kristin-en-1p4m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "kristin-en-1p4m", |
| 4 | "language": "en_US", |
| 5 | "voice": "en_US-kristin-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": 1.08, |
| 11 | "duration_policy": "round(exp(log_duration) * duration_length_scale)" |
| 12 | }, |
| 13 | "weights_file": "weights.fp16.bin", |
| 14 | "weights_size_bytes": 2792302, |
| 15 | "weights_sha256": "4691dc92d21f5b979dafd6d3127bebd2721bc019fbc06ce97034e8fed03d6194", |
| 16 | "total_parameters": 1396151, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "releases/kristin-20260707/web/duration-student-36k.pt", |
| 20 | "checkpoint_sha256": "7ee26b221bfc9e3b5a284a257c55566a53bd628efb3905732d4c8f2515ac6cbc", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 127, |
| 24 | "hidden": 32, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 269, |
| 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": 36164, |
| 44 | "tensors": [ |
| 45 | { |
| 46 | "name": "blocks.0.net.0.bias", |
| 47 | "shape": [ |
| 48 | 32 |
| 49 | ], |
| 50 | "dtype": "float16", |
| 51 | "offset_bytes": 0, |
| 52 | "nbytes": 64, |
| 53 | "sha256": "bbbac684799c5cbc5c9173d6e33663a825dd2b2673683c70461ca57bde89869a" |
| 54 | }, |
| 55 | { |
| 56 | "name": "blocks.0.net.0.weight", |
| 57 | "shape": [ |
| 58 | 32, |
| 59 | 32, |
| 60 | 5 |
| 61 | ], |
| 62 | "dtype": "float16", |
| 63 | "offset_bytes": 64, |
| 64 | "nbytes": 10240, |
| 65 | "sha256": "747124cab7cf609fbdeeadbccb973830530fd4d62dd8c3c41de24a8527f3fd0e" |
| 66 | }, |
| 67 | { |
| 68 | "name": "blocks.0.net.2.bias", |
| 69 | "shape": [ |
| 70 | 32 |
| 71 | ], |
| 72 | "dtype": "float16", |
| 73 | "offset_bytes": 10304, |
| 74 | "nbytes": 64, |
| 75 | "sha256": "690a15ee63e680f150f6a8cfc002e871cf121f3b3fcef75ffc2d905398825959" |
| 76 | }, |
| 77 | { |
| 78 | "name": "blocks.0.net.2.weight", |
| 79 | "shape": [ |
| 80 | 32, |
| 81 | 32, |
| 82 | 5 |
| 83 | ], |
| 84 | "dtype": "float16", |
| 85 | "offset_bytes": 10368, |
| 86 | "nbytes": 10240, |
| 87 | "sha256": "43e98e023b9827633094c2bdb277c23d1cc1a82cb148c4857b9b34e3dbc61802" |
| 88 | }, |
| 89 | { |
| 90 | "name": "blocks.0.scale", |
| 91 | "shape": [ |
| 92 | 1 |
| 93 | ], |
| 94 | "dtype": "float16", |
| 95 | "offset_bytes": 20608, |
| 96 | "nbytes": 2, |
| 97 | "sha256": "f254a1e7d56840a4f2faa65b43f82d21e65add6e7376169cb5a51da30007822a" |
| 98 | }, |
| 99 | { |
| 100 | "name": "blocks.1.net.0.bias", |
| 101 | "shape": [ |
| 102 | 32 |
| 103 | ], |
| 104 | "dtype": "float16", |
| 105 | "offset_bytes": 20610, |
| 106 | "nbytes": 64, |
| 107 | "sha256": "1f1d5a8496103db7cd0bf9b285e33261064e38b15534f5c0cf7460025cd0d104" |
| 108 | }, |
| 109 | { |
| 110 | "name": "blocks.1.net.0.weight", |
| 111 | "shape": [ |
| 112 | 32, |
| 113 | 32, |
| 114 | 5 |
| 115 | ], |
| 116 | "dtype": "float16", |
| 117 | "offset_bytes": 20674, |
| 118 | "nbytes": 10240, |
| 119 | "sha256": "6cc0bb79ed28cbb234ec125e08aaa555d959ae8ba33ebf8ef2de75e1cd3b26c8" |
| 120 | }, |
| 121 | { |
| 122 | "name": "blocks.1.net.2.bias", |
| 123 | "shape": [ |
| 124 | 32 |
| 125 | ], |
| 126 | "dtype": "float16", |
| 127 | "offset_bytes": 30914, |
| 128 | "nbytes": 64, |
| 129 | "sha256": "afd73a7863f9f3c300075a704ba50cf5e33e00997e48da83b9497c93a18de08b" |
| 130 | }, |
| 131 | { |
| 132 | "name": "blocks.1.net.2.weight", |
| 133 | "shape": [ |
| 134 | 32, |
| 135 | 32, |
| 136 | 5 |
| 137 | ], |
| 138 | "dtype": "float16", |
| 139 | "offset_bytes": 30978, |
| 140 | "nbytes": 10240, |
| 141 | "sha256": "2cd42b4f6e96151e0f6b15b20850af31a2cbcb6d76b09fe3b8bd62b6e53d237a" |
| 142 | }, |
| 143 | { |
| 144 | "name": "blocks.1.scale", |
| 145 | "shape": [ |
| 146 | 1 |
| 147 | ], |
| 148 | "dtype": "float16", |
| 149 | "offset_bytes": 41218, |
| 150 | "nbytes": 2, |
| 151 | "sha256": "a174e6c73791db8d2cb8979ae152bcc98391d69f628883fe0587e07a9f461706" |
| 152 | }, |
| 153 | { |
| 154 | "name": "blocks.2.net.0.bias", |
| 155 | "shape": [ |
| 156 | 32 |
| 157 | ], |
| 158 | "dtype": "float16", |
| 159 | "offset_bytes": 41220, |
| 160 | "nbytes": 64, |
| 161 | "sha256": "7da2b215ea0cce48f8381c5faf1e0571ffe65992d1a9697c76f4d15423991771" |
| 162 | }, |
| 163 | { |
| 164 | "name": "blocks.2.net.0.weight", |
| 165 | "shape": [ |
| 166 | 32, |
| 167 | 32, |
| 168 | 5 |
| 169 | ], |
| 170 | "dtype": "float16", |
| 171 | "offset_bytes": 41284, |
| 172 | "nbytes": 10240, |
| 173 | "sha256": "5bb55fce2db8749b87035f3380cb52bbafecc2f43924b07475ed6da19b1f3c27" |
| 174 | }, |
| 175 | { |
| 176 | "name": "blocks.2.net.2.bias", |
| 177 | "shape": [ |
| 178 | 32 |
| 179 | ], |
| 180 | "dtype": "float16", |
| 181 | "offset_bytes": 51524, |
| 182 | "nbytes": 64, |
| 183 | "sha256": "c77b4432fc7c3b68a6c5df571e1702086abc4226d46e3d292076e0dbab512668" |
| 184 | }, |
| 185 | { |
| 186 | "name": "blocks.2.net.2.weight", |
| 187 | "shape": [ |
| 188 | 32, |
| 189 | 32, |
| 190 | 5 |
| 191 | ], |
| 192 | "dtype": "float16", |
| 193 | "offset_bytes": 51588, |
| 194 | "nbytes": 10240, |
| 195 | "sha256": "92c3e3d8892211451a7ada24ace74193b3b32cfd8a77fddc50f02cb083c97fbe" |
| 196 | }, |
| 197 | { |
| 198 | "name": "blocks.2.scale", |
| 199 | "shape": [ |
| 200 | 1 |
| 201 | ], |
| 202 | "dtype": "float16", |
| 203 | "offset_bytes": 61828, |
| 204 | "nbytes": 2, |
| 205 | "sha256": "79f277a9e2ae3f68fec86bec359310859c766561915a37a9b5b519c23321d949" |
| 206 | }, |
| 207 | { |
| 208 | "name": "embedding.weight", |
| 209 | "shape": [ |
| 210 | 127, |
| 211 | 32 |
| 212 | ], |
| 213 | "dtype": "float16", |
| 214 | "offset_bytes": 61830, |
| 215 | "nbytes": 8128, |
