spectral_head_init_summary": null, |
| 861 | "feature_exact_weight": 0.0, |
| 862 | "feature_exact_keys": [ |
| 863 | "pre", |
| 864 | "up0" |
| 865 | ], |
| 866 | "feature_exact_teacher_init": null, |
| 867 | "signature_pack_dir": "artifacts/voices/ru_RU-tiny/signatures", |
| 868 | "exact_feature_pack_dir": null, |
| 869 | "signature_hint_weight": 0.05, |
| 870 | "signature_temporal_weight": 0.4, |
| 871 | "signature_phase_weight": 0.0, |
| 872 | "signature_phase_bins": 0, |
| 873 | "signature_keys": [ |
| 874 | "stage0_mix", |
| 875 | "stage1_mix", |
| 876 | "stage2_mix", |
| 877 | "pre_tanh", |
| 878 | "audio" |
| 879 | ], |
| 880 | "bottleneck_code_checkpoint": null, |
| 881 | "bottleneck_code_weight": 0.0, |
| 882 | "bottleneck_code_keys": [], |
| 883 | "bottleneck_code_training_only_parameters": 0, |
| 884 | "echo_tail_weight": 0.0, |
| 885 | "echo_tail_min_ms": 8.0, |
| 886 | "echo_tail_max_ms": 40.0, |
| 887 | "echo_tail_lags": 9, |
| 888 | "echo_tail_margin": 0.02, |
| 889 | "paired_acoustic_residual_weight": 0.0, |
| 890 | "paired_acoustic_residual_max_scale": 0.25, |
| 891 | "lrc_pred_code_mix_prob": 0.0, |
| 892 | "lrc_pred_code_checkpoint": null, |
| 893 | "lrc_pred_code_residual_prob": 0.0, |
| 894 | "lrc_pred_code_residual_max_scale": 0.25, |
| 895 | "lrc_pred_code_cache": null |
| 896 | }, |
| 897 | "parameters": 314977, |
| 898 | "tensors": [ |
| 899 | { |
| 900 | "name": "post.bias", |
| 901 | "shape": [ |
| 902 | 1 |
| 903 | ], |
| 904 | "dtype": "float16", |
| 905 | "offset_bytes": 394708, |
| 906 | "nbytes": 2, |
| 907 | "sha256": "c05a84ac9d47744ab6852482964647c8bc0100ddfffdbe93833a036eef4542f8" |
| 908 | }, |
| 909 | { |
| 910 | "name": "post.weight", |
| 911 | "shape": [ |
| 912 | 1, |
| 913 | 12, |
| 914 | 7 |
| 915 | ], |
| 916 | "dtype": "float16", |
| 917 | "offset_bytes": 394710, |
| 918 | "nbytes": 168, |
| 919 | "sha256": "132057e7dfd8a30c08fe47d72ddee84674a18c661e1dbed6005e9148601cb4f3" |
| 920 | }, |
| 921 | { |
| 922 | "name": "pre.bias", |
| 923 | "shape": [ |
| 924 | 96 |
| 925 | ], |
| 926 | "dtype": "float16", |
| 927 | "offset_bytes": 394878, |
| 928 | "nbytes": 192, |
| 929 | "sha256": "ee737c218c608f5e88a2e4afc9cc6586e26c9dd15aed71179eaeeaad31101236" |
| 930 | }, |
| 931 | { |
| 932 | "name": "pre.weight", |
| 933 | "shape": [ |
| 934 | 96, |
| 935 | 192, |
| 936 | 7 |
| 937 | ], |
| 938 | "dtype": "float16", |
| 939 | "offset_bytes": 395070, |
| 940 | "nbytes": 258048, |
| 941 | "sha256": "63093459c038943ab00657f60d502f3c01918e411c7c9138ff279ee777be32f5" |
