ro-mihai-1p57m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "ro_RO-flores-20260908", |
| 4 | "language": "ro_RO", |
| 5 | "voice": "ro_RO-mihai-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": 3130328, |
| 15 | "weights_sha256": "670228abd8e93df489362660e32b3f5fb4cbc3984af263aa236cd16c690da52c", |
| 16 | "total_parameters": 1565164, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/ro_RO/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "d7f506679834e2304ad4b52bd53714ee2eb7334830a93a56f24e0be09c5158d5", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 141, |
| 24 | "hidden": 64, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 671, |
| 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": 136708, |
| 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": "3fad775f67c55f5cbb0dc3b0042d4c20f3e5c009682e1d9a44666400fdf28625" |
| 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": "77c279834f4375ef2af6799fb8177fc063791824a697a50b9488b7da27a0943e" |
| 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": "985d7ad920e31cbd9685cedfa45036c8f59675345586e6c57d724e284d868d73" |
| 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": "ba763602ab6b68743823fd242508e5e48476f8cac18f37bcb421bee74be2fd63" |
| 90 | }, |
| 91 | { |
| 92 | "name": "blocks.0.scale", |
| 93 | "shape": [ |
| 94 | 1 |
| 95 | ], |
| 96 | "dtype": "float16", |
| 97 | "offset_bytes": 82176, |
| 98 | "nbytes": 2, |
| 99 | "sha256": "71e7690959239ca065841eba3ebb281072baa78ba0bb31079b9acb4a009a9fe3" |
| 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": "25c618ed4991471df0ae97a27ad93140ac5961f671121ae9d0c9fc8446e0ba42" |
| 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": "900e1d41c00b0701504630f1ad3ec5eef8df242d3e69f53e3a345155715d80a3" |
| 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": "e2f39afabcea0c87a4b8fa326ecb3943c287a8f6341a54239e1e6ffaeb8ffe13" |
| 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": "c47fa7be8c03f4a778f4fcbf70036d504bc1e63068455aee33f49a186dd6396e" |
| 144 | }, |
| 145 | { |
| 146 | "name": "blocks.1.scale", |
| 147 | "shape": [ |
| 148 | 1 |
| 149 | ], |
| 150 | "dtype": "float16", |
| 151 | "offset_bytes": 164354, |
| 152 | "nbytes": 2, |
| 153 | "sha256": "05391a26b7203c55d411f921b934117a61e9798d5937878a68bfa2904fe11215" |
| 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": "273b483c8ab8c7dcf631224e722c8062084edf9075ddb62c9ecf5869237d5989" |
| 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": "9ea0a9dbab0810868766a0bb24e38f812e83aadd9d63cbb723006963c91b761c" |
| 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": "957f388d67100ab43ffb9821565c5b3f60508f32915e7675aaacc1de971c1df0" |
| 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": "30a85dbefad92bbde267553ed431a0f4dbda9c22c06aef940d3161ecd4e0892a" |
| 198 | }, |
| 199 | { |
| 200 | "name": "blocks.2.scale", |
| 201 | "shape": [ |
| 202 | 1 |
| 203 | ], |
| 204 | "dtype": "float16", |
| 205 | "offset_bytes": 246532, |
| 206 | "nbytes": 2, |
| 207 | "sha256": "4e1195df020de59e0d65a33a4279f1183e7ae4e5d980e309f8b55adff2e61c3e" |
| 208 | }, |
| 209 | { |
| 210 | "name": "embedding.weight", |
| 211 | "shape": [ |
| 212 | 141, |
| 213 | 64 |
| 214 | ], |
| 215 | "dtype": "float16", |
| 216 | "offset_bytes": 246534, |
| 217 | "nbytes": 18048, |
| 218 | "sha256": "0544dd048a3ffc1cbc4da8ecaad6f532248a5e73c5c08341c59f9663bc30d44d" |
| 219 | }, |
| 220 | { |
| 221 | "name": "input_proj.bias", |
| 222 | "shape": [ |
| 223 | 64 |
| 224 | ], |
| 225 | "dtype": "float16", |
