ru-irina-1p57m/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": 3133528, |
| 15 | "weights_sha256": "dffff44808e2046e2062426e21818462ee991b8b68e37a4690acd3ab5df87b8d", |
| 16 | "total_parameters": 1566764, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/ru_RU/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "6b795781faa07eb3fe2d91df4cc4cf9800b65106404a4d4b5681378874683771", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 151, |
| 24 | "hidden": 64, |
| 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": 137348, |
| 46 | "tensors": [ |
| 47 | { |
| 48 | "name": "blocks.0.net.0.bias", |
| 49 | "shape": [ |
| 50 | 64 |
| 51 | ], |
| 52 | "dtype": "float16", |
| 53 | "offset_bytes": 0, |
| 54 | "nbytes": 128, |
| 55 | "sha256": "343fd4c3c9054df02991e041a86f298aeb6e575652d5f5baad1d65209bfeeb0e" |
| 56 | }, |
| 57 | { |
| 58 | "name": "blocks.0.net.0.weight", |
| 59 | "shape": [ |
| 60 | 64, |
| 61 | 64, |
| 62 | 5 |
| 63 | ], |
| 64 | "dtype": "float16", |
| 65 | "offset_bytes": 128, |
| 66 | "nbytes": 40960, |
| 67 | "sha256": "e33d2eda9b66c58d21da762f3c34c17536abd641da2a9dc8d68ac24e4d9a41d2" |
| 68 | }, |
| 69 | { |
| 70 | "name": "blocks.0.net.2.bias", |
| 71 | "shape": [ |
| 72 | 64 |
| 73 | ], |
| 74 | "dtype": "float16", |
| 75 | "offset_bytes": 41088, |
| 76 | "nbytes": 128, |
| 77 | "sha256": "c79be016c28e583f5e615fdf4318f130127f3ece3e5ff55a0254241d4e73970b" |
| 78 | }, |
| 79 | { |
| 80 | "name": "blocks.0.net.2.weight", |
| 81 | "shape": [ |
| 82 | 64, |
| 83 | 64, |
| 84 | 5 |
| 85 | ], |
| 86 | "dtype": "float16", |
| 87 | "offset_bytes": 41216, |
| 88 | "nbytes": 40960, |
| 89 | "sha256": "f0c299fdc95be4e1a96cdf901d18ee9760d214b7ca1f8aa009be00682d6d8337" |
| 90 | }, |
| 91 | { |
| 92 | "name": "blocks.0.scale", |
| 93 | "shape": [ |
| 94 | 1 |
| 95 | ], |
| 96 | "dtype": "float16", |
| 97 | "offset_bytes": 82176, |
| 98 | "nbytes": 2, |
| 99 | "sha256": "5394b81627e37da37839dbb47d224ca7ca8185987d6f1f7d6a7ea70d0545c81f" |
| 100 | }, |
| 101 | { |
| 102 | "name": "blocks.1.net.0.bias", |
| 103 | "shape": [ |
| 104 | 64 |
| 105 | ], |
| 106 | "dtype": "float16", |
| 107 | "offset_bytes": 82178, |
| 108 | "nbytes": 128, |
| 109 | "sha256": "7dc26808bbc6ffb5a88daa79f2ba680869adcfe80aecf6b97fc153eb2501fbb7" |
| 110 | }, |
| 111 | { |
| 112 | "name": "blocks.1.net.0.weight", |
| 113 | "shape": [ |
| 114 | 64, |
| 115 | 64, |
| 116 | 5 |
| 117 | ], |
| 118 | "dtype": "float16", |
| 119 | "offset_bytes": 82306, |
| 120 | "nbytes": 40960, |
| 121 | "sha256": "c8c7ddcbe1a723af60ae621e71218c700c2279b608d52602c4bc31a05976d0a7" |
| 122 | }, |
| 123 | { |
| 124 | "name": "blocks.1.net.2.bias", |
| 125 | "shape": [ |
| 126 | 64 |
| 127 | ], |
| 128 | "dtype": "float16", |
| 129 | "offset_bytes": 123266, |
| 130 | "nbytes": 128, |
| 131 | "sha256": "aa31165be6417a63c92f9aa9d1c88b0eb875f747dd0fff7b969b717d832f34fb" |
| 132 | }, |
| 133 | { |
| 134 | "name": "blocks.1.net.2.weight", |
| 135 | "shape": [ |
| 136 | 64, |
| 137 | 64, |
| 138 | 5 |
| 139 | ], |
| 140 | "dtype": "float16", |
| 141 | "offset_bytes": 123394, |
| 142 | "nbytes": 40960, |
| 143 | "sha256": "c68bf888b706256cc39a30aa625c692df71336fc5506cda9ec0aa9f2f96b435e" |
| 144 | }, |
| 145 | { |
| 146 | "name": "blocks.1.scale", |
| 147 | "shape": [ |
| 148 | 1 |
| 149 | ], |
| 150 | "dtype": "float16", |
| 151 | "offset_bytes": 164354, |
| 152 | "nbytes": 2, |
| 153 | "sha256": "771da04c14bc846203f04afa65c93e8ce2c9e6797ab55fa93c35cb141b1d0bd7" |
| 154 | }, |
| 155 | { |
| 156 | "name": "blocks.2.net.0.bias", |
| 157 | "shape": [ |
| 158 | 64 |
| 159 | ], |
| 160 | "dtype": "float16", |
| 161 | "offset_bytes": 164356, |
| 162 | "nbytes": 128, |
| 163 | "sha256": "e293ef9392426c8956ac3ff1a6e3675271ddd28c19296c6a3be722b22181d488" |
| 164 | }, |
| 165 | { |
| 166 | "name": "blocks.2.net.0.weight", |
| 167 | "shape": [ |
| 168 | 64, |
| 169 | 64, |
| 170 | 5 |
| 171 | ], |
| 172 | "dtype": "float16", |
| 173 | "offset_bytes": 164484, |
| 174 | "nbytes": 40960, |
| 175 | "sha256": "66878c361e480bd2af31f30ffc126ca8a1413e8ebec7542683b37c1ed324c98d" |
| 176 | }, |
| 177 | { |
| 178 | "name": "blocks.2.net.2.bias", |
| 179 | "shape": [ |
| 180 | 64 |
| 181 | ], |
| 182 | "dtype": "float16", |
| 183 | "offset_bytes": 205444, |
| 184 | "nbytes": 128, |
| 185 | "sha256": "cf1f43359cd5f7b11718ca6eac1b95b6d004d0c4401fb3dbc047b4ecaae9e2a7" |
| 186 | }, |
| 187 | { |
| 188 | "name": "blocks.2.net.2.weight", |
| 189 | "shape": [ |
| 190 | 64, |
| 191 | 64, |
| 192 | 5 |
| 193 | ], |
| 194 | "dtype": "float16", |
| 195 | "offset_bytes": 205572, |
| 196 | "nbytes": 40960, |
| 197 | "sha256": "cda766b5b830cfc8467c8f98e967d7c463626b6d62da56bd3dabbdc4c8201a2a" |
| 198 | }, |
| 199 | { |
| 200 | "name": "blocks.2.scale", |
| 201 | "shape": [ |
| 202 | 1 |
| 203 | ], |
| 204 | "dtype": "float16", |
| 205 | "offset_bytes": 246532, |
| 206 | "nbytes": 2, |
| 207 | "sha256": "4439e44862b85cc6d4c7f522808581e5b4489ec37c5b7b0ef2f7f0efa1333196" |
| 208 | }, |
| 209 | { |
| 210 | "name": "embedding.weight", |
| 211 | "shape": [ |
| 212 | 151, |
| 213 | 64 |
| 214 | ], |
| 215 | "dtype": "float16", |
| 216 | "offset_bytes": 246534, |
| 217 | "nbytes": 19328, |
| 218 | "sha256": "51ef66eca6c3c4c5438f26a1051e7ac8f0cac60f7d61b8190bfc8031c7964b72" |
| 219 | }, |
| 220 | { |
| 221 | "name": "input_proj.bias", |
| 222 | "shape": [ |
| 223 | 64 |
| 224 | ], |