| 942 | }, |
| 943 | { |
| 944 | "name": "res0.0.blocks.0.conv1.bias", |
| 945 | "shape": [ |
| 946 | 48 |
| 947 | ], |
| 948 | "dtype": "float16", |
| 949 | "offset_bytes": 653118, |
| 950 | "nbytes": 96, |
| 951 | "sha256": "bf1b585d000583ee32df5ee5daeb99e956a5e817267fc8470e4127dbc8d40f52" |
| 952 | }, |
| 953 | { |
| 954 | "name": "res0.0.blocks.0.conv1.weight", |
| 955 | "shape": [ |
| 956 | 48, |
| 957 | 48, |
| 958 | 3 |
| 959 | ], |
| 960 | "dtype": "float16", |
| 961 | "offset_bytes": 653214, |
| 962 | "nbytes": 13824, |
| 963 | "sha256": "bf43906637d512bed55da998a4e73bc793cf008fdab75ae1cd2e91bffee56ffc" |
| 964 | }, |
| 965 | { |
| 966 | "name": "res0.0.blocks.0.conv2.bias", |
| 967 | "shape": [ |
| 968 | 48 |
| 969 | ], |
| 970 | "dtype": "float16", |
| 971 | "offset_bytes": 667038, |
| 972 | "nbytes": 96, |
| 973 | "sha256": "2f2345dc928e134853bbb6d9737aafcaabc61ae5c89c7bd4e0ec24060e6e0b3e" |
| 974 | }, |
| 975 | { |
| 976 | "name": "res0.0.blocks.0.conv2.weight", |
| 977 | "shape": [ |
| 978 | 48, |
| 979 | 48, |
| 980 | 3 |
| 981 | ], |
| 982 | "dtype": "float16", |
| 983 | "offset_bytes": 667134, |
| 984 | "nbytes": 13824, |
| 985 | "sha256": "5b187096679a51b4fd712b265a53291db06bed7c79dc6b400451957dc95aa436" |
| 986 | }, |
| 987 | { |
| 988 | "name": "res0.0.blocks.1.conv1.bias", |
| 989 | "shape": [ |
| 990 | 48 |
| 991 | ], |
| 992 | "dtype": "float16", |
| 993 | "offset_bytes": 680958, |
| 994 | "nbytes": 96, |
| 995 | "sha256": "564687ae2efc0e299b212728e376570d767017727d94292c7305e87bd44cf0c4" |
| 996 | }, |
| 997 | { |
| 998 | "name": "res0.0.blocks.1.conv1.weight", |
| 999 | "shape": [ |
| 1000 | 48, |
| 1001 | 48, |
| 1002 | 5 |
| 1003 | ], |
| 1004 | "dtype": "float16", |
| 1005 | "offset_bytes": 681054, |
| 1006 | "nbytes": 23040, |
| 1007 | "sha256": "0c48973091b3ef414185474b50e8113777dacdd110115754aad90b37e84594d7" |
| 1008 | }, |
| 1009 | { |
| 1010 | "name": "res0.0.blocks.1.conv2.bias", |
| 1011 | "shape": [ |
| 1012 | 48 |
| 1013 | ], |
| 1014 | "dtype": "float16", |
| 1015 | "offset_bytes": 704094, |
| 1016 | "nbytes": 96, |
| 1017 | "sha256": "7fb055e26e0571a3fedca420d53ebca71c566751b946fbba499253702d434af9" |
| 1018 | }, |
| 1019 | { |
| 1020 | "name": "res0.0.blocks.1.conv2.weight", |
| 1021 | "shape": [ |
| 1022 | 48, |
| 1023 | 48, |
| 1024 | 5 |
| 1025 | ], |
| 1026 | "dtype": "float16", |
| 1027 | "offset_bytes": 704190, |
| 1028 | "nbytes": 23040, |
| 1029 | "sha256": "651872afeb484567e8078a24bf2d3cdbe9c16cc89b51552f2e229756e4119d3a" |
| 1030 | }, |