| 226 | "offset_bytes": 264582, |
| 227 | "nbytes": 128, |
| 228 | "sha256": "80e7cac225320148dbeb0d97e1b85f42724e098d9fc50108ff91b3267cb4c57f" |
| 229 | }, |
| 230 | { |
| 231 | "name": "input_proj.weight", |
| 232 | "shape": [ |
| 233 | 64, |
| 234 | 67, |
| 235 | 1 |
| 236 | ], |
| 237 | "dtype": "float16", |
| 238 | "offset_bytes": 264710, |
| 239 | "nbytes": 8576, |
| 240 | "sha256": "56794914a582dc0ef4185f85b4480408fea1a7b59f6342f07f5e2a8d4e6a1a6e" |
| 241 | }, |
| 242 | { |
| 243 | "name": "output.bias", |
| 244 | "shape": [ |
| 245 | 1 |
| 246 | ], |
| 247 | "dtype": "float16", |
| 248 | "offset_bytes": 273286, |
| 249 | "nbytes": 2, |
| 250 | "sha256": "300edaeed889419132fedf47646b190d17eda4c73db596c1ef9d14fc3aedb2b0" |
| 251 | }, |
| 252 | { |
| 253 | "name": "output.weight", |
| 254 | "shape": [ |
| 255 | 1, |
| 256 | 64, |
| 257 | 1 |
| 258 | ], |
| 259 | "dtype": "float16", |
| 260 | "offset_bytes": 273288, |
| 261 | "nbytes": 128, |
| 262 | "sha256": "d63c6d356051fb21e75429566cd480cfe397e680a3ce39c9ec4b16c957980a96" |
| 263 | } |
| 264 | ] |
| 265 | }, |
| 266 | "acoustic": { |
| 267 | "checkpoint": "artifacts/voices/ro_RO/joint/latent-student.pt", |
| 268 | "checkpoint_sha256": "e3b9dd7d863539deb445506aef9d1f8d7e2c44b9c95330d74c6188895a8f14ae", |
| 269 | "config": { |
| 270 | "architecture": "token_context", |
| 271 | "vocab_size": 141, |
| 272 | "hidden": 96, |
| 273 | "depth": 4, |
| 274 | "kernel_size": 5, |
| 275 | "out_channels": 192, |
| 276 | "token_depth": 3 |
| 277 | }, |
| 278 | "parameters": 697735, |
| 279 | "tensors": [ |
| 280 | { |
| 281 | "name": "embedding.weight", |
| 282 | "shape": [ |
| 283 | 141, |
| 284 | 96 |
| 285 | ], |
| 286 | "dtype": "float16", |
| 287 | "offset_bytes": 273416, |
| 288 | "nbytes": 27072, |
| 289 | "sha256": "665ab33571a4c56824496b64ef9dbf728020e049369fe64e10311f4f3fbafe0e" |
| 290 | }, |
| 291 | { |
| 292 | "name": "frame_blocks.0.net.0.bias", |
| 293 | "shape": [ |
| 294 | 96 |
| 295 | ], |
| 296 | "dtype": "float16", |
| 297 | "offset_bytes": 300488, |
| 298 | "nbytes": 192, |
| 299 | "sha256": "901527da4fc6d07b1037363a5278e0656929a0fca7b336ede68e21986635f1d6" |
| 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": 300680, |
| 310 | "nbytes": 92160, |
| 311 | "sha256": "94ff496b9220a5f8cb0c7acc515cb290d0d7930817f811e292e0683982b0b453" |
| 312 | }, |
| 313 | { |
| 314 | "name": "frame_blocks.0.net.2.bias", |
| 315 | "shape": [ |
| 316 | 96 |
| 317 | ], |
| 318 | "dtype": "float16", |
| 319 | "offset_bytes": 392840, |
| 320 | "nbytes": 192, |
| 321 | "sha256": "a7e5cdb71c46c655cc53648b7f6b5acfcd691112f935c2ac90e1dae6128610dd" |
| 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": 393032, |
| 332 | "nbytes": 92160, |
| 333 | "sha256": "e39d9b1c9fbd8d9007224203ebcab5b0fe1ba9ff3bbf3dea050a7e15c321762b" |
| 334 | }, |
| 335 | { |
| 336 | "name": "frame_blocks.0.scale", |
| 337 | "shape": [ |
| 338 | 1 |
| 339 | ], |
| 340 | "dtype": "float16", |
| 341 | "offset_bytes": 485192, |
| 342 | "nbytes": 2, |
| 343 | "sha256": "f053f60f4cd08a493906fee929aebf6136fc87791715a071b9ddffbd9082a135" |
| 344 | }, |
| 345 | { |
| 346 | "name": "frame_blocks.1.net.0.bias", |
| 347 | "shape": [ |
| 348 | 96 |
| 349 | ], |
| 350 | "dtype": "float16", |
| 351 | "offset_bytes": 485194, |
| 352 | "nbytes": 192, |
| 353 | "sha256": "e39d569a8e9ede22dab7c674fb8b540d653957551eb17fe911caf09e26c40b05" |
| 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": 485386, |
| 364 | "nbytes": 92160, |
| 365 | "sha256": "03014239c507363c8655b057d365d13cfed7b8bd905f8a093e5611aa0389ee00" |
