id-newstts-1p46m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "id-newstts-1p46m", |
| 4 | "language": "id_ID", |
| 5 | "voice": "id_ID-news_tts-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": 1.16, |
| 11 | "duration_policy": "round(exp(log_duration) * duration_length_scale)" |
| 12 | }, |
| 13 | "weights_file": "weights.fp16.bin", |
| 14 | "weights_size_bytes": 3124248, |
| 15 | "weights_sha256": "14eb693f8ea7ae61f35c4b3873aaa6c9b402fc10dc98a1df81f640f5146792cd", |
| 16 | "total_parameters": 1562124, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/id_ID-news_tts-medium-retrain-20260712/a5-duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "c4ec508d7f78482aa05f656070b88e15a231fbe036cfae8322e6c0d5792013b6", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 122, |
| 24 | "hidden": 64, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 313, |
| 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 | }, |
| 43 | "parameters": 135492, |
| 44 | "tensors": [ |
| 45 | { |
| 46 | "name": "blocks.0.net.0.bias", |
| 47 | "shape": [ |
| 48 | 64 |
| 49 | ], |
| 50 | "dtype": "float16", |
| 51 | "offset_bytes": 0, |
| 52 | "nbytes": 128, |
| 53 | "sha256": "785d58a958deb6c5072382b109c13e30359941a8d7278ba6b3f397f02211b5b3" |
| 54 | }, |
| 55 | { |
| 56 | "name": "blocks.0.net.0.weight", |
| 57 | "shape": [ |
| 58 | 64, |
| 59 | 64, |
| 60 | 5 |
| 61 | ], |
| 62 | "dtype": "float16", |
| 63 | "offset_bytes": 128, |
| 64 | "nbytes": 40960, |
| 65 | "sha256": "6ee691fa5eb453c963316516b0b2b3fddd7373c014d90a549a29c1f6b0670a07" |
| 66 | }, |
| 67 | { |
| 68 | "name": "blocks.0.net.2.bias", |
| 69 | "shape": [ |
| 70 | 64 |
| 71 | ], |
| 72 | "dtype": "float16", |
| 73 | "offset_bytes": 41088, |
| 74 | "nbytes": 128, |
| 75 | "sha256": "2a7677bef042a73eb9a8ec451990225f062adf802aeea3baa1fbae3df171919f" |
| 76 | }, |
| 77 | { |
| 78 | "name": "blocks.0.net.2.weight", |
| 79 | "shape": [ |
| 80 | 64, |
| 81 | 64, |
| 82 | 5 |
| 83 | ], |
| 84 | "dtype": "float16", |
| 85 | "offset_bytes": 41216, |
| 86 | "nbytes": 40960, |
| 87 | "sha256": "3164138786c2062fbe7c60ded8d4ac4dc33cbb4a17a61421d5c6e51710bc7d84" |
| 88 | }, |
| 89 | { |
| 90 | "name": "blocks.0.scale", |
| 91 | "shape": [ |
| 92 | 1 |
| 93 | ], |
| 94 | "dtype": "float16", |
| 95 | "offset_bytes": 82176, |
| 96 | "nbytes": 2, |
| 97 | "sha256": "34f2596182d6722c5ceed4bc34343ea418e7616c38a3dc2f0e95f11a95331bab" |
| 98 | }, |
| 99 | { |
| 100 | "name": "blocks.1.net.0.bias", |
| 101 | "shape": [ |
| 102 | 64 |
| 103 | ], |
| 104 | "dtype": "float16", |
| 105 | "offset_bytes": 82178, |
| 106 | "nbytes": 128, |
| 107 | "sha256": "13d0c05ad9c1240282454dffe0e85a464d85ff902d734d99d7671dd0338c71b8" |
| 108 | }, |
| 109 | { |
| 110 | "name": "blocks.1.net.0.weight", |
| 111 | "shape": [ |
| 112 | 64, |
| 113 | 64, |
| 114 | 5 |
| 115 | ], |
| 116 | "dtype": "float16", |
| 117 | "offset_bytes": 82306, |
| 118 | "nbytes": 40960, |
| 119 | "sha256": "fcf314793a4f4007933022ef57fa4738c7d56774badc4365c5b41f097eee6063" |
| 120 | }, |
| 121 | { |
| 122 | "name": "blocks.1.net.2.bias", |
| 123 | "shape": [ |
| 124 | 64 |
| 125 | ], |
| 126 | "dtype": "float16", |
| 127 | "offset_bytes": 123266, |
| 128 | "nbytes": 128, |
| 129 | "sha256": "5ddf7d7c23f8185cf474f3f88c28411ee07d62893f80ef16c322eaf9ca2a381f" |
| 130 | }, |
| 131 | { |
| 132 | "name": "blocks.1.net.2.weight", |
| 133 | "shape": [ |
| 134 | 64, |
| 135 | 64, |
| 136 | 5 |
| 137 | ], |
| 138 | "dtype": "float16", |
| 139 | "offset_bytes": 123394, |
| 140 | "nbytes": 40960, |
| 141 | "sha256": "fa879e2cdd5ba85dae3b59df8a2d77ac3007792b10f134073bf52aec05d78898" |
| 142 | }, |
| 143 | { |
| 144 | "name": "blocks.1.scale", |
| 145 | "shape": [ |
| 146 | 1 |
| 147 | ], |
| 148 | "dtype": "float16", |
| 149 | "offset_bytes": 164354, |
| 150 | "nbytes": 2, |
| 151 | "sha256": "00ff325014b60f843eefef9031cb35dc2440623e65f0cabaef783eea4bfa837a" |
| 152 | }, |
| 153 | { |
| 154 | "name": "blocks.2.net.0.bias", |
| 155 | "shape": [ |
| 156 | 64 |
| 157 | ], |
| 158 | "dtype": "float16", |
| 159 | "offset_bytes": 164356, |
| 160 | "nbytes": 128, |
| 161 | "sha256": "c430f8163c4c4bdef3e1e79ddb7e1cc40f85e3ff5ac764f4c2e885d4f366286e" |
| 162 | }, |
| 163 | { |
| 164 | "name": "blocks.2.net.0.weight", |
| 165 | "shape": [ |
| 166 | 64, |
| 167 | 64, |
| 168 | 5 |
| 169 | ], |
| 170 | "dtype": "float16", |
| 171 | "offset_bytes": 164484, |
| 172 | "nbytes": 40960, |
| 173 | "sha256": "4d028c851333dc11a3366caaee5739986035d86e7d6c5d00d1a2650e6f32341a" |
| 174 | }, |
| 175 | { |
| 176 | "name": "blocks.2.net.2.bias", |
| 177 | "shape": [ |
| 178 | 64 |
| 179 | ], |
| 180 | "dtype": "float16", |
| 181 | "offset_bytes": 205444, |
| 182 | "nbytes": 128, |
| 183 | "sha256": "5cb9425f6f03897c393850373cdc5b7315d66f38b5fc8b712d7cb07c43c75736" |
| 184 | }, |
| 185 | { |
| 186 | "name": "blocks.2.net.2.weight", |
| 187 | "shape": [ |
| 188 | 64, |
| 189 | 64, |
| 190 | 5 |
| 191 | ], |
| 192 | "dtype": "float16", |
| 193 | "offset_bytes": 205572, |
| 194 | "nbytes": 40960, |
| 195 | "sha256": "288e52547fc371bbfdaa9ac524402c3952fb86d7443f361181ca94f78009805c" |
| 196 | }, |
| 197 | { |
| 198 | "name": "blocks.2.scale", |
| 199 | "shape": [ |
| 200 | 1 |
| 201 | ], |
| 202 | "dtype": "float16", |