| 216 | "sha256": "dd21182edbe0c76ac39ac8e7b1e2cc14007eef34320ce668d5b408bb2102d556" |
| 217 | }, |
| 218 | { |
| 219 | "name": "input_proj.bias", |
| 220 | "shape": [ |
| 221 | 32 |
| 222 | ], |
| 223 | "dtype": "float16", |
| 224 | "offset_bytes": 69958, |
| 225 | "nbytes": 64, |
| 226 | "sha256": "4e956f5541654de03d6e576e70eb9c1c0e28d684daa71cff2dc3080218b074b0" |
| 227 | }, |
| 228 | { |
| 229 | "name": "input_proj.weight", |
| 230 | "shape": [ |
| 231 | 32, |
| 232 | 35, |
| 233 | 1 |
| 234 | ], |
| 235 | "dtype": "float16", |
| 236 | "offset_bytes": 70022, |
| 237 | "nbytes": 2240, |
| 238 | "sha256": "576720553f2f1408bbf8a94aeb6f30d8b573df235efeeda6ca3ae974bf780002" |
| 239 | }, |
| 240 | { |
| 241 | "name": "output.bias", |
| 242 | "shape": [ |
| 243 | 1 |
| 244 | ], |
| 245 | "dtype": "float16", |
| 246 | "offset_bytes": 72262, |
| 247 | "nbytes": 2, |
| 248 | "sha256": "c9ff6f268b747618589b2c4317d7879bd93469f371b17feb4c2837cd9638dc93" |
| 249 | }, |
| 250 | { |
| 251 | "name": "output.weight", |
| 252 | "shape": [ |
| 253 | 1, |
| 254 | 32, |
| 255 | 1 |
| 256 | ], |
| 257 | "dtype": "float16", |
| 258 | "offset_bytes": 72264, |
| 259 | "nbytes": 64, |
| 260 | "sha256": "7e6210a4d2d9d77c39c3d8de6bd28ac761a1bf733dcf221a16aa03da4d3a54ee" |
| 261 | } |
| 262 | ] |
| 263 | }, |
| 264 | "acoustic": { |
| 265 | "checkpoint": "releases/kristin-20260707/web/acoustic-latent-student-359k.pt", |
| 266 | "checkpoint_sha256": "10e5920e8fa785053d148c6c002d15b7cd27f1a3eff8e7384a7c37a820024922", |
| 267 | "config": { |
| 268 | "architecture": "token_context", |
| 269 | "vocab_size": 145, |
| 270 | "hidden": 64, |
| 271 | "depth": 5, |
| 272 | "kernel_size": 5, |
| 273 | "out_channels": 192, |
| 274 | "token_depth": 3 |
| 275 | }, |
| 276 | "parameters": 359112, |
| 277 | "tensors": [ |
| 278 | { |
| 279 | "name": "embedding.weight", |
| 280 | "shape": [ |
| 281 | 145, |
| 282 | 64 |
| 283 | ], |
| 284 | "dtype": "float16", |
| 285 | "offset_bytes": 72328, |
| 286 | "nbytes": 18560, |
| 287 | "sha256": "9c59f60981feba86ac0e7a8d703110dd7b9e4ee1aa46380a69913ccded6d77bd" |
| 288 | }, |
| 289 | { |
| 290 | "name": "frame_blocks.0.net.0.bias", |
| 291 | "shape": [ |
| 292 | 64 |
| 293 | ], |
| 294 | "dtype": "float16", |
| 295 | "offset_bytes": 90888, |
| 296 | "nbytes": 128, |
| 297 | "sha256": "3027e6d47b3c2c089992eb597c29f87164e6d401366e741775df99794bb5a569" |
| 298 | }, |
| 299 | { |
| 300 | "name": "frame_blocks.0.net.0.weight", |
| 301 | "shape": [ |
| 302 | 64, |
| 303 | 64, |
| 304 | 5 |
| 305 | ], |
| 306 | "dtype": "float16", |
| 307 | "offset_bytes": 91016, |
| 308 | "nbytes": 40960, |
| 309 | "sha256": "9e2617acc81812bc6f0662c0592f62b7cf98c520ba87fb54401705ad3b3eb60c" |
| 310 | }, |
| 311 | { |
| 312 | "name": "frame_blocks.0.net.2.bias", |
| 313 | "shape": [ |
| 314 | 64 |
| 315 | ], |
| 316 | "dtype": "float16", |
| 317 | "offset_bytes": 131976, |
| 318 | "nbytes": 128, |
| 319 | "sha256": "a0cad572345e8fc4011cb96c776c80d1735f451d71345712bfb1b294f96e08c4" |
| 320 | }, |
| 321 | { |
| 322 | "name": "frame_blocks.0.net.2.weight", |
| 323 | "shape": [ |
| 324 | 64, |
| 325 | 64, |
| 326 | 5 |
| 327 | ], |
| 328 | "dtype": "float16", |
| 329 | "offset_bytes": 132104, |
| 330 | "nbytes": 40960, |
| 331 | "sha256": "18f81efc7b282bdd007025a67c8553fe8c9df2d88dcc08976fb6b934696a3443" |
| 332 | }, |
| 333 | { |
| 334 | "name": "frame_blocks.0.scale", |
| 335 | "shape": [ |
| 336 | 1 |
| 337 | ], |
| 338 | "dtype": "float16", |
| 339 | "offset_bytes": 173064, |
| 340 | "nbytes": 2, |
| 341 | "sha256": "fb272fe0bd88ecdcd013cecc19d80779f5b6744c6612ff20a6d7f9e69fdedccb" |
| 342 | }, |
| 343 | { |
| 344 | "name": "frame_blocks.1.net.0.bias", |
| 345 | "shape": [ |
| 346 | 64 |
| 347 | ], |
| 348 | "dtype": "float16", |
| 349 | "offset_bytes": 173066, |
| 350 | "nbytes": 128, |
| 351 | "sha256": "eb5ce89b5ec8810efad02c81d832dba65695784afd18cb26dfcebbfee39e876a" |
| 352 | }, |
| 353 | { |
| 354 | "name": "frame_blocks.1.net.0.weight", |
| 355 | "shape": [ |
| 356 | 64, |
| 357 | 64, |
| 358 | 5 |
| 359 | ], |
| 360 | "dtype": "float16", |
| 361 | "offset_bytes": 173194, |
| 362 | "nbytes": 40960, |
| 363 | "sha256": "63fdab16b95d7bcd86c17ec6aab026109d66029053222336c2b98ad4924990b4" |
| 364 | }, |
| 365 | { |
| 366 | "name": "frame_blocks.1.net.2.bias", |
| 367 | "shape": [ |
| 368 | 64 |
| 369 | ], |
| 370 | "dtype": "float16", |
| 371 | "offset_bytes": 214154, |
| 372 | "nbytes": 128, |
| 373 | "sha256": "e5b1509c8c033c893d39b76f178700eaf5d5ec92530ce0b32cea974234e6d045" |
| 374 | }, |
| 375 | { |
| 376 | "name": "frame_blocks.1.net.2.weight", |
| 377 | "shape": [ |
| 378 | 64, |
| 379 | 64, |
| 380 | 5 |
| 381 | ], |
| 382 | "dtype": "float16", |
| 383 | "offset_bytes": 214282, |
| 384 | "nbytes": 40960, |
| 385 | "sha256": "6e838ff73fb376d1917a6d2e29dcf71bea0a4e04a8291e872d3f70051598953b" |
| 386 | }, |
| 387 | { |
| 388 | "name": "frame_blocks.1.scale", |
| 389 | "shape": [ |
| 390 | 1 |
| 391 | ], |
| 392 | "dtype": "float16", |
| 393 | "offset_bytes": 255242, |
| 394 | "nbytes": 2, |
| 395 | "sha256": "1ef8d7e6cf9a964d8d7a0ee1f3dc126b5600eec4ca5c3b5c65844c6382b6c604" |
| 396 | }, |
| 397 | { |
| 398 | "name": "frame_blocks.2.net.0.bias", |
| 399 | "shape": [ |
| 400 | 64 |
| 401 | ], |
| 402 | "dtype": "float16", |
| 403 | "offset_bytes": 255244, |
| 404 | "nbytes": 128, |
| 405 | "sha256": "e5b6680b0ed35a5a0590723718ab39f2a6934e3eb14ad9f1ecf9131d43289d92" |
| 406 | }, |
| 407 | { |
| 408 | "name": "frame_blocks.2.net.0.weight", |
| 409 | "shape": [ |
| 410 | 64, |
| 411 | 64, |
| 412 | 5 |
| 413 | ], |
