vi-vais1000-1p46m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "vi-vais1000-1p46m", |
| 4 | "language": "vi_VN", |
| 5 | "voice": "vi_VN-vais1000-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": 3130968, |
| 15 | "weights_sha256": "eef846da5cab066f2f60de37cf407709caaaa7a9675a76cb8f78552c5b133da5", |
| 16 | "total_parameters": 1565484, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/vi_VN-vais1000-medium-retrain-20260712/a5-duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "153ed2dc5115ebeeaaef24506cdbca3de94a8715ad77c9245709db747a539f84", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 143, |
| 24 | "hidden": 64, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 333, |
| 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": 136836, |
| 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": "be4672b29acb5b16cb4edd9c53b285d02652f985a7ab5c7fe7ce5f6ec233d04b" |
| 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": "29a22236facc3f3b7c350f6d681414f521020669fc9e1d468a66a0040a91445b" |
| 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": "15d7760f0f90c32bbdb1227f7618ab3d5861fdda1fa9b77977353cf5da5c837e" |
| 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": "625a878646993c3584d922943c9db201b11988b4308d1c7795d1a1db07ac95a9" |
| 88 | }, |
| 89 | { |
| 90 | "name": "blocks.0.scale", |
| 91 | "shape": [ |
| 92 | 1 |
| 93 | ], |
| 94 | "dtype": "float16", |
| 95 | "offset_bytes": 82176, |
| 96 | "nbytes": 2, |
| 97 | "sha256": "0ce8538bbf21a085344eec307a55fd7c288dca920cac8554bd27a0c83bfc5733" |
| 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": "1ba114b0975cccd34434606d9953cc68a0704aca50e79b3c41d162dad256cf0f" |
| 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": "73e94358a6e9c90b912b07d92513bd1cafdb3da338198b5e454cb947046270fb" |
| 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": "4980a37c80089c8faff2422ad0f1c665820ca6512478f010f52f16843ddf8313" |
| 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": "5372a4a53f895f0f8718bd4afc1a6417c3d0f04b6d7993eff0b7c387119dcf5d" |
| 142 | }, |
| 143 | { |
| 144 | "name": "blocks.1.scale", |
| 145 | "shape": [ |
| 146 | 1 |
| 147 | ], |
| 148 | "dtype": "float16", |
| 149 | "offset_bytes": 164354, |
| 150 | "nbytes": 2, |
| 151 | "sha256": "8148981d7167c7733c06893ecf025bbd7730ef7a72df6635d4d1444f358b7513" |
| 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": "f3f8b8e778512e649eb4375708ab8cb7ef4fe59d8747fdcbc5b0a23454014d27" |
| 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": "cdac8d29eb9a0fe2e6522bba6fb1df7dfacc55e7d185ef338546bb0cc5b95a26" |
| 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": "188581c9e6c01def0b21835aa817e9420eef426683b9670ef5eea443809831c2" |
| 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": "b9a0de6703d0f0a82d8d8b83bee1a0c52b42742b0bcaf3f43d105a99d5efe95c" |
| 196 | }, |
| 197 | { |
| 198 | "name": "blocks.2.scale", |
| 199 | "shape": [ |
| 200 | 1 |
| 201 | ], |
| 202 | "dtype": "float16", |
| 203 | "offset_bytes": 246532, |
| 204 | "nbytes": 2, |
| 205 | "sha256": "b424f5e5f5e149649a5d4c88abbf1891195726013b49afe0a9667938a17ce876" |
| 206 | }, |
| 207 | { |
| 208 | "name": "embedding.weight", |
| 209 | "shape": [ |
| 210 | 143, |
| 211 | 64 |
| 212 | ], |
| 213 | "dtype": "float16", |
| 214 | "offset_bytes": 246534, |
| 215 | "nbytes": 18304, |
| 216 | "sha256": "ad614d1e3145186a327b94a0e51f7284458a01adf34daac51f838ab0833c73a5" |
| 217 | }, |
| 218 | { |
| 219 | "name": "input_proj.bias", |
| 220 | "shape": [ |
| 221 | 64 |
| 222 | ], |
| 223 | "dtype": "float16", |