| 203 | "offset_bytes": 246532, |
| 204 | "nbytes": 2, |
| 205 | "sha256": "d6e2bb8cd4629449353a28ae909ef35537271f1c26d386a43201368fe600fee9" |
| 206 | }, |
| 207 | { |
| 208 | "name": "embedding.weight", |
| 209 | "shape": [ |
| 210 | 122, |
| 211 | 64 |
| 212 | ], |
| 213 | "dtype": "float16", |
| 214 | "offset_bytes": 246534, |
| 215 | "nbytes": 15616, |
| 216 | "sha256": "746dff67141cd624d459d736684a395f2243a008a097ea776fd539d40a9520f6" |
| 217 | }, |
| 218 | { |
| 219 | "name": "input_proj.bias", |
| 220 | "shape": [ |
| 221 | 64 |
| 222 | ], |
| 223 | "dtype": "float16", |
| 224 | "offset_bytes": 262150, |
| 225 | "nbytes": 128, |
| 226 | "sha256": "d966bcacffcd0b556c24a1657571d6fc43bec014e0d6a1ea13143c938d44f3d5" |
| 227 | }, |
| 228 | { |
| 229 | "name": "input_proj.weight", |
| 230 | "shape": [ |
| 231 | 64, |
| 232 | 67, |
| 233 | 1 |
| 234 | ], |
| 235 | "dtype": "float16", |
| 236 | "offset_bytes": 262278, |
| 237 | "nbytes": 8576, |
| 238 | "sha256": "efa5b6bfc58c6a4a82de0b01990dc9f924bf79e9b056d5bb2a516003b0b78a5e" |
| 239 | }, |
| 240 | { |
| 241 | "name": "output.bias", |
| 242 | "shape": [ |
| 243 | 1 |
| 244 | ], |
| 245 | "dtype": "float16", |
| 246 | "offset_bytes": 270854, |
| 247 | "nbytes": 2, |
| 248 | "sha256": "4bb48e76f19de6ebed7d59d46800508030aecea46ba56bd19855d94473e28bc0" |
| 249 | }, |
| 250 | { |
| 251 | "name": "output.weight", |
| 252 | "shape": [ |
| 253 | 1, |
| 254 | 64, |
| 255 | 1 |
| 256 | ], |
| 257 | "dtype": "float16", |
| 258 | "offset_bytes": 270856, |
| 259 | "nbytes": 128, |
| 260 | "sha256": "fc64c6bb5488c0167f02a21ce38220f7ee7d290d30cdd8f907b5e7b82943babc" |
| 261 | } |
| 262 | ] |
| 263 | }, |
| 264 | "acoustic": { |
| 265 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/id_ID-news_tts-medium-retrain-20260712/a93-joint/latent-student.pt", |
| 266 | "checkpoint_sha256": "c0ed2bef9f5fd585ee59ce7b18432d828527fb8926441be950bda51603fa887c", |
| 267 | "config": { |
| 268 | "architecture": "token_context", |
| 269 | "vocab_size": 122, |
| 270 | "hidden": 96, |
| 271 | "depth": 4, |
| 272 | "kernel_size": 5, |
| 273 | "out_channels": 192, |
| 274 | "token_depth": 3 |
| 275 | }, |
| 276 | "parameters": 695911, |
| 277 | "tensors": [ |
| 278 | { |
| 279 | "name": "embedding.weight", |
| 280 | "shape": [ |
| 281 | 122, |
| 282 | 96 |
| 283 | ], |
| 284 | "dtype": "float16", |
| 285 | "offset_bytes": 270984, |
| 286 | "nbytes": 23424, |
| 287 | "sha256": "933d1cbfe2b8438db976f083c032a8ec40f7309575e7e97d3ccd1580d893bae0" |
| 288 | }, |
| 289 | { |
| 290 | "name": "frame_blocks.0.net.0.bias", |
| 291 | "shape": [ |
| 292 | 96 |
| 293 | ], |
| 294 | "dtype": "float16", |
| 295 | "offset_bytes": 294408, |
| 296 | "nbytes": 192, |
| 297 | "sha256": "c06f8b73be04840b390ad7e0c3180391156f2e825a2644451cf202a50278fa84" |
| 298 | }, |
| 299 | { |
| 300 | "name": "frame_blocks.0.net.0.weight", |
| 301 | "shape": [ |
| 302 | 96, |
| 303 | 96, |
| 304 | 5 |
| 305 | ], |
| 306 | "dtype": "float16", |
| 307 | "offset_bytes": 294600, |
| 308 | "nbytes": 92160, |
| 309 | "sha256": "19edddc092c44cbbedc04a923f0048ebc6f8aa7f1ade06527d0ccf802556edb2" |
| 310 | }, |
| 311 | { |
| 312 | "name": "frame_blocks.0.net.2.bias", |
| 313 | "shape": [ |
| 314 | 96 |
| 315 | ], |
| 316 | "dtype": "float16", |
| 317 | "offset_bytes": 386760, |
| 318 | "nbytes": 192, |
| 319 | "sha256": "8b429043b20210d4a2c81793cdf718c5b2fb5c55a89ba325e068c6aa1d9c3cf8" |
| 320 | }, |
| 321 | { |
| 322 | "name": "frame_blocks.0.net.2.weight", |
| 323 | "shape": [ |
| 324 | 96, |
| 325 | 96, |
| 326 | 5 |
| 327 | ], |
| 328 | "dtype": "float16", |
| 329 | "offset_bytes": 386952, |
| 330 | "nbytes": 92160, |
| 331 | "sha256": "03412924f7abf3328275c65aee0de80f0439745071ce4225d1dd1b831b5094a6" |
| 332 | }, |
| 333 | { |
| 334 | "name": "frame_blocks.0.scale", |
| 335 | "shape": [ |
| 336 | 1 |
| 337 | ], |
| 338 | "dtype": "float16", |
| 339 | "offset_bytes": 479112, |
| 340 | "nbytes": 2, |
| 341 | "sha256": "31261e1581ee1c0a02c1dc9ce141e039fd3a99a83da3219489233c3652dd9101" |
| 342 | }, |
| 343 | { |
| 344 | "name": "frame_blocks.1.net.0.bias", |
| 345 | "shape": [ |
| 346 | 96 |
| 347 | ], |
| 348 | "dtype": "float16", |
| 349 | "offset_bytes": 479114, |
| 350 | "nbytes": 192, |
| 351 | "sha256": "25174ef47ff14c4e0966863c51ba82ad9925a0794a794c51c30493dfd72e1491" |
| 352 | }, |
| 353 | { |
| 354 | "name": "frame_blocks.1.net.0.weight", |
| 355 | "shape": [ |
| 356 | 96, |
| 357 | 96, |
| 358 | 5 |
| 359 | ], |
| 360 | "dtype": "float16", |
| 361 | "offset_bytes": 479306, |
| 362 | "nbytes": 92160, |
| 363 | "sha256": "b9c0911f846257c43dc4820bcfa30d7b0295ed04731ce0313eeb6de1925e1693" |
| 364 | }, |
| 365 | { |
| 366 | "name": "frame_blocks.1.net.2.bias", |
| 367 | "shape": [ |
| 368 | 96 |
| 369 | ], |
| 370 | "dtype": "float16", |
| 371 | "offset_bytes": 571466, |
| 372 | "nbytes": 192, |
| 373 | "sha256": "60593df55d11f39aaeaa6592427a4ae967df059260b41aa247dc414b907b5cf1" |
| 374 | }, |
| 375 | { |
| 376 | "name": "frame_blocks.1.net.2.weight", |
| 377 | "shape": [ |
| 378 | 96, |
| 379 | 96, |
| 380 | 5 |
| 381 | ], |
| 382 | "dtype": "float16", |
| 383 | "offset_bytes": 571658, |
| 384 | "nbytes": 92160, |
| 385 | "sha256": "644463dd452b986b3ed8d4aac07e1e3ad819684e305753d0f7dbe27a4554d3b4" |