| 366 | }, |
| 367 | { |
| 368 | "name": "frame_blocks.1.net.2.bias", |
| 369 | "shape": [ |
| 370 | 96 |
| 371 | ], |
| 372 | "dtype": "float16", |
| 373 | "offset_bytes": 577546, |
| 374 | "nbytes": 192, |
| 375 | "sha256": "cc3ad5ca29c2e60f5b6dcf6982daa3688d80eba8c5543a439591da136986f6fd" |
| 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": 577738, |
| 386 | "nbytes": 92160, |
| 387 | "sha256": "0f4ef883430221d7176144e21af756251394adbea20ad5583a8c058ae2ee258d" |
| 388 | }, |
| 389 | { |
| 390 | "name": "frame_blocks.1.scale", |
| 391 | "shape": [ |
| 392 | 1 |
| 393 | ], |
| 394 | "dtype": "float16", |
| 395 | "offset_bytes": 669898, |
| 396 | "nbytes": 2, |
| 397 | "sha256": "7c7e4cb01443e1018ebc6d078b72498b3e3bd6135c00d79d12ffa372a8a4f26f" |
| 398 | }, |
| 399 | { |
| 400 | "name": "frame_blocks.2.net.0.bias", |
| 401 | "shape": [ |
| 402 | 96 |
| 403 | ], |
| 404 | "dtype": "float16", |
| 405 | "offset_bytes": 669900, |
| 406 | "nbytes": 192, |
| 407 | "sha256": "bec0a355ee04c1bbe7006fb5d7aa78dc134628ca1ca9d43b036e7d46150e32e9" |
| 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": 670092, |
| 418 | "nbytes": 92160, |
| 419 | "sha256": "e0abf4161e6034784a26f98ae3064f2d6ccd6d829f6ca7d3a519a54c29c210a1" |
| 420 | }, |
| 421 | { |
| 422 | "name": "frame_blocks.2.net.2.bias", |
| 423 | "shape": [ |
| 424 | 96 |
| 425 | ], |
| 426 | "dtype": "float16", |
| 427 | "offset_bytes": 762252, |
| 428 | "nbytes": 192, |
| 429 | "sha256": "597598cf48aed122b2184521d7473cafccf412d9c6b7b6d66276a425a1711091" |
| 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": 762444, |
| 440 | "nbytes": 92160, |
| 441 | "sha256": "81418e5be7762f840d7392184e7b4bbd2f91956eb016a400d390eea1888a114a" |
| 442 | }, |
| 443 | { |
| 444 | "name": "frame_blocks.2.scale", |
| 445 | "shape": [ |
| 446 | 1 |
| 447 | ], |
| 448 | "dtype": "float16", |
| 449 | "offset_bytes": 854604, |
| 450 | "nbytes": 2, |
| 451 | "sha256": "9d2b7d97dbc511591f4865e4dee0f2ece9649852b0b87e3189cac270764b5deb" |
| 452 | }, |
| 453 | { |
| 454 | "name": "frame_blocks.3.net.0.bias", |
| 455 | "shape": [ |
| 456 | 96 |
| 457 | ], |
| 458 | "dtype": "float16", |
| 459 | "offset_bytes": 854606, |
| 460 | "nbytes": 192, |
| 461 | "sha256": "5858cc8c63f24a5627498f869594b4cfef4e8cbcf0500b373e603a083241aa7b" |
| 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": 854798, |
| 472 | "nbytes": 92160, |
| 473 | "sha256": "6eb4fefa8ca5545585b79847d2c9e097139e8d78e5a8c8e0c218ce7de160b744" |
| 474 | }, |
| 475 | { |
| 476 | "name": "frame_blocks.3.net.2.bias", |
| 477 | "shape": [ |
| 478 | 96 |
| 479 | ], |
| 480 | "dtype": "float16", |
| 481 | "offset_bytes": 946958, |
| 482 | "nbytes": 192, |
| 483 | "sha256": "c33e4f9d5a72bb955ce7102b42a3e3751f001af0bf5777003930a977a7ba2800" |
| 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": 947150, |
| 494 | "nbytes": 92160, |
| 495 | "sha256": "09b345ff1b4bc2b30badbf3f314077601454367066885ec22ea3ee03cf50e790" |
| 496 | }, |
| 497 | { |
| 498 | "name": "frame_blocks.3.scale", |
| 499 | "shape": [ |
| 500 | 1 |
| 501 | ], |
| 502 | "dtype": "float16", |
| 503 | "offset_bytes": 1039310, |
| 504 | "nbytes": 2, |
| 505 | "sha256": "7b854d215977337d4605a44b85f6078495d917ff52fcd84f701076b6164bef63" |
| 506 | }, |
| 507 | { |
| 508 | "name": "frame_input_proj.bias", |
| 509 | "shape": [ |
| 510 | 96 |
| 511 | ], |
| 512 | "dtype": "float16", |
| 513 | "offset_bytes": 1039312, |
| 514 | "nbytes": 192, |
| 515 | "sha256": "cdc94c53ecf01b8f918bc36480b17c3219ba9c2100227365ca60af96e5a250c6" |