| 414 | "dtype": "float16", |
| 415 | "offset_bytes": 255372, |
| 416 | "nbytes": 40960, |
| 417 | "sha256": "6986ef68763c148b456e2aea852acf5d566276325192d95a504cfccf6bc76d16" |
| 418 | }, |
| 419 | { |
| 420 | "name": "frame_blocks.2.net.2.bias", |
| 421 | "shape": [ |
| 422 | 64 |
| 423 | ], |
| 424 | "dtype": "float16", |
| 425 | "offset_bytes": 296332, |
| 426 | "nbytes": 128, |
| 427 | "sha256": "c7144362109dccb61daa60714bbe444a5f6cb892936c946796f3af52b9efcebe" |
| 428 | }, |
| 429 | { |
| 430 | "name": "frame_blocks.2.net.2.weight", |
| 431 | "shape": [ |
| 432 | 64, |
| 433 | 64, |
| 434 | 5 |
| 435 | ], |
| 436 | "dtype": "float16", |
| 437 | "offset_bytes": 296460, |
| 438 | "nbytes": 40960, |
| 439 | "sha256": "0da48facc652b6667396e1a0e236347d8597cc556e8479b49ec8b22020b963c6" |
| 440 | }, |
| 441 | { |
| 442 | "name": "frame_blocks.2.scale", |
| 443 | "shape": [ |
| 444 | 1 |
| 445 | ], |
| 446 | "dtype": "float16", |
| 447 | "offset_bytes": 337420, |
| 448 | "nbytes": 2, |
| 449 | "sha256": "2aa242e8be3308a7a4d166d30b08685bf3d6f6f04a48b12148e3489a4eb648a2" |
| 450 | }, |
| 451 | { |
| 452 | "name": "frame_blocks.3.net.0.bias", |
| 453 | "shape": [ |
| 454 | 64 |
| 455 | ], |
| 456 | "dtype": "float16", |
| 457 | "offset_bytes": 337422, |
| 458 | "nbytes": 128, |
| 459 | "sha256": "6e02cf7d92737712b0db26ebf64633d9115d88d22e610e8db78ea3bc1886aaec" |
| 460 | }, |
| 461 | { |
| 462 | "name": "frame_blocks.3.net.0.weight", |
| 463 | "shape": [ |
| 464 | 64, |
| 465 | 64, |
| 466 | 5 |
| 467 | ], |
| 468 | "dtype": "float16", |
| 469 | "offset_bytes": 337550, |
| 470 | "nbytes": 40960, |
| 471 | "sha256": "bc6986704f148b81e17873b0cb6dcfde2a610b58559e942c1922e2415f09d1ea" |
| 472 | }, |
| 473 | { |
| 474 | "name": "frame_blocks.3.net.2.bias", |
| 475 | "shape": [ |
| 476 | 64 |
| 477 | ], |
| 478 | "dtype": "float16", |
| 479 | "offset_bytes": 378510, |
| 480 | "nbytes": 128, |
| 481 | "sha256": "66eb30f2bf237c88a31f09d3f163a440f22892f022144cbd7040690ecdf907f0" |
| 482 | }, |
| 483 | { |
| 484 | "name": "frame_blocks.3.net.2.weight", |
| 485 | "shape": [ |
| 486 | 64, |
| 487 | 64, |
| 488 | 5 |
| 489 | ], |
| 490 | "dtype": "float16", |
| 491 | "offset_bytes": 378638, |
| 492 | "nbytes": 40960, |
| 493 | "sha256": "bf77ee6399837a7f50dda5daf9ef64688916cc115413abc362a313443a6d9b3e" |
| 494 | }, |
| 495 | { |
| 496 | "name": "frame_blocks.3.scale", |
| 497 | "shape": [ |
| 498 | 1 |
| 499 | ], |
| 500 | "dtype": "float16", |
| 501 | "offset_bytes": 419598, |
| 502 | "nbytes": 2, |
| 503 | "sha256": "3609f2b0c343eff3adeb2a1b76bbe045e02e2c3e6c76506e485057738bbd7a47" |
| 504 | }, |
| 505 | { |
| 506 | "name": "frame_blocks.4.net.0.bias", |
| 507 | "shape": [ |
| 508 | 64 |
| 509 | ], |
| 510 | "dtype": "float16", |
| 511 | "offset_bytes": 419600, |
| 512 | "nbytes": 128, |
| 513 | "sha256": "52352d6141c40b80c5f032dcde847c38469a289299bf884c02d5af21f62019f8" |
| 514 | }, |
| 515 | { |
| 516 | "name": "frame_blocks.4.net.0.weight", |
| 517 | "shape": [ |
| 518 | 64, |
| 519 | 64, |
| 520 | 5 |
| 521 | ], |
| 522 | "dtype": "float16", |
| 523 | "offset_bytes": 419728, |
| 524 | "nbytes": 40960, |
| 525 | "sha256": "dcc153f8157a7cd88c1b1a32b551ea56903da0e1ff016432438babb40de22bb6" |
| 526 | }, |
| 527 | { |
| 528 | "name": "frame_blocks.4.net.2.bias", |
| 529 | "shape": [ |
| 530 | 64 |
| 531 | ], |
| 532 | "dtype": "float16", |
| 533 | "offset_bytes": 460688, |
| 534 | "nbytes": 128, |
| 535 | "sha256": "f7238373516f17b75fc01e5b5e1badbd5beb0fc85776ec3141162ab6c8454807" |
| 536 | }, |
| 537 | { |
| 538 | "name": "frame_blocks.4.net.2.weight", |
| 539 | "shape": [ |
| 540 | 64, |
| 541 | 64, |
| 542 | 5 |
| 543 | ], |
| 544 | "dtype": "float16", |
| 545 | "offset_bytes": 460816, |
| 546 | "nbytes": 40960, |
| 547 | "sha256": "6ac750a803abaae8fafc7d06038211c2dbbc59df49cb890421870d78c324aba2" |
| 548 | }, |
| 549 | { |
| 550 | "name": "frame_blocks.4.scale", |
| 551 | "shape": [ |
| 552 | 1 |
| 553 | ], |
| 554 | "dtype": "float16", |
| 555 | "offset_bytes": 501776, |
| 556 | "nbytes": 2, |
| 557 | "sha256": "6033b61668320a028e7e6acaf08b33d175c0135f00795c05a5f70aee0269e78a" |
| 558 | }, |
| 559 | { |
| 560 | "name": "frame_input_proj.bias", |
| 561 | "shape": [ |
| 562 | 64 |
| 563 | ], |
| 564 | "dtype": "float16", |
| 565 | "offset_bytes": 501778, |
| 566 | "nbytes": 128, |
| 567 | "sha256": "52d51d483ccc1a2137427ddb2c104cd75a39ac6bbee03d67d3b8d3794e36a5fc" |
| 568 | }, |
| 569 | { |
| 570 | "name": "frame_input_proj.weight", |
| 571 | "shape": [ |
| 572 | 64, |
| 573 | 67, |
| 574 | 1 |
| 575 | ], |
| 576 | "dtype": "float16", |
| 577 | "offset_bytes": 501906, |
| 578 | "nbytes": 8576, |
| 579 | "sha256": "5d780efcd7bfb6ea6575c21d5b205d506bba693ac712d9925f756c3421983f5f" |
| 580 | }, |
| 581 | { |
| 582 | "name": "output.bias", |
| 583 | "shape": [ |
| 584 | 192 |
| 585 | ], |
| 586 | "dtype": "float16", |
| 587 | "offset_bytes": 510482, |
| 588 | "nbytes": 384, |
| 589 | "sha256": "7a03ed73c4aca8dfebac8de3a015747eae3756e3df150e42d2d7f0daca71c3e3" |
| 590 | }, |
| 591 | { |
| 592 | "name": "output.weight", |
| 593 | "shape": [ |
| 594 | 192, |
| 595 | 64, |
| 596 | 1 |
| 597 | ], |
| 598 | "dtype": "float16", |
| 599 | "offset_bytes": 510866, |
| 600 | "nbytes": 24576, |
| 601 | "sha256": "e7a949db478e7d95ba8fd943589a98d67fd79153ba38e5fe43a239c714516b62" |
| 602 | }, |
| 603 | { |
| 604 | "name": "token_blocks.0.net.0.bias", |
| 605 | "shape": [ |
| 606 | 64 |
| 607 | ], |
| 608 | "dtype": "float16", |
| 609 | "offset_bytes": 535442, |
| 610 | "nbytes": 128, |