| 224 | "offset_bytes": 264838, |
| 225 | "nbytes": 128, |
| 226 | "sha256": "dadaf2e6fa5ad2062bb37c223c9f7f7d81a6237db134d5f6daa6323cfd3a6406" |
| 227 | }, |
| 228 | { |
| 229 | "name": "input_proj.weight", |
| 230 | "shape": [ |
| 231 | 64, |
| 232 | 67, |
| 233 | 1 |
| 234 | ], |
| 235 | "dtype": "float16", |
| 236 | "offset_bytes": 264966, |
| 237 | "nbytes": 8576, |
| 238 | "sha256": "e95ac2e409cdda5ac74790529acd0a74092a89d37ada8a508018eca4bf6b90a4" |
| 239 | }, |
| 240 | { |
| 241 | "name": "output.bias", |
| 242 | "shape": [ |
| 243 | 1 |
| 244 | ], |
| 245 | "dtype": "float16", |
| 246 | "offset_bytes": 273542, |
| 247 | "nbytes": 2, |
| 248 | "sha256": "b6d6d27baddc0c32288b4bcb268418f5c8abcea5b2722ae6e0e839cf08302503" |
| 249 | }, |
| 250 | { |
| 251 | "name": "output.weight", |
| 252 | "shape": [ |
| 253 | 1, |
| 254 | 64, |
| 255 | 1 |
| 256 | ], |
| 257 | "dtype": "float16", |
| 258 | "offset_bytes": 273544, |
| 259 | "nbytes": 128, |
| 260 | "sha256": "cd1d04dbed2da254e7a2a783fc1971c50d484d40cd4efef7b60a60a69e656501" |
| 261 | } |
| 262 | ] |
| 263 | }, |
| 264 | "acoustic": { |
| 265 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/vi_VN-vais1000-medium-retrain-20260712/a93-joint/latent-student.pt", |
| 266 | "checkpoint_sha256": "9baf111117b27f52726c396b0241e598c0eb28625c0c355264a8288fd63b800f", |
| 267 | "config": { |
| 268 | "architecture": "token_context", |
| 269 | "vocab_size": 143, |
| 270 | "hidden": 96, |
| 271 | "depth": 4, |
| 272 | "kernel_size": 5, |
| 273 | "out_channels": 192, |
| 274 | "token_depth": 3 |
| 275 | }, |
| 276 | "parameters": 697927, |
| 277 | "tensors": [ |
| 278 | { |
| 279 | "name": "embedding.weight", |
| 280 | "shape": [ |
| 281 | 143, |
| 282 | 96 |
| 283 | ], |
| 284 | "dtype": "float16", |
| 285 | "offset_bytes": 273672, |
| 286 | "nbytes": 27456, |
| 287 | "sha256": "2fd1b523d46c717803b8a8afdd3e422aa1aa280fcb13f9b89be0f4ab19735218" |
| 288 | }, |
| 289 | { |
| 290 | "name": "frame_blocks.0.net.0.bias", |
| 291 | "shape": [ |
| 292 | 96 |
| 293 | ], |
| 294 | "dtype": "float16", |
| 295 | "offset_bytes": 301128, |
| 296 | "nbytes": 192, |
| 297 | "sha256": "c84f73889b63b08eaaf9ffbf17f1ce655528eb5e291681c05bcaf710dcfe30ca" |
| 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": 301320, |
| 308 | "nbytes": 92160, |
| 309 | "sha256": "07dbbd8530861ae99bd9144a02dc988a613d0b248898e4b102714996a6a1bc3d" |
| 310 | }, |
| 311 | { |
| 312 | "name": "frame_blocks.0.net.2.bias", |
| 313 | "shape": [ |
| 314 | 96 |
| 315 | ], |
| 316 | "dtype": "float16", |
| 317 | "offset_bytes": 393480, |
| 318 | "nbytes": 192, |
| 319 | "sha256": "4304594af2d06bb06e95f6dfbb78d0ed59c6e7edf11b122dc0a3e460c3495aeb" |
| 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": 393672, |
| 330 | "nbytes": 92160, |
| 331 | "sha256": "2ce7da4d78234bd3585d92b99efcbbad2b99bd72102369c37b4599e946574f6c" |
| 332 | }, |
| 333 | { |
| 334 | "name": "frame_blocks.0.scale", |
| 335 | "shape": [ |
| 336 | 1 |
| 337 | ], |
| 338 | "dtype": "float16", |
| 339 | "offset_bytes": 485832, |
| 340 | "nbytes": 2, |
| 341 | "sha256": "9c52113363649628bca20e22f01ff9c2630ec1f856113477c1236144f961bfac" |
| 342 | }, |
| 343 | { |
| 344 | "name": "frame_blocks.1.net.0.bias", |
| 345 | "shape": [ |
| 346 | 96 |
| 347 | ], |
| 348 | "dtype": "float16", |
| 349 | "offset_bytes": 485834, |
| 350 | "nbytes": 192, |
| 351 | "sha256": "99944752444982eb6ac98461b9312f3dcdf9c0f0013abe0deb94db1423de07fb" |
| 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": 486026, |
| 362 | "nbytes": 92160, |