| 516 | }, |
| 517 | { |
| 518 | "name": "frame_input_proj.weight", |
| 519 | "shape": [ |
| 520 | 96, |
| 521 | 99, |
| 522 | 1 |
| 523 | ], |
| 524 | "dtype": "float16", |
| 525 | "offset_bytes": 1039504, |
| 526 | "nbytes": 19008, |
| 527 | "sha256": "9cee6aec95a79a1ffdf50e49adac1c7fd0fff0b9e187f3eab4af389b2129e36c" |
| 528 | }, |
| 529 | { |
| 530 | "name": "output.bias", |
| 531 | "shape": [ |
| 532 | 192 |
| 533 | ], |
| 534 | "dtype": "float16", |
| 535 | "offset_bytes": 1058512, |
| 536 | "nbytes": 384, |
| 537 | "sha256": "5bef967961657b12b0666483b83d1326c2a364d9ddb0fe077cd52930e753858c" |
| 538 | }, |
| 539 | { |
| 540 | "name": "output.weight", |
| 541 | "shape": [ |
| 542 | 192, |
| 543 | 96, |
| 544 | 1 |
| 545 | ], |
| 546 | "dtype": "float16", |
| 547 | "offset_bytes": 1058896, |
| 548 | "nbytes": 36864, |
| 549 | "sha256": "be296d50e3ba71374dce751468e4bbb2b1ef0a7022929c1b86304c426ed66cf0" |
| 550 | }, |
| 551 | { |
| 552 | "name": "token_blocks.0.net.0.bias", |
| 553 | "shape": [ |
| 554 | 96 |
| 555 | ], |
| 556 | "dtype": "float16", |
| 557 | "offset_bytes": 1095760, |
| 558 | "nbytes": 192, |
| 559 | "sha256": "f23f173e7b07c254ffa1eaa9ef08d10ae0ac031f2159cf33f30c223dffcf7ac1" |
| 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": 1095952, |
| 570 | "nbytes": 92160, |
| 571 | "sha256": "8deee89435cfd4bce3c02703061d39374a365bb845aa1da11c6ae4b22bd02baa" |
| 572 | }, |
| 573 | { |
| 574 | "name": "token_blocks.0.net.2.bias", |
| 575 | "shape": [ |
| 576 | 96 |
| 577 | ], |
| 578 | "dtype": "float16", |
| 579 | "offset_bytes": 1188112, |
| 580 | "nbytes": 192, |
| 581 | "sha256": "a36c27c7846409fc70593d9aa4ef223dd0c63d46721157cfc34d4cd6f8565c16" |
| 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": 1188304, |
| 592 | "nbytes": 92160, |
| 593 | "sha256": "6365bfd553d9f95de1d401c1f6f3239f0b60b2639487a04901df3204d4b99647" |
| 594 | }, |
| 595 | { |
| 596 | "name": "token_blocks.0.scale", |
| 597 | "shape": [ |
| 598 | 1 |
| 599 | ], |
| 600 | "dtype": "float16", |
| 601 | "offset_bytes": 1280464, |
| 602 | "nbytes": 2, |
| 603 | "sha256": "d95b3b43f96e1b91cb088c8a91ddd7cfad0cdd355a3d77f3244c6a2167f4fb46" |
| 604 | }, |
| 605 | { |
| 606 | "name": "token_blocks.1.net.0.bias", |
| 607 | "shape": [ |
| 608 | 96 |
| 609 | ], |
| 610 | "dtype": "float16", |
| 611 | "offset_bytes": 1280466, |
| 612 | "nbytes": 192, |
| 613 | "sha256": "d908edcb2cdf7c40df9680b79ab324895361b3d8c2f3956370b5b3c2b6b1c402" |
| 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": 1280658, |
| 624 | "nbytes": 92160, |
| 625 | "sha256": "9eb7ae134fe5f89f408cdab2f33243f8cb070e0cb0ff5f1b50bd295662b8443f" |
| 626 | }, |
| 627 | { |
| 628 | "name": "token_blocks.1.net.2.bias", |
| 629 | "shape": [ |
| 630 | 96 |
| 631 | ], |
| 632 | "dtype": "float16", |
| 633 | "offset_bytes": 1372818, |
| 634 | "nbytes": 192, |
| 635 | "sha256": "44dc7940ff3e8073f0f302aad2c2c6255b125c522f017c3a42db863683c9c905" |
| 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": 1373010, |
| 646 | "nbytes": 92160, |
| 647 | "sha256": "0b31bcb222232ffa10b8c4aa04b03edb8ade7c64127e1dfc475b3e1cbeaaa98a" |
| 648 | }, |
| 649 | { |
| 650 | "name": "token_blocks.1.scale", |
| 651 | "shape": [ |
| 652 | 1 |
| 653 | ], |
| 654 | "dtype": "float16", |
| 655 | "offset_bytes": 1465170, |
| 656 | "nbytes": 2, |
| 657 | "sha256": "338a0e5cc7511540bc81683a71d8d61c44de1399a07bde43bad77d39600f2a85" |
| 658 | }, |
| 659 | { |
| 660 | "name": "token_blocks.2.net.0.bias", |
| 661 | "shape": [ |
| 662 | 96 |