| 225 | "dtype": "float16", |
| 226 | "offset_bytes": 265862, |
| 227 | "nbytes": 128, |
| 228 | "sha256": "7d55fb64b49da835b73c024bb16fc4616c09c40e69b9500faa9690246d0441c1" |
| 229 | }, |
| 230 | { |
| 231 | "name": "input_proj.weight", |
| 232 | "shape": [ |
| 233 | 64, |
| 234 | 67, |
| 235 | 1 |
| 236 | ], |
| 237 | "dtype": "float16", |
| 238 | "offset_bytes": 265990, |
| 239 | "nbytes": 8576, |
| 240 | "sha256": "ec88a31e86fb069d15b10d89207d27a9b0c03bcce74975b16d79433f1cc835e0" |
| 241 | }, |
| 242 | { |
| 243 | "name": "output.bias", |
| 244 | "shape": [ |
| 245 | 1 |
| 246 | ], |
| 247 | "dtype": "float16", |
| 248 | "offset_bytes": 274566, |
| 249 | "nbytes": 2, |
| 250 | "sha256": "3b0bb7c18b6d1339cff222c813ac01a48d4733d2fbf9b75c6a345cc86c0c5bb2" |
| 251 | }, |
| 252 | { |
| 253 | "name": "output.weight", |
| 254 | "shape": [ |
| 255 | 1, |
| 256 | 64, |
| 257 | 1 |
| 258 | ], |
| 259 | "dtype": "float16", |
| 260 | "offset_bytes": 274568, |
| 261 | "nbytes": 128, |
| 262 | "sha256": "1cb880f0c5fe93fece7eca6957fb9e8fece789575111e256b5534854cb7ac3a1" |
| 263 | } |
| 264 | ] |
| 265 | }, |
| 266 | "acoustic": { |
| 267 | "checkpoint": "artifacts/voices/ru_RU/joint/latent-student.pt", |
| 268 | "checkpoint_sha256": "278b669b4c4b111e3da1c1de7608204a12fe6968894789861a0f1cddfeb3124b", |
| 269 | "config": { |
| 270 | "architecture": "token_context", |
| 271 | "vocab_size": 151, |
| 272 | "hidden": 96, |
| 273 | "depth": 4, |
| 274 | "kernel_size": 5, |
| 275 | "out_channels": 192, |
| 276 | "token_depth": 3 |
| 277 | }, |
| 278 | "parameters": 698695, |
| 279 | "tensors": [ |
| 280 | { |
| 281 | "name": "embedding.weight", |
| 282 | "shape": [ |
| 283 | 151, |
| 284 | 96 |
| 285 | ], |
| 286 | "dtype": "float16", |
| 287 | "offset_bytes": 274696, |
| 288 | "nbytes": 28992, |
| 289 | "sha256": "cc562f6cd54e2694e09268b4dbed71c2f20cbf4fbb9125bbcaf36da0fd41dbde" |
| 290 | }, |
| 291 | { |
| 292 | "name": "frame_blocks.0.net.0.bias", |
| 293 | "shape": [ |
| 294 | 96 |
| 295 | ], |
| 296 | "dtype": "float16", |
| 297 | "offset_bytes": 303688, |
| 298 | "nbytes": 192, |
| 299 | "sha256": "f8a8ba2a2100b67a3c3ffd090a34d633b26e5712a0707480ab539331cd14595c" |
| 300 | }, |
| 301 | { |
| 302 | "name": "frame_blocks.0.net.0.weight", |
| 303 | "shape": [ |
| 304 | 96, |
| 305 | 96, |
| 306 | 5 |
| 307 | ], |
| 308 | "dtype": "float16", |
| 309 | "offset_bytes": 303880, |
| 310 | "nbytes": 92160, |
| 311 | "sha256": "3ab0a5f8c1668392cbd3797d82b3b90feadaceac0acb1b0da32d2c0bee92e137" |
| 312 | }, |
| 313 | { |
| 314 | "name": "frame_blocks.0.net.2.bias", |
| 315 | "shape": [ |
| 316 | 96 |
| 317 | ], |
| 318 | "dtype": "float16", |
| 319 | "offset_bytes": 396040, |
| 320 | "nbytes": 192, |
| 321 | "sha256": "75bc52f4b1c0e38419d4999b175e47f8f1ca6b7ddd5ad83c3d0d53450656ecac" |
| 322 | }, |
| 323 | { |
| 324 | "name": "frame_blocks.0.net.2.weight", |
| 325 | "shape": [ |
| 326 | 96, |
| 327 | 96, |
| 328 | 5 |
| 329 | ], |
| 330 | "dtype": "float16", |
| 331 | "offset_bytes": 396232, |
| 332 | "nbytes": 92160, |
| 333 | "sha256": "a07dd23dba304d9bf605d2b07077bb6ccb28d599223d09d1039f44c69b4d5582" |
| 334 | }, |
| 335 | { |
| 336 | "name": "frame_blocks.0.scale", |
| 337 | "shape": [ |
| 338 | 1 |
| 339 | ], |
| 340 | "dtype": "float16", |
| 341 | "offset_bytes": 488392, |
| 342 | "nbytes": 2, |
| 343 | "sha256": "cd9867ec09e26fe8111919a66f2007b7f2b8e8bcef1d09df3769236b0a1e7e82" |
| 344 | }, |
| 345 | { |
| 346 | "name": "frame_blocks.1.net.0.bias", |
| 347 | "shape": [ |
| 348 | 96 |
| 349 | ], |
| 350 | "dtype": "float16", |
| 351 | "offset_bytes": 488394, |
| 352 | "nbytes": 192, |
| 353 | "sha256": "561ec09554e41d61cc4386f6c69d494dadbc1ac7861846abec7a1da53e656a39" |
| 354 | }, |
| 355 | { |
| 356 | "name": "frame_blocks.1.net.0.weight", |
| 357 | "shape": [ |
| 358 | 96, |
| 359 | 96, |
| 360 | 5 |
| 361 | ], |
| 362 | "dtype": "float16", |
| 363 | "offset_bytes": 488586, |
| 364 | "nbytes": 92160, |
| 365 | "sha256": "397213573371a179872ac088deacde958e54843559f60a14682a5c8acfc9017e" |
| 366 | }, |
| 367 | { |
| 368 | "name": "frame_blocks.1.net.2.bias", |
| 369 | "shape": [ |
| 370 | 96 |
| 371 | ], |
| 372 | "dtype": "float16", |
| 373 | "offset_bytes": 580746, |
| 374 | "nbytes": 192, |
| 375 | "sha256": "6787addb65e2485df925eaf7779f08edebe613b3d94cf8a176f7c592077b3303" |
| 376 | }, |
| 377 | { |
| 378 | "name": "frame_blocks.1.net.2.weight", |
| 379 | "shape": [ |
| 380 | 96, |
| 381 | 96, |
| 382 | 5 |
| 383 | ], |
| 384 | "dtype": "float16", |
| 385 | "offset_bytes": 580938, |
| 386 | "nbytes": 92160, |
| 387 | "sha256": "7d54e020cca7049ca5a37911787f8f7398cd097f8c5f7336d255eea30606c7e5" |
| 388 | }, |
| 389 | { |
| 390 | "name": "frame_blocks.1.scale", |
| 391 | "shape": [ |
| 392 | 1 |
| 393 | ], |
| 394 | "dtype": "float16", |
| 395 | "offset_bytes": 673098, |
| 396 | "nbytes": 2, |
| 397 | "sha256": "84a0936ff1fba6d2811d094757231685673447aed2170738872305076026a316" |
| 398 | }, |
| 399 | { |
| 400 | "name": "frame_blocks.2.net.0.bias", |
| 401 | "shape": [ |
| 402 | 96 |
| 403 | ], |
| 404 | "dtype": "float16", |
| 405 | "offset_bytes": 673100, |
| 406 | "nbytes": 192, |
| 407 | "sha256": "84d2d789c0be401d8ed6f5ce1238843c1d090cd4c11cfc63b53bce9912bf6a9a" |
| 408 | }, |
| 409 | { |
| 410 | "name": "frame_blocks.2.net.0.weight", |
| 411 | "shape": [ |
| 412 | 96, |
| 413 | 96, |
| 414 | 5 |
| 415 | ], |
