tr-dfki-1p56m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "tr_TR-flores-20260908", |
| 4 | "language": "tr_TR", |
| 5 | "voice": "tr_TR-dfki-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": 3124568, |
| 15 | "weights_sha256": "15334771721194574725b78f8834b8f0c2fe9e12eb8e449b68c81305bb76de41", |
| 16 | "total_parameters": 1562284, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/tr_TR/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "71d3c6c0e6f7186281b0f5313d36b360f1a035cde25036dc3aa85f832b4a4f6b", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 123, |
| 24 | "hidden": 64, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 639, |
| 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": 135556, |
| 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": "ad76683090f40e34d424c0dfb9e4daa74a873a4ee2f6aee3024e52c96b139b77" |
| 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": "6a69470b9d7c0580254bd7b8910a33484a01110ac461154ea85b61b9753f7194" |
| 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": "1a11dcdb2d62d17617526986770043669279ee4a5c5703c546a7914ada224fde" |
| 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": "9b51dc4d650407dec1b2d18b8c0070de55ecc0976ee975baa1432ac9ff4a6902" |
| 90 | }, |
| 91 | { |
| 92 | "name": "blocks.0.scale", |
| 93 | "shape": [ |
| 94 | 1 |
| 95 | ], |
| 96 | "dtype": "float16", |
| 97 | "offset_bytes": 82176, |
| 98 | "nbytes": 2, |
| 99 | "sha256": "e4f95d56d90df35eca464869e14c2ff08c7c7cc86dc10a5c3083fc811a3c0952" |
| 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": "163ef72dc8ab8f985760fcd41b53d1a792082c47ffc128835ecb08fb753c4d5b" |
| 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": "35fff80c80918889c148543a79110b8949a078e647eada33239383c1577d4984" |
| 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": "0255208173da2a3408b94c8d9aa950ffa670294b111b8ea3f2aaf1a8fcbf6545" |
| 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": "057af80722170df26f619313f7d2368d292db3c35330a50282e7b2ba78b3fd7b" |
| 144 | }, |
| 145 | { |
| 146 | "name": "blocks.1.scale", |
| 147 | "shape": [ |
| 148 | 1 |
| 149 | ], |
| 150 | "dtype": "float16", |
| 151 | "offset_bytes": 164354, |
| 152 | "nbytes": 2, |
| 153 | "sha256": "fbff3de6e2e3dc94a34f0cb32cdfda81e3f354b4b9ec4f5c70f50de28d8ccb2e" |
| 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": "cf259d8332ea32bc42ee31e768e8924460aed4498fca0511da05b41359751b49" |
| 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": "56ca349180616616fdfd55c944d64572e0d3766d4eac93c5735290b7d6e6fcf7" |
| 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": "435d6ca376acead11c921a271fd158b0ac70b298f598f59adc3270669b46eb5f" |
| 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": "19087d0c77d31225147d6ade895ab6b6ccc960fc378aa0ede59baa92948f3053" |
| 198 | }, |
| 199 | { |
| 200 | "name": "blocks.2.scale", |
| 201 | "shape": [ |
| 202 | 1 |
| 203 | ], |
| 204 | "dtype": "float16", |
| 205 | "offset_bytes": 246532, |
| 206 | "nbytes": 2, |
| 207 | "sha256": "bb4e05654fb22d092b8e6347a377838551da9783bf0cedd362614ddf9145d51d" |
| 208 | }, |
| 209 | { |
| 210 | "name": "embedding.weight", |
| 211 | "shape": [ |
| 212 | 123, |
| 213 | 64 |
| 214 | ], |
| 215 | "dtype": "float16", |
| 216 | "offset_bytes": 246534, |
| 217 | "nbytes": 15744, |
| 218 | "sha256": "78c1ecce82a76781c7c49b32a6daa09edbbd82c6ea3ed849acf3a8f840fe2f40" |
| 219 | }, |
| 220 | { |
| 221 | "name": "input_proj.bias", |
| 222 | "shape": [ |
| 223 | 64 |
| 224 | ], |
| 225 | "dtype": "float16", |
| 226 | "offset_bytes": 262278, |
