ar-kareem-1p57m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "ar_JO-flores-20260908", |
| 4 | "language": "ar_JO", |
| 5 | "voice": "ar_JO-kareem-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": 3132248, |
| 15 | "weights_sha256": "7239f6dd4a8247e79238257811b298ba6559546f429e4be5ca9805dea2089ae0", |
| 16 | "total_parameters": 1566124, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/ar_JO/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "87812b32e4b5b6b57feb9195da05e7b94a776d478dc1f87dbd2aae4de31c7b36", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 147, |
| 24 | "hidden": 64, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 1361, |
| 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": 137092, |
| 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": "d6cf03356276f9985a583b6e89896461339cc6540d897b260f7086fddf0ffb36" |
| 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": "08abf671930ddb94b0619c741c9337d5e63ef3cefb5a6d4244df57f398de864a" |
| 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": "3bebdea262e2181bdad1562f93565db93a9d6ca4d1310e8cee70795aecafbe13" |
| 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": "21e26f4093e500bb94daf182783f163552977b865b703b6628cd22027a4741b6" |
| 90 | }, |
| 91 | { |
| 92 | "name": "blocks.0.scale", |
| 93 | "shape": [ |
| 94 | 1 |
| 95 | ], |
| 96 | "dtype": "float16", |
| 97 | "offset_bytes": 82176, |
| 98 | "nbytes": 2, |
| 99 | "sha256": "bc54334f1fbe08b8f21c3c737a947567806406d94f1348f69dcdefe37199914c" |
| 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": "0d276b1e8b3c0faaa351db078072f36e34496b9292192a47b16d51ff83cb6f72" |
| 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": "d41e2cc4ec083757c215e8a3114258162f592f93fd1a28494c5a3126da060a98" |
| 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": "2d5cd6a2db4692f8d9549da934c7ba8fc1dc873811769fe00d06452dc8d4d0f9" |
| 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": "ec790a99edf0920e7491a3878d2f70202ce4f63c64c7a107ecab99a45fbe684f" |
| 144 | }, |
| 145 | { |
| 146 | "name": "blocks.1.scale", |
| 147 | "shape": [ |
| 148 | 1 |
| 149 | ], |
| 150 | "dtype": "float16", |
| 151 | "offset_bytes": 164354, |
| 152 | "nbytes": 2, |
| 153 | "sha256": "228f90542699528c5c3f1f62c47a40d1fe802ca6cc02d4a5b43af66e2eed273e" |
| 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": "22758fd07e7f11567ee9f67fb63a2b41d5a4f0491ab9a3c8dab7510324de74f9" |
| 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": "54324c075e917f0f2d4ecca41c6ed162705f98b0548aa5e4ccac7f4f84f56739" |
| 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": "b6bfb1098239117d70a59ffc0c3bb588ac8520f8100a3b279cae2b446e8ef527" |
| 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": "0ba7f155ad528dfba24249dd614901f8227236e6e92e300d042e2ca155233155" |
| 198 | }, |
| 199 | { |
| 200 | "name": "blocks.2.scale", |
| 201 | "shape": [ |
| 202 | 1 |
| 203 | ], |
| 204 | "dtype": "float16", |
| 205 | "offset_bytes": 246532, |
| 206 | "nbytes": 2, |
| 207 | "sha256": "025eeddefafbdbf60ec7a07cd1439b318a336f3afd8d2962eb344f57075a7d2f" |
| 208 | }, |
| 209 | { |
| 210 | "name": "embedding.weight", |
| 211 | "shape": [ |
| 212 | 147, |
| 213 | 64 |
| 214 | ], |
| 215 | "dtype": "float16", |
| 216 | "offset_bytes": 246534, |
| 217 | "nbytes": 18816, |
| 218 | "sha256": "b913f162c4cf189d6eede439d16e9a22a0d137b214a0068011883ee37d365406" |
| 219 | }, |
| 220 | { |
| 221 | "name": "input_proj.bias", |
| 222 | "shape": [ |
| 223 | 64 |
| 224 | ], |
| 225 | "dtype": "float16", |