| 416 | "dtype": "float16", |
| 417 | "offset_bytes": 673292, |
| 418 | "nbytes": 92160, |
| 419 | "sha256": "46e881196954a033d1dcd4ae0287a5d010c8ea0f1fc9ffcf2f12d6f974a0a439" |
| 420 | }, |
| 421 | { |
| 422 | "name": "frame_blocks.2.net.2.bias", |
| 423 | "shape": [ |
| 424 | 96 |
| 425 | ], |
| 426 | "dtype": "float16", |
| 427 | "offset_bytes": 765452, |
| 428 | "nbytes": 192, |
| 429 | "sha256": "fb2d4d647168faca73322db33ea758ada8a2f762fae1eb6059ceb58dd06d7c02" |
| 430 | }, |
| 431 | { |
| 432 | "name": "frame_blocks.2.net.2.weight", |
| 433 | "shape": [ |
| 434 | 96, |
| 435 | 96, |
| 436 | 5 |
| 437 | ], |
| 438 | "dtype": "float16", |
| 439 | "offset_bytes": 765644, |
| 440 | "nbytes": 92160, |
| 441 | "sha256": "7f16b37c8125a4d71d96a1a113a27948c32fb1fac7906c6580852890d1dc80af" |
| 442 | }, |
| 443 | { |
| 444 | "name": "frame_blocks.2.scale", |
| 445 | "shape": [ |
| 446 | 1 |
| 447 | ], |
| 448 | "dtype": "float16", |
| 449 | "offset_bytes": 857804, |
| 450 | "nbytes": 2, |
| 451 | "sha256": "dc017890c0c6e947d9a787bd48e77c5b527336cdca4e00b790707bec35f9eda4" |
| 452 | }, |
| 453 | { |
| 454 | "name": "frame_blocks.3.net.0.bias", |
| 455 | "shape": [ |
| 456 | 96 |
| 457 | ], |
| 458 | "dtype": "float16", |
| 459 | "offset_bytes": 857806, |
| 460 | "nbytes": 192, |
| 461 | "sha256": "2d9c7556c890d09bcd5731e6f7772772fb66d3d3e8b1be4226ace06d1a8d5252" |
| 462 | }, |
| 463 | { |
| 464 | "name": "frame_blocks.3.net.0.weight", |
| 465 | "shape": [ |
| 466 | 96, |
| 467 | 96, |
| 468 | 5 |
| 469 | ], |
| 470 | "dtype": "float16", |
| 471 | "offset_bytes": 857998, |
| 472 | "nbytes": 92160, |
| 473 | "sha256": "b2a3844717e6835769678e02c9689f5a733a341bc33f088d820f005b0aac1caa" |
| 474 | }, |
| 475 | { |
| 476 | "name": "frame_blocks.3.net.2.bias", |
| 477 | "shape": [ |
| 478 | 96 |
| 479 | ], |
| 480 | "dtype": "float16", |
| 481 | "offset_bytes": 950158, |
| 482 | "nbytes": 192, |
| 483 | "sha256": "537db764c1fdc4dcc7c1b5e380f1fd655c031dbb7e32dbda315c6584b46fa01d" |
| 484 | }, |
| 485 | { |
| 486 | "name": "frame_blocks.3.net.2.weight", |
| 487 | "shape": [ |
| 488 | 96, |
| 489 | 96, |
| 490 | 5 |
| 491 | ], |
| 492 | "dtype": "float16", |
| 493 | "offset_bytes": 950350, |
| 494 | "nbytes": 92160, |
| 495 | "sha256": "e5f81f3afea7d78f0f7d84abf005254eeeab8c4e460f493129e962f5ca0098e5" |
| 496 | }, |
| 497 | { |
| 498 | "name": "frame_blocks.3.scale", |
| 499 | "shape": [ |
| 500 | 1 |
| 501 | ], |
| 502 | "dtype": "float16", |
| 503 | "offset_bytes": 1042510, |
| 504 | "nbytes": 2, |
| 505 | "sha256": "a8bd50ef6068567a410dff9231c7e4691bb55ddf698c6355e209142991ceb996" |
| 506 | }, |
| 507 | { |
| 508 | "name": "frame_input_proj.bias", |
| 509 | "shape": [ |
| 510 | 96 |
| 511 | ], |
| 512 | "dtype": "float16", |
| 513 | "offset_bytes": 1042512, |
| 514 | "nbytes": 192, |
| 515 | "sha256": "268e454c285f41ba10fe53c8d3a61cc2762cdddd9e5e3f77325dc228327b8ded" |
| 516 | }, |
| 517 | { |
| 518 | "name": "frame_input_proj.weight", |
| 519 | "shape": [ |
| 520 | 96, |
| 521 | 99, |
| 522 | 1 |
| 523 | ], |
| 524 | "dtype": "float16", |
| 525 | "offset_bytes": 1042704, |
| 526 | "nbytes": 19008, |
| 527 | "sha256": "5b8fe2650b1456f3e8c22e7d63a258e456eba498df58d1dd0f32096ab739f789" |
| 528 | }, |
| 529 | { |
| 530 | "name": "output.bias", |
| 531 | "shape": [ |
| 532 | 192 |
| 533 | ], |
| 534 | "dtype": "float16", |
| 535 | "offset_bytes": 1061712, |
| 536 | "nbytes": 384, |
| 537 | "sha256": "7495e82e2a4850fe188e68cbbd9c3a0a1d6e77aea273db93a43275ca036ddca2" |
| 538 | }, |
| 539 | { |
| 540 | "name": "output.weight", |
| 541 | "shape": [ |
| 542 | 192, |
| 543 | 96, |
| 544 | 1 |
| 545 | ], |
| 546 | "dtype": "float16", |
| 547 | "offset_bytes": 1062096, |
| 548 | "nbytes": 36864, |
| 549 | "sha256": "c2773d305aa31f85ab1d8b75b14a18f04b968bc7fc3f08a43cf478b9a8825f5c" |
| 550 | }, |
| 551 | { |
| 552 | "name": "token_blocks.0.net.0.bias", |
| 553 | "shape": [ |
| 554 | 96 |
| 555 | ], |
| 556 | "dtype": "float16", |
| 557 | "offset_bytes": 1098960, |
| 558 | "nbytes": 192, |
| 559 | "sha256": "4c658dba281110fced8c5c7dee0edb33f1236ddef6480a41eba2fd209e97ff01" |
| 560 | }, |
| 561 | { |
| 562 | "name": "token_blocks.0.net.0.weight", |
| 563 | "shape": [ |
| 564 | 96, |
| 565 | 96, |
| 566 | 5 |
| 567 | ], |
| 568 | "dtype": "float16", |
| 569 | "offset_bytes": 1099152, |
| 570 | "nbytes": 92160, |
| 571 | "sha256": "841a40ea6c2ea45f477ece771367da96b2804c1c4cf23dbccf305b62bf4c56b2" |
| 572 | }, |
| 573 | { |
| 574 | "name": "token_blocks.0.net.2.bias", |
| 575 | "shape": [ |
| 576 | 96 |
| 577 | ], |
| 578 | "dtype": "float16", |
| 579 | "offset_bytes": 1191312, |
| 580 | "nbytes": 192, |
| 581 | "sha256": "5a23297a06fae3359691dd7cc4470364cc2a051332b9ee2f33998ddf3c3d1481" |
| 582 | }, |
| 583 | { |
| 584 | "name": "token_blocks.0.net.2.weight", |
| 585 | "shape": [ |
| 586 | 96, |
| 587 | 96, |
| 588 | 5 |
| 589 | ], |
| 590 | "dtype": "float16", |
| 591 | "offset_bytes": 1191504, |
| 592 | "nbytes": 92160, |
| 593 | "sha256": "e3fba96f742777fa357d1c92a94673f5081b61e83e70a1d5e6ed0c3164050f52" |
| 594 | }, |
| 595 | { |
| 596 | "name": "token_blocks.0.scale", |
| 597 | "shape": [ |
| 598 | 1 |
| 599 | ], |
| 600 | "dtype": "float16", |
| 601 | "offset_bytes": 1283664, |
| 602 | "nbytes": 2, |
| 603 | "sha256": "b53dbd428c1b769823a27b8bfd6982c184229540409bed8954b6462bb0edf48e" |
| 604 | }, |
| 605 | { |