| 227 | "nbytes": 128, |
| 228 | "sha256": "88a92c531681206caf90ff475520df61987c59982af198c48b5187c4e7b47b17" |
| 229 | }, |
| 230 | { |
| 231 | "name": "input_proj.weight", |
| 232 | "shape": [ |
| 233 | 64, |
| 234 | 67, |
| 235 | 1 |
| 236 | ], |
| 237 | "dtype": "float16", |
| 238 | "offset_bytes": 262406, |
| 239 | "nbytes": 8576, |
| 240 | "sha256": "c0adb64248c440aebc7a461a47fdfe7a87f871733e26c2df994dc3fe4f00de7d" |
| 241 | }, |
| 242 | { |
| 243 | "name": "output.bias", |
| 244 | "shape": [ |
| 245 | 1 |
| 246 | ], |
| 247 | "dtype": "float16", |
| 248 | "offset_bytes": 270982, |
| 249 | "nbytes": 2, |
| 250 | "sha256": "7e410e0ca11b4271dbd55fa5898d449fb7de5a1d826257162ed12176f69693ca" |
| 251 | }, |
| 252 | { |
| 253 | "name": "output.weight", |
| 254 | "shape": [ |
| 255 | 1, |
| 256 | 64, |
| 257 | 1 |
| 258 | ], |
| 259 | "dtype": "float16", |
| 260 | "offset_bytes": 270984, |
| 261 | "nbytes": 128, |
| 262 | "sha256": "afbca2ea46279f2a9706e34ad9c317ed7f63d64e3d34d451338254fdc7106f85" |
| 263 | } |
| 264 | ] |
| 265 | }, |
| 266 | "acoustic": { |
| 267 | "checkpoint": "artifacts/voices/tr_TR/joint/latent-student.pt", |
| 268 | "checkpoint_sha256": "0418c99f21cd250aa3c287e5e41f932aaf41fd4f17167392b1da5ecd196ab963", |
| 269 | "config": { |
| 270 | "architecture": "token_context", |
| 271 | "vocab_size": 123, |
| 272 | "hidden": 96, |
| 273 | "depth": 4, |
| 274 | "kernel_size": 5, |
| 275 | "out_channels": 192, |
| 276 | "token_depth": 3 |
| 277 | }, |
| 278 | "parameters": 696007, |
| 279 | "tensors": [ |
| 280 | { |
| 281 | "name": "embedding.weight", |
| 282 | "shape": [ |
| 283 | 123, |
| 284 | 96 |
| 285 | ], |
| 286 | "dtype": "float16", |
| 287 | "offset_bytes": 271112, |
| 288 | "nbytes": 23616, |
| 289 | "sha256": "26f41ef23655268df6a7b0023e60bffaa8e5937cf2b85cdf1a0856079d079d52" |
| 290 | }, |
| 291 | { |
| 292 | "name": "frame_blocks.0.net.0.bias", |
| 293 | "shape": [ |
| 294 | 96 |
| 295 | ], |
| 296 | "dtype": "float16", |
| 297 | "offset_bytes": 294728, |
| 298 | "nbytes": 192, |
| 299 | "sha256": "e14772f3dfce10383722a0d2f006d6c323da9d8de2711e1f8a46ecbfd2ce134a" |
| 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": 294920, |
| 310 | "nbytes": 92160, |
| 311 | "sha256": "ba58a00079e969fa68f03f73fbe42f8697801a646f93d1b917bbad9ce5536bd9" |
| 312 | }, |
| 313 | { |
| 314 | "name": "frame_blocks.0.net.2.bias", |
| 315 | "shape": [ |
| 316 | 96 |
| 317 | ], |
| 318 | "dtype": "float16", |
| 319 | "offset_bytes": 387080, |
| 320 | "nbytes": 192, |
| 321 | "sha256": "677aac3ca4f61a947311d2e2bdfc69990fed2f757e324264d5e8a41ea5a4a1ca" |
| 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": 387272, |
| 332 | "nbytes": 92160, |
| 333 | "sha256": "1df934c8257c1f209f5be327f31448eaee59e41dba1d16f8a260a90a7af246be" |
| 334 | }, |
| 335 | { |
| 336 | "name": "frame_blocks.0.scale", |
| 337 | "shape": [ |
| 338 | 1 |
| 339 | ], |
| 340 | "dtype": "float16", |
| 341 | "offset_bytes": 479432, |
| 342 | "nbytes": 2, |
| 343 | "sha256": "9552ab2ce7779ee6e4186875c37c5d3e870ce0d4b471a7a82f3a92c54f72d2f6" |
| 344 | }, |
| 345 | { |
| 346 | "name": "frame_blocks.1.net.0.bias", |
| 347 | "shape": [ |
| 348 | 96 |
| 349 | ], |
| 350 | "dtype": "float16", |
| 351 | "offset_bytes": 479434, |
| 352 | "nbytes": 192, |
| 353 | "sha256": "f097953ced4b187ae62b80253736f78e72e684e861f44b6376242f929638f227" |
| 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": 479626, |
| 364 | "nbytes": 92160, |
| 365 | "sha256": "458c03de1a4bf56e5cd0c6a2df4014a1cb071e570d4ad17668de4c065a4582d0" |