| 363 | "sha256": "f10026f6ed9d404d55556b99717e291ed08d7faa8cc0b60c0a2284a85e18a1c0" |
| 364 | }, |
| 365 | { |
| 366 | "name": "frame_blocks.1.net.2.bias", |
| 367 | "shape": [ |
| 368 | 96 |
| 369 | ], |
| 370 | "dtype": "float16", |
| 371 | "offset_bytes": 578186, |
| 372 | "nbytes": 192, |
| 373 | "sha256": "937afe056eb1438f7a5c31bb3c51504e005dae75b982490e293e29175166cd9f" |
| 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": 578378, |
| 384 | "nbytes": 92160, |
| 385 | "sha256": "d9a4aa2677b9e1035cae60dadda81344ceec0dad32197a8f5398e7781af9d6f4" |
| 386 | }, |
| 387 | { |
| 388 | "name": "frame_blocks.1.scale", |
| 389 | "shape": [ |
| 390 | 1 |
| 391 | ], |
| 392 | "dtype": "float16", |
| 393 | "offset_bytes": 670538, |
| 394 | "nbytes": 2, |
| 395 | "sha256": "72c512b89fe7ebd31cf8d8f6d160087d170498924d4d3b1cf27903d85647472f" |
| 396 | }, |
| 397 | { |
| 398 | "name": "frame_blocks.2.net.0.bias", |
| 399 | "shape": [ |
| 400 | 96 |
| 401 | ], |
| 402 | "dtype": "float16", |
| 403 | "offset_bytes": 670540, |
| 404 | "nbytes": 192, |
| 405 | "sha256": "696b7978c345f2d2285ea7f8d4701ab5ab9bf3b8e7120bd4b7bc3b7d8f400d28" |
| 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": 670732, |
| 416 | "nbytes": 92160, |
| 417 | "sha256": "f4ee5cbf9f877b91372e7501a80046d13872c55148aa057e287ec24f3aea186d" |
| 418 | }, |
| 419 | { |
| 420 | "name": "frame_blocks.2.net.2.bias", |
| 421 | "shape": [ |
| 422 | 96 |
| 423 | ], |
| 424 | "dtype": "float16", |
| 425 | "offset_bytes": 762892, |
| 426 | "nbytes": 192, |
| 427 | "sha256": "e59a72151470f5737209261d71b6641cca7a044f5926803664e0cfa04e5c1fdd" |
| 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": 763084, |
| 438 | "nbytes": 92160, |
| 439 | "sha256": "ad28c1539f9782b554871cee3237bd5f55b73fe2369824230626d555bbbcaddc" |
| 440 | }, |
| 441 | { |
| 442 | "name": "frame_blocks.2.scale", |
| 443 | "shape": [ |
| 444 | 1 |
| 445 | ], |
| 446 | "dtype": "float16", |
| 447 | "offset_bytes": 855244, |
| 448 | "nbytes": 2, |
| 449 | "sha256": "b45190bb84eba9c350921495650a16d83fd5ebc922e7c1793b99b368c939e728" |
| 450 | }, |
| 451 | { |
| 452 | "name": "frame_blocks.3.net.0.bias", |
| 453 | "shape": [ |
| 454 | 96 |
| 455 | ], |
| 456 | "dtype": "float16", |
| 457 | "offset_bytes": 855246, |
| 458 | "nbytes": 192, |
| 459 | "sha256": "c2a94e852385120d14d6a6e519993327dcc59313eb76590f6f93491cb329877e" |
| 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": 855438, |
| 470 | "nbytes": 92160, |
| 471 | "sha256": "cd9909b2199b5f675b203ae71bdbe3c53c507ca4d4ff29a8c5738b1d0370e283" |
| 472 | }, |
| 473 | { |
| 474 | "name": "frame_blocks.3.net.2.bias", |
| 475 | "shape": [ |
| 476 | 96 |
| 477 | ], |
| 478 | "dtype": "float16", |
| 479 | "offset_bytes": 947598, |
| 480 | "nbytes": 192, |
| 481 | "sha256": "e0fef7d0c0278b9effb3b4967ccf4deeaea74551d25868446a7516bf55d10c93" |
| 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": 947790, |
| 492 | "nbytes": 92160, |
| 493 | "sha256": "5ce4239c31d52f9a97fa6d60af56139153e5c4300c41c208068f0ed691c36ef3" |
| 494 | }, |
| 495 | { |
| 496 | "name": "frame_blocks.3.scale", |
| 497 | "shape": [ |
| 498 | 1 |
| 499 | ], |
| 500 | "dtype": "float16", |
| 501 | "offset_bytes": 1039950, |
| 502 | "nbytes": 2, |
| 503 | "sha256": "233958247b4c6bca6a75ae2fbf7b2c2e75c95ba109bc8c3e53daf12b4284ad6c" |
| 504 | }, |
| 505 | { |
| 506 | "name": "frame_input_proj.bias", |
| 507 | "shape": [ |
| 508 | 96 |
| 509 | ], |
| 510 | "dtype": "float16", |
| 511 | "offset_bytes": 1039952, |
| 512 | "nbytes": 192, |
| 513 | "sha256": "39b559bda55ec8acbd554311e29c84b68f35b1d6b72567c6fd4f2b24ad76da1e" |