| 606 | "name": "token_blocks.1.net.0.bias", |
| 607 | "shape": [ |
| 608 | 96 |
| 609 | ], |
| 610 | "dtype": "float16", |
| 611 | "offset_bytes": 1283666, |
| 612 | "nbytes": 192, |
| 613 | "sha256": "73e4d052d052f1515985cc5ac8475425913b099b660dd9c070104126fb11f577" |
| 614 | }, |
| 615 | { |
| 616 | "name": "token_blocks.1.net.0.weight", |
| 617 | "shape": [ |
| 618 | 96, |
| 619 | 96, |
| 620 | 5 |
| 621 | ], |
| 622 | "dtype": "float16", |
| 623 | "offset_bytes": 1283858, |
| 624 | "nbytes": 92160, |
| 625 | "sha256": "b7aaac273555ccb58d6f3606654f6665bfbf5a75c59c01bfa236b4969fbd495d" |
| 626 | }, |
| 627 | { |
| 628 | "name": "token_blocks.1.net.2.bias", |
| 629 | "shape": [ |
| 630 | 96 |
| 631 | ], |
| 632 | "dtype": "float16", |
| 633 | "offset_bytes": 1376018, |
| 634 | "nbytes": 192, |
| 635 | "sha256": "b3e8bd744f85af83edebc1a339f8bf360f0ea48211519fe4c535ed630672207d" |
| 636 | }, |
| 637 | { |
| 638 | "name": "token_blocks.1.net.2.weight", |
| 639 | "shape": [ |
| 640 | 96, |
| 641 | 96, |
| 642 | 5 |
| 643 | ], |
| 644 | "dtype": "float16", |
| 645 | "offset_bytes": 1376210, |
| 646 | "nbytes": 92160, |
| 647 | "sha256": "6bb961aec8e661a508ba3fea3ed78f5a095f2c2825a33aef6a162ae1bae0b7bd" |
| 648 | }, |
| 649 | { |
| 650 | "name": "token_blocks.1.scale", |
| 651 | "shape": [ |
| 652 | 1 |
| 653 | ], |
| 654 | "dtype": "float16", |
| 655 | "offset_bytes": 1468370, |
| 656 | "nbytes": 2, |
| 657 | "sha256": "830df1b84d111d66c61cdb9bbf760adcc4ee011b1ae5b2245b0e94f064540afe" |
| 658 | }, |
| 659 | { |
| 660 | "name": "token_blocks.2.net.0.bias", |
| 661 | "shape": [ |
| 662 | 96 |
| 663 | ], |
| 664 | "dtype": "float16", |
| 665 | "offset_bytes": 1468372, |
| 666 | "nbytes": 192, |
| 667 | "sha256": "f352b43143f516c4651911aae01eac5e53af2aca1fe595263b7bc36a98cf13a4" |
| 668 | }, |
| 669 | { |
| 670 | "name": "token_blocks.2.net.0.weight", |
| 671 | "shape": [ |
| 672 | 96, |
| 673 | 96, |
| 674 | 5 |
| 675 | ], |
| 676 | "dtype": "float16", |
| 677 | "offset_bytes": 1468564, |
| 678 | "nbytes": 92160, |
| 679 | "sha256": "19fb0ccf5337998607e5a2a6316f6bfdd9fad99cba9a364a810e8deefea52bf5" |
| 680 | }, |
| 681 | { |
| 682 | "name": "token_blocks.2.net.2.bias", |
| 683 | "shape": [ |
| 684 | 96 |
| 685 | ], |
| 686 | "dtype": "float16", |
| 687 | "offset_bytes": 1560724, |
| 688 | "nbytes": 192, |
| 689 | "sha256": "4351cca54ea09b725ad6165be76043c7a03d9027ae187a48a3753991c29157f1" |
| 690 | }, |
| 691 | { |
| 692 | "name": "token_blocks.2.net.2.weight", |
| 693 | "shape": [ |
| 694 | 96, |
| 695 | 96, |
| 696 | 5 |
| 697 | ], |
| 698 | "dtype": "float16", |
| 699 | "offset_bytes": 1560916, |
| 700 | "nbytes": 92160, |
| 701 | "sha256": "aef82c0976f8f9339560fd708921f12fb3ec5ee20c56efec97111668753e6bcc" |
| 702 | }, |
| 703 | { |
| 704 | "name": "token_blocks.2.scale", |
| 705 | "shape": [ |
| 706 | 1 |
| 707 | ], |
| 708 | "dtype": "float16", |
| 709 | "offset_bytes": 1653076, |
| 710 | "nbytes": 2, |
| 711 | "sha256": "73d22a21ed5e854dd507c716817fe94e9f6fa009d254593344346e83385bec01" |
| 712 | }, |
| 713 | { |
| 714 | "name": "token_input_proj.bias", |
| 715 | "shape": [ |
| 716 | 96 |
| 717 | ], |
| 718 | "dtype": "float16", |
| 719 | "offset_bytes": 1653078, |
| 720 | "nbytes": 192, |
| 721 | "sha256": "4bf835d27a503081d1ae1bce5951654f35275bcde190b885323ec66da11b72ba" |
| 722 | }, |
| 723 | { |
| 724 | "name": "token_input_proj.weight", |
| 725 | "shape": [ |
| 726 | 96, |
| 727 | 98, |
| 728 | 1 |
| 729 | ], |
| 730 | "dtype": "float16", |
| 731 | "offset_bytes": 1653270, |
| 732 | "nbytes": 18816, |
| 733 | "sha256": "f3e9499e12bffe30db850dd934c76ed83dc5bd2c30b2f64235738a81a41dee08" |
| 734 | } |
| 735 | ] |
| 736 | }, |
| 737 | "decoder": { |
| 738 | "checkpoint": "artifacts/voices/ru_RU/joint/decoder-student.pt", |
| 739 | "checkpoint_sha256": "bc13c4000f151cfb5190e8ca4de1d4e460c5fcc4c1f9226c86e219d982b62ec0", |
| 740 | "config": { |
| 741 | "in_channels": 192, |
| 742 | "source_in_channels": 192, |
| 743 | "channels": [ |
| 744 | 160, |
| 745 | 80, |
| 746 | 40, |
| 747 | 20 |
| 748 | ], |
| 749 | "res_layers": 1, |
| 750 | "variant": "piperlite", |
| 751 | "lrc_code_dim": 40, |
| 752 | "lrc_encoder_hidden": 64, |
| 753 | "rank_ratio": 0.5, |
| 754 | "activation": "leaky_relu", |
| 755 | "stage_affine": false, |
| 756 | "factorized_pre_rank": 0, |
| 757 | "piper_res_factor_rank_ratio": 0.0, |
| 758 | "res_bank_scale_mode": "kept", |
| 759 | "stage0_branches": [ |
| 760 | 0, |
| 761 | 1, |
| 762 | 2 |
| 763 | ], |
| 764 | "stage1_branches": [ |
| 765 | 0, |
| 766 | 1, |
| 767 | 2 |
| 768 | ], |
| 769 | "stage2_branches": [ |
| 770 | 0, |
| 771 | 1, |
| 772 | 2 |
| 773 | ], |
| 774 | "stage3_branches": [ |
| 775 | 0, |
| 776 | 1, |
| 777 | 2 |
| 778 | ], |
| 779 | "fsd_dim": 72, |
| 780 | "fsd_blocks": 5, |
| 781 | "fsd_film_rank": 12, |
| 782 | "fsd_head_rank": 48, |
| 783 | "wavehax_channels": 32, |
| 784 | "wavehax_blocks": 8, |
| 785 | "wavehax_mult_channels": 2, |
| 786 | "wavehax_kernel_freq": 7, |
| 787 | "wavehax_kernel_time": 7, |
| 788 | "wavehax_f0_channel": 80, |
| 789 | "wavehax_voiced_channel": 81, |
| 790 | "wavehax_sample_rate": 24000, |
| 791 | "wavehax_prior_power": 0.1, |
| 792 | "wavehax_prior_noise": 0.01, |
| 793 | "whx_channels": 32, |
| 794 | "whx_hidden": 64, |
| 795 | "whx_blocks": 8, |
| 796 | "whx_n_fft": 512, |
| 797 | "whx_f0_channel": 80, |
| 798 | "whx_voiced_channel": 81, |