| 366 | }, |
| 367 | { |
| 368 | "name": "frame_blocks.1.net.2.bias", |
| 369 | "shape": [ |
| 370 | 96 |
| 371 | ], |
| 372 | "dtype": "float16", |
| 373 | "offset_bytes": 571786, |
| 374 | "nbytes": 192, |
| 375 | "sha256": "26b2df9fc737cee237796e229d1a2d66de411db0596d00ffcb8d9ad9fff51868" |
| 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": 571978, |
| 386 | "nbytes": 92160, |
| 387 | "sha256": "46275cdec527e80c7f7e1c24636704695a4807086ced0b400a37ff2f4a9e34bb" |
| 388 | }, |
| 389 | { |
| 390 | "name": "frame_blocks.1.scale", |
| 391 | "shape": [ |
| 392 | 1 |
| 393 | ], |
| 394 | "dtype": "float16", |
| 395 | "offset_bytes": 664138, |
| 396 | "nbytes": 2, |
| 397 | "sha256": "b1ecffb387d0734451c7c02048e4b2759293c028482492665be7d4ae651f37f5" |
| 398 | }, |
| 399 | { |
| 400 | "name": "frame_blocks.2.net.0.bias", |
| 401 | "shape": [ |
| 402 | 96 |
| 403 | ], |
| 404 | "dtype": "float16", |
| 405 | "offset_bytes": 664140, |
| 406 | "nbytes": 192, |
| 407 | "sha256": "e12b1af7d81aecc6d3645fc674328a2af60f2f138e7a1a2e9579ef36ea417d7a" |
| 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": 664332, |
| 418 | "nbytes": 92160, |
| 419 | "sha256": "3e9ace0fc03ed9f0cd8ced70e00f659a62f6556398dc4b4df85585e9d9c0bf40" |
| 420 | }, |
| 421 | { |
| 422 | "name": "frame_blocks.2.net.2.bias", |
| 423 | "shape": [ |
| 424 | 96 |
| 425 | ], |
| 426 | "dtype": "float16", |
| 427 | "offset_bytes": 756492, |
| 428 | "nbytes": 192, |
| 429 | "sha256": "6184550fc713b862cf48d24f24fcdeb23909cfacbe4826ea27d94f461eccfc6a" |
| 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": 756684, |
| 440 | "nbytes": 92160, |
| 441 | "sha256": "707fe39c4e28485ba137ef6182999f3fc446fdd39d84a46da29264c95fdc66a7" |
| 442 | }, |
| 443 | { |
| 444 | "name": "frame_blocks.2.scale", |
| 445 | "shape": [ |
| 446 | 1 |
| 447 | ], |
| 448 | "dtype": "float16", |
| 449 | "offset_bytes": 848844, |
| 450 | "nbytes": 2, |
| 451 | "sha256": "5831f059c5d971b9fff65750e0c69690a720ed93432896615a36539a53fd86d3" |
| 452 | }, |
| 453 | { |
| 454 | "name": "frame_blocks.3.net.0.bias", |
| 455 | "shape": [ |
| 456 | 96 |
| 457 | ], |
| 458 | "dtype": "float16", |
| 459 | "offset_bytes": 848846, |
| 460 | "nbytes": 192, |
| 461 | "sha256": "362abbd93551f5abb4dde095db52441e9c0e0036b594801afc0d8bd475d74a1a" |
| 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": 849038, |
| 472 | "nbytes": 92160, |
| 473 | "sha256": "cd0fcbd027ed129f78d89b6237a38227d6265a0b67ea2742ffbcaa7a635f4831" |
| 474 | }, |
| 475 | { |
| 476 | "name": "frame_blocks.3.net.2.bias", |
| 477 | "shape": [ |
| 478 | 96 |
| 479 | ], |
| 480 | "dtype": "float16", |
| 481 | "offset_bytes": 941198, |
| 482 | "nbytes": 192, |
| 483 | "sha256": "3e67df1deb5e9dab098d673b3aede3f0efbcf6481768f493b37a230cfa079013" |
| 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": 941390, |
| 494 | "nbytes": 92160, |
| 495 | "sha256": "df2d6265ba225e399d4bdc850de3d4289d02d64a608fad91192265631eeb6034" |
| 496 | }, |
| 497 | { |
| 498 | "name": "frame_blocks.3.scale", |
| 499 | "shape": [ |
| 500 | 1 |
| 501 | ], |
| 502 | "dtype": "float16", |
| 503 | "offset_bytes": 1033550, |
| 504 | "nbytes": 2, |
| 505 | "sha256": "1bbda0bcafb2ce49544d205783344a4cbab020a241aaff4059fac6991171387a" |
| 506 | }, |
| 507 | { |
| 508 | "name": "frame_input_proj.bias", |
| 509 | "shape": [ |
| 510 | 96 |
| 511 | ], |
| 512 | "dtype": "float16", |
| 513 | "offset_bytes": 1033552, |
| 514 | "nbytes": 192, |
| 515 | "sha256": "423f9964eaa215af47f2f7b7b64cdedc057df48b24b992d867cdfc94d6f08190" |