| 611 | "sha256": "09fb0dec223e88d6c1d3d171e4b3e53a7a286d3132f76201e6f9e17d11ff6568" |
| 612 | }, |
| 613 | { |
| 614 | "name": "token_blocks.0.net.0.weight", |
| 615 | "shape": [ |
| 616 | 64, |
| 617 | 64, |
| 618 | 5 |
| 619 | ], |
| 620 | "dtype": "float16", |
| 621 | "offset_bytes": 535570, |
| 622 | "nbytes": 40960, |
| 623 | "sha256": "09973cf7f35844a00935f53ceee41c097564cebaeef714055ea5c2b7ae020c30" |
| 624 | }, |
| 625 | { |
| 626 | "name": "token_blocks.0.net.2.bias", |
| 627 | "shape": [ |
| 628 | 64 |
| 629 | ], |
| 630 | "dtype": "float16", |
| 631 | "offset_bytes": 576530, |
| 632 | "nbytes": 128, |
| 633 | "sha256": "dff2480f13e1627fe4168350282fd4d0ef3e679bb12eeecf8eed5afc11f234aa" |
| 634 | }, |
| 635 | { |
| 636 | "name": "token_blocks.0.net.2.weight", |
| 637 | "shape": [ |
| 638 | 64, |
| 639 | 64, |
| 640 | 5 |
| 641 | ], |
| 642 | "dtype": "float16", |
| 643 | "offset_bytes": 576658, |
| 644 | "nbytes": 40960, |
| 645 | "sha256": "206cf1d732e83fedf87dd14bb837cde16a297c4df2ed616c0cc7f968bcb2e68e" |
| 646 | }, |
| 647 | { |
| 648 | "name": "token_blocks.0.scale", |
| 649 | "shape": [ |
| 650 | 1 |
| 651 | ], |
| 652 | "dtype": "float16", |
| 653 | "offset_bytes": 617618, |
| 654 | "nbytes": 2, |
| 655 | "sha256": "00ddb76b63f203d7513604725cd560fe4477055548440ace86323b844583c8d7" |
| 656 | }, |
| 657 | { |
| 658 | "name": "token_blocks.1.net.0.bias", |
| 659 | "shape": [ |
| 660 | 64 |
| 661 | ], |
| 662 | "dtype": "float16", |
| 663 | "offset_bytes": 617620, |
| 664 | "nbytes": 128, |
| 665 | "sha256": "c418b4f10db3b7c1c28bc107582f9415c0a521a5d3a25758b5c838a7430e3b8f" |
| 666 | }, |
| 667 | { |
| 668 | "name": "token_blocks.1.net.0.weight", |
| 669 | "shape": [ |
| 670 | 64, |
| 671 | 64, |
| 672 | 5 |
| 673 | ], |
| 674 | "dtype": "float16", |
| 675 | "offset_bytes": 617748, |
| 676 | "nbytes": 40960, |
| 677 | "sha256": "073a5cdbb9a6ca1dc6457a37d450ca3fc9480391f983b6b24904beea7d25a33f" |
| 678 | }, |
| 679 | { |
| 680 | "name": "token_blocks.1.net.2.bias", |
| 681 | "shape": [ |
| 682 | 64 |
| 683 | ], |
| 684 | "dtype": "float16", |
| 685 | "offset_bytes": 658708, |
| 686 | "nbytes": 128, |
| 687 | "sha256": "6d757213a05ff7a72336c67ee1ac630985d04c18a4c3e10be8554c574ecd09bc" |
| 688 | }, |
| 689 | { |
| 690 | "name": "token_blocks.1.net.2.weight", |
| 691 | "shape": [ |
| 692 | 64, |
| 693 | 64, |
| 694 | 5 |
| 695 | ], |
| 696 | "dtype": "float16", |
| 697 | "offset_bytes": 658836, |
| 698 | "nbytes": 40960, |
| 699 | "sha256": "4c91495f6445683ab56ccf48728f5433e83c58dbc46c575873b7659ab25718c8" |
| 700 | }, |
| 701 | { |
| 702 | "name": "token_blocks.1.scale", |
| 703 | "shape": [ |
| 704 | 1 |
| 705 | ], |
| 706 | "dtype": "float16", |
| 707 | "offset_bytes": 699796, |
| 708 | "nbytes": 2, |
| 709 | "sha256": "b588f9b12891d2313ffec07d46e1e82723506893e6343d74391fa799ad36d4a1" |
| 710 | }, |
| 711 | { |
| 712 | "name": "token_blocks.2.net.0.bias", |
| 713 | "shape": [ |
| 714 | 64 |
| 715 | ], |
| 716 | "dtype": "float16", |
| 717 | "offset_bytes": 699798, |
| 718 | "nbytes": 128, |
| 719 | "sha256": "ec6b87f74d1bba297668334bcff6e83cd74e329c0a3a24331adb15a89b92ce5b" |
| 720 | }, |
| 721 | { |
| 722 | "name": "token_blocks.2.net.0.weight", |
| 723 | "shape": [ |
| 724 | 64, |
| 725 | 64, |
| 726 | 5 |
| 727 | ], |
| 728 | "dtype": "float16", |
| 729 | "offset_bytes": 699926, |
| 730 | "nbytes": 40960, |
| 731 | "sha256": "25dbef0fcaaf742df6a10cbb3bac468aa93652fe7152cda743861da43149a959" |
| 732 | }, |
| 733 | { |
| 734 | "name": "token_blocks.2.net.2.bias", |
| 735 | "shape": [ |
| 736 | 64 |
| 737 | ], |
| 738 | "dtype": "float16", |
| 739 | "offset_bytes": 740886, |
| 740 | "nbytes": 128, |
| 741 | "sha256": "ffac566427a0cb280554406aa89de13b08339869d6a0d49b688f0497b3b9d21f" |
| 742 | }, |
| 743 | { |
| 744 | "name": "token_blocks.2.net.2.weight", |
| 745 | "shape": [ |
| 746 | 64, |
| 747 | 64, |
| 748 | 5 |
| 749 | ], |
| 750 | "dtype": "float16", |
| 751 | "offset_bytes": 741014, |
| 752 | "nbytes": 40960, |
| 753 | "sha256": "77dd4de018986592b40c0facfdf809a9899cc1e6e5e1ca975fef1c2bd9a39256" |
| 754 | }, |
| 755 | { |
| 756 | "name": "token_blocks.2.scale", |
| 757 | "shape": [ |
| 758 | 1 |
| 759 | ], |
| 760 | "dtype": "float16", |
| 761 | "offset_bytes": 781974, |
| 762 | "nbytes": 2, |
| 763 | "sha256": "76702b32875e7d3af7368727a331ed16e20815cde7a73ee8682497bd487261a8" |
| 764 | }, |
| 765 | { |
| 766 | "name": "token_input_proj.bias", |
| 767 | "shape": [ |
| 768 | 64 |
| 769 | ], |
| 770 | "dtype": "float16", |
| 771 | "offset_bytes": 781976, |
| 772 | "nbytes": 128, |
| 773 | "sha256": "a3a3602cef0b51dc4951f97a0f52dd78590526d5a333cdd1dba4fe38e5862294" |
| 774 | }, |
| 775 | { |
| 776 | "name": "token_input_proj.weight", |
| 777 | "shape": [ |
| 778 | 64, |
| 779 | 66, |
| 780 | 1 |
| 781 | ], |
| 782 | "dtype": "float16", |
| 783 | "offset_bytes": 782104, |
| 784 | "nbytes": 8448, |
| 785 | "sha256": "33ed6a6a8f2909ab4b5e3e8cd7e906b0755dc88d61a6dbe7a53f9ec5634485e1" |
| 786 | } |
| 787 | ] |
| 788 | }, |
| 789 | "decoder": { |
| 790 | "checkpoint": "releases/kristin-20260707/web/decoder-student-1p0m-demetalB.pt", |
| 791 | "checkpoint_sha256": "fe8a77f049a82fdea2f726c05dc40448edcf6698da31359256e38696dffd2c29", |
| 792 | "config": { |
| 793 | "in_channels": 192, |
| 794 | "source_in_channels": 192, |
| 795 | "channels": [ |
| 796 | 192, |
| 797 | 96, |
| 798 | 48, |
| 799 | 24 |
| 800 | ], |
| 801 | "res_layers": 1, |
| 802 | "variant": "piperlite", |
| 803 | "lrc_code_dim": 40, |
| 804 | "lrc_encoder_hidden": 64, |
| 805 | "rank_ratio": 0.5, |
| 806 | "activation": "leaky_relu", |
| 807 | "stage_affine": false, |