| 663 | ], |
| 664 | "dtype": "float16", |
| 665 | "offset_bytes": 1465172, |
| 666 | "nbytes": 192, |
| 667 | "sha256": "d5bfd3722fe66c97aeb763da861999be051c6b9f557776c5816f597cc295ab6d" |
| 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": 1465364, |
| 678 | "nbytes": 92160, |
| 679 | "sha256": "f5019bc78280e10d0b9e717461e9d41599b02a6425cb8e204d74d5d620ab206a" |
| 680 | }, |
| 681 | { |
| 682 | "name": "token_blocks.2.net.2.bias", |
| 683 | "shape": [ |
| 684 | 96 |
| 685 | ], |
| 686 | "dtype": "float16", |
| 687 | "offset_bytes": 1557524, |
| 688 | "nbytes": 192, |
| 689 | "sha256": "da8f72fdde324eb617d5d1f778d87d6e3f67f4eda9411270f079eba4fdd69a29" |
| 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": 1557716, |
| 700 | "nbytes": 92160, |
| 701 | "sha256": "0baf3ae22a3aae890426043d3fcf265deb59301d969b8c39174a558fd1a4d3da" |
| 702 | }, |
| 703 | { |
| 704 | "name": "token_blocks.2.scale", |
| 705 | "shape": [ |
| 706 | 1 |
| 707 | ], |
| 708 | "dtype": "float16", |
| 709 | "offset_bytes": 1649876, |
| 710 | "nbytes": 2, |
| 711 | "sha256": "54cd1dd9d255a588ffc4e53586af5a57582259eb34a3db722420a67e941cd18a" |
| 712 | }, |
| 713 | { |
| 714 | "name": "token_input_proj.bias", |
| 715 | "shape": [ |
| 716 | 96 |
| 717 | ], |
| 718 | "dtype": "float16", |
| 719 | "offset_bytes": 1649878, |
| 720 | "nbytes": 192, |
| 721 | "sha256": "87708e16446981385ed5ddd54e70ff8494ea5eda6bb666b453d9419b79981e7f" |
| 722 | }, |
| 723 | { |
| 724 | "name": "token_input_proj.weight", |
| 725 | "shape": [ |
| 726 | 96, |
| 727 | 98, |
| 728 | 1 |
| 729 | ], |
| 730 | "dtype": "float16", |
| 731 | "offset_bytes": 1650070, |
| 732 | "nbytes": 18816, |
| 733 | "sha256": "2be3e83990b214a72fb6333a76c28f9e1aca596bcb399db984eea2d00cf55f72" |
| 734 | } |
| 735 | ] |
| 736 | }, |
| 737 | "decoder": { |
| 738 | "checkpoint": "artifacts/voices/ro_RO/joint/decoder-student.pt", |
| 739 | "checkpoint_sha256": "ad7b6a1a34475dda9462e4813d1b01b0c4b8b6ad8717704bdc35abe227e3c272", |
| 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/ro_RO/decoder/decoder-student.pt", |
| 812 | "decoder_init": { |
| 813 | "checkpoint": "artifacts/voices/ro_RO/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/ro_RO/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": 1668886, |
| 960 | "nbytes": 2, |
| 961 | "sha256": "0529eccfe63b4cd57380152f69d4e9bcec23081b7c091c9220024c2931a344c1" |
| 962 | }, |
| 963 | { |
| 964 | "name": "post.weight", |
| 965 | "shape": [ |
| 966 | 1, |
| 967 | 20, |
| 968 | 7 |
| 969 | ], |
| 970 | "dtype": "float16", |
| 971 | "offset_bytes": 1668888, |
| 972 | "nbytes": 280, |
| 973 | "sha256": "752245567b2caace2f9a2d1167b5fc3dd7666d278ca836baf1212cf12d502cee" |
| 974 | }, |
| 975 | { |
| 976 | "name": "pre.bias", |
| 977 | "shape": [ |
| 978 | 160 |
| 979 | ], |
| 980 | "dtype": "float16", |
| 981 | "offset_bytes": 1669168, |
| 982 | "nbytes": 320, |
| 983 | "sha256": "ab983ed7e382e72b6987bb3540deb2ac1ff543b9acb8b0cee6cf4ff3e26240e0" |
| 984 | }, |
| 985 | { |
| 986 | "name": "pre.weight", |
| 987 | "shape": [ |
| 988 | 160, |
| 989 | 192, |
| 990 | 7 |
| 991 | ], |
| 992 | "dtype": "float16", |
| 993 | "offset_bytes": 1669488, |
| 994 | "nbytes": 430080, |
| 995 | "sha256": "6cd5beb0cdcb9d6733da3314a4a0d0f9d46240a4ab44b0e1f5723ae77e45d456" |
| 996 | }, |
| 997 | { |
| 998 | "name": "res0.0.blocks.0.conv1.bias", |
| 999 | "shape": [ |
| 1000 | 80 |
| 1001 | ], |
| 1002 | "dtype": "float16", |
| 1003 | "offset_bytes": 2099568, |
| 1004 | "nbytes": 160, |