| 516 | }, |
| 517 | { |
| 518 | "name": "frame_input_proj.weight", |
| 519 | "shape": [ |
| 520 | 96, |
| 521 | 99, |
| 522 | 1 |
| 523 | ], |
| 524 | "dtype": "float16", |
| 525 | "offset_bytes": 1033744, |
| 526 | "nbytes": 19008, |
| 527 | "sha256": "6e96c8789c4514b5ef6a81cb232476267689410aa89d274b3c191ff4c5301e42" |
| 528 | }, |
| 529 | { |
| 530 | "name": "output.bias", |
| 531 | "shape": [ |
| 532 | 192 |
| 533 | ], |
| 534 | "dtype": "float16", |
| 535 | "offset_bytes": 1052752, |
| 536 | "nbytes": 384, |
| 537 | "sha256": "47733cd51af895e71b8451958ef07803678c407153128641a63bf085ede6c4e5" |
| 538 | }, |
| 539 | { |
| 540 | "name": "output.weight", |
| 541 | "shape": [ |
| 542 | 192, |
| 543 | 96, |
| 544 | 1 |
| 545 | ], |
| 546 | "dtype": "float16", |
| 547 | "offset_bytes": 1053136, |
| 548 | "nbytes": 36864, |
| 549 | "sha256": "388572d7a3f94cf036f63c2ca601dc8cfd79a83062477f8c4951e498cea5546a" |
| 550 | }, |
| 551 | { |
| 552 | "name": "token_blocks.0.net.0.bias", |
| 553 | "shape": [ |
| 554 | 96 |
| 555 | ], |
| 556 | "dtype": "float16", |
| 557 | "offset_bytes": 1090000, |
| 558 | "nbytes": 192, |
| 559 | "sha256": "5b3d684897a441718a890566fa91a20a944c0022de00cf289f0798ced7cd7c2b" |
| 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": 1090192, |
| 570 | "nbytes": 92160, |
| 571 | "sha256": "d390c96c5d62b88735f1ce26c359e970afb54dbdfc4747182a6786dcdb39210c" |
| 572 | }, |
| 573 | { |
| 574 | "name": "token_blocks.0.net.2.bias", |
| 575 | "shape": [ |
| 576 | 96 |
| 577 | ], |
| 578 | "dtype": "float16", |
| 579 | "offset_bytes": 1182352, |
| 580 | "nbytes": 192, |
| 581 | "sha256": "d949b8d376b1743a76d6cc31684cdd2c87542ffcfccc4106a3e7702a351d991e" |
| 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": 1182544, |
| 592 | "nbytes": 92160, |
| 593 | "sha256": "7c15207b38100cc052f0a444daa0033360ea0e653318f5a70395214ca7347c90" |
| 594 | }, |
| 595 | { |
| 596 | "name": "token_blocks.0.scale", |
| 597 | "shape": [ |
| 598 | 1 |
| 599 | ], |
| 600 | "dtype": "float16", |
| 601 | "offset_bytes": 1274704, |
| 602 | "nbytes": 2, |
| 603 | "sha256": "795ac1af62df1338b7274d456d775f170cf0353325e5708211c6bb568b181302" |
| 604 | }, |
| 605 | { |
| 606 | "name": "token_blocks.1.net.0.bias", |
| 607 | "shape": [ |
| 608 | 96 |
| 609 | ], |
| 610 | "dtype": "float16", |
| 611 | "offset_bytes": 1274706, |
| 612 | "nbytes": 192, |
| 613 | "sha256": "b406d143a2de7fd4187790c26535cde7c7046e203c3be80c02320bcbff3eb21a" |
| 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": 1274898, |
| 624 | "nbytes": 92160, |
| 625 | "sha256": "1cce544493c020860d202e3c7978b31c251e49f0c0a5d85d18e15dcd7d76bb1c" |
| 626 | }, |
| 627 | { |
| 628 | "name": "token_blocks.1.net.2.bias", |
| 629 | "shape": [ |
| 630 | 96 |
| 631 | ], |
| 632 | "dtype": "float16", |
| 633 | "offset_bytes": 1367058, |
| 634 | "nbytes": 192, |
| 635 | "sha256": "adac9ee9798890703fe17890cebf74420c03d44646330bcb0537003c2cfdbe35" |
| 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": 1367250, |
| 646 | "nbytes": 92160, |
| 647 | "sha256": "9ec4d342675ef84726aaf87ea444106fbb0e4f4f0a456cdf94e41796c9c644ed" |
| 648 | }, |
| 649 | { |
| 650 | "name": "token_blocks.1.scale", |
| 651 | "shape": [ |
| 652 | 1 |
| 653 | ], |
| 654 | "dtype": "float16", |
| 655 | "offset_bytes": 1459410, |
| 656 | "nbytes": 2, |
| 657 | "sha256": "d95b3b43f96e1b91cb088c8a91ddd7cfad0cdd355a3d77f3244c6a2167f4fb46" |
| 658 | }, |
| 659 | { |
| 660 | "name": "token_blocks.2.net.0.bias", |
| 661 | "shape": [ |
| 662 | 96 |
| 663 | ], |
| 664 | "dtype": "float16", |