| 226 | "offset_bytes": 265350, |
| 227 | "nbytes": 128, |
| 228 | "sha256": "5e71141e485fca2d31b379bf0efd664b40f753a20668cdc01de587cf2c80697f" |
| 229 | }, |
| 230 | { |
| 231 | "name": "input_proj.weight", |
| 232 | "shape": [ |
| 233 | 64, |
| 234 | 67, |
| 235 | 1 |
| 236 | ], |
| 237 | "dtype": "float16", |
| 238 | "offset_bytes": 265478, |
| 239 | "nbytes": 8576, |
| 240 | "sha256": "2036fc6f4102dbd184f78c409a62dedbd8f66afc12d1386555801b8b754cf9db" |
| 241 | }, |
| 242 | { |
| 243 | "name": "output.bias", |
| 244 | "shape": [ |
| 245 | 1 |
| 246 | ], |
| 247 | "dtype": "float16", |
| 248 | "offset_bytes": 274054, |
| 249 | "nbytes": 2, |
| 250 | "sha256": "78503765357b731abf3a3f70167c8efba880ad3bedf78aa32f1cebbe983ad16e" |
| 251 | }, |
| 252 | { |
| 253 | "name": "output.weight", |
| 254 | "shape": [ |
| 255 | 1, |
| 256 | 64, |
| 257 | 1 |
| 258 | ], |
| 259 | "dtype": "float16", |
| 260 | "offset_bytes": 274056, |
| 261 | "nbytes": 128, |
| 262 | "sha256": "7eb9090c65a76be4af63df8db10b8cf4dd81767f4eed9ff3bd0750ee7cec82d9" |
| 263 | } |
| 264 | ] |
| 265 | }, |
| 266 | "acoustic": { |
| 267 | "checkpoint": "artifacts/voices/ar_JO/joint/latent-student.pt", |
| 268 | "checkpoint_sha256": "4cec3e5fb2080e6ab0c59043063a43be1a941a94a83f1e57e1cabf3987bd19dc", |
| 269 | "config": { |
| 270 | "architecture": "token_context", |
| 271 | "vocab_size": 147, |
| 272 | "hidden": 96, |
| 273 | "depth": 4, |
| 274 | "kernel_size": 5, |
| 275 | "out_channels": 192, |
| 276 | "token_depth": 3 |
| 277 | }, |
| 278 | "parameters": 698311, |
| 279 | "tensors": [ |
| 280 | { |
| 281 | "name": "embedding.weight", |
| 282 | "shape": [ |
| 283 | 147, |
| 284 | 96 |
| 285 | ], |
| 286 | "dtype": "float16", |
| 287 | "offset_bytes": 274184, |
| 288 | "nbytes": 28224, |
| 289 | "sha256": "0ed5fe6e37de93baf2af82207d39b52628130efc01f35268c92710dccc0304ba" |
| 290 | }, |
| 291 | { |
| 292 | "name": "frame_blocks.0.net.0.bias", |
| 293 | "shape": [ |
| 294 | 96 |
| 295 | ], |
| 296 | "dtype": "float16", |
| 297 | "offset_bytes": 302408, |
| 298 | "nbytes": 192, |
| 299 | "sha256": "82ca74229781f8e8198780442e0a9b2c7eebc8fbc9383fe71107a1f83cfe96fe" |
| 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": 302600, |
| 310 | "nbytes": 92160, |
| 311 | "sha256": "647ee881c8bc0aafc9395bba89f505733e81dd2e07bc8056245381feb76dcbe5" |
| 312 | }, |
| 313 | { |
| 314 | "name": "frame_blocks.0.net.2.bias", |
| 315 | "shape": [ |
| 316 | 96 |
| 317 | ], |
| 318 | "dtype": "float16", |
| 319 | "offset_bytes": 394760, |
| 320 | "nbytes": 192, |
| 321 | "sha256": "4eb75226b0c8a349d92ae97874a93ba95bd35509662fcd046f61ce8ce1d205c5" |
| 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": 394952, |
| 332 | "nbytes": 92160, |
| 333 | "sha256": "3ddf93f7ba624147843a865f4bdd7faa6268f58473f23df002145a60b390a68d" |
| 334 | }, |
| 335 | { |
| 336 | "name": "frame_blocks.0.scale", |
| 337 | "shape": [ |
| 338 | 1 |
| 339 | ], |
| 340 | "dtype": "float16", |
| 341 | "offset_bytes": 487112, |
| 342 | "nbytes": 2, |
| 343 | "sha256": "61c5caba2f046b2ca39e5d59ee82a283aad168ee72e937e7de348ac7e513b2e6" |
| 344 | }, |
| 345 | { |
| 346 | "name": "frame_blocks.1.net.0.bias", |
| 347 | "shape": [ |
| 348 | 96 |
| 349 | ], |
| 350 | "dtype": "float16", |
| 351 | "offset_bytes": 487114, |
| 352 | "nbytes": 192, |
| 353 | "sha256": "ec978602b326fa3156740d3a3569ff235f15b6c6bef9ef5449c9ec44c248b057" |
| 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": 487306, |
| 364 | "nbytes": 92160, |
| 365 | "sha256": "7142b5f424b02511920207460a820949fc7dffa7fcdbd4e18942e948ed1f2d76" |