| 799 | "whx_sample_rate": 24000, |
| 800 | "hift_width": 224, |
| 801 | "hift_gen_channels": 192, |
| 802 | "hift_style_dim": 64, |
| 803 | "hift_body_blocks": 3, |
| 804 | "hift_f0_channel": 80, |
| 805 | "hift_voiced_channel": 81, |
| 806 | "hift_sample_rate": 24000, |
| 807 | "stage_projection_bottlenecks": [], |
| 808 | "teacher_init_checkpoint": null, |
| 809 | "teacher_init_method": "first", |
| 810 | "teacher_init": null, |
| 811 | "init_decoder_checkpoint": "artifacts/voices/ru_RU/decoder/decoder-student.pt", |
| 812 | "decoder_init": { |
| 813 | "checkpoint": "artifacts/voices/ru_RU/decoder/decoder-student.pt", |
| 814 | "decoder_parameters": 730721, |
| 815 | "matched_config": { |
| 816 | "in_channels": 192, |
| 817 | "channels": [ |
| 818 | 160, |
| 819 | 80, |
| 820 | 40, |
| 821 | 20 |
| 822 | ], |
| 823 | "res_layers": 1, |
| 824 | "variant": "piperlite", |
| 825 | "rank_ratio": 0.5, |
| 826 | "activation": "leaky_relu", |
| 827 | "stage_affine": false, |
| 828 | "factorized_pre_rank": 0, |
| 829 | "piper_res_factor_rank_ratio": 0.0, |
| 830 | "res_bank_scale_mode": "kept", |
| 831 | "stage0_branches": [ |
| 832 | 0, |
| 833 | 1, |
| 834 | 2 |
| 835 | ], |
| 836 | "stage1_branches": [ |
| 837 | 0, |
| 838 | 1, |
| 839 | 2 |
| 840 | ], |
| 841 | "stage2_branches": [ |
| 842 | 0, |
| 843 | 1, |
| 844 | 2 |
| 845 | ], |
| 846 | "stage3_branches": [ |
| 847 | 0, |
| 848 | 1, |
| 849 | 2 |
| 850 | ], |
| 851 | "fsd_dim": 72, |
| 852 | "fsd_blocks": 5, |
| 853 | "fsd_film_rank": 12, |
| 854 | "fsd_head_rank": 48, |
| 855 | "wavehax_channels": 32, |
| 856 | "wavehax_blocks": 8, |
| 857 | "wavehax_mult_channels": 2, |
| 858 | "wavehax_kernel_freq": 7, |
| 859 | "wavehax_kernel_time": 7, |
| 860 | "wavehax_f0_channel": 80, |
| 861 | "wavehax_voiced_channel": 81, |
| 862 | "wavehax_sample_rate": 24000, |
| 863 | "wavehax_prior_power": 0.1, |
| 864 | "wavehax_prior_noise": 0.01, |
| 865 | "whx_channels": 32, |
| 866 | "whx_hidden": 64, |
| 867 | "whx_blocks": 8, |
| 868 | "whx_n_fft": 512, |
| 869 | "whx_f0_channel": 80, |
| 870 | "whx_voiced_channel": 81, |
| 871 | "whx_sample_rate": 24000, |
| 872 | "hift_width": 224, |
| 873 | "hift_gen_channels": 192, |
| 874 | "hift_style_dim": 64, |
| 875 | "hift_body_blocks": 3, |
| 876 | "hift_f0_channel": 80, |
| 877 | "hift_voiced_channel": 81, |
| 878 | "hift_sample_rate": 24000, |
| 879 | "stage_projection_bottlenecks": [], |
| 880 | "post_filter_channels": 0, |
| 881 | "post_filter_layers": 0, |
| 882 | "post_filter_kernel": 9, |
| 883 | "post_filter_scale": 0.25, |
| 884 | "pre_tanh_repair_channels": 0, |
| 885 | "pre_tanh_repair_layers": 0, |
| 886 | "pre_tanh_repair_kernel": 7, |
| 887 | "pre_tanh_repair_scale": 0.15, |
| 888 | "istft_n_fft": 512, |
| 889 | "istft_tail_head_rank": 20 |
| 890 | }, |
| 891 | "allowed_post_filter_mismatches": [], |
| 892 | "allowed_pre_tanh_repair_mismatches": [], |
| 893 | "allowed_activation_mismatches": [], |
| 894 | "missing_keys": [], |
| 895 | "unexpected_keys": [] |
| 896 | }, |
| 897 | "post_filter_channels": 0, |
| 898 | "post_filter_layers": 0, |
| 899 | "post_filter_kernel": 9, |
| 900 | "post_filter_scale": 0.25, |
| 901 | "pre_tanh_repair_channels": 0, |
| 902 | "pre_tanh_repair_layers": 0, |
| 903 | "pre_tanh_repair_kernel": 7, |
| 904 | "pre_tanh_repair_scale": 0.15, |
| 905 | "istft_n_fft": 512, |
| 906 | "istft_tail_head_rank": 20, |
| 907 | "ap_amplitude_weight": 0.0, |
| 908 | "ap_phase_weight": 0.0, |
| 909 | "ap_complex_weight": 0.0, |
| 910 | "spectral_head_init": "default", |
| 911 | "spectral_head_init_scale": 0.001, |
| 912 | "ap_amp_init_bias": -6.0, |
| 913 | "ap_phase_real_init_bias": 1.0, |
| 914 | "spectral_head_init_summary": null, |
| 915 | "feature_exact_weight": 0.0, |
| 916 | "feature_exact_keys": [ |
| 917 | "pre", |
| 918 | "up0" |
| 919 | ], |
| 920 | "feature_exact_teacher_init": null, |
| 921 | "signature_pack_dir": "artifacts/voices/ru_RU/signatures", |
| 922 | "exact_feature_pack_dir": null, |
| 923 | "signature_hint_weight": 0.05, |
| 924 | "signature_temporal_weight": 0.4, |
| 925 | "signature_phase_weight": 0.0, |
| 926 | "signature_phase_bins": 0, |
| 927 | "signature_keys": [ |
| 928 | "stage0_mix", |
| 929 | "stage1_mix", |
| 930 | "stage2_mix", |
| 931 | "pre_tanh", |
| 932 | "audio" |
| 933 | ], |
| 934 | "bottleneck_code_checkpoint": null, |
| 935 | "bottleneck_code_weight": 0.0, |
| 936 | "bottleneck_code_keys": [], |
| 937 | "bottleneck_code_training_only_parameters": 0, |
| 938 | "echo_tail_weight": 0.0, |
| 939 | "echo_tail_min_ms": 8.0, |
| 940 | "echo_tail_max_ms": 40.0, |
| 941 | "echo_tail_lags": 9, |
| 942 | "echo_tail_margin": 0.02, |
| 943 | "paired_acoustic_residual_weight": 0.0, |
| 944 | "paired_acoustic_residual_max_scale": 0.25, |
| 945 | "lrc_pred_code_mix_prob": 0.0, |
| 946 | "lrc_pred_code_checkpoint": null, |
| 947 | "lrc_pred_code_residual_prob": 0.0, |
| 948 | "lrc_pred_code_residual_max_scale": 0.25, |
| 949 | "lrc_pred_code_cache": null |
| 950 | }, |
| 951 | "parameters": 730721, |
| 952 | "tensors": [ |
| 953 | { |
| 954 | "name": "post.bias", |
| 955 | "shape": [ |
| 956 | 1 |
| 957 | ], |
| 958 | "dtype": "float16", |
| 959 | "offset_bytes": 1672086, |
| 960 | "nbytes": 2, |
| 961 | "sha256": "3996cbbc2923614229cf038d298debe70d48bfb0bd3ce4bf8de11a5c385080ab" |
| 962 | }, |
| 963 | { |
| 964 | "name": "post.weight", |