| 1005 | "sha256": "d1cd982501834f5558b1b85e45c425686c4451344c70ac6b82aefa40d3298376" |
| 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": 2099728, |
| 1016 | "nbytes": 38400, |
| 1017 | "sha256": "18db3a16c99fd5849c191ec6c766ed43d5414e22a7ba2bdda5e121260a9da7b7" |
| 1018 | }, |
| 1019 | { |
| 1020 | "name": "res0.0.blocks.0.conv2.bias", |
| 1021 | "shape": [ |
| 1022 | 80 |
| 1023 | ], |
| 1024 | "dtype": "float16", |
| 1025 | "offset_bytes": 2138128, |
| 1026 | "nbytes": 160, |
| 1027 | "sha256": "1ee7b3777df0a0cd07684d3bc49e33c6eae602b1f418a0a71df5c8eae1ed79a8" |
| 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": 2138288, |
| 1038 | "nbytes": 38400, |
| 1039 | "sha256": "28bc1259da8c3c6ea15a1cba5276c9ad73516b1e6dc25d699de136343aa81ac7" |
| 1040 | }, |
| 1041 | { |
| 1042 | "name": "res0.0.blocks.1.conv1.bias", |
| 1043 | "shape": [ |
| 1044 | 80 |
| 1045 | ], |
| 1046 | "dtype": "float16", |
| 1047 | "offset_bytes": 2176688, |
| 1048 | "nbytes": 160, |
| 1049 | "sha256": "da0f6a9431c144eedf49b476c8e9c62d78b49b7544c8c47726adfb9812fc532f" |
| 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": 2176848, |
| 1060 | "nbytes": 64000, |
| 1061 | "sha256": "3486dc01370c0554e345622786ee5cbb164a773ef0a7f243697f0e7509b35bd7" |
| 1062 | }, |
| 1063 | { |
| 1064 | "name": "res0.0.blocks.1.conv2.bias", |
| 1065 | "shape": [ |
| 1066 | 80 |
| 1067 | ], |
| 1068 | "dtype": "float16", |
| 1069 | "offset_bytes": 2240848, |
| 1070 | "nbytes": 160, |
| 1071 | "sha256": "dbe456f5d6301f87e82ccbb5bbdeec9059b41752bb0e26b4d203807ba68c00a5" |
| 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": 2241008, |
| 1082 | "nbytes": 64000, |
| 1083 | "sha256": "82cc939de768eeb9fcbf354dc3845a81b11f72271bd58fb07abe99a4a5acddfd" |
| 1084 | }, |
| 1085 | { |
| 1086 | "name": "res0.0.blocks.2.conv1.bias", |
| 1087 | "shape": [ |
| 1088 | 80 |
| 1089 | ], |
| 1090 | "dtype": "float16", |
| 1091 | "offset_bytes": 2305008, |
| 1092 | "nbytes": 160, |
| 1093 | "sha256": "5a9d93c4f487a84203e030755ea107a81ff65460efb905233bb92f65e1a13871" |
| 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": 2305168, |
| 1104 | "nbytes": 89600, |
| 1105 | "sha256": "0064e0a66a76adb102b85c050ec3688f2a1ac3932f08b513223b0d0bcbb7bc21" |
| 1106 | }, |
| 1107 | { |
| 1108 | "name": "res0.0.blocks.2.conv2.bias", |
| 1109 | "shape": [ |
| 1110 | 80 |
| 1111 | ], |
| 1112 | "dtype": "float16", |
| 1113 | "offset_bytes": 2394768, |
| 1114 | "nbytes": 160, |
| 1115 | "sha256": "254d15f2613143214094d60b8f01e35931a87100bb78e8076f0b15bd4963b92a" |
| 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": 2394928, |
| 1126 | "nbytes": 89600, |
| 1127 | "sha256": "854a8174581952dfe55d5aea54d19b394bee4159073a400ef17354fb31c1ac13" |
| 1128 | }, |
| 1129 | { |
| 1130 | "name": "res1.0.blocks.0.conv1.bias", |
| 1131 | "shape": [ |
| 1132 | 40 |
| 1133 | ], |
| 1134 | "dtype": "float16", |
| 1135 | "offset_bytes": 2484528, |
| 1136 | "nbytes": 80, |
| 1137 | "sha256": "ec4eefcdfc2dbe27bee911774ef185fbd9345221d4938d6e47e1854f572f80e9" |
| 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": 2484608, |
| 1148 | "nbytes": 9600, |
| 1149 | "sha256": "0a09aadf32df958cad63a9a9bafe6e3419b3ad0c0b1bb18e33419d446bfa721a" |
| 1150 | }, |
| 1151 | { |
| 1152 | "name": "res1.0.blocks.0.conv2.bias", |
| 1153 | "shape": [ |
| 1154 | 40 |
| 1155 | ], |
| 1156 | "dtype": "float16", |
| 1157 | "offset_bytes": 2494208, |
| 1158 | "nbytes": 80, |
| 1159 | "sha256": "e30c9b45de4742179853d2e25fddb8a957c7c4f1dac7eb751fbfbc188dcfabfe" |
| 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": 2494288, |