| 366 | }, |
| 367 | { |
| 368 | "name": "frame_blocks.1.net.2.bias", |
| 369 | "shape": [ |
| 370 | 96 |
| 371 | ], |
| 372 | "dtype": "float16", |
| 373 | "offset_bytes": 579466, |
| 374 | "nbytes": 192, |
| 375 | "sha256": "e63a178b9d1b2f02f5c7a602ae79d23b81a779b319f3824d40ff539cef1d3e41" |
| 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": 579658, |
| 386 | "nbytes": 92160, |
| 387 | "sha256": "dbb63eb3b27f47f69226ba293c574d5c379b4fec0351024d43beb93b6c4ac1d3" |
| 388 | }, |
| 389 | { |
| 390 | "name": "frame_blocks.1.scale", |
| 391 | "shape": [ |
| 392 | 1 |
| 393 | ], |
| 394 | "dtype": "float16", |
| 395 | "offset_bytes": 671818, |
| 396 | "nbytes": 2, |
| 397 | "sha256": "b3bc69c97c12111e04362db1cbb2de16ac26a59c7704b2a358b56df68b3b89b3" |
| 398 | }, |
| 399 | { |
| 400 | "name": "frame_blocks.2.net.0.bias", |
| 401 | "shape": [ |
| 402 | 96 |
| 403 | ], |
| 404 | "dtype": "float16", |
| 405 | "offset_bytes": 671820, |
| 406 | "nbytes": 192, |
| 407 | "sha256": "5fdd2afb22209bbed4799af3af1941725c1de1b214957bd5cb619df826717e74" |
| 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": 672012, |
| 418 | "nbytes": 92160, |
| 419 | "sha256": "3a865557ff883f25a4246e604ed284a6d86702734868057ac829e30f351ecefa" |
| 420 | }, |
| 421 | { |
| 422 | "name": "frame_blocks.2.net.2.bias", |
| 423 | "shape": [ |
| 424 | 96 |
| 425 | ], |
| 426 | "dtype": "float16", |
| 427 | "offset_bytes": 764172, |
| 428 | "nbytes": 192, |
| 429 | "sha256": "25755dfc550f5bbe7f27b73d40b4be90a43a6265de4ad0c7584a670b4fa95a2e" |
| 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": 764364, |
| 440 | "nbytes": 92160, |
| 441 | "sha256": "fc65ba75241efa32bdc411d6b0d73aca0984e678bff5202ee12404effa327754" |
| 442 | }, |
| 443 | { |
| 444 | "name": "frame_blocks.2.scale", |
| 445 | "shape": [ |
| 446 | 1 |
| 447 | ], |
| 448 | "dtype": "float16", |
| 449 | "offset_bytes": 856524, |
| 450 | "nbytes": 2, |
| 451 | "sha256": "e9586b616536950f82cad86972b68f4f99d3fb3cdf4013c0d1a0a09394ca96b5" |
| 452 | }, |
| 453 | { |
| 454 | "name": "frame_blocks.3.net.0.bias", |
| 455 | "shape": [ |
| 456 | 96 |
| 457 | ], |
| 458 | "dtype": "float16", |
| 459 | "offset_bytes": 856526, |
| 460 | "nbytes": 192, |
| 461 | "sha256": "6172557a32526bc92a3dae6d2910b5ecc869d05f015df9826937db8b7c5a8bc3" |
| 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": 856718, |
| 472 | "nbytes": 92160, |
| 473 | "sha256": "95fd478046b9558ae97a041ba4b067d894e9eb93e238e7ed1b0d409a9139c74a" |
| 474 | }, |
| 475 | { |
| 476 | "name": "frame_blocks.3.net.2.bias", |
| 477 | "shape": [ |
| 478 | 96 |
| 479 | ], |
| 480 | "dtype": "float16", |
| 481 | "offset_bytes": 948878, |
| 482 | "nbytes": 192, |
| 483 | "sha256": "2b5e9da436ece38b273039aaa292674512b07b730a29de8c8e9314b5f6a3c805" |
| 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": 949070, |
| 494 | "nbytes": 92160, |
| 495 | "sha256": "1ac49f5316efc344819b2feabe5b9479a6a18d1f75b3f2c66bf88885d01784eb" |
| 496 | }, |
| 497 | { |
| 498 | "name": "frame_blocks.3.scale", |
| 499 | "shape": [ |
| 500 | 1 |
| 501 | ], |
| 502 | "dtype": "float16", |
| 503 | "offset_bytes": 1041230, |
| 504 | "nbytes": 2, |
| 505 | "sha256": "dd5862237d2dda48a843f52f9459959ea8d6dc7d335812b02cfbcdcf368fec0d" |
| 506 | }, |
| 507 | { |
| 508 | "name": "frame_input_proj.bias", |
| 509 | "shape": [ |
| 510 | 96 |
| 511 | ], |
| 512 | "dtype": "float16", |
| 513 | "offset_bytes": 1041232, |
| 514 | "nbytes": 192, |
| 515 | "sha256": "e22d54876b95bb96a663c355d75b9eff09ea4068c3539c49e764399afe25bbcf" |