| 808 | "factorized_pre_rank": 0, |
| 809 | "piper_res_factor_rank_ratio": 0.0, |
| 810 | "res_bank_scale_mode": "kept", |
| 811 | "stage0_branches": [ |
| 812 | 0, |
| 813 | 1, |
| 814 | 2 |
| 815 | ], |
| 816 | "stage1_branches": [ |
| 817 | 0, |
| 818 | 1, |
| 819 | 2 |
| 820 | ], |
| 821 | "stage2_branches": [ |
| 822 | 0, |
| 823 | 1, |
| 824 | 2 |
| 825 | ], |
| 826 | "stage3_branches": [ |
| 827 | 0, |
| 828 | 1, |
| 829 | 2 |
| 830 | ], |
| 831 | "fsd_dim": 72, |
| 832 | "fsd_blocks": 5, |
| 833 | "fsd_film_rank": 12, |
| 834 | "fsd_head_rank": 48, |
| 835 | "stage_projection_bottlenecks": [], |
| 836 | "teacher_init_checkpoint": null, |
| 837 | "teacher_init_method": "first", |
| 838 | "teacher_init": null, |
| 839 | "init_decoder_checkpoint": "artifacts/sub10m-search/root-a-piper-vits/en_US-kristin-u600-dmB-20260707/dmB-1/decoder-student.pt", |
| 840 | "decoder_init": { |
| 841 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/en_US-kristin-u600-dmB-20260707/dmB-1/decoder-student.pt", |
| 842 | "decoder_parameters": 1000875, |
| 843 | "matched_config": { |
| 844 | "in_channels": 192, |
| 845 | "channels": [ |
| 846 | 192, |
| 847 | 96, |
| 848 | 48, |
| 849 | 24 |
| 850 | ], |
| 851 | "res_layers": 1, |
| 852 | "variant": "piperlite", |
| 853 | "rank_ratio": 0.5, |
| 854 | "activation": "leaky_relu", |
| 855 | "stage_affine": false, |
| 856 | "factorized_pre_rank": 0, |
| 857 | "piper_res_factor_rank_ratio": 0.0, |
| 858 | "res_bank_scale_mode": "kept", |
| 859 | "stage0_branches": [ |
| 860 | 0, |
| 861 | 1, |
| 862 | 2 |
| 863 | ], |
| 864 | "stage1_branches": [ |
| 865 | 0, |
| 866 | 1, |
| 867 | 2 |
| 868 | ], |
| 869 | "stage2_branches": [ |
| 870 | 0, |
| 871 | 1, |
| 872 | 2 |
| 873 | ], |
| 874 | "stage3_branches": [ |
| 875 | 0, |
| 876 | 1, |
| 877 | 2 |
| 878 | ], |
| 879 | "fsd_dim": 72, |
| 880 | "fsd_blocks": 5, |
| 881 | "fsd_film_rank": 12, |
| 882 | "fsd_head_rank": 48, |
| 883 | "stage_projection_bottlenecks": [], |
| 884 | "post_filter_channels": 8, |
| 885 | "post_filter_layers": 1, |
| 886 | "post_filter_kernel": 9, |
| 887 | "post_filter_scale": 0.1, |
| 888 | "pre_tanh_repair_channels": 0, |
| 889 | "pre_tanh_repair_layers": 0, |
| 890 | "pre_tanh_repair_kernel": 7, |
| 891 | "pre_tanh_repair_scale": 0.15, |
| 892 | "istft_n_fft": 512 |
| 893 | }, |
| 894 | "allowed_post_filter_mismatches": [], |
| 895 | "allowed_pre_tanh_repair_mismatches": [], |
| 896 | "allowed_activation_mismatches": [], |
| 897 | "missing_keys": [], |
| 898 | "unexpected_keys": [] |
| 899 | }, |
| 900 | "post_filter_channels": 8, |
| 901 | "post_filter_layers": 1, |
| 902 | "post_filter_kernel": 9, |
| 903 | "post_filter_scale": 0.1, |
| 904 | "pre_tanh_repair_channels": 0, |
| 905 | "pre_tanh_repair_layers": 0, |
| 906 | "pre_tanh_repair_kernel": 7, |
| 907 | "pre_tanh_repair_scale": 0.15, |
| 908 | "istft_n_fft": 512, |
| 909 | "ap_amplitude_weight": 0.0, |
| 910 | "ap_phase_weight": 0.0, |
| 911 | "ap_complex_weight": 0.0, |
| 912 | "spectral_head_init": "default", |
| 913 | "spectral_head_init_scale": 0.001, |
| 914 | "ap_amp_init_bias": -6.0, |
| 915 | "ap_phase_real_init_bias": 1.0, |
| 916 | "spectral_head_init_summary": null, |
| 917 | "feature_exact_weight": 0.0, |
| 918 | "feature_exact_keys": [ |
| 919 | "pre", |
| 920 | "up0" |
| 921 | ], |
| 922 | "feature_exact_teacher_init": null, |
| 923 | "signature_pack_dir": null, |
| 924 | "exact_feature_pack_dir": null, |
| 925 | "signature_hint_weight": 0.0, |
| 926 | "signature_temporal_weight": 0.0, |
| 927 | "signature_phase_weight": 0.0, |
| 928 | "signature_phase_bins": 0, |
| 929 | "signature_keys": [ |
| 930 | "stage0_mix", |
| 931 | "stage1_mix", |
| 932 | "stage2_mix", |
| 933 | "pre_tanh", |
| 934 | "audio" |
| 935 | ], |
| 936 | "bottleneck_code_checkpoint": null, |
| 937 | "bottleneck_code_weight": 0.0, |
| 938 | "bottleneck_code_keys": [], |
| 939 | "bottleneck_code_training_only_parameters": 0, |
| 940 | "echo_tail_weight": 0.0, |
| 941 | "echo_tail_min_ms": 8.0, |
| 942 | "echo_tail_max_ms": 40.0, |
| 943 | "echo_tail_lags": 9, |
| 944 | "echo_tail_margin": 0.02, |
| 945 | "paired_acoustic_residual_weight": 0.0, |
| 946 | "paired_acoustic_residual_max_scale": 0.25, |
| 947 | "lrc_pred_code_mix_prob": 0.0, |
| 948 | "lrc_pred_code_checkpoint": null, |
| 949 | "lrc_pred_code_residual_prob": 0.0, |
| 950 | "lrc_pred_code_residual_max_scale": 0.25, |
| 951 | "lrc_pred_code_cache": null |
| 952 | }, |
| 953 | "parameters": 1000875, |
| 954 | "tensors": [ |
| 955 | { |
| 956 | "name": "post.bias", |
| 957 | "shape": [ |
| 958 | 1 |
| 959 | ], |
| 960 | "dtype": "float16", |
| 961 | "offset_bytes": 790552, |
| 962 | "nbytes": 2, |
| 963 | "sha256": "15a7c867f962425124e4392b1ac24693c02c3f4d22a2a88a390196d696c574fc" |
| 964 | }, |
| 965 | { |
| 966 | "name": "post.weight", |
| 967 | "shape": [ |
| 968 | 1, |
| 969 | 24, |
| 970 | 7 |
| 971 | ], |
| 972 | "dtype": "float16", |
| 973 | "offset_bytes": 790554, |
| 974 | "nbytes": 336, |
| 975 | "sha256": "e663098c58a3058f4b0ad7e80cd2a79c133b64c3db04917f90307602f3b8d7e3" |
| 976 | }, |
| 977 | { |
| 978 | "name": "post_filter.in_conv.bias", |
| 979 | "shape": [ |
| 980 | 8 |
| 981 | ], |
| 982 | "dtype": "float16", |
| 983 | "offset_bytes": 790890, |
| 984 | "nbytes": 16, |
| 985 | "sha256": "4eadbfb2a18f3fe5d4b47a7c9701dd5319043baf88862d06d1824cdf571ff589" |
| 986 | }, |
| 987 | { |
| 988 | "name": "post_filter.in_conv.weight", |
| 989 | "shape": [ |
| 990 | 8, |
| 991 | 1, |
| 992 | 9 |
| 993 | ], |
| 994 | "dtype": "float16", |
| 995 | "offset_bytes": 790906, |