| 514 | }, |
| 515 | { |
| 516 | "name": "frame_input_proj.weight", |
| 517 | "shape": [ |
| 518 | 96, |
| 519 | 99, |
| 520 | 1 |
| 521 | ], |
| 522 | "dtype": "float16", |
| 523 | "offset_bytes": 1040144, |
| 524 | "nbytes": 19008, |
| 525 | "sha256": "145933a68e78be33d8622ff689e51cceb89d788f63d0d8dbbe5b97f036e22e75" |
| 526 | }, |
| 527 | { |
| 528 | "name": "output.bias", |
| 529 | "shape": [ |
| 530 | 192 |
| 531 | ], |
| 532 | "dtype": "float16", |
| 533 | "offset_bytes": 1059152, |
| 534 | "nbytes": 384, |
| 535 | "sha256": "e100b4364778c41ae0341423216da1e7de6b5af3477f78c15f0d4395da2391dc" |
| 536 | }, |
| 537 | { |
| 538 | "name": "output.weight", |
| 539 | "shape": [ |
| 540 | 192, |
| 541 | 96, |
| 542 | 1 |
| 543 | ], |
| 544 | "dtype": "float16", |
| 545 | "offset_bytes": 1059536, |
| 546 | "nbytes": 36864, |
| 547 | "sha256": "afc2d870b2068a148d268f59e5c1deec83a984c5acfe14054422a3192dac4c35" |
| 548 | }, |
| 549 | { |
| 550 | "name": "token_blocks.0.net.0.bias", |
| 551 | "shape": [ |
| 552 | 96 |
| 553 | ], |
| 554 | "dtype": "float16", |
| 555 | "offset_bytes": 1096400, |
| 556 | "nbytes": 192, |
| 557 | "sha256": "261ea0dfa62b8d97bdbba21824e8265a52976b914cfeb8d52a71e15205f33aba" |
| 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": 1096592, |
| 568 | "nbytes": 92160, |
| 569 | "sha256": "a56bd7bb60e5aa961e14d857c9e426fc327cad842b10b07469e6f2d1a957632b" |
| 570 | }, |
| 571 | { |
| 572 | "name": "token_blocks.0.net.2.bias", |
| 573 | "shape": [ |
| 574 | 96 |
| 575 | ], |
| 576 | "dtype": "float16", |
| 577 | "offset_bytes": 1188752, |
| 578 | "nbytes": 192, |
| 579 | "sha256": "a31e88a5535e1561bb2860bd048b2068873dfaed14774e35f7fd10b193cda491" |
| 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": 1188944, |
| 590 | "nbytes": 92160, |
| 591 | "sha256": "5fddd164b69221fb2590728bf7211fa850c51caaf0ad3da53e835616a7bdaff7" |
| 592 | }, |
| 593 | { |
| 594 | "name": "token_blocks.0.scale", |
| 595 | "shape": [ |
| 596 | 1 |
| 597 | ], |
| 598 | "dtype": "float16", |
| 599 | "offset_bytes": 1281104, |
| 600 | "nbytes": 2, |
| 601 | "sha256": "14af05c14f85808d5086c9b765699132eb5779c69667dd325b64d33a3c942d3b" |
| 602 | }, |
| 603 | { |
| 604 | "name": "token_blocks.1.net.0.bias", |
| 605 | "shape": [ |
| 606 | 96 |
| 607 | ], |
| 608 | "dtype": "float16", |
| 609 | "offset_bytes": 1281106, |
| 610 | "nbytes": 192, |
| 611 | "sha256": "fff092cfb73c15d2adfc9c847a9569c31d0092a709e9f6d02460fbb20b75d79f" |
| 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": 1281298, |
| 622 | "nbytes": 92160, |
| 623 | "sha256": "2b95ab3a2bf5b14577f87e21b5397167c0f8e5073d5c1ed2f1cb47fa6be6ac44" |
| 624 | }, |
| 625 | { |
| 626 | "name": "token_blocks.1.net.2.bias", |
| 627 | "shape": [ |
| 628 | 96 |
| 629 | ], |
| 630 | "dtype": "float16", |
| 631 | "offset_bytes": 1373458, |
| 632 | "nbytes": 192, |
| 633 | "sha256": "ec02fdc6055729b29054b1a95203da378b8820f92a147e720e26d8a384e7b2de" |
| 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": 1373650, |
| 644 | "nbytes": 92160, |
| 645 | "sha256": "e26be4c3187c40c5aa45eb01f95c1c5f8281b32bb8097719d9d147f1f2558aa5" |
| 646 | }, |
| 647 | { |
| 648 | "name": "token_blocks.1.scale", |
| 649 | "shape": [ |
| 650 | 1 |
| 651 | ], |
| 652 | "dtype": "float16", |
| 653 | "offset_bytes": 1465810, |
| 654 | "nbytes": 2, |
| 655 | "sha256": "b712f985f8f595900e35707ed2edc1aa8c6cf3dd9df0b675260fc8758fca1f24" |
| 656 | }, |
| 657 | { |
| 658 | "name": "token_blocks.2.net.0.bias", |
| 659 | "shape": [ |
| 660 | 96 |
| 661 | ], |
| 662 | "dtype": "float16", |