| 516 | }, |
| 517 | { |
| 518 | "name": "frame_input_proj.weight", |
| 519 | "shape": [ |
| 520 | 96, |
| 521 | 99, |
| 522 | 1 |
| 523 | ], |
| 524 | "dtype": "float16", |
| 525 | "offset_bytes": 1041424, |
| 526 | "nbytes": 19008, |
| 527 | "sha256": "40b1e5fc420cbc0334f55514ab931da390be9ac7e6b953fd4dfa6de088fc527d" |
| 528 | }, |
| 529 | { |
| 530 | "name": "output.bias", |
| 531 | "shape": [ |
| 532 | 192 |
| 533 | ], |
| 534 | "dtype": "float16", |
| 535 | "offset_bytes": 1060432, |
| 536 | "nbytes": 384, |
| 537 | "sha256": "2a9096e2beee11e7c69e748dee881f43716d4134ad7dab407f3c1b78c66b890e" |
| 538 | }, |
| 539 | { |
| 540 | "name": "output.weight", |
| 541 | "shape": [ |
| 542 | 192, |
| 543 | 96, |
| 544 | 1 |
| 545 | ], |
| 546 | "dtype": "float16", |
| 547 | "offset_bytes": 1060816, |
| 548 | "nbytes": 36864, |
| 549 | "sha256": "56d002c6f45bd4f2b790f54e4969e9d8bcaaa1177fe48931e34dd0f92b8a749c" |
| 550 | }, |
| 551 | { |
| 552 | "name": "token_blocks.0.net.0.bias", |
| 553 | "shape": [ |
| 554 | 96 |
| 555 | ], |
| 556 | "dtype": "float16", |
| 557 | "offset_bytes": 1097680, |
| 558 | "nbytes": 192, |
| 559 | "sha256": "9cef868688ccc614f0ca814bafbfcbb4149e4eb67731c53fe2e16b5e2b9fd47b" |
| 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": 1097872, |
| 570 | "nbytes": 92160, |
| 571 | "sha256": "c2c5fa0cbfbc63a2bfd819a425c00cec31ac9c0f613aae5db4ce4acfb990e29e" |
| 572 | }, |
| 573 | { |
| 574 | "name": "token_blocks.0.net.2.bias", |
| 575 | "shape": [ |
| 576 | 96 |
| 577 | ], |
| 578 | "dtype": "float16", |
| 579 | "offset_bytes": 1190032, |
| 580 | "nbytes": 192, |
| 581 | "sha256": "fecf48ac3448267677f5a3b98281ab08fa80c7241dd1db6429b08e3e2f431eef" |
| 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": 1190224, |
| 592 | "nbytes": 92160, |
| 593 | "sha256": "8c817b12f30de80b34c2533bc98b4f27ac8d2fbedf11f19fde429eb743e8f799" |
| 594 | }, |
| 595 | { |
| 596 | "name": "token_blocks.0.scale", |
| 597 | "shape": [ |
| 598 | 1 |
| 599 | ], |
| 600 | "dtype": "float16", |
| 601 | "offset_bytes": 1282384, |
| 602 | "nbytes": 2, |
| 603 | "sha256": "d92eb408bf6938e072dfb1597873dc5587787367e0e97a94437ac71785b2fbe2" |
| 604 | }, |
| 605 | { |
| 606 | "name": "token_blocks.1.net.0.bias", |
| 607 | "shape": [ |
| 608 | 96 |
| 609 | ], |
| 610 | "dtype": "float16", |
| 611 | "offset_bytes": 1282386, |
| 612 | "nbytes": 192, |
| 613 | "sha256": "db1acd9282d5eba9f6446c951a624482e8ead3f40285d6745e3f03a83b009887" |
| 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": 1282578, |
| 624 | "nbytes": 92160, |
| 625 | "sha256": "774e5f5554baf419057b50035c06b7d9d0fe0ab225bbaac42bcb113de2ec256c" |
| 626 | }, |
| 627 | { |
| 628 | "name": "token_blocks.1.net.2.bias", |
| 629 | "shape": [ |
| 630 | 96 |
| 631 | ], |
| 632 | "dtype": "float16", |
| 633 | "offset_bytes": 1374738, |
| 634 | "nbytes": 192, |
| 635 | "sha256": "2a55bb25cee0fb1f4dec13529416a1e492de37c42bd4b52646eef1c22af24295" |
| 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": 1374930, |
| 646 | "nbytes": 92160, |
| 647 | "sha256": "e693b25f58ba3361ace75827043de5ecaa8495bd2bc8824883f6416ef0d4ca2b" |
| 648 | }, |
| 649 | { |
| 650 | "name": "token_blocks.1.scale", |
| 651 | "shape": [ |
| 652 | 1 |
| 653 | ], |
| 654 | "dtype": "float16", |
| 655 | "offset_bytes": 1467090, |
| 656 | "nbytes": 2, |
| 657 | "sha256": "d64e30a190ba3a58151d9c7ea551a1046b8a556afa89f754a34b59db96b81f7a" |
| 658 | }, |
| 659 | { |
| 660 | "name": "token_blocks.2.net.0.bias", |
| 661 | "shape": [ |
| 662 | 96 |
| 663 | ], |