| 386 | }, |
| 387 | { |
| 388 | "name": "frame_blocks.1.scale", |
| 389 | "shape": [ |
| 390 | 1 |
| 391 | ], |
| 392 | "dtype": "float16", |
| 393 | "offset_bytes": 663818, |
| 394 | "nbytes": 2, |
| 395 | "sha256": "68a814a74f17b8c0617aa421b2e71891ed7f1a13524432d5c07d9389932c1bc4" |
| 396 | }, |
| 397 | { |
| 398 | "name": "frame_blocks.2.net.0.bias", |
| 399 | "shape": [ |
| 400 | 96 |
| 401 | ], |
| 402 | "dtype": "float16", |
| 403 | "offset_bytes": 663820, |
| 404 | "nbytes": 192, |
| 405 | "sha256": "85e2c52d26052ae2338376c1e930df86bc89db416bf0fa3def1f0f3e3f7ce943" |
| 406 | }, |
| 407 | { |
| 408 | "name": "frame_blocks.2.net.0.weight", |
| 409 | "shape": [ |
| 410 | 96, |
| 411 | 96, |
| 412 | 5 |
| 413 | ], |
| 414 | "dtype": "float16", |
| 415 | "offset_bytes": 664012, |
| 416 | "nbytes": 92160, |
| 417 | "sha256": "20b263ff883749c714a57805b594d4696274fff7fc91eedd4104e809d1783d13" |
| 418 | }, |
| 419 | { |
| 420 | "name": "frame_blocks.2.net.2.bias", |
| 421 | "shape": [ |
| 422 | 96 |
| 423 | ], |
| 424 | "dtype": "float16", |
| 425 | "offset_bytes": 756172, |
| 426 | "nbytes": 192, |
| 427 | "sha256": "0d6796e662e1cfef7e153240e8d5249326dc87129eed72b5d73953df16618ad3" |
| 428 | }, |
| 429 | { |
| 430 | "name": "frame_blocks.2.net.2.weight", |
| 431 | "shape": [ |
| 432 | 96, |
| 433 | 96, |
| 434 | 5 |
| 435 | ], |
| 436 | "dtype": "float16", |
| 437 | "offset_bytes": 756364, |
| 438 | "nbytes": 92160, |
| 439 | "sha256": "de122d4f538b20e4ad208add93031bb88897df0be5ab56d8ec37b9b456ea9942" |
| 440 | }, |
| 441 | { |
| 442 | "name": "frame_blocks.2.scale", |
| 443 | "shape": [ |
| 444 | 1 |
| 445 | ], |
| 446 | "dtype": "float16", |
| 447 | "offset_bytes": 848524, |
| 448 | "nbytes": 2, |
| 449 | "sha256": "ad0961600ed3c5762ee0bad47000e0b6ffcc2b1cedaecfa9d328409c83a43197" |
| 450 | }, |
| 451 | { |
| 452 | "name": "frame_blocks.3.net.0.bias", |
| 453 | "shape": [ |
| 454 | 96 |
| 455 | ], |
| 456 | "dtype": "float16", |
| 457 | "offset_bytes": 848526, |
| 458 | "nbytes": 192, |
| 459 | "sha256": "d39d60d8fa81716448d0c0440d92fe0fad0428b99dc634d5ef82ae6b3712ad30" |
| 460 | }, |
| 461 | { |
| 462 | "name": "frame_blocks.3.net.0.weight", |
| 463 | "shape": [ |
| 464 | 96, |
| 465 | 96, |
| 466 | 5 |
| 467 | ], |
| 468 | "dtype": "float16", |
| 469 | "offset_bytes": 848718, |
| 470 | "nbytes": 92160, |
| 471 | "sha256": "623790f75065fa5ab0ee68c186560fa4d14c1104dca0b7d2c54e472904dbbf70" |
| 472 | }, |
| 473 | { |
| 474 | "name": "frame_blocks.3.net.2.bias", |
| 475 | "shape": [ |
| 476 | 96 |
| 477 | ], |
| 478 | "dtype": "float16", |
| 479 | "offset_bytes": 940878, |
| 480 | "nbytes": 192, |
| 481 | "sha256": "f8cfd8551b85434979adf95d678ed1e081aac8e6de0f4ef679168b26324a7faa" |
| 482 | }, |
| 483 | { |
| 484 | "name": "frame_blocks.3.net.2.weight", |
| 485 | "shape": [ |
| 486 | 96, |
| 487 | 96, |
| 488 | 5 |
| 489 | ], |
| 490 | "dtype": "float16", |
| 491 | "offset_bytes": 941070, |
| 492 | "nbytes": 92160, |
| 493 | "sha256": "344fa466bf052b9ac482f2d0545625d453d3fcfa460ef1b686ab5ba89a7c9484" |
| 494 | }, |
| 495 | { |
| 496 | "name": "frame_blocks.3.scale", |
| 497 | "shape": [ |
| 498 | 1 |
| 499 | ], |
| 500 | "dtype": "float16", |
| 501 | "offset_bytes": 1033230, |
| 502 | "nbytes": 2, |
| 503 | "sha256": "71b8f10cba2f9ed87ee5d358ffe7e99663012c04a7c7d63f69e455013f514f0e" |
| 504 | }, |
| 505 | { |
| 506 | "name": "frame_input_proj.bias", |
| 507 | "shape": [ |
| 508 | 96 |
| 509 | ], |
| 510 | "dtype": "float16", |
| 511 | "offset_bytes": 1033232, |
| 512 | "nbytes": 192, |
| 513 | "sha256": "17ad81093cce62818c2cfd29ff26d115604325b123d17ecc992c8b566c0ae272" |
| 514 | }, |
| 515 | { |
| 516 | "name": "frame_input_proj.weight", |
| 517 | "shape": [ |
| 518 | 96, |
| 519 | 99, |
| 520 | 1 |
| 521 | ], |
| 522 | "dtype": "float16", |
| 523 | "offset_bytes": 1033424, |
| 524 | "nbytes": 19008, |
| 525 | "sha256": "fde4b4dc88bf36130d7160f8cbe0fc8b23729e252297444858344aa1ffb71cca" |
| 526 | }, |
| 527 | { |
| 528 | "name": "output.bias", |
| 529 | "shape": [ |
| 530 | 192 |
| 531 | ], |
| 532 | "dtype": "float16", |
| 533 | "offset_bytes": 1052432, |
| 534 | "nbytes": 384, |
| 535 | "sha256": "1319b9c004477ae3ac575edceb1d9f72a2d7bb0a2764ff3c8421872649a9924d" |
| 536 | }, |
| 537 | { |
| 538 | "name": "output.weight", |
| 539 | "shape": [ |
| 540 | 192, |
| 541 | 96, |
| 542 | 1 |
| 543 | ], |
| 544 | "dtype": "float16", |
| 545 | "offset_bytes": 1052816, |
| 546 | "nbytes": 36864, |
| 547 | "sha256": "a0bd205594b36b8ff713e744be58122c5e95ba9cb6486b5200daf48acb533bff" |
| 548 | }, |
| 549 | { |
| 550 | "name": "token_blocks.0.net.0.bias", |
| 551 | "shape": [ |
| 552 | 96 |
| 553 | ], |
| 554 | "dtype": "float16", |
| 555 | "offset_bytes": 1089680, |
| 556 | "nbytes": 192, |
| 557 | "sha256": "0142ba35c9ec65aa6d5a7cce7d002759b49a27a1d745221d65d9faad5464a7a7" |
| 558 | }, |
| 559 | { |
| 560 | "name": "token_blocks.0.net.0.weight", |
| 561 | "shape": [ |
| 562 | 96, |
| 563 | 96, |
| 564 | 5 |
| 565 | ], |
| 566 | "dtype": "float16", |
| 567 | "offset_bytes": 1089872, |
| 568 | "nbytes": 92160, |
| 569 | "sha256": "63777a784a254882831a3b496a36659ade469923a2ed09a3ba33ef14ccb9b1e1" |
| 570 | }, |
| 571 | { |
| 572 | "name": "token_blocks.0.net.2.bias", |