| 663 | "offset_bytes": 1465812, |
| 664 | "nbytes": 192, |
| 665 | "sha256": "e80be4895c2d722caf1b9742841f81a3316e91fe60e941266a363704ceeae3df" |
| 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": 1466004, |
| 676 | "nbytes": 92160, |
| 677 | "sha256": "224be7a9d0f134284c7c6e85a666e4a0b6b474fe057b20613541907848344e48" |
| 678 | }, |
| 679 | { |
| 680 | "name": "token_blocks.2.net.2.bias", |
| 681 | "shape": [ |
| 682 | 96 |
| 683 | ], |
| 684 | "dtype": "float16", |
| 685 | "offset_bytes": 1558164, |
| 686 | "nbytes": 192, |
| 687 | "sha256": "80513ce10f2e20c1088c939a1140837affde470fc279eded44409eab970b69b9" |
| 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": 1558356, |
| 698 | "nbytes": 92160, |
| 699 | "sha256": "896d6b0c057c7642cdc7cbe8f055ef190e2e913dce7c50441e599d199f66a6c9" |
| 700 | }, |
| 701 | { |
| 702 | "name": "token_blocks.2.scale", |
| 703 | "shape": [ |
| 704 | 1 |
| 705 | ], |
| 706 | "dtype": "float16", |
| 707 | "offset_bytes": 1650516, |
| 708 | "nbytes": 2, |
| 709 | "sha256": "bf91c899184eb0ca745d11008ce80aaf94de6b201cfa4ac7c66fc3911421d53e" |
| 710 | }, |
| 711 | { |
| 712 | "name": "token_input_proj.bias", |
| 713 | "shape": [ |
| 714 | 96 |
| 715 | ], |
| 716 | "dtype": "float16", |
| 717 | "offset_bytes": 1650518, |
| 718 | "nbytes": 192, |
| 719 | "sha256": "d5695e684a410a6a3093d29aab6469a8811c6b09afd9cbac1fb983b6fe0c09bf" |
| 720 | }, |
| 721 | { |
| 722 | "name": "token_input_proj.weight", |
| 723 | "shape": [ |
| 724 | 96, |
| 725 | 98, |
| 726 | 1 |
| 727 | ], |
| 728 | "dtype": "float16", |
| 729 | "offset_bytes": 1650710, |
| 730 | "nbytes": 18816, |
| 731 | "sha256": "ba2f7910f552d406a12e2ca2bac9b35b89d0a0a1bd5aa31c1362e0adb2a67763" |
| 732 | } |
| 733 | ] |
| 734 | }, |
| 735 | "decoder": { |
| 736 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/vi_VN-vais1000-medium-retrain-20260712/a93-joint/decoder-student.pt", |
| 737 | "checkpoint_sha256": "f178f387439d9e6bb51f1440cdee4e21863801c32873c876bd6c95d204eec20c", |
| 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/vi_VN-vais1000-medium-retrain-20260712/a91-decoder-piperlite/decoder-student.pt", |
| 786 | "decoder_init": { |
| 787 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/vi_VN-vais1000-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/vi_VN-vais1000-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": 1669526, |
| 908 | "nbytes": 2, |
| 909 | "sha256": "6ef6b31980d1683fafa1542c79a506b3fab6497e23e618f500d33ad65507eabb" |
| 910 | }, |
| 911 | { |
| 912 | "name": "post.weight", |
| 913 | "shape": [ |
| 914 | 1, |
| 915 | 20, |
| 916 | 7 |
| 917 | ], |
| 918 | "dtype": "float16", |
| 919 | "offset_bytes": 1669528, |
| 920 | "nbytes": 280, |
| 921 | "sha256": "9a05fb89896a6c9d43e1b43cb3c206bd9ee525b9d8209dc4d36efdaf97c1d727" |
| 922 | }, |
| 923 | { |
| 924 | "name": "pre.bias", |
| 925 | "shape": [ |
| 926 | 160 |
| 927 | ], |
| 928 | "dtype": "float16", |
| 929 | "offset_bytes": 1669808, |
| 930 | "nbytes": 320, |
| 931 | "sha256": "fe0eef310eeacc89050fdbdc78b9b52bc21830f9032684e09e680052d63b6b6f" |
| 932 | }, |
| 933 | { |
| 934 | "name": "pre.weight", |
| 935 | "shape": [ |
| 936 | 160, |
| 937 | 192, |
| 938 | 7 |
| 939 | ], |
| 940 | "dtype": "float16", |
| 941 | "offset_bytes": 1670128, |
| 942 | "nbytes": 430080, |
| 943 | "sha256": "7b8921780a927b2f6fd13fe6b2b27df6dca161a3c094f67bff418aa5a4c21007" |
| 944 | }, |
| 945 | { |
| 946 | "name": "res0.0.blocks.0.conv1.bias", |
| 947 | "shape": [ |
| 948 | 80 |
| 949 | ], |
| 950 | "dtype": "float16", |
| 951 | "offset_bytes": 2100208, |