| 664 | "dtype": "float16", |
| 665 | "offset_bytes": 1467092, |
| 666 | "nbytes": 192, |
| 667 | "sha256": "57feda9cfe3c9a5719726650eae2130d3fb4cbd672b88a5103e751defab9d133" |
| 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": 1467284, |
| 678 | "nbytes": 92160, |
| 679 | "sha256": "14c11bd7635a5e067f59ce497b161816ea14423ee0818a965d0a599393d18981" |
| 680 | }, |
| 681 | { |
| 682 | "name": "token_blocks.2.net.2.bias", |
| 683 | "shape": [ |
| 684 | 96 |
| 685 | ], |
| 686 | "dtype": "float16", |
| 687 | "offset_bytes": 1559444, |
| 688 | "nbytes": 192, |
| 689 | "sha256": "bd3a45f70b1d20e83d1a8d66ec07d851aab61f8b5a1eca52d717add312b7d503" |
| 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": 1559636, |
| 700 | "nbytes": 92160, |
| 701 | "sha256": "4aef64970eb90ab5c5cb6a295c172d9e1a2133f18cf5ad0ca4c37395426e1dce" |
| 702 | }, |
| 703 | { |
| 704 | "name": "token_blocks.2.scale", |
| 705 | "shape": [ |
| 706 | 1 |
| 707 | ], |
| 708 | "dtype": "float16", |
| 709 | "offset_bytes": 1651796, |
| 710 | "nbytes": 2, |
| 711 | "sha256": "1b87ff8e20019cb79d8bcc1feca340f86cfdf253ee729c0e31bd1f47dd2dd232" |
| 712 | }, |
| 713 | { |
| 714 | "name": "token_input_proj.bias", |
| 715 | "shape": [ |
| 716 | 96 |
| 717 | ], |
| 718 | "dtype": "float16", |
| 719 | "offset_bytes": 1651798, |
| 720 | "nbytes": 192, |
| 721 | "sha256": "dbf6db92fac14b3306e21fd115c7bd39c87e27e4f5769b27652570d11aadd114" |
| 722 | }, |
| 723 | { |
| 724 | "name": "token_input_proj.weight", |
| 725 | "shape": [ |
| 726 | 96, |
| 727 | 98, |
| 728 | 1 |
| 729 | ], |
| 730 | "dtype": "float16", |
| 731 | "offset_bytes": 1651990, |
| 732 | "nbytes": 18816, |
| 733 | "sha256": "3f57a85d5705e0b55e4edc7c7856ec51ecfb045d348636c230a67b8b2455639b" |
| 734 | } |
| 735 | ] |
| 736 | }, |
| 737 | "decoder": { |
| 738 | "checkpoint": "artifacts/voices/ar_JO/joint/decoder-student.pt", |
| 739 | "checkpoint_sha256": "31c09870dea55817c977fe80d56ca384252eab8196fcfd697c40bda2cf0d4340", |
| 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/ar_JO/decoder/decoder-student.pt", |
| 812 | "decoder_init": { |
| 813 | "checkpoint": "artifacts/voices/ar_JO/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/ar_JO/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": 1670806, |
| 960 | "nbytes": 2, |
| 961 | "sha256": "8cc9ce257862fcf6e3697359b34e49114d1bde37e43cbdcf62755a282ba90bad" |
| 962 | }, |
| 963 | { |
| 964 | "name": "post.weight", |
| 965 | "shape": [ |
| 966 | 1, |
| 967 | 20, |
| 968 | 7 |
| 969 | ], |
| 970 | "dtype": "float16", |
| 971 | "offset_bytes": 1670808, |
| 972 | "nbytes": 280, |
| 973 | "sha256": "e6566fbfe32feeea7d096490bb34e8833537eb1544bfc0455f38ec691f239927" |
| 974 | }, |
| 975 | { |
| 976 | "name": "pre.bias", |
| 977 | "shape": [ |
| 978 | 160 |
| 979 | ], |
| 980 | "dtype": "float16", |
| 981 | "offset_bytes": 1671088, |
| 982 | "nbytes": 320, |
| 983 | "sha256": "99216c12baf7b8c9a0f14a231b7aa060f374d1a7bb30cfeeca0c13ef0d6145a8" |
| 984 | }, |
| 985 | { |
| 986 | "name": "pre.weight", |
| 987 | "shape": [ |
| 988 | 160, |
| 989 | 192, |
| 990 | 7 |
| 991 | ], |
| 992 | "dtype": "float16", |
| 993 | "offset_bytes": 1671408, |
| 994 | "nbytes": 430080, |
| 995 | "sha256": "897c24a8d416805c7d18e62244315f5310d756256586a0f756f8dc60d5a14e7d" |
| 996 | }, |
| 997 | { |
| 998 | "name": "res0.0.blocks.0.conv1.bias", |
| 999 | "shape": [ |
| 1000 | 80 |
| 1001 | ], |
| 1002 | "dtype": "float16", |
| 1003 | "offset_bytes": 2101488, |
| 1004 | "nbytes": 160, |
| 1005 | "sha256": "1dec6c7656ddde44f601bfdf1f7268f8f9cc897cd2195ce0180d181a2fa530ab" |