| 996 | "nbytes": 144, |
| 997 | "sha256": "51122c09961e5e3880d82275a80a082bf79d0c5afd8b75bfcdb8be79f4e998b5" |
| 998 | }, |
| 999 | { |
| 1000 | "name": "post_filter.out_conv.bias", |
| 1001 | "shape": [ |
| 1002 | 1 |
| 1003 | ], |
| 1004 | "dtype": "float16", |
| 1005 | "offset_bytes": 791050, |
| 1006 | "nbytes": 2, |
| 1007 | "sha256": "e4fefdba96b45754620484022e18a1d241446963db270ed0c405626188285dce" |
| 1008 | }, |
| 1009 | { |
| 1010 | "name": "post_filter.out_conv.weight", |
| 1011 | "shape": [ |
| 1012 | 1, |
| 1013 | 8, |
| 1014 | 9 |
| 1015 | ], |
| 1016 | "dtype": "float16", |
| 1017 | "offset_bytes": 791052, |
| 1018 | "nbytes": 144, |
| 1019 | "sha256": "ac10a5ce3614931ba3a3f5e1b00248375b5b987f4aeacb820d2a0ddfbf8a05cf" |
| 1020 | }, |
| 1021 | { |
| 1022 | "name": "post_filter.units.0.conv1.bias", |
| 1023 | "shape": [ |
| 1024 | 8 |
| 1025 | ], |
| 1026 | "dtype": "float16", |
| 1027 | "offset_bytes": 791196, |
| 1028 | "nbytes": 16, |
| 1029 | "sha256": "c1537b7d83780939db1399d959a2ef05b1af92f218dfe2c255255a4fe676836e" |
| 1030 | }, |
| 1031 | { |
| 1032 | "name": "post_filter.units.0.conv1.weight", |
| 1033 | "shape": [ |
| 1034 | 8, |
| 1035 | 8, |
| 1036 | 3 |
| 1037 | ], |
| 1038 | "dtype": "float16", |
| 1039 | "offset_bytes": 791212, |
| 1040 | "nbytes": 384, |
| 1041 | "sha256": "2c745e858b70a7771a2a32a0c0ec9da7a247fa8f74e8a80150d36cfb0c9ca8c4" |
| 1042 | }, |
| 1043 | { |
| 1044 | "name": "post_filter.units.0.conv2.bias", |
| 1045 | "shape": [ |
| 1046 | 8 |
| 1047 | ], |
| 1048 | "dtype": "float16", |
| 1049 | "offset_bytes": 791596, |
| 1050 | "nbytes": 16, |
| 1051 | "sha256": "b0441de2c88e8f6b8c5942e53e212005cfbefbaa19db6490f39bf8ef3e23f369" |
| 1052 | }, |
| 1053 | { |
| 1054 | "name": "post_filter.units.0.conv2.weight", |
| 1055 | "shape": [ |
| 1056 | 8, |
| 1057 | 8, |
| 1058 | 3 |
| 1059 | ], |
| 1060 | "dtype": "float16", |
| 1061 | "offset_bytes": 791612, |
| 1062 | "nbytes": 384, |
| 1063 | "sha256": "e53c88d1bf56a94372956f5a9a675161195a8b620413b074293e59d72b0f85f4" |
| 1064 | }, |
| 1065 | { |
| 1066 | "name": "post_filter.units.0.scale", |
| 1067 | "shape": [ |
| 1068 | 1 |
| 1069 | ], |
| 1070 | "dtype": "float16", |
| 1071 | "offset_bytes": 791996, |
| 1072 | "nbytes": 2, |
| 1073 | "sha256": "fa61e3dec3439589f4784c893bf321d0084f04c572c7af2b68e3f3360a35b486" |
| 1074 | }, |
| 1075 | { |
| 1076 | "name": "pre.bias", |
| 1077 | "shape": [ |
| 1078 | 192 |
| 1079 | ], |
| 1080 | "dtype": "float16", |
| 1081 | "offset_bytes": 791998, |
| 1082 | "nbytes": 384, |
| 1083 | "sha256": "d3fa57352ab42484ffc432e9ed821b40670c5032d50bc263dfae81c7cd6c2593" |
| 1084 | }, |
| 1085 | { |
| 1086 | "name": "pre.weight", |
| 1087 | "shape": [ |
| 1088 | 192, |
| 1089 | 192, |
| 1090 | 7 |
| 1091 | ], |
| 1092 | "dtype": "float16", |
| 1093 | "offset_bytes": 792382, |
| 1094 | "nbytes": 516096, |
| 1095 | "sha256": "1cc7ed63e93a14a5408c0acc59f82b820f1b179549fcb7784fea4d7da683dc33" |
| 1096 | }, |
| 1097 | { |
| 1098 | "name": "res0.0.blocks.0.conv1.bias", |
| 1099 | "shape": [ |
| 1100 | 96 |
| 1101 | ], |
| 1102 | "dtype": "float16", |
| 1103 | "offset_bytes": 1308478, |
| 1104 | "nbytes": 192, |
| 1105 | "sha256": "c6b3f579f167368e06ce176b3725a60f232417cc63209b59b91b08a965224ead" |
| 1106 | }, |
| 1107 | { |
| 1108 | "name": "res0.0.blocks.0.conv1.weight", |
| 1109 | "shape": [ |
| 1110 | 96, |
| 1111 | 96, |
| 1112 | 3 |
| 1113 | ], |
| 1114 | "dtype": "float16", |
| 1115 | "offset_bytes": 1308670, |
| 1116 | "nbytes": 55296, |
| 1117 | "sha256": "5cb168665726d62fbf9cfbe79cc7f2519b19463422f6a9788d2a4df7b29b2e88" |
| 1118 | }, |
| 1119 | { |
| 1120 | "name": "res0.0.blocks.0.conv2.bias", |
| 1121 | "shape": [ |
| 1122 | 96 |
| 1123 | ], |
| 1124 | "dtype": "float16", |
| 1125 | "offset_bytes": 1363966, |
| 1126 | "nbytes": 192, |
| 1127 | "sha256": "d94031112853fb58518b94414be1866ff7cb1e5c478d8ee224c04b57d627ee8e" |
| 1128 | }, |
| 1129 | { |
| 1130 | "name": "res0.0.blocks.0.conv2.weight", |
| 1131 | "shape": [ |
| 1132 | 96, |
| 1133 | 96, |
| 1134 | 3 |
| 1135 | ], |
| 1136 | "dtype": "float16", |
| 1137 | "offset_bytes": 1364158, |
| 1138 | "nbytes": 55296, |
| 1139 | "sha256": "2953545c5cafa7644beaa6ad292ee1f3979a4cd4d9c379c61f07b01eeff48cc5" |
| 1140 | }, |
| 1141 | { |
| 1142 | "name": "res0.0.blocks.1.conv1.bias", |
| 1143 | "shape": [ |
| 1144 | 96 |
| 1145 | ], |
| 1146 | "dtype": "float16", |
| 1147 | "offset_bytes": 1419454, |
| 1148 | "nbytes": 192, |
| 1149 | "sha256": "e84a800b6cfbcfb90a8d59a1bd9f33b444798ec64288a7953e82ef906fcd1fd1" |
| 1150 | }, |
| 1151 | { |
| 1152 | "name": "res0.0.blocks.1.conv1.weight", |
| 1153 | "shape": [ |
| 1154 | 96, |
| 1155 | 96, |
| 1156 | 5 |
| 1157 | ], |
| 1158 | "dtype": "float16", |
| 1159 | "offset_bytes": 1419646, |
| 1160 | "nbytes": 92160, |
| 1161 | "sha256": "1a3d399bb0286491cb3ef5616c7c2f8a08b21668217e839c449234ac4824f1db" |
| 1162 | }, |
| 1163 | { |
| 1164 | "name": "res0.0.blocks.1.conv2.bias", |
| 1165 | "shape": [ |
| 1166 | 96 |
| 1167 | ], |
| 1168 | "dtype": "float16", |
| 1169 | "offset_bytes": 1511806, |
| 1170 | "nbytes": 192, |
| 1171 | "sha256": "e0dba88b4b97bb092990069f23aef2025a09981cc3e8a5c2d4d0b5829c563f08" |
| 1172 | }, |
| 1173 | { |
| 1174 | "name": "res0.0.blocks.1.conv2.weight", |
| 1175 | "shape": [ |
| 1176 | 96, |
| 1177 | 96, |
| 1178 | 5 |
| 1179 | ], |
| 1180 | "dtype": "float16", |
| 1181 | "offset_bytes": 1511998, |
| 1182 | "nbytes": 92160, |
| 1183 | "sha256": "29dae893895180aed2c4a1d83541de3a3e53052c1b58869130059f4a98b8f032" |
| 1184 | }, |
| 1185 | { |