| 665 | "offset_bytes": 1459412, |
| 666 | "nbytes": 192, |
| 667 | "sha256": "37ea5cc45cc325c2c27ebe4e486f703e40c37716b948bc6b52371109aad6cf52" |
| 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": 1459604, |
| 678 | "nbytes": 92160, |
| 679 | "sha256": "5df1870832ae8556c19974e4fb9aef5ff9f5c035e35df6fb3dcb8bcfd3a4cc2a" |
| 680 | }, |
| 681 | { |
| 682 | "name": "token_blocks.2.net.2.bias", |
| 683 | "shape": [ |
| 684 | 96 |
| 685 | ], |
| 686 | "dtype": "float16", |
| 687 | "offset_bytes": 1551764, |
| 688 | "nbytes": 192, |
| 689 | "sha256": "3a9b53323b419ab19804b4608472538756279d6daf651960f3085db516c39f69" |
| 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": 1551956, |
| 700 | "nbytes": 92160, |
| 701 | "sha256": "0695c398774bce2f1ebc9397e3c83ea07a290602ba7820d5f20a2198f03371f5" |
| 702 | }, |
| 703 | { |
| 704 | "name": "token_blocks.2.scale", |
| 705 | "shape": [ |
| 706 | 1 |
| 707 | ], |
| 708 | "dtype": "float16", |
| 709 | "offset_bytes": 1644116, |
| 710 | "nbytes": 2, |
| 711 | "sha256": "969662d35a1a46457f99e6b206204c59ec4a0b228493b31501207ca0513dcf41" |
| 712 | }, |
| 713 | { |
| 714 | "name": "token_input_proj.bias", |
| 715 | "shape": [ |
| 716 | 96 |
| 717 | ], |
| 718 | "dtype": "float16", |
| 719 | "offset_bytes": 1644118, |
| 720 | "nbytes": 192, |
| 721 | "sha256": "b9f03167c3f34b897347ad5f152ced2f006b26feae12353148a580749a2b168f" |
| 722 | }, |
| 723 | { |
| 724 | "name": "token_input_proj.weight", |
| 725 | "shape": [ |
| 726 | 96, |
| 727 | 98, |
| 728 | 1 |
| 729 | ], |
| 730 | "dtype": "float16", |
| 731 | "offset_bytes": 1644310, |
| 732 | "nbytes": 18816, |
| 733 | "sha256": "23e2a6aaa09b2e95de36998281c48748f5594b067ca4605a8db0c8b7a2b913ad" |
| 734 | } |
| 735 | ] |
| 736 | }, |
| 737 | "decoder": { |
| 738 | "checkpoint": "artifacts/voices/tr_TR/joint/decoder-student.pt", |
| 739 | "checkpoint_sha256": "412b9d0ba87e8b9aeffdd4804bd9d3db8673c29e5593f2714cb4afb0af5bd551", |
| 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/tr_TR/decoder/decoder-student.pt", |
| 812 | "decoder_init": { |
| 813 | "checkpoint": "artifacts/voices/tr_TR/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/tr_TR/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": 1663126, |
| 960 | "nbytes": 2, |
| 961 | "sha256": "14377f03874ab133da79e2bf69883ffd08d50c917814dea06ca2a87bfb1a073f" |
| 962 | }, |
| 963 | { |
| 964 | "name": "post.weight", |
| 965 | "shape": [ |
| 966 | 1, |
| 967 | 20, |
| 968 | 7 |
| 969 | ], |
| 970 | "dtype": "float16", |
| 971 | "offset_bytes": 1663128, |
| 972 | "nbytes": 280, |
| 973 | "sha256": "e6874e22c2f266face5e12bf9e44ceef80df40e83e2a02c0d24e03bdf557ad71" |
| 974 | }, |
| 975 | { |
| 976 | "name": "pre.bias", |
| 977 | "shape": [ |
| 978 | 160 |
| 979 | ], |
| 980 | "dtype": "float16", |
| 981 | "offset_bytes": 1663408, |
| 982 | "nbytes": 320, |
| 983 | "sha256": "7168a19641e210ee8bdfd4a3bf3e379dc53a9f2475669c72fab78d4bce4dcdd0" |
| 984 | }, |
| 985 | { |
| 986 | "name": "pre.weight", |
| 987 | "shape": [ |
| 988 | 160, |
| 989 | 192, |
| 990 | 7 |
| 991 | ], |
| 992 | "dtype": "float16", |
| 993 | "offset_bytes": 1663728, |
| 994 | "nbytes": 430080, |
| 995 | "sha256": "fa99761a5ef9edfbe8dc598e0d09460d84e34b4a6f195adee312f1ffba261e6d" |
| 996 | }, |
| 997 | { |
| 998 | "name": "res0.0.blocks.0.conv1.bias", |
| 999 | "shape": [ |
| 1000 | 80 |
| 1001 | ], |
| 1002 | "dtype": "float16", |
| 1003 | "offset_bytes": 2093808, |
| 1004 | "nbytes": 160, |
| 1005 | "sha256": "9ba70b643574bc6789b6417d0cbfb956d4922b267f3b8ad9442c8c264ebf9a26" |