| 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": 2101648, |
| 1016 | "nbytes": 38400, |
| 1017 | "sha256": "15942a8ba96e12b00fd5b48504b08c26374f9a9e577f7ca67debfc267830fdcc" |
| 1018 | }, |
| 1019 | { |
| 1020 | "name": "res0.0.blocks.0.conv2.bias", |
| 1021 | "shape": [ |
| 1022 | 80 |
| 1023 | ], |
| 1024 | "dtype": "float16", |
| 1025 | "offset_bytes": 2140048, |
| 1026 | "nbytes": 160, |
| 1027 | "sha256": "612ef04558d2cec535297e61839574a82dca784528594b4f106cd62eab817410" |
| 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": 2140208, |
| 1038 | "nbytes": 38400, |
| 1039 | "sha256": "34bb12b360d7c1a560d85e17b6c5f9b91cad6a444fa5146c3635b933a38ccd0b" |
| 1040 | }, |
| 1041 | { |
| 1042 | "name": "res0.0.blocks.1.conv1.bias", |
| 1043 | "shape": [ |
| 1044 | 80 |
| 1045 | ], |
| 1046 | "dtype": "float16", |
| 1047 | "offset_bytes": 2178608, |
| 1048 | "nbytes": 160, |
| 1049 | "sha256": "5d3196237c968acd9d7f701e839639cc56d3a369e3299d1bbd5205a1b8cdc121" |
| 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": 2178768, |
| 1060 | "nbytes": 64000, |
| 1061 | "sha256": "18e8d055b491934c27955434661c47458e3f5218f374cceb0edcc8f7e293ab9f" |
| 1062 | }, |
| 1063 | { |
| 1064 | "name": "res0.0.blocks.1.conv2.bias", |
| 1065 | "shape": [ |
| 1066 | 80 |
| 1067 | ], |
| 1068 | "dtype": "float16", |
| 1069 | "offset_bytes": 2242768, |
| 1070 | "nbytes": 160, |
| 1071 | "sha256": "7671d01263a980633d729c805abed2c4eb9a269ff7b8435869a16db5adc6d785" |
| 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": 2242928, |
| 1082 | "nbytes": 64000, |
| 1083 | "sha256": "3c15746c7937278f952c8e49ea9bf6da4a9bc28acbeb8561705ff2c8d669ba0d" |
| 1084 | }, |
| 1085 | { |
| 1086 | "name": "res0.0.blocks.2.conv1.bias", |
| 1087 | "shape": [ |
| 1088 | 80 |
| 1089 | ], |
| 1090 | "dtype": "float16", |
| 1091 | "offset_bytes": 2306928, |
| 1092 | "nbytes": 160, |
| 1093 | "sha256": "7391d0bcc9afa96f1f3a86df8638ed0b5de542c90bd6202845d30ae9f2244840" |
| 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": 2307088, |
| 1104 | "nbytes": 89600, |
| 1105 | "sha256": "fa79cb718e400b2150db0dfb54b7756241d72d6bb0d97b5384c0d4457b4cf2bc" |
| 1106 | }, |
| 1107 | { |
| 1108 | "name": "res0.0.blocks.2.conv2.bias", |
| 1109 | "shape": [ |
| 1110 | 80 |
| 1111 | ], |
| 1112 | "dtype": "float16", |
| 1113 | "offset_bytes": 2396688, |
| 1114 | "nbytes": 160, |
| 1115 | "sha256": "79e49acf6175fdcf0e1f13a3a4fd1c1113bd87b757135bfd13015abb9934deb3" |
| 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": 2396848, |
| 1126 | "nbytes": 89600, |
| 1127 | "sha256": "f9b61dbd849a9c46a9b368ab60c92b9a22c5b1eb78b793cc941b3ed73a9606e5" |
| 1128 | }, |
| 1129 | { |
| 1130 | "name": "res1.0.blocks.0.conv1.bias", |
| 1131 | "shape": [ |
| 1132 | 40 |
| 1133 | ], |
| 1134 | "dtype": "float16", |
| 1135 | "offset_bytes": 2486448, |
| 1136 | "nbytes": 80, |
| 1137 | "sha256": "b2183358a56efabfb0c3a6dcb29a11665d0faaf5f71a47f5121e85cffaa3dabd" |
| 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": 2486528, |
| 1148 | "nbytes": 9600, |
| 1149 | "sha256": "d5554e56c9a7eb0378b980af65f3cdf9a0cf2bfb371e0bc6dfb8890c9fe9ef1c" |
| 1150 | }, |
| 1151 | { |
| 1152 | "name": "res1.0.blocks.0.conv2.bias", |
| 1153 | "shape": [ |
| 1154 | 40 |
| 1155 | ], |
| 1156 | "dtype": "float16", |
| 1157 | "offset_bytes": 2496128, |
| 1158 | "nbytes": 80, |
| 1159 | "sha256": "44a4d0436d79a7857000a6b7ffa4f893d66ed4faa60a814b9afc4e1b02a1b195" |
| 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": 2496208, |