| 573 | "shape": [ |
| 574 | 96 |
| 575 | ], |
| 576 | "dtype": "float16", |
| 577 | "offset_bytes": 1182032, |
| 578 | "nbytes": 192, |
| 579 | "sha256": "3d84d43d9843454009090e2c34f44093abbc409937b2dde78f7fdbd10147f21f" |
| 580 | }, |
| 581 | { |
| 582 | "name": "token_blocks.0.net.2.weight", |
| 583 | "shape": [ |
| 584 | 96, |
| 585 | 96, |
| 586 | 5 |
| 587 | ], |
| 588 | "dtype": "float16", |
| 589 | "offset_bytes": 1182224, |
| 590 | "nbytes": 92160, |
| 591 | "sha256": "ae01e2726737562a5821f08fc8a17fd49429b93dabf660c294ede84259d988ef" |
| 592 | }, |
| 593 | { |
| 594 | "name": "token_blocks.0.scale", |
| 595 | "shape": [ |
| 596 | 1 |
| 597 | ], |
| 598 | "dtype": "float16", |
| 599 | "offset_bytes": 1274384, |
| 600 | "nbytes": 2, |
| 601 | "sha256": "99d4986930dec7f3a5a7912ed2ef5bae945623c54a7672299264f3b35ba209fe" |
| 602 | }, |
| 603 | { |
| 604 | "name": "token_blocks.1.net.0.bias", |
| 605 | "shape": [ |
| 606 | 96 |
| 607 | ], |
| 608 | "dtype": "float16", |
| 609 | "offset_bytes": 1274386, |
| 610 | "nbytes": 192, |
| 611 | "sha256": "c689e746c2bc7282c91561110206c860e5219c03fe1ecc3199e57a8f2a8bf6e6" |
| 612 | }, |
| 613 | { |
| 614 | "name": "token_blocks.1.net.0.weight", |
| 615 | "shape": [ |
| 616 | 96, |
| 617 | 96, |
| 618 | 5 |
| 619 | ], |
| 620 | "dtype": "float16", |
| 621 | "offset_bytes": 1274578, |
| 622 | "nbytes": 92160, |
| 623 | "sha256": "25c2607216201a9ac41309b568cf220efb3cf31a379a23e0a7373bd3856255d7" |
| 624 | }, |
| 625 | { |
| 626 | "name": "token_blocks.1.net.2.bias", |
| 627 | "shape": [ |
| 628 | 96 |
| 629 | ], |
| 630 | "dtype": "float16", |
| 631 | "offset_bytes": 1366738, |
| 632 | "nbytes": 192, |
| 633 | "sha256": "6e459849010d7e20fc2985008c3d62678bd77792c98d7adec70ea59cb1b5dbf8" |
| 634 | }, |
| 635 | { |
| 636 | "name": "token_blocks.1.net.2.weight", |
| 637 | "shape": [ |
| 638 | 96, |
| 639 | 96, |
| 640 | 5 |
| 641 | ], |
| 642 | "dtype": "float16", |
| 643 | "offset_bytes": 1366930, |
| 644 | "nbytes": 92160, |
| 645 | "sha256": "9bd71ee4027867ad6efadc33609abc5fb0783b8e6b513f23bf4729ce7a72772f" |
| 646 | }, |
| 647 | { |
| 648 | "name": "token_blocks.1.scale", |
| 649 | "shape": [ |
| 650 | 1 |
| 651 | ], |
| 652 | "dtype": "float16", |
| 653 | "offset_bytes": 1459090, |
| 654 | "nbytes": 2, |
| 655 | "sha256": "b239265bcb8ba182c1833cf0367e33b428a32405b62bd90189f49010a1137ba2" |
| 656 | }, |
| 657 | { |
| 658 | "name": "token_blocks.2.net.0.bias", |
| 659 | "shape": [ |
| 660 | 96 |
| 661 | ], |
| 662 | "dtype": "float16", |
| 663 | "offset_bytes": 1459092, |
| 664 | "nbytes": 192, |
| 665 | "sha256": "974aba49fbe8922f41d5a1937aa863853efbfc2bcef1efd9786df356d46d24a3" |
| 666 | }, |
| 667 | { |
| 668 | "name": "token_blocks.2.net.0.weight", |
| 669 | "shape": [ |
| 670 | 96, |
| 671 | 96, |
| 672 | 5 |
| 673 | ], |
| 674 | "dtype": "float16", |
| 675 | "offset_bytes": 1459284, |
| 676 | "nbytes": 92160, |
| 677 | "sha256": "661574a262a51f631f9cb6e95b06e126fb5a9ab50bbf96ac3f1db8789a2eb0ee" |
| 678 | }, |
| 679 | { |
| 680 | "name": "token_blocks.2.net.2.bias", |
| 681 | "shape": [ |
| 682 | 96 |
| 683 | ], |
| 684 | "dtype": "float16", |
| 685 | "offset_bytes": 1551444, |
| 686 | "nbytes": 192, |
| 687 | "sha256": "54aa32514ebffdac8033cc1feb0f403fc4fad96394b6219dcd083fc2d17fbdbe" |
| 688 | }, |
| 689 | { |
| 690 | "name": "token_blocks.2.net.2.weight", |
| 691 | "shape": [ |
| 692 | 96, |
| 693 | 96, |
| 694 | 5 |
| 695 | ], |
| 696 | "dtype": "float16", |
| 697 | "offset_bytes": 1551636, |
| 698 | "nbytes": 92160, |
| 699 | "sha256": "fce28049a88dfdefec139a874ba118fa57974a40b1ed526eeb19e9e370d19b90" |
| 700 | }, |
| 701 | { |
| 702 | "name": "token_blocks.2.scale", |
| 703 | "shape": [ |
| 704 | 1 |
| 705 | ], |
| 706 | "dtype": "float16", |
| 707 | "offset_bytes": 1643796, |
| 708 | "nbytes": 2, |
| 709 | "sha256": "ae199b5a1c746d3fc448f5b96aea00666a8015e1dfd75408dea28eb4f17d668a" |
| 710 | }, |
| 711 | { |
| 712 | "name": "token_input_proj.bias", |
| 713 | "shape": [ |
| 714 | 96 |
| 715 | ], |
| 716 | "dtype": "float16", |
| 717 | "offset_bytes": 1643798, |
| 718 | "nbytes": 192, |
| 719 | "sha256": "f073228567c9d4f558885c23d3c28aff30863cdc7f510950e1a8573171236393" |
| 720 | }, |
| 721 | { |
| 722 | "name": "token_input_proj.weight", |
| 723 | "shape": [ |
| 724 | 96, |
| 725 | 98, |
| 726 | 1 |
| 727 | ], |
| 728 | "dtype": "float16", |
| 729 | "offset_bytes": 1643990, |
| 730 | "nbytes": 18816, |
| 731 | "sha256": "bb5ceafec148c1390001b7a826af18af5e5bfb1820f87ad7ec1edd1565545658" |
| 732 | } |
| 733 | ] |
| 734 | }, |
| 735 | "decoder": { |
| 736 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/id_ID-news_tts-medium-retrain-20260712/a93-joint/decoder-student.pt", |
| 737 | "checkpoint_sha256": "c62ebf6e599a4687d9cb43456a68526a9032d098f8640bb8af51a68a5e8b4c4b", |
| 738 | "config": { |
| 739 | "in_channels": 192, |
| 740 | "source_in_channels": 192, |
| 741 | "channels": [ |
| 742 | 160, |
| 743 | 80, |
| 744 | 40, |
| 745 | 20 |
| 746 | ], |
| 747 | "res_layers": 1, |
| 748 | "variant": "piperlite", |
| 749 | "lrc_code_dim": 40, |
| 750 | "lrc_encoder_hidden": 64, |
| 751 | "rank_ratio": 0.5, |
| 752 | "activation": "leaky_relu", |
| 753 | "stage_affine": false, |
| 754 | "factorized_pre_rank": 0, |