| 1031 | { |
| 1032 | "name": "res0.0.blocks.2.conv1.bias", |
| 1033 | "shape": [ |
| 1034 | 48 |
| 1035 | ], |
| 1036 | "dtype": "float16", |
| 1037 | "offset_bytes": 727230, |
| 1038 | "nbytes": 96, |
| 1039 | "sha256": "a9fdf964c802204279bee918d2aa9e8bd2de37cfaa886d07212d238f963e1d43" |
| 1040 | }, |
| 1041 | { |
| 1042 | "name": "res0.0.blocks.2.conv1.weight", |
| 1043 | "shape": [ |
| 1044 | 48, |
| 1045 | 48, |
| 1046 | 7 |
| 1047 | ], |
| 1048 | "dtype": "float16", |
| 1049 | "offset_bytes": 727326, |
| 1050 | "nbytes": 32256, |
| 1051 | "sha256": "093a3c36ddcf4e5e0fce83d2bb367734b52fe6f8de0586bdfeb46060f1f03d36" |
| 1052 | }, |
| 1053 | { |
| 1054 | "name": "res0.0.blocks.2.conv2.bias", |
| 1055 | "shape": [ |
| 1056 | 48 |
| 1057 | ], |
| 1058 | "dtype": "float16", |
| 1059 | "offset_bytes": 759582, |
| 1060 | "nbytes": 96, |
| 1061 | "sha256": "4dca799e65484e9de8737d26df631c42f8b533e2c1231c738eeb2da5de876d55" |
| 1062 | }, |
| 1063 | { |
| 1064 | "name": "res0.0.blocks.2.conv2.weight", |
| 1065 | "shape": [ |
| 1066 | 48, |
| 1067 | 48, |
| 1068 | 7 |
| 1069 | ], |
| 1070 | "dtype": "float16", |
| 1071 | "offset_bytes": 759678, |
| 1072 | "nbytes": 32256, |
| 1073 | "sha256": "514617a4082dd44e09ea84a01e40ec7c1dfc74bc7bbb2a5172312edd2ce5a0eb" |
| 1074 | }, |
| 1075 | { |
| 1076 | "name": "res1.0.blocks.0.conv1.bias", |
| 1077 | "shape": [ |
| 1078 | 24 |
| 1079 | ], |
| 1080 | "dtype": "float16", |
| 1081 | "offset_bytes": 791934, |
| 1082 | "nbytes": 48, |
| 1083 | "sha256": "902f717cc8213bada9f86913388cfee7e237d9baeb0c68d7ca4d7da7e843de11" |
| 1084 | }, |
| 1085 | { |
| 1086 | "name": "res1.0.blocks.0.conv1.weight", |
| 1087 | "shape": [ |
| 1088 | 24, |
| 1089 | 24, |
| 1090 | 3 |
| 1091 | ], |
| 1092 | "dtype": "float16", |
| 1093 | "offset_bytes": 791982, |
| 1094 | "nbytes": 3456, |
| 1095 | "sha256": "8a56706d06108a5892dd44a7690b5e78190bfddb3771354b97cccb399cef34ac" |
| 1096 | }, |
| 1097 | { |
| 1098 | "name": "res1.0.blocks.0.conv2.bias", |
| 1099 | "shape": [ |
| 1100 | 24 |
| 1101 | ], |
| 1102 | "dtype": "float16", |
| 1103 | "offset_bytes": 795438, |
| 1104 | "nbytes": 48, |
| 1105 | "sha256": "94aac5f892f682d0eb829ff0359e741b9acff101236bf25f7d03bdd2a38453f4" |
| 1106 | }, |
| 1107 | { |
| 1108 | "name": "res1.0.blocks.0.conv2.weight", |
| 1109 | "shape": [ |
| 1110 | 24, |
| 1111 | 24, |
| 1112 | 3 |
| 1113 | ], |
| 1114 | "dtype": "float16", |
| 1115 | "offset_bytes": 795486, |
| 1116 | "nbytes": 3456, |
| 1117 | "sha256": "8f41b41a9f6c8afe3a05ca5e357984467dfe7c83517db703ab27b684866756cf" |