| 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": 2093968, |
| 1016 | "nbytes": 38400, |
| 1017 | "sha256": "fd658deb68047644a7e48ec96cf149bab78a13ae8b6b3d16a7a81023095f6a3e" |
| 1018 | }, |
| 1019 | { |
| 1020 | "name": "res0.0.blocks.0.conv2.bias", |
| 1021 | "shape": [ |
| 1022 | 80 |
| 1023 | ], |
| 1024 | "dtype": "float16", |
| 1025 | "offset_bytes": 2132368, |
| 1026 | "nbytes": 160, |
| 1027 | "sha256": "ac9e6f064b25a2c7abf6ac5e52fd6b335816c5591f9140a5b9dd49391de1e400" |
| 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": 2132528, |
| 1038 | "nbytes": 38400, |
| 1039 | "sha256": "3ef9859ceb37ed34ab1d95edccf2d838927094c8edd1b2eb6668b9471000bf58" |
| 1040 | }, |
| 1041 | { |
| 1042 | "name": "res0.0.blocks.1.conv1.bias", |
| 1043 | "shape": [ |
| 1044 | 80 |
| 1045 | ], |
| 1046 | "dtype": "float16", |
| 1047 | "offset_bytes": 2170928, |
| 1048 | "nbytes": 160, |
| 1049 | "sha256": "27c11a96f16c20aea30557c51408d3f8a051fb683eb799d5e53445080076683c" |
| 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": 2171088, |
| 1060 | "nbytes": 64000, |
| 1061 | "sha256": "4fd1f4d3ac84f2fb15d8e03f9fd43279d25c8d82152f1c065b13d0305ebc8425" |
| 1062 | }, |
| 1063 | { |
| 1064 | "name": "res0.0.blocks.1.conv2.bias", |
| 1065 | "shape": [ |
| 1066 | 80 |
| 1067 | ], |
| 1068 | "dtype": "float16", |
| 1069 | "offset_bytes": 2235088, |
| 1070 | "nbytes": 160, |
| 1071 | "sha256": "a714d2384530429150d2cff8667e4ecf50a1f547cdbc8c3e9541ae1007eecd60" |
| 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": 2235248, |
| 1082 | "nbytes": 64000, |
| 1083 | "sha256": "b2c7bd8927435203e367ecb77620c6d038a01abcdbd6e89039d5392b2e1c34fb" |
| 1084 | }, |
| 1085 | { |
| 1086 | "name": "res0.0.blocks.2.conv1.bias", |
| 1087 | "shape": [ |
| 1088 | 80 |
| 1089 | ], |
| 1090 | "dtype": "float16", |
| 1091 | "offset_bytes": 2299248, |
| 1092 | "nbytes": 160, |
| 1093 | "sha256": "19e1182219ca43d98a8d50dd853fd3fac7ef4a6711c22b0cecf9578ee17a5688" |
| 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": 2299408, |
| 1104 | "nbytes": 89600, |
| 1105 | "sha256": "16a80618ee57dcc0d8b0448f7b52e6b8fbcf7663d74aad0d2c4bfe917b39b62d" |
| 1106 | }, |
| 1107 | { |
| 1108 | "name": "res0.0.blocks.2.conv2.bias", |
| 1109 | "shape": [ |
| 1110 | 80 |
| 1111 | ], |
| 1112 | "dtype": "float16", |
| 1113 | "offset_bytes": 2389008, |
| 1114 | "nbytes": 160, |
| 1115 | "sha256": "cd7ccc8b917a9a00bcb0bcc996bab74a5b9ec3f14dcb3adb9ebf1ef6636c7633" |
| 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": 2389168, |
| 1126 | "nbytes": 89600, |
| 1127 | "sha256": "19c32600e8f300787892413f4dd6ebaf075f54132d1822f6d290e3fb756faa6e" |
| 1128 | }, |
| 1129 | { |
| 1130 | "name": "res1.0.blocks.0.conv1.bias", |
| 1131 | "shape": [ |
| 1132 | 40 |
| 1133 | ], |
| 1134 | "dtype": "float16", |
| 1135 | "offset_bytes": 2478768, |
| 1136 | "nbytes": 80, |
| 1137 | "sha256": "18b37d7d37e6906656f000c1852ba37813572ed5e3202b183ce2342da69cd417" |
| 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": 2478848, |
| 1148 | "nbytes": 9600, |
| 1149 | "sha256": "6ea84e13349203aa386aa11b708a60384af828dccd43bbc50d4227ea18c7c9f6" |
| 1150 | }, |
| 1151 | { |
| 1152 | "name": "res1.0.blocks.0.conv2.bias", |
| 1153 | "shape": [ |
| 1154 | 40 |
| 1155 | ], |
| 1156 | "dtype": "float16", |
| 1157 | "offset_bytes": 2488448, |
| 1158 | "nbytes": 80, |
| 1159 | "sha256": "0a0bb4ff736ec583d57c6fa2e76a63025dc02d9fe8e743abb50c5171cf8b1083" |
| 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": 2488528, |
| 1170 | "nbytes": 9600, |