| 755 | "piper_res_factor_rank_ratio": 0.0, |
| 756 | "res_bank_scale_mode": "kept", |
| 757 | "stage0_branches": [ |
| 758 | 0, |
| 759 | 1, |
| 760 | 2 |
| 761 | ], |
| 762 | "stage1_branches": [ |
| 763 | 0, |
| 764 | 1, |
| 765 | 2 |
| 766 | ], |
| 767 | "stage2_branches": [ |
| 768 | 0, |
| 769 | 1, |
| 770 | 2 |
| 771 | ], |
| 772 | "stage3_branches": [ |
| 773 | 0, |
| 774 | 1, |
| 775 | 2 |
| 776 | ], |
| 777 | "fsd_dim": 72, |
| 778 | "fsd_blocks": 5, |
| 779 | "fsd_film_rank": 12, |
| 780 | "fsd_head_rank": 48, |
| 781 | "stage_projection_bottlenecks": [], |
| 782 | "teacher_init_checkpoint": null, |
| 783 | "teacher_init_method": "first", |
| 784 | "teacher_init": null, |
| 785 | "init_decoder_checkpoint": "artifacts/sub10m-search/root-a-piper-vits/id_ID-news_tts-medium-retrain-20260712/a91-decoder-piperlite/decoder-student.pt", |
| 786 | "decoder_init": { |
| 787 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/id_ID-news_tts-medium-retrain-20260712/a91-decoder-piperlite/decoder-student.pt", |
| 788 | "decoder_parameters": 730721, |
| 789 | "matched_config": { |
| 790 | "in_channels": 192, |
| 791 | "channels": [ |
| 792 | 160, |
| 793 | 80, |
| 794 | 40, |
| 795 | 20 |
| 796 | ], |
| 797 | "res_layers": 1, |
| 798 | "variant": "piperlite", |
| 799 | "rank_ratio": 0.5, |
| 800 | "activation": "leaky_relu", |
| 801 | "stage_affine": false, |
| 802 | "factorized_pre_rank": 0, |
| 803 | "piper_res_factor_rank_ratio": 0.0, |
| 804 | "res_bank_scale_mode": "kept", |
| 805 | "stage0_branches": [ |
| 806 | 0, |
| 807 | 1, |
| 808 | 2 |
| 809 | ], |
| 810 | "stage1_branches": [ |
| 811 | 0, |
| 812 | 1, |
| 813 | 2 |
| 814 | ], |
| 815 | "stage2_branches": [ |
| 816 | 0, |
| 817 | 1, |
| 818 | 2 |
| 819 | ], |
| 820 | "stage3_branches": [ |
| 821 | 0, |
| 822 | 1, |
| 823 | 2 |
| 824 | ], |
| 825 | "fsd_dim": 72, |
| 826 | "fsd_blocks": 5, |
| 827 | "fsd_film_rank": 12, |
| 828 | "fsd_head_rank": 48, |
| 829 | "stage_projection_bottlenecks": [], |
| 830 | "post_filter_channels": 0, |
| 831 | "post_filter_layers": 0, |
| 832 | "post_filter_kernel": 9, |
| 833 | "post_filter_scale": 0.25, |
| 834 | "pre_tanh_repair_channels": 0, |
| 835 | "pre_tanh_repair_layers": 0, |
| 836 | "pre_tanh_repair_kernel": 7, |
| 837 | "pre_tanh_repair_scale": 0.15, |
| 838 | "istft_n_fft": 512 |
| 839 | }, |
| 840 | "allowed_post_filter_mismatches": [], |
| 841 | "allowed_pre_tanh_repair_mismatches": [], |
| 842 | "allowed_activation_mismatches": [], |
| 843 | "missing_keys": [], |
| 844 | "unexpected_keys": [] |
| 845 | }, |
| 846 | "post_filter_channels": 0, |
| 847 | "post_filter_layers": 0, |
| 848 | "post_filter_kernel": 9, |
| 849 | "post_filter_scale": 0.25, |
| 850 | "pre_tanh_repair_channels": 0, |
| 851 | "pre_tanh_repair_layers": 0, |
| 852 | "pre_tanh_repair_kernel": 7, |
| 853 | "pre_tanh_repair_scale": 0.15, |
| 854 | "istft_n_fft": 512, |
| 855 | "ap_amplitude_weight": 0.0, |
| 856 | "ap_phase_weight": 0.0, |
| 857 | "ap_complex_weight": 0.0, |
| 858 | "spectral_head_init": "default", |
| 859 | "spectral_head_init_scale": 0.001, |
| 860 | "ap_amp_init_bias": -6.0, |
| 861 | "ap_phase_real_init_bias": 1.0, |
| 862 | "spectral_head_init_summary": null, |
| 863 | "feature_exact_weight": 0.0, |
| 864 | "feature_exact_keys": [ |
| 865 | "pre", |
| 866 | "up0" |
| 867 | ], |
| 868 | "feature_exact_teacher_init": null, |
| 869 | "signature_pack_dir": "artifacts/sub10m-search/root-a-piper-vits/id_ID-news_tts-medium-retrain-20260712/a26-signatures-train512", |
| 870 | "exact_feature_pack_dir": null, |
| 871 | "signature_hint_weight": 0.05, |
| 872 | "signature_temporal_weight": 0.4, |
| 873 | "signature_phase_weight": 0.0, |
| 874 | "signature_phase_bins": 0, |
| 875 | "signature_keys": [ |
| 876 | "stage0_mix", |
| 877 | "stage1_mix", |
| 878 | "stage2_mix", |
| 879 | "pre_tanh", |
| 880 | "audio" |
| 881 | ], |
| 882 | "bottleneck_code_checkpoint": null, |
| 883 | "bottleneck_code_weight": 0.0, |
| 884 | "bottleneck_code_keys": [], |
| 885 | "bottleneck_code_training_only_parameters": 0, |
| 886 | "echo_tail_weight": 0.0, |
| 887 | "echo_tail_min_ms": 8.0, |
| 888 | "echo_tail_max_ms": 40.0, |
| 889 | "echo_tail_lags": 9, |
| 890 | "echo_tail_margin": 0.02, |
| 891 | "paired_acoustic_residual_weight": 0.0, |
| 892 | "paired_acoustic_residual_max_scale": 0.25, |
| 893 | "lrc_pred_code_mix_prob": 0.0, |
| 894 | "lrc_pred_code_checkpoint": null, |
| 895 | "lrc_pred_code_residual_prob": 0.0, |
| 896 | "lrc_pred_code_residual_max_scale": 0.25, |
| 897 | "lrc_pred_code_cache": null |
| 898 | }, |
| 899 | "parameters": 730721, |
| 900 | "tensors": [ |
| 901 | { |
| 902 | "name": "post.bias", |
| 903 | "shape": [ |
| 904 | 1 |
| 905 | ], |
| 906 | "dtype": "float16", |
| 907 | "offset_bytes": 1662806, |
| 908 | "nbytes": 2, |
| 909 | "sha256": "8d6dd9f330421e48e73e97c0819d55a9965bac90b1f2f36057b4c906d5791f05" |
| 910 | }, |
| 911 | { |
| 912 | "name": "post.weight", |
| 913 | "shape": [ |
| 914 | 1, |
| 915 | 20, |
| 916 | 7 |
| 917 | ], |
| 918 | "dtype": "float16", |
| 919 | "offset_bytes": 1662808, |
| 920 | "nbytes": 280, |
| 921 | "sha256": "95e795bb3446b9cecd8680730cba8a58879fc71a8398a87f3c15eb691f0d436b" |
| 922 | }, |
| 923 | { |
| 924 | "name": "pre.bias", |
| 925 | "shape": [ |