| 965 | "shape": [ |
| 966 | 1, |
| 967 | 20, |
| 968 | 7 |
| 969 | ], |
| 970 | "dtype": "float16", |
| 971 | "offset_bytes": 1672088, |
| 972 | "nbytes": 280, |
| 973 | "sha256": "d96be0e4a701d62f77e6f76d35ea9636377ee09ce375114aeb7bce8cb214eebc" |
| 974 | }, |
| 975 | { |
| 976 | "name": "pre.bias", |
| 977 | "shape": [ |
| 978 | 160 |
| 979 | ], |
| 980 | "dtype": "float16", |
| 981 | "offset_bytes": 1672368, |
| 982 | "nbytes": 320, |
| 983 | "sha256": "030e790c5621018e30da37048469719367f7b65b5f638d62ca0f52b3806dc0e0" |
| 984 | }, |
| 985 | { |
| 986 | "name": "pre.weight", |
| 987 | "shape": [ |
| 988 | 160, |
| 989 | 192, |
| 990 | 7 |
| 991 | ], |
| 992 | "dtype": "float16", |
| 993 | "offset_bytes": 1672688, |
| 994 | "nbytes": 430080, |
| 995 | "sha256": "01e7ffe82c67e6e292fe4574e8338b835869aba703348891730b30b8510fc0d8" |
| 996 | }, |
| 997 | { |
| 998 | "name": "res0.0.blocks.0.conv1.bias", |
| 999 | "shape": [ |
| 1000 | 80 |
| 1001 | ], |
| 1002 | "dtype": "float16", |
| 1003 | "offset_bytes": 2102768, |
| 1004 | "nbytes": 160, |
| 1005 | "sha256": "4413feb2528f81856013d1db3ea0984ca26d8b93a03d101a5bf2f5f14ce04c92" |
| 1006 | }, |
| 1007 | { |
| 1008 | "name": "res0.0.blocks.0.conv1.weight", |
| 1009 | "shape": [ |
| 1010 | 80, |
| 1011 | 80, |
| 1012 | 3 |
| 1013 | ], |
| 1014 | "dtype": "float16", |
| 1015 | "offset_bytes": 2102928, |
| 1016 | "nbytes": 38400, |
| 1017 | "sha256": "f1b054a08800fabfe62ea1b61f5cb450d7817827870ad094bbc1065298147474" |
| 1018 | }, |
| 1019 | { |
| 1020 | "name": "res0.0.blocks.0.conv2.bias", |
| 1021 | "shape": [ |
| 1022 | 80 |
| 1023 | ], |
| 1024 | "dtype": "float16", |
| 1025 | "offset_bytes": 2141328, |
| 1026 | "nbytes": 160, |
| 1027 | "sha256": "f58eb0a0fe015e7d99c9c6ee1422aaff9b44cdcad6aed033fa28c90167b2e2dd" |
| 1028 | }, |
| 1029 | { |
| 1030 | "name": "res0.0.blocks.0.conv2.weight", |
| 1031 | "shape": [ |
| 1032 | 80, |
| 1033 | 80, |
| 1034 | 3 |
| 1035 | ], |
| 1036 | "dtype": "float16", |
| 1037 | "offset_bytes": 2141488, |
| 1038 | "nbytes": 38400, |
| 1039 | "sha256": "22831a35a66ca617d376fe05efb56948d8d5083d76887e2af994c59cc276c2be" |
| 1040 | }, |
| 1041 | { |
| 1042 | "name": "res0.0.blocks.1.conv1.bias", |
| 1043 | "shape": [ |
| 1044 | 80 |
| 1045 | ], |
| 1046 | "dtype": "float16", |
| 1047 | "offset_bytes": 2179888, |
| 1048 | "nbytes": 160, |
| 1049 | "sha256": "e4c29abe7309f2e391b280ea5befeb7bbab14a5fbdbf3e8c1c2c3792fd65c07b" |
| 1050 | }, |
| 1051 | { |
| 1052 | "name": "res0.0.blocks.1.conv1.weight", |
| 1053 | "shape": [ |
| 1054 | 80, |
| 1055 | 80, |
| 1056 | 5 |
| 1057 | ], |
| 1058 | "dtype": "float16", |
| 1059 | "offset_bytes": 2180048, |
| 1060 | "nbytes": 64000, |
| 1061 | "sha256": "ef1ced4b1b51fada242e2047a82cfa2b7d9be0177a9813415fa733ea225b1f10" |
| 1062 | }, |
| 1063 | { |
| 1064 | "name": "res0.0.blocks.1.conv2.bias", |
| 1065 | "shape": [ |
| 1066 | 80 |
| 1067 | ], |
| 1068 | "dtype": "float16", |
| 1069 | "offset_bytes": 2244048, |
| 1070 | "nbytes": 160, |
| 1071 | "sha256": "aa9678c2913fac854b61cefe6de068681bc0a70f0cc7008bf7f6591bcca64db0" |
| 1072 | }, |
| 1073 | { |
| 1074 | "name": "res0.0.blocks.1.conv2.weight", |
| 1075 | "shape": [ |
| 1076 | 80, |
| 1077 | 80, |
| 1078 | 5 |
| 1079 | ], |
| 1080 | "dtype": "float16", |
| 1081 | "offset_bytes": 2244208, |
| 1082 | "nbytes": 64000, |
| 1083 | "sha256": "e33cf26c8c81dccc33247420890d8912492fa928df6cf62a2f8624d4c5ff373a" |
| 1084 | }, |
| 1085 | { |
| 1086 | "name": "res0.0.blocks.2.conv1.bias", |
| 1087 | "shape": [ |
| 1088 | 80 |
| 1089 | ], |
| 1090 | "dtype": "float16", |
| 1091 | "offset_bytes": 2308208, |
| 1092 | "nbytes": 160, |
| 1093 | "sha256": "cfae178403172e07d9bdaba37519b25c32e3c6589f2b365a0840827de636c0b5" |
| 1094 | }, |
| 1095 | { |
| 1096 | "name": "res0.0.blocks.2.conv1.weight", |
| 1097 | "shape": [ |
| 1098 | 80, |
| 1099 | 80, |
| 1100 | 7 |
| 1101 | ], |
| 1102 | "dtype": "float16", |
| 1103 | "offset_bytes": 2308368, |
| 1104 | "nbytes": 89600, |
| 1105 | "sha256": "1c6df19d0fda6db2689f86636f6dacbd40953ab0b40a61082d319f7ab9d364cc" |
| 1106 | }, |
| 1107 | { |
| 1108 | "name": "res0.0.blocks.2.conv2.bias", |
| 1109 | "shape": [ |
| 1110 | 80 |
| 1111 | ], |
| 1112 | "dtype": "float16", |
| 1113 | "offset_bytes": 2397968, |
| 1114 | "nbytes": 160, |
| 1115 | "sha256": "31116f9c584a61e076c1d8bfc84efe04f94174793c8790541ee373aea6d24158" |
| 1116 | }, |
| 1117 | { |
| 1118 | "name": "res0.0.blocks.2.conv2.weight", |
| 1119 | "shape": [ |
| 1120 | 80, |
| 1121 | 80, |
| 1122 | 7 |
| 1123 | ], |
| 1124 | "dtype": "float16", |
| 1125 | "offset_bytes": 2398128, |
| 1126 | "nbytes": 89600, |
| 1127 | "sha256": "08bc9ade8322b7be5fc46cfd396de747c7f3d01eb64efcb00c769f530b82fc9f" |
| 1128 | }, |
| 1129 | { |
| 1130 | "name": "res1.0.blocks.0.conv1.bias", |
| 1131 | "shape": [ |
| 1132 | 40 |
| 1133 | ], |
| 1134 | "dtype": "float16", |
| 1135 | "offset_bytes": 2487728, |
| 1136 | "nbytes": 80, |
| 1137 | "sha256": "86824965b4729053cff03ad3bcafd8e1be3208a38f729f1d71f4af9f80cb476c" |
| 1138 | }, |
| 1139 | { |
| 1140 | "name": "res1.0.blocks.0.conv1.weight", |
| 1141 | "shape": [ |
| 1142 | 40, |
| 1143 | 40, |
| 1144 | 3 |
| 1145 | ], |
| 1146 | "dtype": "float16", |
| 1147 | "offset_bytes": 2487808, |
| 1148 | "nbytes": 9600, |
| 1149 | "sha256": "e027ee8a3ee635fad111f358bc275bf611da8675504f97e4bb1e23d8697e39ae" |
| 1150 | }, |
| 1151 | { |