| 952 | "nbytes": 160, |
| 953 | "sha256": "97ce66e619f5af163c1a717d39aa4a699e04fde92ba569c6300c13ce84d62501" |
| 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": 2100368, |
| 964 | "nbytes": 38400, |
| 965 | "sha256": "9584ead382c160c56ce9cb6ea7edc45d28e96fe78c73a6c399bf74ba9d9f48f6" |
| 966 | }, |
| 967 | { |
| 968 | "name": "res0.0.blocks.0.conv2.bias", |
| 969 | "shape": [ |
| 970 | 80 |
| 971 | ], |
| 972 | "dtype": "float16", |
| 973 | "offset_bytes": 2138768, |
| 974 | "nbytes": 160, |
| 975 | "sha256": "a55d93dc461e609de114f83fe6e336b86387e7a7d07ecdea98fce0bb82b8f0c8" |
| 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": 2138928, |
| 986 | "nbytes": 38400, |
| 987 | "sha256": "dcd4d70d2c485d3a0b9422b43ea2d22fadaeb8d699647aaa2ae11fbaf6d2d533" |
| 988 | }, |
| 989 | { |
| 990 | "name": "res0.0.blocks.1.conv1.bias", |
| 991 | "shape": [ |
| 992 | 80 |
| 993 | ], |
| 994 | "dtype": "float16", |
| 995 | "offset_bytes": 2177328, |
| 996 | "nbytes": 160, |
| 997 | "sha256": "bec83d09a3c8f48dca7c0f64f6f321402a1192e5595f81f269d43356f07576ec" |
| 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": 2177488, |
| 1008 | "nbytes": 64000, |
| 1009 | "sha256": "8c24b10107c648ffa4b16d84009a4a431d6ee7c7b4b0bdd2289606785f253c0a" |
| 1010 | }, |
| 1011 | { |
| 1012 | "name": "res0.0.blocks.1.conv2.bias", |
| 1013 | "shape": [ |
| 1014 | 80 |
| 1015 | ], |
| 1016 | "dtype": "float16", |
| 1017 | "offset_bytes": 2241488, |
| 1018 | "nbytes": 160, |
| 1019 | "sha256": "8f88a2ff5680de596383454e0931544c093c90328c522b53af462399aba5a560" |
| 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": 2241648, |
| 1030 | "nbytes": 64000, |
| 1031 | "sha256": "02d44016af5479998a42a3461343433472579793d5c433858ab88904e24a744f" |
| 1032 | }, |
| 1033 | { |
| 1034 | "name": "res0.0.blocks.2.conv1.bias", |
| 1035 | "shape": [ |
| 1036 | 80 |
| 1037 | ], |
| 1038 | "dtype": "float16", |
| 1039 | "offset_bytes": 2305648, |
| 1040 | "nbytes": 160, |
| 1041 | "sha256": "ff7db24aeff812cb3a3c2e5756c9c5fd046f79cf12730827ec5b5449d592dfea" |
| 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": 2305808, |
| 1052 | "nbytes": 89600, |
| 1053 | "sha256": "5af96a92769d857f927ef3ab6b87b97d98c4e6818559119e108488b6138c5955" |
| 1054 | }, |
| 1055 | { |
| 1056 | "name": "res0.0.blocks.2.conv2.bias", |
| 1057 | "shape": [ |
| 1058 | 80 |
| 1059 | ], |
| 1060 | "dtype": "float16", |
| 1061 | "offset_bytes": 2395408, |
| 1062 | "nbytes": 160, |
| 1063 | "sha256": "4b29c87de11566353dd226bfbee1040edbdf6341ce1d29b741c85eadecddf827" |
| 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": 2395568, |
| 1074 | "nbytes": 89600, |
| 1075 | "sha256": "841d2c4240f2a8092bfb2d82a3c47b326196fb0f8f61cc7ba32be4a94e4691b5" |
| 1076 | }, |
| 1077 | { |
| 1078 | "name": "res1.0.blocks.0.conv1.bias", |
| 1079 | "shape": [ |
| 1080 | 40 |
| 1081 | ], |
| 1082 | "dtype": "float16", |
| 1083 | "offset_bytes": 2485168, |
| 1084 | "nbytes": 80, |
| 1085 | "sha256": "8c8b821454731b4deb729b3711d4778d44a847e24ca2a48ac459ad391871df70" |
| 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": 2485248, |
| 1096 | "nbytes": 9600, |
| 1097 | "sha256": "15933fc1ee375486a64d22e030e1955f7b7e60bcc9262be517602dd118fbdb62" |
| 1098 | }, |
| 1099 | { |
| 1100 | "name": "res1.0.blocks.0.conv2.bias", |
| 1101 | "shape": [ |
| 1102 | 40 |
| 1103 | ], |
| 1104 | "dtype": "float16", |
| 1105 | "offset_bytes": 2494848, |
| 1106 | "nbytes": 80, |
| 1107 | "sha256": "6e4b3bfce2ecd32a8e0f3769bb68aa0ba7ab243a6f3eaf8f1973484a3e3126d4" |
| 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": 2494928, |