| 1170 | "nbytes": 9600, |
| 1171 | "sha256": "b48cd82476f4dffdfceee37300678e98f35a0537a2f6a3188cec7d8d41ce941a" |
| 1172 | }, |
| 1173 | { |
| 1174 | "name": "res1.0.blocks.1.conv1.bias", |
| 1175 | "shape": [ |
| 1176 | 40 |
| 1177 | ], |
| 1178 | "dtype": "float16", |
| 1179 | "offset_bytes": 2505808, |
| 1180 | "nbytes": 80, |
| 1181 | "sha256": "813fc7ef390220cf500eacdd14e058b14c00e25f4e64342fb19eb9c0c01ef21d" |
| 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": 2505888, |
| 1192 | "nbytes": 16000, |
| 1193 | "sha256": "e7c9711d47d05e343936234adf5ab92c9a178546c00d190f1e18edbf93c465c1" |
| 1194 | }, |
| 1195 | { |
| 1196 | "name": "res1.0.blocks.1.conv2.bias", |
| 1197 | "shape": [ |
| 1198 | 40 |
| 1199 | ], |
| 1200 | "dtype": "float16", |
| 1201 | "offset_bytes": 2521888, |
| 1202 | "nbytes": 80, |
| 1203 | "sha256": "142b042bf986bcbf0bc13f2151acb7b106a26b09eb270e622252f1c69749b297" |
| 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": 2521968, |
| 1214 | "nbytes": 16000, |
| 1215 | "sha256": "31097ddeb29ced98020dfa3aa7f977c77ab5dc733c9e55b14afc6ba942c8b057" |
| 1216 | }, |
| 1217 | { |
| 1218 | "name": "res1.0.blocks.2.conv1.bias", |
| 1219 | "shape": [ |
| 1220 | 40 |
| 1221 | ], |
| 1222 | "dtype": "float16", |
| 1223 | "offset_bytes": 2537968, |
| 1224 | "nbytes": 80, |
| 1225 | "sha256": "e51e356f6df42f7c6bd55b3a2dbf81c7714e5ddcba104258fabaa59bce6b3a1b" |
| 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": 2538048, |
| 1236 | "nbytes": 22400, |
| 1237 | "sha256": "d58ea596891c286ad756a437436d25ce522d5c10d791388837cd11cabf42af8b" |
| 1238 | }, |
| 1239 | { |
| 1240 | "name": "res1.0.blocks.2.conv2.bias", |
| 1241 | "shape": [ |
| 1242 | 40 |
| 1243 | ], |
| 1244 | "dtype": "float16", |
| 1245 | "offset_bytes": 2560448, |
| 1246 | "nbytes": 80, |
| 1247 | "sha256": "022db8e7e2dc1967d0119385b3e32d0f75c1e6bd8ccb671adfdee583d85d3f30" |
| 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": 2560528, |
| 1258 | "nbytes": 22400, |
| 1259 | "sha256": "a29d9d3e9ea3b0d4aa73bdea069ba1422efbacf5d86be2b2625d3d66556ad45d" |
| 1260 | }, |
| 1261 | { |
| 1262 | "name": "res2.0.blocks.0.conv1.bias", |
| 1263 | "shape": [ |
| 1264 | 20 |
| 1265 | ], |
| 1266 | "dtype": "float16", |
| 1267 | "offset_bytes": 2582928, |
| 1268 | "nbytes": 40, |
| 1269 | "sha256": "f4a87c99340932a783c13be80705eb1783990a1e9d6bd1f91807b2eec6c77d2d" |
| 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": 2582968, |
| 1280 | "nbytes": 2400, |
| 1281 | "sha256": "86831ec67945c106c674683aa8f130abc010dafd25eb7fc04e3e5bffea93378a" |
| 1282 | }, |
| 1283 | { |
| 1284 | "name": "res2.0.blocks.0.conv2.bias", |
| 1285 | "shape": [ |
| 1286 | 20 |
| 1287 | ], |
| 1288 | "dtype": "float16", |
| 1289 | "offset_bytes": 2585368, |
| 1290 | "nbytes": 40, |
| 1291 | "sha256": "6b0417523d954430d25ffe8d1eaac82cedbd28e9eaabb13ce1915d8c12448eef" |
| 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": 2585408, |
| 1302 | "nbytes": 2400, |
| 1303 | "sha256": "45d0d41d2a4e7599fad1328699f87492f43249e594c43d85d63c0ebb10bbce05" |
| 1304 | }, |
| 1305 | { |
| 1306 | "name": "res2.0.blocks.1.conv1.bias", |
| 1307 | "shape": [ |
| 1308 | 20 |
| 1309 | ], |
| 1310 | "dtype": "float16", |
| 1311 | "offset_bytes": 2587808, |
| 1312 | "nbytes": 40, |
| 1313 | "sha256": "4ebe10675d3d1e00b9ac6d5c3949f5215a450ddd9eedf1c97585f25489300510" |
| 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": 2587848, |
| 1324 | "nbytes": 4000, |
| 1325 | "sha256": "c3fc5bd66771b02c0dd16e7f89bba65660a96d448757930a28e9fe932c30a50a" |
| 1326 | }, |