| 1152 | "name": "res1.0.blocks.0.conv2.bias", |
| 1153 | "shape": [ |
| 1154 | 40 |
| 1155 | ], |
| 1156 | "dtype": "float16", |
| 1157 | "offset_bytes": 2497408, |
| 1158 | "nbytes": 80, |
| 1159 | "sha256": "cb8f69fff2cf423109b183f37c1ea687172004f2261204bd84cd18b3520f6a6e" |
| 1160 | }, |
| 1161 | { |
| 1162 | "name": "res1.0.blocks.0.conv2.weight", |
| 1163 | "shape": [ |
| 1164 | 40, |
| 1165 | 40, |
| 1166 | 3 |
| 1167 | ], |
| 1168 | "dtype": "float16", |
| 1169 | "offset_bytes": 2497488, |
| 1170 | "nbytes": 9600, |
| 1171 | "sha256": "1c78affa38d8345ddb36889d2722b41700086add22ef687647cec1cccc1b2a7c" |
| 1172 | }, |
| 1173 | { |
| 1174 | "name": "res1.0.blocks.1.conv1.bias", |
| 1175 | "shape": [ |
| 1176 | 40 |
| 1177 | ], |
| 1178 | "dtype": "float16", |
| 1179 | "offset_bytes": 2507088, |
| 1180 | "nbytes": 80, |
| 1181 | "sha256": "16bcffb37ed21a0868705f9766b8298bce2e249882a1602c86bba7fbaae04203" |
| 1182 | }, |
| 1183 | { |
| 1184 | "name": "res1.0.blocks.1.conv1.weight", |
| 1185 | "shape": [ |
| 1186 | 40, |
| 1187 | 40, |
| 1188 | 5 |
| 1189 | ], |
| 1190 | "dtype": "float16", |
| 1191 | "offset_bytes": 2507168, |
| 1192 | "nbytes": 16000, |
| 1193 | "sha256": "55c3ea66ce1525fb11bf77a66faba6dc85a5165f1e006c89d9115cb3927895cd" |
| 1194 | }, |
| 1195 | { |
| 1196 | "name": "res1.0.blocks.1.conv2.bias", |
| 1197 | "shape": [ |
| 1198 | 40 |
| 1199 | ], |
| 1200 | "dtype": "float16", |
| 1201 | "offset_bytes": 2523168, |
| 1202 | "nbytes": 80, |
| 1203 | "sha256": "323e527c1024949be69eec6e0a8e530fd1a3567a42a64d0549d8802c13272c6e" |
| 1204 | }, |
| 1205 | { |
| 1206 | "name": "res1.0.blocks.1.conv2.weight", |
| 1207 | "shape": [ |
| 1208 | 40, |
| 1209 | 40, |
| 1210 | 5 |
| 1211 | ], |
| 1212 | "dtype": "float16", |
| 1213 | "offset_bytes": 2523248, |
| 1214 | "nbytes": 16000, |
| 1215 | "sha256": "ee501766f0ae87ef5f1d320ec610189140a0d62eb9f974b60bd316231ee8ec34" |
| 1216 | }, |
| 1217 | { |
| 1218 | "name": "res1.0.blocks.2.conv1.bias", |
| 1219 | "shape": [ |
| 1220 | 40 |
| 1221 | ], |
| 1222 | "dtype": "float16", |
| 1223 | "offset_bytes": 2539248, |
| 1224 | "nbytes": 80, |
| 1225 | "sha256": "4b3a9e0fb64c4366af5fdb0114d9beda2b5a6e92243fdaf3587feea2baeb4138" |
| 1226 | }, |
| 1227 | { |
| 1228 | "name": "res1.0.blocks.2.conv1.weight", |
| 1229 | "shape": [ |
| 1230 | 40, |
| 1231 | 40, |
| 1232 | 7 |
| 1233 | ], |
| 1234 | "dtype": "float16", |
| 1235 | "offset_bytes": 2539328, |
| 1236 | "nbytes": 22400, |
| 1237 | "sha256": "037290cae29270b88d7fff646b7b074c0d4c352f644e362a8102de585bdeca7a" |
| 1238 | }, |
| 1239 | { |
| 1240 | "name": "res1.0.blocks.2.conv2.bias", |
| 1241 | "shape": [ |
| 1242 | 40 |
| 1243 | ], |
| 1244 | "dtype": "float16", |
| 1245 | "offset_bytes": 2561728, |
| 1246 | "nbytes": 80, |
| 1247 | "sha256": "ee94f76f6f04876226764ed8106b451bb3b95a871e2bcfb466c25f1673c90035" |
| 1248 | }, |
| 1249 | { |
| 1250 | "name": "res1.0.blocks.2.conv2.weight", |
| 1251 | "shape": [ |
| 1252 | 40, |
| 1253 | 40, |
| 1254 | 7 |
| 1255 | ], |
| 1256 | "dtype": "float16", |
| 1257 | "offset_bytes": 2561808, |
| 1258 | "nbytes": 22400, |
| 1259 | "sha256": "7539ca6cef0150793026e8d9f10299bb45107ddabac25e28c5a4df2840532d84" |
| 1260 | }, |
| 1261 | { |
| 1262 | "name": "res2.0.blocks.0.conv1.bias", |
| 1263 | "shape": [ |
| 1264 | 20 |
| 1265 | ], |
| 1266 | "dtype": "float16", |
| 1267 | "offset_bytes": 2584208, |
| 1268 | "nbytes": 40, |
| 1269 | "sha256": "85cddbbddde1e42984979e0eb21b55c85f1e8007b02119a561a6b0f824ed6a8a" |
| 1270 | }, |
| 1271 | { |
| 1272 | "name": "res2.0.blocks.0.conv1.weight", |
| 1273 | "shape": [ |
| 1274 | 20, |
| 1275 | 20, |
| 1276 | 3 |
| 1277 | ], |
| 1278 | "dtype": "float16", |
| 1279 | "offset_bytes": 2584248, |
| 1280 | "nbytes": 2400, |
| 1281 | "sha256": "dadedb22c1cf2550c012823ffe06b9bb1d7548718d5e8bebbfe2583072261c1b" |
| 1282 | }, |
| 1283 | { |
| 1284 | "name": "res2.0.blocks.0.conv2.bias", |
| 1285 | "shape": [ |
| 1286 | 20 |
| 1287 | ], |
| 1288 | "dtype": "float16", |
| 1289 | "offset_bytes": 2586648, |
| 1290 | "nbytes": 40, |
| 1291 | "sha256": "a25b67c71a1deb5ec01097f743d78056acf96f2448d0cbcc49ac774991635f73" |
| 1292 | }, |
| 1293 | { |
| 1294 | "name": "res2.0.blocks.0.conv2.weight", |
| 1295 | "shape": [ |
| 1296 | 20, |
| 1297 | 20, |
| 1298 | 3 |
| 1299 | ], |
| 1300 | "dtype": "float16", |
| 1301 | "offset_bytes": 2586688, |
| 1302 | "nbytes": 2400, |
| 1303 | "sha256": "954f9591c6d9da8f21e108e36b5a82de49db217c7809c4f74da01c88a756a864" |
| 1304 | }, |
| 1305 | { |
| 1306 | "name": "res2.0.blocks.1.conv1.bias", |
| 1307 | "shape": [ |
| 1308 | 20 |
| 1309 | ], |
| 1310 | "dtype": "float16", |
| 1311 | "offset_bytes": 2589088, |
| 1312 | "nbytes": 40, |
| 1313 | "sha256": "1f74f387a0f793ec5a9291460e25c6fb7d736b1b57b5accf79dd1ef3726e194f" |
| 1314 | }, |
| 1315 | { |
| 1316 | "name": "res2.0.blocks.1.conv1.weight", |
| 1317 | "shape": [ |
| 1318 | 20, |
| 1319 | 20, |
| 1320 | 5 |
| 1321 | ], |
| 1322 | "dtype": "float16", |
| 1323 | "offset_bytes": 2589128, |
| 1324 | "nbytes": 4000, |
| 1325 | "sha256": "52bdfd5d58aeed7e0622fa6d21348411dc627661953cd6f37c37e32c106d4598" |
| 1326 | }, |
| 1327 | { |
| 1328 | "name": "res2.0.blocks.1.conv2.bias", |
| 1329 | "shape": [ |
| 1330 | 20 |
| 1331 | ], |
| 1332 | "dtype": "float16", |
| 1333 | "offset_bytes": 2593128, |
| 1334 | "nbytes": 40, |
| 1335 | "sha256": "c60ecee1fd6b416b401150fd2629731e471ad49ca37da594b790ce5c71d6fa5b" |