| 1118 | }, |
| 1119 | { |
| 1120 | "name": "res1.0.blocks.1.conv1.bias", |
| 1121 | "shape": [ |
| 1122 | 24 |
| 1123 | ], |
| 1124 | "dtype": "float16", |
| 1125 | "offset_bytes": 798942, |
| 1126 | "nbytes": 48, |
| 1127 | "sha256": "2cdd452cdc1667fb1a00bdbb7ce0d26ac47926d65d8c3c6f1ac887fcb0868474" |
| 1128 | }, |
| 1129 | { |
| 1130 | "name": "res1.0.blocks.1.conv1.weight", |
| 1131 | "shape": [ |
| 1132 | 24, |
| 1133 | 24, |
| 1134 | 5 |
| 1135 | ], |
| 1136 | "dtype": "float16", |
| 1137 | "offset_bytes": 798990, |
| 1138 | "nbytes": 5760, |
| 1139 | "sha256": "e794258c54adff492a33f54eb4757a96109c1a2827de5ff568bf076eb5ca26df" |
| 1140 | }, |
| 1141 | { |
| 1142 | "name": "res1.0.blocks.1.conv2.bias", |
| 1143 | "shape": [ |
| 1144 | 24 |
| 1145 | ], |
| 1146 | "dtype": "float16", |
| 1147 | "offset_bytes": 804750, |
| 1148 | "nbytes": 48, |
| 1149 | "sha256": "3389357c56b135e6b771fd9d5ed10f8ba3548ed76c193b39cfbe32c483ec86f6" |
| 1150 | }, |
| 1151 | { |
| 1152 | "name": "res1.0.blocks.1.conv2.weight", |
| 1153 | "shape": [ |
| 1154 | 24, |
| 1155 | 24, |
| 1156 | 5 |
| 1157 | ], |
| 1158 | "dtype": "float16", |
| 1159 | "offset_bytes": 804798, |
| 1160 | "nbytes": 5760, |
| 1161 | "sha256": "e474579d773841382fe6a265133728fff2a37e2b6f6055e8e13d598a05dac338" |
| 1162 | }, |
| 1163 | { |
| 1164 | "name": "res1.0.blocks.2.conv1.bias", |
| 1165 | "shape": [ |
| 1166 | 24 |
| 1167 | ], |
| 1168 | "dtype": "float16", |
| 1169 | "offset_bytes": 810558, |
| 1170 | "nbytes": 48, |
| 1171 | "sha256": "04d5f8f4ec4113d5bd02310b87f6f4a68500723c9d278ec4c7dba3ff00de6f3c" |
| 1172 | }, |
| 1173 | { |
| 1174 | "name": "res1.0.blocks.2.conv1.weight", |
| 1175 | "shape": [ |
| 1176 | 24, |
| 1177 | 24, |
| 1178 | 7 |
| 1179 | ], |
| 1180 | "dtype": "float16", |
| 1181 | "offset_bytes": 810606, |
| 1182 | "nbytes": 8064, |
| 1183 | "sha256": "2d42e96d4443da484b144a79a6244e1550c96052fa33e18a40a3190dcd8b8af3" |
| 1184 | }, |
| 1185 | { |
| 1186 | "name": "res1.0.blocks.2.conv2.bias", |
| 1187 | "shape": [ |
| 1188 | 24 |
| 1189 | ], |
| 1190 | "dtype": "float16", |
| 1191 | "offset_bytes": 818670, |
| 1192 | "nbytes": 48, |
| 1193 | "sha256": "688df0276ef118dc36891861ffc72d555c3e8881d00d68198b5dc6ede99d1513" |
| 1194 | }, |
| 1195 | { |
| 1196 | "name": "res1.0.blocks.2.conv2.weight", |
| 1197 | "shape": [ |
| 1198 | 24, |
| 1199 | 24, |
| 1200 | 7 |
| 1201 | ], |
| 1202 | "dtype": "float16", |
| 1203 | "offset_bytes": 818718, |
| 1204 | "nbytes": 8064, |