| 1186 | "name": "res0.0.blocks.2.conv1.bias", |
| 1187 | "shape": [ |
| 1188 | 96 |
| 1189 | ], |
| 1190 | "dtype": "float16", |
| 1191 | "offset_bytes": 1604158, |
| 1192 | "nbytes": 192, |
| 1193 | "sha256": "5104c759b842622aececc83f782a6638d2fff3b25cf52489aed9afe2810efb1e" |
| 1194 | }, |
| 1195 | { |
| 1196 | "name": "res0.0.blocks.2.conv1.weight", |
| 1197 | "shape": [ |
| 1198 | 96, |
| 1199 | 96, |
| 1200 | 7 |
| 1201 | ], |
| 1202 | "dtype": "float16", |
| 1203 | "offset_bytes": 1604350, |
| 1204 | "nbytes": 129024, |
| 1205 | "sha256": "ac10a963d8878ddd2c0b8f19306ceee12c021ad93a44bdbcaa2c8d6e632907ee" |
| 1206 | }, |
| 1207 | { |
| 1208 | "name": "res0.0.blocks.2.conv2.bias", |
| 1209 | "shape": [ |
| 1210 | 96 |
| 1211 | ], |
| 1212 | "dtype": "float16", |
| 1213 | "offset_bytes": 1733374, |
| 1214 | "nbytes": 192, |
| 1215 | "sha256": "b6efe9b0a7e243119146ccc455de3c298dfb2934239ce6eb52c8a14ead999cce" |
| 1216 | }, |
| 1217 | { |
| 1218 | "name": "res0.0.blocks.2.conv2.weight", |
| 1219 | "shape": [ |
| 1220 | 96, |
| 1221 | 96, |
| 1222 | 7 |
| 1223 | ], |
| 1224 | "dtype": "float16", |
| 1225 | "offset_bytes": 1733566, |
| 1226 | "nbytes": 129024, |
| 1227 | "sha256": "288375c8d779d5dd55769415c6372399a6d827e0af9ec256ae65d009dd290c7c" |
| 1228 | }, |
| 1229 | { |
| 1230 | "name": "res1.0.blocks.0.conv1.bias", |
| 1231 | "shape": [ |
| 1232 | 48 |
| 1233 | ], |
| 1234 | "dtype": "float16", |
| 1235 | "offset_bytes": 1862590, |
| 1236 | "nbytes": 96, |
| 1237 | "sha256": "7adc8ae9cb438ecd943c859c0bf9a237dfb582ffc521aeb4c548a04f5a68df95" |
| 1238 | }, |
| 1239 | { |
| 1240 | "name": "res1.0.blocks.0.conv1.weight", |
| 1241 | "shape": [ |
| 1242 | 48, |
| 1243 | 48, |
| 1244 | 3 |
| 1245 | ], |
| 1246 | "dtype": "float16", |
| 1247 | "offset_bytes": 1862686, |
| 1248 | "nbytes": 13824, |
| 1249 | "sha256": "e15099bf06e5fbdeae1f5cdfc2c9d656727a1f544e5f117d1c8a424363e6397e" |
| 1250 | }, |
| 1251 | { |
| 1252 | "name": "res1.0.blocks.0.conv2.bias", |
| 1253 | "shape": [ |
| 1254 | 48 |
| 1255 | ], |
| 1256 | "dtype": "float16", |
| 1257 | "offset_bytes": 1876510, |
| 1258 | "nbytes": 96, |
| 1259 | "sha256": "7c0b7a32edc30142a71fb6847c2982ff4f793f6a20f298ef74b3ca8dc64e58ca" |
| 1260 | }, |
| 1261 | { |
| 1262 | "name": "res1.0.blocks.0.conv2.weight", |
| 1263 | "shape": [ |
| 1264 | 48, |
| 1265 | 48, |
| 1266 | 3 |
| 1267 | ], |
| 1268 | "dtype": "float16", |
| 1269 | "offset_bytes": 1876606, |
| 1270 | "nbytes": 13824, |
| 1271 | "sha256": "9fbee617801b03ee0bade3e384a869e9a215000fcb1816d8c3828807bd3d43d1" |
| 1272 | }, |
| 1273 | { |
| 1274 | "name": "res1.0.blocks.1.conv1.bias", |
| 1275 | "shape": [ |
| 1276 | 48 |
| 1277 | ], |
| 1278 | "dtype": "float16", |
| 1279 | "offset_bytes": 1890430, |
| 1280 | "nbytes": 96, |
| 1281 | "sha256": "f658d2f3df709f34efaf71da91cabd23fd04d2ca3578aed72240b49f45db94b4" |
| 1282 | }, |
| 1283 | { |
| 1284 | "name": "res1.0.blocks.1.conv1.weight", |
| 1285 | "shape": [ |
| 1286 | 48, |
| 1287 | 48, |
| 1288 | 5 |
| 1289 | ], |
| 1290 | "dtype": "float16", |
| 1291 | "offset_bytes": 1890526, |
| 1292 | "nbytes": 23040, |
| 1293 | "sha256": "d3398324b6710e1977c0757b4b2d075d7bd584ef46dae5a2f9cfefd645d1011d" |
| 1294 | }, |
| 1295 | { |
| 1296 | "name": "res1.0.blocks.1.conv2.bias", |
| 1297 | "shape": [ |
| 1298 | 48 |
| 1299 | ], |
| 1300 | "dtype": "float16", |
| 1301 | "offset_bytes": 1913566, |
| 1302 | "nbytes": 96, |
| 1303 | "sha256": "7dd7de6672e7c158b24a0c9f50990744e13101a005bd06cc5b038cceb7197504" |
| 1304 | }, |
| 1305 | { |
| 1306 | "name": "res1.0.blocks.1.conv2.weight", |
| 1307 | "shape": [ |
| 1308 | 48, |
| 1309 | 48, |
| 1310 | 5 |
| 1311 | ], |
| 1312 | "dtype": "float16", |
| 1313 | "offset_bytes": 1913662, |
| 1314 | "nbytes": 23040, |
| 1315 | "sha256": "4c5f5ebd78f49e15c695ba9f9ad901e9549c16ffdaa43e24b33f6f6df71f4fd6" |
| 1316 | }, |
| 1317 | { |
| 1318 | "name": "res1.0.blocks.2.conv1.bias", |
| 1319 | "shape": [ |
| 1320 | 48 |
| 1321 | ], |
| 1322 | "dtype": "float16", |
| 1323 | "offset_bytes": 1936702, |
| 1324 | "nbytes": 96, |
| 1325 | "sha256": "19ce03e68126ebbe0929f22859cc5bc753e0bd36a538006a43e2e199491dd1d3" |
| 1326 | }, |
| 1327 | { |
| 1328 | "name": "res1.0.blocks.2.conv1.weight", |
| 1329 | "shape": [ |
| 1330 | 48, |
| 1331 | 48, |
| 1332 | 7 |
| 1333 | ], |
| 1334 | "dtype": "float16", |
| 1335 | "offset_bytes": 1936798, |
| 1336 | "nbytes": 32256, |
| 1337 | "sha256": "76b2cf92820c5732c4f51455e74458010f61418ec6f9d9d5693249578bf7c30d" |
| 1338 | }, |
| 1339 | { |
| 1340 | "name": "res1.0.blocks.2.conv2.bias", |
| 1341 | "shape": [ |
| 1342 | 48 |
| 1343 | ], |
| 1344 | "dtype": "float16", |
| 1345 | "offset_bytes": 1969054, |
| 1346 | "nbytes": 96, |
| 1347 | "sha256": "22631004e59ddbf0b8de8e0b5941d77c73e12cedfd729b34b1da6fff6a594d41" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "res1.0.blocks.2.conv2.weight", |
| 1351 | "shape": [ |
| 1352 | 48, |
| 1353 | 48, |
| 1354 | 7 |
| 1355 | ], |
| 1356 | "dtype": "float16", |
| 1357 | "offset_bytes": 1969150, |
| 1358 | "nbytes": 32256, |
| 1359 | "sha256": "7b03a58626f749c25aefd6e6fd06760141448e85b2891b12bafd6368d4adec6d" |
| 1360 | }, |
| 1361 | { |
| 1362 | "name": "res2.0.blocks.0.conv1.bias", |
| 1363 | "shape": [ |
| 1364 | 24 |
| 1365 | ], |
| 1366 | "dtype": "float16", |
| 1367 | "offset_bytes": 2001406, |
| 1368 | "nbytes": 48, |
| 1369 | "sha256": "0dcec62d57f0be7f3914d2b15f0139fcdac3c2be7e33b00be8a27ba1a054a2ae" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "res2.0.blocks.0.conv1.weight", |
| 1373 | "shape": [ |
| 1374 | 24, |
| 1375 | 24, |
| 1376 | 3 |