| 1118 | "nbytes": 9600, |
| 1119 | "sha256": "2d78b7143859260f0bc0436110b6ba0aec1386486110f4adc34b75560391f9dc" |
| 1120 | }, |
| 1121 | { |
| 1122 | "name": "res1.0.blocks.1.conv1.bias", |
| 1123 | "shape": [ |
| 1124 | 40 |
| 1125 | ], |
| 1126 | "dtype": "float16", |
| 1127 | "offset_bytes": 2504528, |
| 1128 | "nbytes": 80, |
| 1129 | "sha256": "726f9cb01fdb39b2fea6d8575fec088dde5dcf8646c0802fee4f9ff91db2608a" |
| 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": 2504608, |
| 1140 | "nbytes": 16000, |
| 1141 | "sha256": "df352e07be292537e4474feb18c22cc6814b9f7673f111b12e8cb12507cb4d82" |
| 1142 | }, |
| 1143 | { |
| 1144 | "name": "res1.0.blocks.1.conv2.bias", |
| 1145 | "shape": [ |
| 1146 | 40 |
| 1147 | ], |
| 1148 | "dtype": "float16", |
| 1149 | "offset_bytes": 2520608, |
| 1150 | "nbytes": 80, |
| 1151 | "sha256": "1599e60dcdb8d7f90a896597f170832dd850c80956453ec29c338686f0700cf8" |
| 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": 2520688, |
| 1162 | "nbytes": 16000, |
| 1163 | "sha256": "3662ee7440c76f5cf28382362dbcd9bd9238dbbaf8501bccb941815be31d0a85" |
| 1164 | }, |
| 1165 | { |
| 1166 | "name": "res1.0.blocks.2.conv1.bias", |
| 1167 | "shape": [ |
| 1168 | 40 |
| 1169 | ], |
| 1170 | "dtype": "float16", |
| 1171 | "offset_bytes": 2536688, |
| 1172 | "nbytes": 80, |
| 1173 | "sha256": "6fbc34e68f81006a405025a6a6279fa73406adcea80b2ccdce1c9fa8941ddb25" |
| 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": 2536768, |
| 1184 | "nbytes": 22400, |
| 1185 | "sha256": "18135d5a5936626ac70e82236c4584273f9d3011d8fdff72bdcfb63a05e58797" |
| 1186 | }, |
| 1187 | { |
| 1188 | "name": "res1.0.blocks.2.conv2.bias", |
| 1189 | "shape": [ |
| 1190 | 40 |
| 1191 | ], |
| 1192 | "dtype": "float16", |
| 1193 | "offset_bytes": 2559168, |
| 1194 | "nbytes": 80, |
| 1195 | "sha256": "9f635aa797f0d39f1ffb3043fa77afd133037c9c58ec12606e093e9014fec38b" |
| 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": 2559248, |
| 1206 | "nbytes": 22400, |
| 1207 | "sha256": "84d0166f6dfeb8ffe87c6a0b4f92d3ca77993f76cabfc1cd8a7d2ec600f8ec92" |
| 1208 | }, |
| 1209 | { |
| 1210 | "name": "res2.0.blocks.0.conv1.bias", |
| 1211 | "shape": [ |
| 1212 | 20 |
| 1213 | ], |
| 1214 | "dtype": "float16", |
| 1215 | "offset_bytes": 2581648, |
| 1216 | "nbytes": 40, |
| 1217 | "sha256": "134faa502b7efcf9c675c63d9680959bf3e295be27dec8065f07d4d150545410" |
| 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": 2581688, |
| 1228 | "nbytes": 2400, |
| 1229 | "sha256": "47ee48f703cabb84d3f09f69a2a005443c997374df2ff6784f13e01619e18e08" |
| 1230 | }, |
| 1231 | { |
| 1232 | "name": "res2.0.blocks.0.conv2.bias", |
| 1233 | "shape": [ |
| 1234 | 20 |
| 1235 | ], |
| 1236 | "dtype": "float16", |
| 1237 | "offset_bytes": 2584088, |
| 1238 | "nbytes": 40, |
| 1239 | "sha256": "12e5a584c28e0e9e62b5f94cb6686880139973d353e98102badcb998b38bbe2c" |
| 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": 2584128, |
| 1250 | "nbytes": 2400, |
| 1251 | "sha256": "8ea86240c192da70d57ecbe7125dea3a36a339ade5e7b8a6c886c6a4a6f7b0ed" |
| 1252 | }, |
| 1253 | { |
| 1254 | "name": "res2.0.blocks.1.conv1.bias", |
| 1255 | "shape": [ |
| 1256 | 20 |
| 1257 | ], |
| 1258 | "dtype": "float16", |
| 1259 | "offset_bytes": 2586528, |
| 1260 | "nbytes": 40, |
| 1261 | "sha256": "569bc1a525b0be7fac79d2a1b6149391cf9c1a0a30380f227e911ea92562b109" |
| 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": 2586568, |
| 1272 | "nbytes": 4000, |
| 1273 | "sha256": "6fb15a0841b73feb6f9327fc2a28119199ac5d7fe88e126a2f7848997399c1ba" |