| 1205 | "sha256": "a4a6d9aa157c1bb7251d26ebd4e67959d0d64b466f9ca00e41b1d66d6a300bc7" |
| 1206 | }, |
| 1207 | { |
| 1208 | "name": "res2.0.blocks.0.conv1.bias", |
| 1209 | "shape": [ |
| 1210 | 12 |
| 1211 | ], |
| 1212 | "dtype": "float16", |
| 1213 | "offset_bytes": 826782, |
| 1214 | "nbytes": 24, |
| 1215 | "sha256": "98a0e517368383b9bd777704aaf1fa0d47561f8ed33ca750601a7c58a7aefc28" |
| 1216 | }, |
| 1217 | { |
| 1218 | "name": "res2.0.blocks.0.conv1.weight", |
| 1219 | "shape": [ |
| 1220 | 12, |
| 1221 | 12, |
| 1222 | 3 |
| 1223 | ], |
| 1224 | "dtype": "float16", |
| 1225 | "offset_bytes": 826806, |
| 1226 | "nbytes": 864, |
| 1227 | "sha256": "817520dd0a10ade6538e113412416954a8c95a9152c43205013980dbb826981b" |
| 1228 | }, |
| 1229 | { |
| 1230 | "name": "res2.0.blocks.0.conv2.bias", |
| 1231 | "shape": [ |
| 1232 | 12 |
| 1233 | ], |
| 1234 | "dtype": "float16", |
| 1235 | "offset_bytes": 827670, |
| 1236 | "nbytes": 24, |
| 1237 | "sha256": "2927f5024c35ba64828a4531798454f4d98bb99d9849deee4d758a653f44a0c8" |
| 1238 | }, |
| 1239 | { |
| 1240 | "name": "res2.0.blocks.0.conv2.weight", |
| 1241 | "shape": [ |
| 1242 | 12, |
| 1243 | 12, |
| 1244 | 3 |
| 1245 | ], |
| 1246 | "dtype": "float16", |
| 1247 | "offset_bytes": 827694, |
| 1248 | "nbytes": 864, |
| 1249 | "sha256": "3dc0986c1fca4f469a333a8ff7a653f30f761431e8c877929e89e851e1ac35a2" |
| 1250 | }, |
| 1251 | { |
| 1252 | "name": "res2.0.blocks.1.conv1.bias", |
| 1253 | "shape": [ |
| 1254 | 12 |
| 1255 | ], |
| 1256 | "dtype": "float16", |
| 1257 | "offset_bytes": 828558, |
| 1258 | "nbytes": 24, |
| 1259 | "sha256": "b407138bbc34c67e28f08e64c19a2a1875f3786956c5e1444d4ae0af95e08cb4" |
| 1260 | }, |
| 1261 | { |
| 1262 | "name": "res2.0.blocks.1.conv1.weight", |
| 1263 | "shape": [ |
| 1264 | 12, |
| 1265 | 12, |
| 1266 | 5 |
| 1267 | ], |
| 1268 | "dtype": "float16", |
| 1269 | "offset_bytes": 828582, |
| 1270 | "nbytes": 1440, |
| 1271 | "sha256": "743c37b409286e6e3430f4978071640e3a140f82963a1dcbf4129fe267d6b86d" |
| 1272 | }, |
| 1273 | { |
| 1274 | "name": "res2.0.blocks.1.conv2.bias", |
| 1275 | "shape": [ |
| 1276 | 12 |
| 1277 | ], |
| 1278 | "dtype": "float16", |
| 1279 | "offset_bytes": 830022, |
| 1280 | "nbytes": 24, |
| 1281 | "sha256": "b5d89d342a310caf8c1214be4b024b483febd1ec5fa35b3a8114d0e8709d0072" |
| 1282 | }, |
| 1283 | { |
| 1284 | "name": "res2.0.blocks.1.conv2.weight", |
| 1285 | "shape": [ |
| 1286 | 12, |
| 1287 | 12, |
| 1288 | 5 |
| 1289 | ], |
| 1290 | "dtype": "float16", |
| 1291 | "offset_bytes": 830046, |
| 1292 | "nbytes": 1440, |