| 926 | 160 |
| 927 | ], |
| 928 | "dtype": "float16", |
| 929 | "offset_bytes": 1663088, |
| 930 | "nbytes": 320, |
| 931 | "sha256": "9f5780bcdbfac77dad36cd8e4fc6f3181508fe4115d1a5d87aaa5339274f3db3" |
| 932 | }, |
| 933 | { |
| 934 | "name": "pre.weight", |
| 935 | "shape": [ |
| 936 | 160, |
| 937 | 192, |
| 938 | 7 |
| 939 | ], |
| 940 | "dtype": "float16", |
| 941 | "offset_bytes": 1663408, |
| 942 | "nbytes": 430080, |
| 943 | "sha256": "8eb3b78b5d8d3e1e6deebf21025a4a504e61acba400248d90de74f63a531c277" |
| 944 | }, |
| 945 | { |
| 946 | "name": "res0.0.blocks.0.conv1.bias", |
| 947 | "shape": [ |
| 948 | 80 |
| 949 | ], |
| 950 | "dtype": "float16", |
| 951 | "offset_bytes": 2093488, |
| 952 | "nbytes": 160, |
| 953 | "sha256": "6b5b3dd80927bfad818f454de2e2608b3d5b288175632a65cfdaccd49f9da1ed" |
| 954 | }, |
| 955 | { |
| 956 | "name": "res0.0.blocks.0.conv1.weight", |
| 957 | "shape": [ |
| 958 | 80, |
| 959 | 80, |
| 960 | 3 |
| 961 | ], |
| 962 | "dtype": "float16", |
| 963 | "offset_bytes": 2093648, |
| 964 | "nbytes": 38400, |
| 965 | "sha256": "340fd5ce89610a16e904c38a5034199e5f23ad5075fe04b03bbfbe1bf9900e3c" |
| 966 | }, |
| 967 | { |
| 968 | "name": "res0.0.blocks.0.conv2.bias", |
| 969 | "shape": [ |
| 970 | 80 |
| 971 | ], |
| 972 | "dtype": "float16", |
| 973 | "offset_bytes": 2132048, |
| 974 | "nbytes": 160, |
| 975 | "sha256": "3c59b75cdbbaf2d83e5cc29185ecfcce90f621baef2bb4b4f0dcd7a94af94e63" |
| 976 | }, |
| 977 | { |
| 978 | "name": "res0.0.blocks.0.conv2.weight", |
| 979 | "shape": [ |
| 980 | 80, |
| 981 | 80, |
| 982 | 3 |
| 983 | ], |
| 984 | "dtype": "float16", |
| 985 | "offset_bytes": 2132208, |
| 986 | "nbytes": 38400, |
| 987 | "sha256": "79cefadd8074b12f87b812458991ebde7ca3f630249bcb75c91c805deb906834" |
| 988 | }, |
| 989 | { |
| 990 | "name": "res0.0.blocks.1.conv1.bias", |
| 991 | "shape": [ |
| 992 | 80 |
| 993 | ], |
| 994 | "dtype": "float16", |
| 995 | "offset_bytes": 2170608, |
| 996 | "nbytes": 160, |
| 997 | "sha256": "6cf04972bc74e4398e6ef216dea52eb78b15c881740348ce420bc902643f762f" |
| 998 | }, |
| 999 | { |
| 1000 | "name": "res0.0.blocks.1.conv1.weight", |
| 1001 | "shape": [ |
| 1002 | 80, |
| 1003 | 80, |
| 1004 | 5 |
| 1005 | ], |
| 1006 | "dtype": "float16", |
| 1007 | "offset_bytes": 2170768, |
| 1008 | "nbytes": 64000, |
| 1009 | "sha256": "5022d23f5742cd1adc84d91dc865c91a9b9ba5853e2276cb8b5c953798afa504" |
| 1010 | }, |
| 1011 | { |
| 1012 | "name": "res0.0.blocks.1.conv2.bias", |
| 1013 | "shape": [ |
| 1014 | 80 |
| 1015 | ], |
| 1016 | "dtype": "float16", |
| 1017 | "offset_bytes": 2234768, |
| 1018 | "nbytes": 160, |
| 1019 | "sha256": "2626718d2537de72ea266c826865a83648e13cd152ba6a3f59f326f8e0e2ef8d" |
| 1020 | }, |
| 1021 | { |
| 1022 | "name": "res0.0.blocks.1.conv2.weight", |
| 1023 | "shape": [ |
| 1024 | 80, |
| 1025 | 80, |
| 1026 | 5 |
| 1027 | ], |
| 1028 | "dtype": "float16", |
| 1029 | "offset_bytes": 2234928, |
| 1030 | "nbytes": 64000, |
| 1031 | "sha256": "cb644242e7439a06cf09a9be09d89ae290b292e08d42d994f30408902e11cadc" |
| 1032 | }, |
| 1033 | { |
| 1034 | "name": "res0.0.blocks.2.conv1.bias", |
| 1035 | "shape": [ |
| 1036 | 80 |
| 1037 | ], |
| 1038 | "dtype": "float16", |
| 1039 | "offset_bytes": 2298928, |
| 1040 | "nbytes": 160, |
| 1041 | "sha256": "e332940751a2cc7acdb95122b78da0cd2297cc236d1d6a456ee10e1ffea520ab" |
| 1042 | }, |
| 1043 | { |
| 1044 | "name": "res0.0.blocks.2.conv1.weight", |
| 1045 | "shape": [ |
| 1046 | 80, |
| 1047 | 80, |
| 1048 | 7 |
| 1049 | ], |
| 1050 | "dtype": "float16", |
| 1051 | "offset_bytes": 2299088, |
| 1052 | "nbytes": 89600, |
| 1053 | "sha256": "ffc5a169f2cdb5ab5b8ab3c26671ced864420199816c11e1ed3055589f0d7239" |
| 1054 | }, |
| 1055 | { |
| 1056 | "name": "res0.0.blocks.2.conv2.bias", |
| 1057 | "shape": [ |
| 1058 | 80 |
| 1059 | ], |
| 1060 | "dtype": "float16", |
| 1061 | "offset_bytes": 2388688, |
| 1062 | "nbytes": 160, |
| 1063 | "sha256": "a21b2ae06076c10f3a887f6473b270019b9012f287bb5e73a4bc1c491744a7d0" |
| 1064 | }, |
| 1065 | { |
| 1066 | "name": "res0.0.blocks.2.conv2.weight", |
| 1067 | "shape": [ |
| 1068 | 80, |
| 1069 | 80, |
| 1070 | 7 |
| 1071 | ], |
| 1072 | "dtype": "float16", |
| 1073 | "offset_bytes": 2388848, |
| 1074 | "nbytes": 89600, |
| 1075 | "sha256": "fd763f5b0b1ad8fc6e168e76a959b6f72d2cf4d24fedd3d780e478fb1c33222e" |
| 1076 | }, |
| 1077 | { |
| 1078 | "name": "res1.0.blocks.0.conv1.bias", |
| 1079 | "shape": [ |
| 1080 | 40 |
| 1081 | ], |
| 1082 | "dtype": "float16", |
| 1083 | "offset_bytes": 2478448, |
| 1084 | "nbytes": 80, |
| 1085 | "sha256": "a2455db7e4d7075b1a74f7622c26620203591973aa709a413e6e3838ca652469" |
| 1086 | }, |
| 1087 | { |
| 1088 | "name": "res1.0.blocks.0.conv1.weight", |
| 1089 | "shape": [ |
| 1090 | 40, |
| 1091 | 40, |
| 1092 | 3 |
| 1093 | ], |
| 1094 | "dtype": "float16", |
| 1095 | "offset_bytes": 2478528, |
| 1096 | "nbytes": 9600, |
| 1097 | "sha256": "e25f43d2cbea1a7636c0a2e83155b8aa35b61a87951467ab305e8821d451ada5" |
| 1098 | }, |
| 1099 | { |
| 1100 | "name": "res1.0.blocks.0.conv2.bias", |
| 1101 | "shape": [ |
| 1102 | 40 |
| 1103 | ], |
| 1104 | "dtype": "float16", |
| 1105 | "offset_bytes": 2488128, |
| 1106 | "nbytes": 80, |
| 1107 | "sha256": "eb10d59f6e127cbc87c803c7957df8e1dfee970782add70dea6e0c24899623c6" |