| 1336 | }, |
| 1337 | { |
| 1338 | "name": "res2.0.blocks.1.conv2.weight", |
| 1339 | "shape": [ |
| 1340 | 20, |
| 1341 | 20, |
| 1342 | 5 |
| 1343 | ], |
| 1344 | "dtype": "float16", |
| 1345 | "offset_bytes": 2593168, |
| 1346 | "nbytes": 4000, |
| 1347 | "sha256": "fd3c38cd46fe111a05f514091a710038d589eaa06d96446075026e514f73e143" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "res2.0.blocks.2.conv1.bias", |
| 1351 | "shape": [ |
| 1352 | 20 |
| 1353 | ], |
| 1354 | "dtype": "float16", |
| 1355 | "offset_bytes": 2597168, |
| 1356 | "nbytes": 40, |
| 1357 | "sha256": "472e61ab64a344724252fc0a1f3c8f7c081dd21f2b27bc36454bead2d55e13d5" |
| 1358 | }, |
| 1359 | { |
| 1360 | "name": "res2.0.blocks.2.conv1.weight", |
| 1361 | "shape": [ |
| 1362 | 20, |
| 1363 | 20, |
| 1364 | 7 |
| 1365 | ], |
| 1366 | "dtype": "float16", |
| 1367 | "offset_bytes": 2597208, |
| 1368 | "nbytes": 5600, |
| 1369 | "sha256": "c239261a6ebee2e9e646f9015e6d3d5b077528407bd1f3585e5eeee9a601a931" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "res2.0.blocks.2.conv2.bias", |
| 1373 | "shape": [ |
| 1374 | 20 |
| 1375 | ], |
| 1376 | "dtype": "float16", |
| 1377 | "offset_bytes": 2602808, |
| 1378 | "nbytes": 40, |
| 1379 | "sha256": "9194fe36bb699ac8d056d6e28d03271de9431d642d6af636980010a91cd5a500" |
| 1380 | }, |
| 1381 | { |
| 1382 | "name": "res2.0.blocks.2.conv2.weight", |
| 1383 | "shape": [ |
| 1384 | 20, |
| 1385 | 20, |
| 1386 | 7 |
| 1387 | ], |
| 1388 | "dtype": "float16", |
| 1389 | "offset_bytes": 2602848, |
| 1390 | "nbytes": 5600, |
| 1391 | "sha256": "40c25e9d050bdd56fa33fe49a81376c6a01a8966a9e3b4a596bdb29a86e9f1a3" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "up0.bias", |
| 1395 | "shape": [ |
| 1396 | 80 |
| 1397 | ], |
| 1398 | "dtype": "float16", |
| 1399 | "offset_bytes": 2608448, |
| 1400 | "nbytes": 160, |
| 1401 | "sha256": "21d52d7fb004cdfaf6a7193d9950c1679a6fa78fb4a9a8d02da66799dcfe1c8c" |
| 1402 | }, |
| 1403 | { |
| 1404 | "name": "up0.weight", |
| 1405 | "shape": [ |
| 1406 | 160, |
| 1407 | 80, |
| 1408 | 16 |
| 1409 | ], |
| 1410 | "dtype": "float16", |
| 1411 | "offset_bytes": 2608608, |
| 1412 | "nbytes": 409600, |
| 1413 | "sha256": "83f22df1b15b95c2f9d08522df098a2843e1c3449572d25424e745ff1dbbf452" |
| 1414 | }, |
| 1415 | { |
| 1416 | "name": "up1.bias", |
| 1417 | "shape": [ |
| 1418 | 40 |
| 1419 | ], |
| 1420 | "dtype": "float16", |
| 1421 | "offset_bytes": 3018208, |
| 1422 | "nbytes": 80, |
| 1423 | "sha256": "c53fb3ccb173b42544b724c9425f32a71eebca37384b1afddeb52e0a3fded194" |
| 1424 | }, |
| 1425 | { |
| 1426 | "name": "up1.weight", |
| 1427 | "shape": [ |
| 1428 | 80, |
| 1429 | 40, |
| 1430 | 16 |
| 1431 | ], |
| 1432 | "dtype": "float16", |
| 1433 | "offset_bytes": 3018288, |
| 1434 | "nbytes": 102400, |
| 1435 | "sha256": "99416379e66bdfaef92bc82c35405547bb1b0efb0bba8908b3bc5727e488a8dc" |
| 1436 | }, |
| 1437 | { |
| 1438 | "name": "up2.bias", |
| 1439 | "shape": [ |
| 1440 | 20 |
| 1441 | ], |
| 1442 | "dtype": "float16", |
| 1443 | "offset_bytes": 3120688, |
| 1444 | "nbytes": 40, |
| 1445 | "sha256": "058ff72cf1dc300fb6b24c9f469c5749b7ccebf357de8626397ab873ea9c2d9d" |
| 1446 | }, |
| 1447 | { |
| 1448 | "name": "up2.weight", |
| 1449 | "shape": [ |
| 1450 | 40, |
| 1451 | 20, |
| 1452 | 8 |
| 1453 | ], |
| 1454 | "dtype": "float16", |
| 1455 | "offset_bytes": 3120728, |
| 1456 | "nbytes": 12800, |
| 1457 | "sha256": "74a2596d9bd1667d5db75f94a63e16e4267ff562ba00ee1a35a6542bf673a215" |
| 1458 | } |
| 1459 | ] |
| 1460 | } |
| 1461 | }, |
| 1462 | "runtime_ops": [ |
| 1463 | "embedding_lookup", |
| 1464 | "conv1d_same", |
| 1465 | "conv1d_1x1", |
| 1466 | "conv_transpose1d", |
| 1467 | "silu", |
| 1468 | "leaky_relu", |
| 1469 | "tanh", |
| 1470 | "residual_add", |
| 1471 | "repeat_interleave", |
| 1472 | "masked_conv1d", |
| 1473 | "depthwise_conv1d", |
| 1474 | "low_rank_linear_or_conv1d" |
| 1475 | ], |
| 1476 | "runtime_boundary": { |
| 1477 | "neural_stack_self_contained": true, |
| 1478 | "arbitrary_text_self_contained": false, |
| 1479 | "accepted_runtime_input": "Piper phoneme ID sequence matching piper-phoneme-config.json", |
| 1480 | "output": "mono float waveform at manifest.sample_rate" |
| 1481 | }, |
| 1482 | "frontend": { |
| 1483 | "status": "not fully self-contained for arbitrary text yet", |
| 1484 | "input_contract": "runtime may accept Piper phoneme ID sequences directly", |
| 1485 | "included_config": "piper-phoneme-config.json", |
| 1486 | "piper_phoneme_type": null, |
| 1487 | "piper_espeak_voice": "ru", |
| 1488 | "phoneme_id_map_entries": 151, |
| 1489 | "next_work": "embed or rewrite the Piper/eSpeak-compatible frontend, or keep a strict phoneme-ID API" |
| 1490 | }, |
| 1491 | "size_budget": { |
| 1492 | "target_max_bytes": 4000000, |
| 1493 | "target_max_mib": 3.814697265625, |
| 1494 | "actual_model_asset_bytes": 3138293, |
| 1495 | "target_max_parameters": 1500000, |
| 1496 | "under_parameter_budget": false, |
| 1497 | "under_4mb_asset_budget": true |
| 1498 | }, |
| 1499 | "limitations": [ |
| 1500 | "This is a raw neural asset package, not a compiled standalone runtime binary.", |
| 1501 | "No PyTorch checkpoint, ONNX graph, or training state is included.", |
| 1502 | "Arbitrary text still requires a Piper/eSpeak-compatible frontend outside this package." |
| 1503 | ] |
| 1504 | } |
| 1505 | |