| 1293 | "sha256": "33fcc5631ebf0f5c773415713342979a6e0918e620250c70ee1e64b0ea27024d" |
| 1294 | }, |
| 1295 | { |
| 1296 | "name": "res2.0.blocks.2.conv1.bias", |
| 1297 | "shape": [ |
| 1298 | 12 |
| 1299 | ], |
| 1300 | "dtype": "float16", |
| 1301 | "offset_bytes": 831486, |
| 1302 | "nbytes": 24, |
| 1303 | "sha256": "698f764d28a0b1fdf5f2773d2ee0ed52e95c31357d7a8eaed3bd788d0cc97f7e" |
| 1304 | }, |
| 1305 | { |
| 1306 | "name": "res2.0.blocks.2.conv1.weight", |
| 1307 | "shape": [ |
| 1308 | 12, |
| 1309 | 12, |
| 1310 | 7 |
| 1311 | ], |
| 1312 | "dtype": "float16", |
| 1313 | "offset_bytes": 831510, |
| 1314 | "nbytes": 2016, |
| 1315 | "sha256": "25c73d5d6e6ea1ae98d46768e19e87a53d6b3a0b3c8a422afcf198610dc1af6b" |
| 1316 | }, |
| 1317 | { |
| 1318 | "name": "res2.0.blocks.2.conv2.bias", |
| 1319 | "shape": [ |
| 1320 | 12 |
| 1321 | ], |
| 1322 | "dtype": "float16", |
| 1323 | "offset_bytes": 833526, |
| 1324 | "nbytes": 24, |
| 1325 | "sha256": "a49018eda243d873cdcb9f4bfe7eeb6c87d238a9eac2836f752652effb1d6cdc" |
| 1326 | }, |
| 1327 | { |
| 1328 | "name": "res2.0.blocks.2.conv2.weight", |
| 1329 | "shape": [ |
| 1330 | 12, |
| 1331 | 12, |
| 1332 | 7 |
| 1333 | ], |
| 1334 | "dtype": "float16", |
| 1335 | "offset_bytes": 833550, |
| 1336 | "nbytes": 2016, |
| 1337 | "sha256": "47661037be63497bfd747b60e37a0863a79438aa12cf58f383a36c0e9e49371a" |
| 1338 | }, |
| 1339 | { |
| 1340 | "name": "up0.bias", |
| 1341 | "shape": [ |
| 1342 | 48 |
| 1343 | ], |
| 1344 | "dtype": "float16", |
| 1345 | "offset_bytes": 835566, |
| 1346 | "nbytes": 96, |
| 1347 | "sha256": "0f99b10e91567ea28b8a789016843b9b8d63d621d44080c74d9797f271f4b6cf" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "up0.weight", |
| 1351 | "shape": [ |
| 1352 | 96, |
| 1353 | 48, |
| 1354 | 16 |
| 1355 | ], |
| 1356 | "dtype": "float16", |
| 1357 | "offset_bytes": 835662, |
| 1358 | "nbytes": 147456, |
| 1359 | "sha256": "98f65ed45e15224b11ae12de2140861a2beb760a5ecd2c4b094d5196c9dd15aa" |
| 1360 | }, |
| 1361 | { |
| 1362 | "name": "up1.bias", |
| 1363 | "shape": [ |
| 1364 | 24 |
| 1365 | ], |
| 1366 | "dtype": "float16", |
| 1367 | "offset_bytes": 983118, |
| 1368 | "nbytes": 48, |
| 1369 | "sha256": "095ae9b7d92483ced9d94fc8c0344c08e5f047664d31af0edfc55f48fa326e5b" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "up1.weight", |
| 1373 | "shape": [ |
| 1374 | 48, |
| 1375 | 24, |
| 1376 | 16 |
| 1377 | ], |
| 1378 | "dtype": "float16", |
| 1379 | "offset_bytes": 983166, |
| 1380 | "nbytes": 36864, |