| 1327 | { |
| 1328 | "name": "res2.0.blocks.1.conv2.bias", |
| 1329 | "shape": [ |
| 1330 | 20 |
| 1331 | ], |
| 1332 | "dtype": "float16", |
| 1333 | "offset_bytes": 2591848, |
| 1334 | "nbytes": 40, |
| 1335 | "sha256": "be2580569553dcb99e5f50a80fc128b38b9afe15e63380afb26201dde819278b" |
| 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": 2591888, |
| 1346 | "nbytes": 4000, |
| 1347 | "sha256": "327fe0fb66dc2d4ef2e244dcb9db79421e4e7948f95f26ecb3267c174cd4380d" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "res2.0.blocks.2.conv1.bias", |
| 1351 | "shape": [ |
| 1352 | 20 |
| 1353 | ], |
| 1354 | "dtype": "float16", |
| 1355 | "offset_bytes": 2595888, |
| 1356 | "nbytes": 40, |
| 1357 | "sha256": "1985074fcf7078ca404bfd15edd9ff9264c6778cfe6ae516336bbdd7917d9efd" |
| 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": 2595928, |
| 1368 | "nbytes": 5600, |
| 1369 | "sha256": "bc9c651da4bb0abaa28ace8dc524a3ef5acf0f01d3df1f9ee2b69fb2fa37f546" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "res2.0.blocks.2.conv2.bias", |
| 1373 | "shape": [ |
| 1374 | 20 |
| 1375 | ], |
| 1376 | "dtype": "float16", |
| 1377 | "offset_bytes": 2601528, |
| 1378 | "nbytes": 40, |
| 1379 | "sha256": "47446140d7aec326f3d68ae2b977f683371a2e1297983823b10c42eb9b139885" |
| 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": 2601568, |
| 1390 | "nbytes": 5600, |
| 1391 | "sha256": "e01c0442f42d8ea6b967d28291030fdff89e0ff8b7297afa83081d2f6bd77992" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "up0.bias", |
| 1395 | "shape": [ |
| 1396 | 80 |
| 1397 | ], |
| 1398 | "dtype": "float16", |
| 1399 | "offset_bytes": 2607168, |
| 1400 | "nbytes": 160, |
| 1401 | "sha256": "7eb668eb0730d1e4a75a0c53cf54be76f3b8102ac4db26d682a81e5b30b87e6c" |
| 1402 | }, |
| 1403 | { |
| 1404 | "name": "up0.weight", |
| 1405 | "shape": [ |
| 1406 | 160, |
| 1407 | 80, |
| 1408 | 16 |
| 1409 | ], |
| 1410 | "dtype": "float16", |
| 1411 | "offset_bytes": 2607328, |
| 1412 | "nbytes": 409600, |
| 1413 | "sha256": "2551dbd86ad6d74b8ccaa28fb720e5382ad14a512fe9c5e193d4ae76a124ed87" |
| 1414 | }, |
| 1415 | { |
| 1416 | "name": "up1.bias", |
| 1417 | "shape": [ |
| 1418 | 40 |
| 1419 | ], |
| 1420 | "dtype": "float16", |
| 1421 | "offset_bytes": 3016928, |
| 1422 | "nbytes": 80, |
| 1423 | "sha256": "8f875476fb7e8868d6790f77b062f86effa518aa55fc24b739a9c53428221b45" |
| 1424 | }, |
| 1425 | { |
| 1426 | "name": "up1.weight", |
| 1427 | "shape": [ |
| 1428 | 80, |
| 1429 | 40, |
| 1430 | 16 |
| 1431 | ], |
| 1432 | "dtype": "float16", |
| 1433 | "offset_bytes": 3017008, |
| 1434 | "nbytes": 102400, |
| 1435 | "sha256": "1ff2f1f7a5af58d6af7e3797f1bc1de751036d72e6128be237ed579dd354acb7" |
| 1436 | }, |
| 1437 | { |
| 1438 | "name": "up2.bias", |
| 1439 | "shape": [ |
| 1440 | 20 |
| 1441 | ], |
| 1442 | "dtype": "float16", |
| 1443 | "offset_bytes": 3119408, |
| 1444 | "nbytes": 40, |
| 1445 | "sha256": "f82c305da30a80eb2c069ed25b583b6230f389da640e0a2bf4fe2fee3979f4d9" |
| 1446 | }, |
| 1447 | { |
| 1448 | "name": "up2.weight", |
| 1449 | "shape": [ |
| 1450 | 40, |
| 1451 | 20, |
| 1452 | 8 |
| 1453 | ], |
| 1454 | "dtype": "float16", |
| 1455 | "offset_bytes": 3119448, |
| 1456 | "nbytes": 12800, |
| 1457 | "sha256": "5a3f15429b426fe65dc73f43657bdf4812cc3ce56dda03fc5e83f82ea6551169" |
| 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": "ar", |
| 1488 | "phoneme_id_map_entries": 159, |
| 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": 3137272, |
| 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 | |