| 1170 | "nbytes": 9600, |
| 1171 | "sha256": "3525b9755aae5a149fbccbbda823957d2bd7c2c7a0100d85233fec64ca88f01b" |
| 1172 | }, |
| 1173 | { |
| 1174 | "name": "res1.0.blocks.1.conv1.bias", |
| 1175 | "shape": [ |
| 1176 | 40 |
| 1177 | ], |
| 1178 | "dtype": "float16", |
| 1179 | "offset_bytes": 2503888, |
| 1180 | "nbytes": 80, |
| 1181 | "sha256": "a5a052bad035a7ff94f8f355b2f090af63098acb1a8a91bc26391f938d534e1a" |
| 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": 2503968, |
| 1192 | "nbytes": 16000, |
| 1193 | "sha256": "7a460797643543f9d4b05dcc732c816b02d930dd24fe447426fcfa5e4a22d46f" |
| 1194 | }, |
| 1195 | { |
| 1196 | "name": "res1.0.blocks.1.conv2.bias", |
| 1197 | "shape": [ |
| 1198 | 40 |
| 1199 | ], |
| 1200 | "dtype": "float16", |
| 1201 | "offset_bytes": 2519968, |
| 1202 | "nbytes": 80, |
| 1203 | "sha256": "6e6c99b0ee43cc5c2688acd10e5cd2dbd90f8bec62a928c4a3e11ddadc09e479" |
| 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": 2520048, |
| 1214 | "nbytes": 16000, |
| 1215 | "sha256": "e3ba0c21d14971231110296ea3dc17c2560bdfa6463168a981fbeda8abdd064d" |
| 1216 | }, |
| 1217 | { |
| 1218 | "name": "res1.0.blocks.2.conv1.bias", |
| 1219 | "shape": [ |
| 1220 | 40 |
| 1221 | ], |
| 1222 | "dtype": "float16", |
| 1223 | "offset_bytes": 2536048, |
| 1224 | "nbytes": 80, |
| 1225 | "sha256": "d3c9c34b61661c75c2cec8ef0568f371cc46e02914014df92e71c21328967476" |
| 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": 2536128, |
| 1236 | "nbytes": 22400, |
| 1237 | "sha256": "efd840326d1a005866ad7816ee222fdc810d632ef0180eb2d830376324b4483d" |
| 1238 | }, |
| 1239 | { |
| 1240 | "name": "res1.0.blocks.2.conv2.bias", |
| 1241 | "shape": [ |
| 1242 | 40 |
| 1243 | ], |
| 1244 | "dtype": "float16", |
| 1245 | "offset_bytes": 2558528, |
| 1246 | "nbytes": 80, |
| 1247 | "sha256": "a41dbec1e479048c05f0d56f204179b8a36c0370c4aa706084d1ce45ce72a527" |
| 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": 2558608, |
| 1258 | "nbytes": 22400, |
| 1259 | "sha256": "4a842aaea10f110b5988311702ce5a8822ca1291a6884f062b22fd30a567eec1" |
| 1260 | }, |
| 1261 | { |
| 1262 | "name": "res2.0.blocks.0.conv1.bias", |
| 1263 | "shape": [ |
| 1264 | 20 |
| 1265 | ], |
| 1266 | "dtype": "float16", |
| 1267 | "offset_bytes": 2581008, |
| 1268 | "nbytes": 40, |
| 1269 | "sha256": "1face69c34751c10034b8affc50486575f32a5edd1cfee4d70cdd45781bb08a5" |
| 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": 2581048, |
| 1280 | "nbytes": 2400, |
| 1281 | "sha256": "ee2d372e27f4aed612db442228023402026d9d312c91168493bf67153a1c14e5" |
| 1282 | }, |
| 1283 | { |
| 1284 | "name": "res2.0.blocks.0.conv2.bias", |
| 1285 | "shape": [ |
| 1286 | 20 |
| 1287 | ], |
| 1288 | "dtype": "float16", |
| 1289 | "offset_bytes": 2583448, |
| 1290 | "nbytes": 40, |
| 1291 | "sha256": "67b35bb975f5363c94b9b34b449b38d60ad75b7a328e6ad2f9a39322afa155d3" |
| 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": 2583488, |
| 1302 | "nbytes": 2400, |
| 1303 | "sha256": "943ec181da53027144fdcf089219c1c95910c56c803108bd3a83476c33c0091b" |
| 1304 | }, |
| 1305 | { |
| 1306 | "name": "res2.0.blocks.1.conv1.bias", |
| 1307 | "shape": [ |
| 1308 | 20 |
| 1309 | ], |
| 1310 | "dtype": "float16", |
| 1311 | "offset_bytes": 2585888, |
| 1312 | "nbytes": 40, |
| 1313 | "sha256": "c4e7c7e8db0e181aa1bfc57f0339c162a41fe455879922cad7d8d5beaa57258f" |
| 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": 2585928, |
| 1324 | "nbytes": 4000, |
| 1325 | "sha256": "05d0a0234ab230b48d2393a477094bb06a2acf5605f2cda558c11cc8ecf089f0" |