| 1108 | }, |
| 1109 | { |
| 1110 | "name": "res1.0.blocks.0.conv2.weight", |
| 1111 | "shape": [ |
| 1112 | 40, |
| 1113 | 40, |
| 1114 | 3 |
| 1115 | ], |
| 1116 | "dtype": "float16", |
| 1117 | "offset_bytes": 2488208, |
| 1118 | "nbytes": 9600, |
| 1119 | "sha256": "c05a14f7795e6d4436218d8983af872212707696245409d39f98300c14f55a92" |
| 1120 | }, |
| 1121 | { |
| 1122 | "name": "res1.0.blocks.1.conv1.bias", |
| 1123 | "shape": [ |
| 1124 | 40 |
| 1125 | ], |
| 1126 | "dtype": "float16", |
| 1127 | "offset_bytes": 2497808, |
| 1128 | "nbytes": 80, |
| 1129 | "sha256": "fc55bd6c17ca013dcf8d23ac546d27ec26dce04317f46e84c141347b79b02701" |
| 1130 | }, |
| 1131 | { |
| 1132 | "name": "res1.0.blocks.1.conv1.weight", |
| 1133 | "shape": [ |
| 1134 | 40, |
| 1135 | 40, |
| 1136 | 5 |
| 1137 | ], |
| 1138 | "dtype": "float16", |
| 1139 | "offset_bytes": 2497888, |
| 1140 | "nbytes": 16000, |
| 1141 | "sha256": "1b0a9cd6a14a4eaf19e80d32bfcc034b46ec1f9b752c5944d2050827f4498c95" |
| 1142 | }, |
| 1143 | { |
| 1144 | "name": "res1.0.blocks.1.conv2.bias", |
| 1145 | "shape": [ |
| 1146 | 40 |
| 1147 | ], |
| 1148 | "dtype": "float16", |
| 1149 | "offset_bytes": 2513888, |
| 1150 | "nbytes": 80, |
| 1151 | "sha256": "194542def7cc25f2e2378fac6a1a7044ce4962e4cc5ad8834c3b9fe631256529" |
| 1152 | }, |
| 1153 | { |
| 1154 | "name": "res1.0.blocks.1.conv2.weight", |
| 1155 | "shape": [ |
| 1156 | 40, |
| 1157 | 40, |
| 1158 | 5 |
| 1159 | ], |
| 1160 | "dtype": "float16", |
| 1161 | "offset_bytes": 2513968, |
| 1162 | "nbytes": 16000, |
| 1163 | "sha256": "c719a7832c9a37e0af66707228880fc19c3f87725ab5324597f9209ccb1f65fe" |
| 1164 | }, |
| 1165 | { |
| 1166 | "name": "res1.0.blocks.2.conv1.bias", |
| 1167 | "shape": [ |
| 1168 | 40 |
| 1169 | ], |
| 1170 | "dtype": "float16", |
| 1171 | "offset_bytes": 2529968, |
| 1172 | "nbytes": 80, |
| 1173 | "sha256": "918756f2bac8afa95e82cbc45a454aa845057aa2990ea72d3c8dbd052bb89234" |
| 1174 | }, |
| 1175 | { |
| 1176 | "name": "res1.0.blocks.2.conv1.weight", |
| 1177 | "shape": [ |
| 1178 | 40, |
| 1179 | 40, |
| 1180 | 7 |
| 1181 | ], |
| 1182 | "dtype": "float16", |
| 1183 | "offset_bytes": 2530048, |
| 1184 | "nbytes": 22400, |
| 1185 | "sha256": "bb267c7392e5b95ec427072e450f4c56a426fdd1449beae902234e3129e93aca" |
| 1186 | }, |
| 1187 | { |
| 1188 | "name": "res1.0.blocks.2.conv2.bias", |
| 1189 | "shape": [ |
| 1190 | 40 |
| 1191 | ], |
| 1192 | "dtype": "float16", |
| 1193 | "offset_bytes": 2552448, |
| 1194 | "nbytes": 80, |
| 1195 | "sha256": "27b511b1c651a391aefc5e20f23acd7f0c1fcac2198dbe702a771898c8033541" |
| 1196 | }, |
| 1197 | { |
| 1198 | "name": "res1.0.blocks.2.conv2.weight", |
| 1199 | "shape": [ |
| 1200 | 40, |
| 1201 | 40, |
| 1202 | 7 |
| 1203 | ], |
| 1204 | "dtype": "float16", |
| 1205 | "offset_bytes": 2552528, |
| 1206 | "nbytes": 22400, |
| 1207 | "sha256": "e65f3c7cd4b34c9466e99e55214577c29d57fa1ff81dd54a77c7c7b4d95cb132" |
| 1208 | }, |
| 1209 | { |
| 1210 | "name": "res2.0.blocks.0.conv1.bias", |
| 1211 | "shape": [ |
| 1212 | 20 |
| 1213 | ], |
| 1214 | "dtype": "float16", |
| 1215 | "offset_bytes": 2574928, |
| 1216 | "nbytes": 40, |
| 1217 | "sha256": "128b01fa275a89ae7eee75462301d1e9453f7753cfb150a9d08c82f52970a08f" |
| 1218 | }, |
| 1219 | { |
| 1220 | "name": "res2.0.blocks.0.conv1.weight", |
| 1221 | "shape": [ |
| 1222 | 20, |
| 1223 | 20, |
| 1224 | 3 |
| 1225 | ], |
| 1226 | "dtype": "float16", |
| 1227 | "offset_bytes": 2574968, |
| 1228 | "nbytes": 2400, |
| 1229 | "sha256": "40ce192ff5f0f82ecd61c1926b8d65cab27984ae3d3853df899aa6ccc47ff1df" |
| 1230 | }, |
| 1231 | { |
| 1232 | "name": "res2.0.blocks.0.conv2.bias", |
| 1233 | "shape": [ |
| 1234 | 20 |
| 1235 | ], |
| 1236 | "dtype": "float16", |
| 1237 | "offset_bytes": 2577368, |
| 1238 | "nbytes": 40, |
| 1239 | "sha256": "337eead432798de42fcd99cf288f1d3b0688a4c549e4e56ce51c25c467bdf7a9" |
| 1240 | }, |
| 1241 | { |
| 1242 | "name": "res2.0.blocks.0.conv2.weight", |
| 1243 | "shape": [ |
| 1244 | 20, |
| 1245 | 20, |
| 1246 | 3 |
| 1247 | ], |
| 1248 | "dtype": "float16", |
| 1249 | "offset_bytes": 2577408, |
| 1250 | "nbytes": 2400, |
| 1251 | "sha256": "3f475af26285df7d8ccd54af9a992c1e84de6438f8fd5e0cdc0bccc010458767" |
| 1252 | }, |
| 1253 | { |
| 1254 | "name": "res2.0.blocks.1.conv1.bias", |
| 1255 | "shape": [ |
| 1256 | 20 |
| 1257 | ], |
| 1258 | "dtype": "float16", |
| 1259 | "offset_bytes": 2579808, |
| 1260 | "nbytes": 40, |
| 1261 | "sha256": "e44080dffb229cf16ee7864947765de5958987087a58690d0b276a8b74fd7aff" |
| 1262 | }, |
| 1263 | { |
| 1264 | "name": "res2.0.blocks.1.conv1.weight", |
| 1265 | "shape": [ |
| 1266 | 20, |
| 1267 | 20, |
| 1268 | 5 |
| 1269 | ], |
| 1270 | "dtype": "float16", |
| 1271 | "offset_bytes": 2579848, |
| 1272 | "nbytes": 4000, |
| 1273 | "sha256": "da33d1cd128dbdb0dacbda8c5d1a31cefaa2f6a11a22af4ab654f02ee8392a6a" |
| 1274 | }, |
| 1275 | { |
| 1276 | "name": "res2.0.blocks.1.conv2.bias", |
| 1277 | "shape": [ |
| 1278 | 20 |
| 1279 | ], |
| 1280 | "dtype": "float16", |
| 1281 | "offset_bytes": 2583848, |
| 1282 | "nbytes": 40, |
| 1283 | "sha256": "36abb74c335fafb14fa1a2ba6fcfdc0fb18346b981d19d310cae6aadf0672a3c" |
| 1284 | }, |
| 1285 | { |
| 1286 | "name": "res2.0.blocks.1.conv2.weight", |