| 1381 | "sha256": "3ae138120923cd2cb453364b8f5101d2cba6c254479edab9329df6947088352e" |
| 1382 | }, |
| 1383 | { |
| 1384 | "name": "up2.bias", |
| 1385 | "shape": [ |
| 1386 | 12 |
| 1387 | ], |
| 1388 | "dtype": "float16", |
| 1389 | "offset_bytes": 1020030, |
| 1390 | "nbytes": 24, |
| 1391 | "sha256": "b5d4c5e186f26c5372ec55862fe3980cb8a409b24801591d9cd5352f44b6e0af" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "up2.weight", |
| 1395 | "shape": [ |
| 1396 | 24, |
| 1397 | 12, |
| 1398 | 8 |
| 1399 | ], |
| 1400 | "dtype": "float16", |
| 1401 | "offset_bytes": 1020054, |
| 1402 | "nbytes": 4608, |
| 1403 | "sha256": "23553a823b3a14867f48ce912dadbff8c8e14e19b89710b03f2bc59a58a22578" |
| 1404 | } |
| 1405 | ] |
| 1406 | } |
| 1407 | }, |
| 1408 | "runtime_ops": [ |
| 1409 | "embedding_lookup", |
| 1410 | "conv1d_same", |
| 1411 | "conv1d_1x1", |
| 1412 | "conv_transpose1d", |
| 1413 | "silu", |
| 1414 | "leaky_relu", |
| 1415 | "tanh", |
| 1416 | "residual_add", |
| 1417 | "repeat_interleave", |
| 1418 | "masked_conv1d", |
| 1419 | "depthwise_conv1d", |
| 1420 | "low_rank_linear_or_conv1d" |
| 1421 | ], |
| 1422 | "runtime_boundary": { |
| 1423 | "neural_stack_self_contained": true, |
| 1424 | "arbitrary_text_self_contained": false, |
| 1425 | "accepted_runtime_input": "Piper phoneme ID sequence matching piper-phoneme-config.json", |
| 1426 | "output": "mono float waveform at manifest.sample_rate" |
| 1427 | }, |
| 1428 | "frontend": { |
| 1429 | "status": "not fully self-contained for arbitrary text yet", |
| 1430 | "input_contract": "runtime may accept Piper phoneme ID sequences directly", |
| 1431 | "included_config": "piper-phoneme-config.json", |
| 1432 | "piper_phoneme_type": null, |
| 1433 | "piper_espeak_voice": "ru", |
| 1434 | "phoneme_id_map_entries": 151, |
| 1435 | "next_work": "embed or rewrite the Piper/eSpeak-compatible frontend, or keep a strict phoneme-ID API" |
| 1436 | }, |
| 1437 | "size_budget": { |
| 1438 | "target_max_bytes": 4000000, |
| 1439 | "target_max_mib": 3.814697265625, |
| 1440 | "actual_model_asset_bytes": 1029427, |
| 1441 | "target_max_parameters": 1500000, |
| 1442 | "under_parameter_budget": true, |
| 1443 | "under_4mb_asset_budget": true |
| 1444 | }, |
| 1445 | "limitations": [ |
| 1446 | "This is a raw neural asset package, not a compiled standalone runtime binary.", |
| 1447 | "No PyTorch checkpoint, ONNX graph, or training state is included.", |
| 1448 | "Arbitrary text still requires a Piper/eSpeak-compatible frontend outside this package." |
| 1449 | ] |
| 1450 | } |
| 1451 | |