| 1326 | }, |
| 1327 | { |
| 1328 | "name": "res2.0.blocks.1.conv2.bias", |
| 1329 | "shape": [ |
| 1330 | 20 |
| 1331 | ], |
| 1332 | "dtype": "float16", |
| 1333 | "offset_bytes": 2589928, |
| 1334 | "nbytes": 40, |
| 1335 | "sha256": "c6f79c86c15780fce03c004df3209e4197e03c2fe4cc6c81c2bd15efae462aea" |
| 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": 2589968, |
| 1346 | "nbytes": 4000, |
| 1347 | "sha256": "51e9f4ca0866d9d9ef3f615d32409e01ede0c85172c1d3211a90a961114931c5" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "res2.0.blocks.2.conv1.bias", |
| 1351 | "shape": [ |
| 1352 | 20 |
| 1353 | ], |
| 1354 | "dtype": "float16", |
| 1355 | "offset_bytes": 2593968, |
| 1356 | "nbytes": 40, |
| 1357 | "sha256": "522ebc09e18ab6c6526f4adfba2657ddf44d1f2bba82bdd84270b92353295a41" |
| 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": 2594008, |
| 1368 | "nbytes": 5600, |
| 1369 | "sha256": "e01045ef38fdd9a1739d4dadba221f0399c9992f2dc5a452d8579ac74cb1cf1c" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "res2.0.blocks.2.conv2.bias", |
| 1373 | "shape": [ |
| 1374 | 20 |
| 1375 | ], |
| 1376 | "dtype": "float16", |
| 1377 | "offset_bytes": 2599608, |
| 1378 | "nbytes": 40, |
| 1379 | "sha256": "06a9a39a6066665ef5c40f079327a08877bd4fbdff2511076c0fd4443f061045" |
| 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": 2599648, |
| 1390 | "nbytes": 5600, |
| 1391 | "sha256": "69657ad9e7bf953c297f2884431106221ce93aad5fa6f26db91edae3843fadb3" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "up0.bias", |
| 1395 | "shape": [ |
| 1396 | 80 |
| 1397 | ], |
| 1398 | "dtype": "float16", |
| 1399 | "offset_bytes": 2605248, |
| 1400 | "nbytes": 160, |
| 1401 | "sha256": "3eccfae295793e83478a6d0f83fe36d1f7bec928b5767027afda6903a53a8b28" |
| 1402 | }, |
| 1403 | { |
| 1404 | "name": "up0.weight", |
| 1405 | "shape": [ |
| 1406 | 160, |
| 1407 | 80, |
| 1408 | 16 |
| 1409 | ], |
| 1410 | "dtype": "float16", |
| 1411 | "offset_bytes": 2605408, |
| 1412 | "nbytes": 409600, |
| 1413 | "sha256": "df559bd0c2554a360d93e3c83e37243b2bfe35d6d3a7f2c2307f38854af9f77a" |
| 1414 | }, |
| 1415 | { |
| 1416 | "name": "up1.bias", |
| 1417 | "shape": [ |
| 1418 | 40 |
| 1419 | ], |
| 1420 | "dtype": "float16", |
| 1421 | "offset_bytes": 3015008, |
| 1422 | "nbytes": 80, |
| 1423 | "sha256": "9425763b85700d5a8abe2e604fed46975699e80ece90bafb341e4f2e4965dc87" |
| 1424 | }, |
| 1425 | { |
| 1426 | "name": "up1.weight", |
| 1427 | "shape": [ |
| 1428 | 80, |
| 1429 | 40, |
| 1430 | 16 |
| 1431 | ], |
| 1432 | "dtype": "float16", |
| 1433 | "offset_bytes": 3015088, |
| 1434 | "nbytes": 102400, |
| 1435 | "sha256": "3a9aa1bfe53166bdcacadc2c6cfb183513256ea7991be89025e2833c84e064d5" |
| 1436 | }, |
| 1437 | { |
| 1438 | "name": "up2.bias", |
| 1439 | "shape": [ |
| 1440 | 20 |
| 1441 | ], |
| 1442 | "dtype": "float16", |
| 1443 | "offset_bytes": 3117488, |
| 1444 | "nbytes": 40, |
| 1445 | "sha256": "fb989429dbafba2368e70db877009c0fb91c5e27280a3ac08f2499a63e1946b8" |
| 1446 | }, |
| 1447 | { |
| 1448 | "name": "up2.weight", |
| 1449 | "shape": [ |
| 1450 | 40, |
| 1451 | 20, |
| 1452 | 8 |
| 1453 | ], |
| 1454 | "dtype": "float16", |
| 1455 | "offset_bytes": 3117528, |
| 1456 | "nbytes": 12800, |
| 1457 | "sha256": "1c6d8f93aa87f86827630dfc7ca5493d9e5832974a9d44297b57dedded887064" |
| 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": "ro", |
| 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": 3135205, |
| 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 | |