| 1171 | "sha256": "39703474c929c4fe4099f2b86f167f53a070ac063009c141946d86c3bd077fbf" |
| 1172 | }, |
| 1173 | { |
| 1174 | "name": "res1.0.blocks.1.conv1.bias", |
| 1175 | "shape": [ |
| 1176 | 40 |
| 1177 | ], |
| 1178 | "dtype": "float16", |
| 1179 | "offset_bytes": 2498128, |
| 1180 | "nbytes": 80, |
| 1181 | "sha256": "55ca47796b990823155e26402b0388191a6facb5e24aa062f92736cf299aae57" |
| 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": 2498208, |
| 1192 | "nbytes": 16000, |
| 1193 | "sha256": "ff47d142ed731a8adb09052dbb0e60f92879b4ab76a3d093b7a974392350de99" |
| 1194 | }, |
| 1195 | { |
| 1196 | "name": "res1.0.blocks.1.conv2.bias", |
| 1197 | "shape": [ |
| 1198 | 40 |
| 1199 | ], |
| 1200 | "dtype": "float16", |
| 1201 | "offset_bytes": 2514208, |
| 1202 | "nbytes": 80, |
| 1203 | "sha256": "a55cb55907a5e71765b7c8d698ebacafc50ae17c3223f77387e79821258eff2d" |
| 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": 2514288, |
| 1214 | "nbytes": 16000, |
| 1215 | "sha256": "5f2e9975a6abdb80c7ce16f0da7b48740b5717b4e80d6fefa6d002ee060dcefb" |
| 1216 | }, |
| 1217 | { |
| 1218 | "name": "res1.0.blocks.2.conv1.bias", |
| 1219 | "shape": [ |
| 1220 | 40 |
| 1221 | ], |
| 1222 | "dtype": "float16", |
| 1223 | "offset_bytes": 2530288, |
| 1224 | "nbytes": 80, |
| 1225 | "sha256": "06df978d84d25ab5088f7816ce425ceb109ec886dbb82160ef5c2751be4789db" |
| 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": 2530368, |
| 1236 | "nbytes": 22400, |
| 1237 | "sha256": "d7cc7cf084036e11c917861ef65f47838a0adef9e06e23034811c3db17c5651b" |
| 1238 | }, |
| 1239 | { |
| 1240 | "name": "res1.0.blocks.2.conv2.bias", |
| 1241 | "shape": [ |
| 1242 | 40 |
| 1243 | ], |
| 1244 | "dtype": "float16", |
| 1245 | "offset_bytes": 2552768, |
| 1246 | "nbytes": 80, |
| 1247 | "sha256": "b51bd7638d43d8459ba7ef92e380bf5750cbea13d2f7d87467d6f8b233f8b9ac" |
| 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": 2552848, |
| 1258 | "nbytes": 22400, |
| 1259 | "sha256": "5cf0b1b1918f10381ec6e52f75880a1b6c2f61ede46271c7d9c4d2d8448cc695" |
| 1260 | }, |
| 1261 | { |
| 1262 | "name": "res2.0.blocks.0.conv1.bias", |
| 1263 | "shape": [ |
| 1264 | 20 |
| 1265 | ], |
| 1266 | "dtype": "float16", |
| 1267 | "offset_bytes": 2575248, |
| 1268 | "nbytes": 40, |
| 1269 | "sha256": "aa1ff0d40567ce28ac808a1260f7d4a365710db75d596d3bff13c0b1f3c1cc53" |
| 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": 2575288, |
| 1280 | "nbytes": 2400, |
| 1281 | "sha256": "03806c498fca90db403b7031704e47afc2d15ea4420c23d301c0e23eb8a910c3" |
| 1282 | }, |
| 1283 | { |
| 1284 | "name": "res2.0.blocks.0.conv2.bias", |
| 1285 | "shape": [ |
| 1286 | 20 |
| 1287 | ], |
| 1288 | "dtype": "float16", |
| 1289 | "offset_bytes": 2577688, |
| 1290 | "nbytes": 40, |
| 1291 | "sha256": "982d64112d130a22e1556980379960b2a0cb39b5f2e9579eb2ff20c18fe8cf07" |
| 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": 2577728, |
| 1302 | "nbytes": 2400, |
| 1303 | "sha256": "ef4cede570a2c66696b92666f3c78aaa56190d772c1c103f9ddc1241fe719a4f" |
| 1304 | }, |
| 1305 | { |
| 1306 | "name": "res2.0.blocks.1.conv1.bias", |
| 1307 | "shape": [ |
| 1308 | 20 |
| 1309 | ], |
| 1310 | "dtype": "float16", |
| 1311 | "offset_bytes": 2580128, |
| 1312 | "nbytes": 40, |
| 1313 | "sha256": "b1c414811c99d4d7d30398e6a3f37f7ed5f5e05e00d6039f2fd0785e48dfe1d7" |
| 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": 2580168, |
| 1324 | "nbytes": 4000, |
| 1325 | "sha256": "492647267d7d3f9000812bd313ec55d74217b4f0ecd3dcbd520292d5daa362d5" |
| 1326 | }, |
| 1327 | { |