| 1274 | }, |
| 1275 | { |
| 1276 | "name": "res2.0.blocks.1.conv2.bias", |
| 1277 | "shape": [ |
| 1278 | 20 |
| 1279 | ], |
| 1280 | "dtype": "float16", |
| 1281 | "offset_bytes": 2590568, |
| 1282 | "nbytes": 40, |
| 1283 | "sha256": "4fce4b930ccd53f442c794642d0e1361277c0b61f41850fe9114880f191e7bd9" |
| 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": 2590608, |
| 1294 | "nbytes": 4000, |
| 1295 | "sha256": "2d95b097299e1ed7a3415866a9975acd6763924e3f80149b2d178107a1039afe" |
| 1296 | }, |
| 1297 | { |
| 1298 | "name": "res2.0.blocks.2.conv1.bias", |
| 1299 | "shape": [ |
| 1300 | 20 |
| 1301 | ], |
| 1302 | "dtype": "float16", |
| 1303 | "offset_bytes": 2594608, |
| 1304 | "nbytes": 40, |
| 1305 | "sha256": "3752b33c97b2460ccb2870e872207842b51afb44515e63864a069a16322d15d7" |
| 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": 2594648, |
| 1316 | "nbytes": 5600, |
| 1317 | "sha256": "0082da53c9f8a6a8ab2a2e7d1b99f878b39695a50912d346b55d818ed4a88129" |
| 1318 | }, |
| 1319 | { |
| 1320 | "name": "res2.0.blocks.2.conv2.bias", |
| 1321 | "shape": [ |
| 1322 | 20 |
| 1323 | ], |
| 1324 | "dtype": "float16", |
| 1325 | "offset_bytes": 2600248, |
| 1326 | "nbytes": 40, |
| 1327 | "sha256": "aab683aa3627822bccb44a098482302840bab2dcab92fdfb797f7ad30363caea" |
| 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": 2600288, |
| 1338 | "nbytes": 5600, |
| 1339 | "sha256": "fb657f0c3f79a3cbb8dd83b5dc0dad0940cdee70532e67ef47f29b2a26bbad0a" |
| 1340 | }, |
| 1341 | { |
| 1342 | "name": "up0.bias", |
| 1343 | "shape": [ |
| 1344 | 80 |
| 1345 | ], |
| 1346 | "dtype": "float16", |
| 1347 | "offset_bytes": 2605888, |
| 1348 | "nbytes": 160, |
| 1349 | "sha256": "1e5b256eba5f42edf0e3aa643f0f781db5afd728a00a51bbbd86741a9d0d99b0" |
| 1350 | }, |
| 1351 | { |
| 1352 | "name": "up0.weight", |
| 1353 | "shape": [ |
| 1354 | 160, |
| 1355 | 80, |
| 1356 | 16 |
| 1357 | ], |
| 1358 | "dtype": "float16", |
| 1359 | "offset_bytes": 2606048, |
| 1360 | "nbytes": 409600, |
| 1361 | "sha256": "af48e4cea5156aa5f885b09d9393d0ba2ccfe67b7f59b82a8357098f52841fff" |
| 1362 | }, |
| 1363 | { |
| 1364 | "name": "up1.bias", |
| 1365 | "shape": [ |
| 1366 | 40 |
| 1367 | ], |
| 1368 | "dtype": "float16", |
| 1369 | "offset_bytes": 3015648, |
| 1370 | "nbytes": 80, |
| 1371 | "sha256": "13dcaab5a4ce07d3c2d8e766e98f23e0c39d83216edcc0054c747f9346659800" |
| 1372 | }, |
| 1373 | { |
| 1374 | "name": "up1.weight", |
| 1375 | "shape": [ |
| 1376 | 80, |
| 1377 | 40, |
| 1378 | 16 |
| 1379 | ], |
| 1380 | "dtype": "float16", |
| 1381 | "offset_bytes": 3015728, |
| 1382 | "nbytes": 102400, |
| 1383 | "sha256": "9a9a7aa309d7b0a3b99654419d3270590ebe03f60b2c8b142e796eb8e529b2b8" |
| 1384 | }, |
| 1385 | { |
| 1386 | "name": "up2.bias", |
| 1387 | "shape": [ |
| 1388 | 20 |
| 1389 | ], |
| 1390 | "dtype": "float16", |
| 1391 | "offset_bytes": 3118128, |
| 1392 | "nbytes": 40, |
| 1393 | "sha256": "e0336eb36896b4f5a92ab0ec29f287d873f2765ec0b7750f7ec09d1baee7ba71" |
| 1394 | }, |
| 1395 | { |
| 1396 | "name": "up2.weight", |
| 1397 | "shape": [ |
| 1398 | 40, |
| 1399 | 20, |
| 1400 | 8 |
| 1401 | ], |
| 1402 | "dtype": "float16", |
| 1403 | "offset_bytes": 3118168, |
| 1404 | "nbytes": 12800, |
| 1405 | "sha256": "b3948d0f9749d9a2a719a652cebec08cf0d586ecc2a6e5369a7b60228e6813ed" |
| 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": null, |
| 1435 | "piper_espeak_voice": "vi", |
| 1436 | "phoneme_id_map_entries": 154, |
| 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": 3135828, |
| 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 | |