| 1287 | "shape": [ |
| 1288 | 20, |
| 1289 | 20, |
| 1290 | 5 |
| 1291 | ], |
| 1292 | "dtype": "float16", |
| 1293 | "offset_bytes": 2583888, |
| 1294 | "nbytes": 4000, |
| 1295 | "sha256": "4fbc9697cdd61bdd53d595aa7fff23e80f9fb518a808e9125ac530c983067b11" |
| 1296 | }, |
| 1297 | { |
| 1298 | "name": "res2.0.blocks.2.conv1.bias", |
| 1299 | "shape": [ |
| 1300 | 20 |
| 1301 | ], |
| 1302 | "dtype": "float16", |
| 1303 | "offset_bytes": 2587888, |
| 1304 | "nbytes": 40, |
| 1305 | "sha256": "6c252da0f62cb8f60885f0e2321452b2880777c38ba7fa9898d300a6e3b7045b" |
| 1306 | }, |
| 1307 | { |
| 1308 | "name": "res2.0.blocks.2.conv1.weight", |
| 1309 | "shape": [ |
| 1310 | 20, |
| 1311 | 20, |
| 1312 | 7 |
| 1313 | ], |
| 1314 | "dtype": "float16", |
| 1315 | "offset_bytes": 2587928, |
| 1316 | "nbytes": 5600, |
| 1317 | "sha256": "4dd0aed6f9dcb28519a2d12559928a0d5285bae6ea4f832f5b75b71a1865336e" |
| 1318 | }, |
| 1319 | { |
| 1320 | "name": "res2.0.blocks.2.conv2.bias", |
| 1321 | "shape": [ |
| 1322 | 20 |
| 1323 | ], |
| 1324 | "dtype": "float16", |
| 1325 | "offset_bytes": 2593528, |
| 1326 | "nbytes": 40, |
| 1327 | "sha256": "07c76f7bf048c80345614a6a4fd1c3cd1e89be3bec9b1c0692f67731ecbec55f" |
| 1328 | }, |
| 1329 | { |
| 1330 | "name": "res2.0.blocks.2.conv2.weight", |
| 1331 | "shape": [ |
| 1332 | 20, |
| 1333 | 20, |
| 1334 | 7 |
| 1335 | ], |
| 1336 | "dtype": "float16", |
| 1337 | "offset_bytes": 2593568, |
| 1338 | "nbytes": 5600, |
| 1339 | "sha256": "ef8e9b1b74372e3067cbaa73dbb5feca903beb05b67cfe34b8a27fb48b6b2975" |
| 1340 | }, |
| 1341 | { |
| 1342 | "name": "up0.bias", |
| 1343 | "shape": [ |
| 1344 | 80 |
| 1345 | ], |
| 1346 | "dtype": "float16", |
| 1347 | "offset_bytes": 2599168, |
| 1348 | "nbytes": 160, |
| 1349 | "sha256": "9bf76fd4316d660ad7cc0792ea32048525d6e5f0b004c0f531f55c12374046b5" |
| 1350 | }, |
| 1351 | { |
| 1352 | "name": "up0.weight", |
| 1353 | "shape": [ |
| 1354 | 160, |
| 1355 | 80, |
| 1356 | 16 |
| 1357 | ], |
| 1358 | "dtype": "float16", |
| 1359 | "offset_bytes": 2599328, |
| 1360 | "nbytes": 409600, |
| 1361 | "sha256": "cf6741eeef527ad746c8c10ae0ba452ff231e6bb0b2344722ad2fe06c2b8b7ed" |
| 1362 | }, |
| 1363 | { |
| 1364 | "name": "up1.bias", |
| 1365 | "shape": [ |
| 1366 | 40 |
| 1367 | ], |
| 1368 | "dtype": "float16", |
| 1369 | "offset_bytes": 3008928, |
| 1370 | "nbytes": 80, |
| 1371 | "sha256": "41a10a67426484d7fdfd461fb095314c677c5faed6bac488b7c8cef23b1ec349" |
| 1372 | }, |
| 1373 | { |
| 1374 | "name": "up1.weight", |
| 1375 | "shape": [ |
| 1376 | 80, |
| 1377 | 40, |
| 1378 | 16 |
| 1379 | ], |
| 1380 | "dtype": "float16", |
| 1381 | "offset_bytes": 3009008, |
| 1382 | "nbytes": 102400, |
| 1383 | "sha256": "84d50f35022a24f18af6b8e02affeddbce9b780a53a2e62a310e9669a42c5486" |
| 1384 | }, |
| 1385 | { |
| 1386 | "name": "up2.bias", |
| 1387 | "shape": [ |
| 1388 | 20 |
| 1389 | ], |
| 1390 | "dtype": "float16", |
| 1391 | "offset_bytes": 3111408, |
| 1392 | "nbytes": 40, |
| 1393 | "sha256": "58136944d94e6e9caead6ce59b54bb045d2cef9f293df4822f2a7b9e99728e0a" |
| 1394 | }, |
| 1395 | { |
| 1396 | "name": "up2.weight", |
| 1397 | "shape": [ |
| 1398 | 40, |
| 1399 | 20, |
| 1400 | 8 |
| 1401 | ], |
| 1402 | "dtype": "float16", |
| 1403 | "offset_bytes": 3111448, |
| 1404 | "nbytes": 12800, |
| 1405 | "sha256": "1a67966aa1e3f326492629a52926c7982808fa6b540767799d98077bb1438a71" |
| 1406 | } |
| 1407 | ] |
| 1408 | } |
| 1409 | }, |
| 1410 | "runtime_ops": [ |
| 1411 | "embedding_lookup", |
| 1412 | "conv1d_same", |
| 1413 | "conv1d_1x1", |
| 1414 | "conv_transpose1d", |
| 1415 | "silu", |
| 1416 | "leaky_relu", |
| 1417 | "tanh", |
| 1418 | "residual_add", |
| 1419 | "repeat_interleave", |
| 1420 | "masked_conv1d", |
| 1421 | "depthwise_conv1d", |
| 1422 | "low_rank_linear_or_conv1d" |
| 1423 | ], |
| 1424 | "runtime_boundary": { |
| 1425 | "neural_stack_self_contained": true, |
| 1426 | "arbitrary_text_self_contained": false, |
| 1427 | "accepted_runtime_input": "Piper phoneme ID sequence matching piper-phoneme-config.json", |
| 1428 | "output": "mono float waveform at manifest.sample_rate" |
| 1429 | }, |
| 1430 | "frontend": { |
| 1431 | "status": "not fully self-contained for arbitrary text yet", |
| 1432 | "input_contract": "runtime may accept Piper phoneme ID sequences directly", |
| 1433 | "included_config": "piper-phoneme-config.json", |
| 1434 | "piper_phoneme_type": "espeak", |
| 1435 | "piper_espeak_voice": "id", |
| 1436 | "phoneme_id_map_entries": 161, |
| 1437 | "next_work": "embed or rewrite the Piper/eSpeak-compatible frontend, or keep a strict phoneme-ID API" |
| 1438 | }, |
| 1439 | "size_budget": { |
| 1440 | "target_max_bytes": 4000000, |
| 1441 | "target_max_mib": 3.814697265625, |
| 1442 | "actual_model_asset_bytes": 3129298, |
| 1443 | "target_max_parameters": 1500000, |
| 1444 | "under_parameter_budget": false, |
| 1445 | "under_4mb_asset_budget": true |
| 1446 | }, |
| 1447 | "limitations": [ |
| 1448 | "This is a raw neural asset package, not a compiled standalone runtime binary.", |
| 1449 | "No PyTorch checkpoint, ONNX graph, or training state is included.", |
| 1450 | "Arbitrary text still requires a Piper/eSpeak-compatible frontend outside this package." |
| 1451 | ] |
| 1452 | } |
| 1453 | |