| 1328 | "name": "res2.0.blocks.1.conv2.bias", |
| 1329 | "shape": [ |
| 1330 | 20 |
| 1331 | ], |
| 1332 | "dtype": "float16", |
| 1333 | "offset_bytes": 2584168, |
| 1334 | "nbytes": 40, |
| 1335 | "sha256": "9edb83eaeed0a89ccad488d0ca7ee3d6e20aac3566060889663931e05366398e" |
| 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": 2584208, |
| 1346 | "nbytes": 4000, |
| 1347 | "sha256": "c992af2e97c8c29971c27e52aa2c3f5d8809d0b95331c9cb72c62da4bef884dc" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "res2.0.blocks.2.conv1.bias", |
| 1351 | "shape": [ |
| 1352 | 20 |
| 1353 | ], |
| 1354 | "dtype": "float16", |
| 1355 | "offset_bytes": 2588208, |
| 1356 | "nbytes": 40, |
| 1357 | "sha256": "fa1ce3c234d7187ca404ac9851899ae129925fe6bcb480f925dc5deaba3dcbca" |
| 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": 2588248, |
| 1368 | "nbytes": 5600, |
| 1369 | "sha256": "b654105d39b2d18729a3d7a91e017412e22565c449d972201b102a76e59a1e0d" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "res2.0.blocks.2.conv2.bias", |
| 1373 | "shape": [ |
| 1374 | 20 |
| 1375 | ], |
| 1376 | "dtype": "float16", |
| 1377 | "offset_bytes": 2593848, |
| 1378 | "nbytes": 40, |
| 1379 | "sha256": "30f653075a7257dc4475822a5eb2515801c97219027eec5dc845c7626f370bba" |
| 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": 2593888, |
| 1390 | "nbytes": 5600, |
| 1391 | "sha256": "7f9a13ff20f543b08567b322a47fe8470eb9692edce39b0b459d7f94c55fa49f" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "up0.bias", |
| 1395 | "shape": [ |
| 1396 | 80 |
| 1397 | ], |
| 1398 | "dtype": "float16", |
| 1399 | "offset_bytes": 2599488, |
| 1400 | "nbytes": 160, |
| 1401 | "sha256": "343efd1367f6e46cc1aeed14d5da0789486032bf1eb5daa53ddf3e72cbb6f4cb" |
| 1402 | }, |
| 1403 | { |
| 1404 | "name": "up0.weight", |
| 1405 | "shape": [ |
| 1406 | 160, |
| 1407 | 80, |
| 1408 | 16 |
| 1409 | ], |
| 1410 | "dtype": "float16", |
| 1411 | "offset_bytes": 2599648, |
| 1412 | "nbytes": 409600, |
| 1413 | "sha256": "133ad8b2a13fd50607263b952fc1900548d47a89ade8f78b65170c0827fab325" |
| 1414 | }, |
| 1415 | { |
| 1416 | "name": "up1.bias", |
| 1417 | "shape": [ |
| 1418 | 40 |
| 1419 | ], |
| 1420 | "dtype": "float16", |
| 1421 | "offset_bytes": 3009248, |
| 1422 | "nbytes": 80, |
| 1423 | "sha256": "601d6a18a699ab58140f064ea34e0053a698f3624b0270fe9a74b6e3d372c4b8" |
| 1424 | }, |
| 1425 | { |
| 1426 | "name": "up1.weight", |
| 1427 | "shape": [ |
| 1428 | 80, |
| 1429 | 40, |
| 1430 | 16 |
| 1431 | ], |
| 1432 | "dtype": "float16", |
| 1433 | "offset_bytes": 3009328, |
| 1434 | "nbytes": 102400, |
| 1435 | "sha256": "cc5875da3dc0bc07192075b111cfacb223f8096475690b1d2b5d59891c4b05e6" |
| 1436 | }, |
| 1437 | { |
| 1438 | "name": "up2.bias", |
| 1439 | "shape": [ |
| 1440 | 20 |
| 1441 | ], |
| 1442 | "dtype": "float16", |
| 1443 | "offset_bytes": 3111728, |
| 1444 | "nbytes": 40, |
| 1445 | "sha256": "84bc1151b0371c32eadbbfc9eb97d6b8c50f8eb3fb436e3bc4aaea77162dd8e8" |
| 1446 | }, |
| 1447 | { |
| 1448 | "name": "up2.weight", |
| 1449 | "shape": [ |
| 1450 | 40, |
| 1451 | 20, |
| 1452 | 8 |
| 1453 | ], |
| 1454 | "dtype": "float16", |
| 1455 | "offset_bytes": 3111768, |
| 1456 | "nbytes": 12800, |
| 1457 | "sha256": "8ee053c87edbce12d5122fbdee7d7c5fb98e82c3e5abc6b1d3b7845854289b58" |
| 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": "espeak", |
| 1487 | "piper_espeak_voice": "tr", |
| 1488 | "phoneme_id_map_entries": 154, |
| 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": 3129528, |
| 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 | |