| 1377 | ], |
| 1378 | "dtype": "float16", |
| 1379 | "offset_bytes": 2001454, |
| 1380 | "nbytes": 3456, |
| 1381 | "sha256": "4c0972b336ac75eecee18b39218fa5da9595ae9b395cc32becc13dea5bc865ab" |
| 1382 | }, |
| 1383 | { |
| 1384 | "name": "res2.0.blocks.0.conv2.bias", |
| 1385 | "shape": [ |
| 1386 | 24 |
| 1387 | ], |
| 1388 | "dtype": "float16", |
| 1389 | "offset_bytes": 2004910, |
| 1390 | "nbytes": 48, |
| 1391 | "sha256": "ab3d639940baa5e582b4fdfb098036cabe14df1f78358c86ddd63644bf6f742a" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "res2.0.blocks.0.conv2.weight", |
| 1395 | "shape": [ |
| 1396 | 24, |
| 1397 | 24, |
| 1398 | 3 |
| 1399 | ], |
| 1400 | "dtype": "float16", |
| 1401 | "offset_bytes": 2004958, |
| 1402 | "nbytes": 3456, |
| 1403 | "sha256": "e684629d9231eb23115876e39815fc19a5b17d68f85798e10492d7203df0d026" |
| 1404 | }, |
| 1405 | { |
| 1406 | "name": "res2.0.blocks.1.conv1.bias", |
| 1407 | "shape": [ |
| 1408 | 24 |
| 1409 | ], |
| 1410 | "dtype": "float16", |
| 1411 | "offset_bytes": 2008414, |
| 1412 | "nbytes": 48, |
| 1413 | "sha256": "fa8404f472d8d60e5eaebb84cd653bb36600db09a6e55a7b185b04786580b523" |
| 1414 | }, |
| 1415 | { |
| 1416 | "name": "res2.0.blocks.1.conv1.weight", |
| 1417 | "shape": [ |
| 1418 | 24, |
| 1419 | 24, |
| 1420 | 5 |
| 1421 | ], |
| 1422 | "dtype": "float16", |
| 1423 | "offset_bytes": 2008462, |
| 1424 | "nbytes": 5760, |
| 1425 | "sha256": "0d921b9b17170693d758f5f2136436af99c81a2af1190c23a7c97b44be6dcc7d" |
| 1426 | }, |
| 1427 | { |
| 1428 | "name": "res2.0.blocks.1.conv2.bias", |
| 1429 | "shape": [ |
| 1430 | 24 |
| 1431 | ], |
| 1432 | "dtype": "float16", |
| 1433 | "offset_bytes": 2014222, |
| 1434 | "nbytes": 48, |
| 1435 | "sha256": "b6ea268cd443c73d79eee4f2231f48b7fe55369a043e58eb445cccba7b52502a" |
| 1436 | }, |
| 1437 | { |
| 1438 | "name": "res2.0.blocks.1.conv2.weight", |
| 1439 | "shape": [ |
| 1440 | 24, |
| 1441 | 24, |
| 1442 | 5 |
| 1443 | ], |
| 1444 | "dtype": "float16", |
| 1445 | "offset_bytes": 2014270, |
| 1446 | "nbytes": 5760, |
| 1447 | "sha256": "495ffee9bc67ab4d797a294451d26f210daabece93920841883d5c90cfa8f15c" |
| 1448 | }, |
| 1449 | { |
| 1450 | "name": "res2.0.blocks.2.conv1.bias", |
| 1451 | "shape": [ |
| 1452 | 24 |
| 1453 | ], |
| 1454 | "dtype": "float16", |
| 1455 | "offset_bytes": 2020030, |
| 1456 | "nbytes": 48, |
| 1457 | "sha256": "dc7c14d3beef5f99483074b51e866ec4791d362e383323ded741ed02f06b2f94" |
| 1458 | }, |
| 1459 | { |
| 1460 | "name": "res2.0.blocks.2.conv1.weight", |
| 1461 | "shape": [ |
| 1462 | 24, |
| 1463 | 24, |
| 1464 | 7 |
| 1465 | ], |
| 1466 | "dtype": "float16", |
| 1467 | "offset_bytes": 2020078, |
| 1468 | "nbytes": 8064, |
| 1469 | "sha256": "7b0ba08dfd0243ce14a47f3819cbdc5ec15f4818b4bd9eb88aefc605ca3ef8e2" |
| 1470 | }, |
| 1471 | { |
| 1472 | "name": "res2.0.blocks.2.conv2.bias", |
| 1473 | "shape": [ |
| 1474 | 24 |
| 1475 | ], |
| 1476 | "dtype": "float16", |
| 1477 | "offset_bytes": 2028142, |
| 1478 | "nbytes": 48, |
| 1479 | "sha256": "a6c385a046b34452c7ffdd2231653703cf4a53ed6d2af6632a139e80df040c0a" |
| 1480 | }, |
| 1481 | { |
| 1482 | "name": "res2.0.blocks.2.conv2.weight", |
| 1483 | "shape": [ |
| 1484 | 24, |
| 1485 | 24, |
| 1486 | 7 |
| 1487 | ], |
| 1488 | "dtype": "float16", |
| 1489 | "offset_bytes": 2028190, |
| 1490 | "nbytes": 8064, |
| 1491 | "sha256": "0b4a8bc56a40625b9100837835efed2486f280f807a3fb0b0cc9fe07e1c12d4b" |
| 1492 | }, |
| 1493 | { |
| 1494 | "name": "up0.bias", |
| 1495 | "shape": [ |
| 1496 | 96 |
| 1497 | ], |
| 1498 | "dtype": "float16", |
| 1499 | "offset_bytes": 2036254, |
| 1500 | "nbytes": 192, |
| 1501 | "sha256": "0d55f96c642757742634bf9e1c92cedaf95f7640cfd28f306a2f590d49410740" |
| 1502 | }, |
| 1503 | { |
| 1504 | "name": "up0.weight", |
| 1505 | "shape": [ |
| 1506 | 192, |
| 1507 | 96, |
| 1508 | 16 |
| 1509 | ], |
| 1510 | "dtype": "float16", |
| 1511 | "offset_bytes": 2036446, |
| 1512 | "nbytes": 589824, |
| 1513 | "sha256": "5b740650aa510d052954047db213d00d4cc621e8610bab0729a83ca4a5c9a5c9" |
| 1514 | }, |
| 1515 | { |
| 1516 | "name": "up1.bias", |
| 1517 | "shape": [ |
| 1518 | 48 |
| 1519 | ], |
| 1520 | "dtype": "float16", |
| 1521 | "offset_bytes": 2626270, |
| 1522 | "nbytes": 96, |
| 1523 | "sha256": "b419d777427b61c7e8f4254e8491e9f4fff889fdc3f5ac6dfbb5f00144f15675" |
| 1524 | }, |
| 1525 | { |
| 1526 | "name": "up1.weight", |
| 1527 | "shape": [ |
| 1528 | 96, |
| 1529 | 48, |
| 1530 | 16 |
| 1531 | ], |
| 1532 | "dtype": "float16", |
| 1533 | "offset_bytes": 2626366, |
| 1534 | "nbytes": 147456, |
| 1535 | "sha256": "3aa4dd65465522bcb3405620140eacdfd52638f92d2c260e72f17fde0fe55f28" |
| 1536 | }, |
| 1537 | { |
| 1538 | "name": "up2.bias", |
| 1539 | "shape": [ |
| 1540 | 24 |
| 1541 | ], |
| 1542 | "dtype": "float16", |
| 1543 | "offset_bytes": 2773822, |
| 1544 | "nbytes": 48, |
| 1545 | "sha256": "b133816abe2c95fa10490524c89302678f055efae7cf783a60aff63c83b8c497" |
| 1546 | }, |
| 1547 | { |
| 1548 | "name": "up2.weight", |
| 1549 | "shape": [ |
| 1550 | 48, |
| 1551 | 24, |
| 1552 | 8 |
| 1553 | ], |
| 1554 | "dtype": "float16", |
| 1555 | "offset_bytes": 2773870, |
| 1556 | "nbytes": 18432, |
| 1557 | "sha256": "939b3a19317842acaf1af2ae93ddeb2c17b13e8ebbbcdd1179b29e9203b00876" |
| 1558 | } |
| 1559 | ] |
| 1560 | } |
| 1561 | }, |
| 1562 | "runtime_ops": [ |
| 1563 | "embedding_lookup", |
| 1564 | "conv1d_same", |
| 1565 | "conv1d_1x1", |
| 1566 | "conv_transpose1d", |
| 1567 | "silu", |
| 1568 | "leaky_relu", |
| 1569 | "tanh", |
| 1570 | "residual_add", |
| 1571 | "repeat_interleave", |
| 1572 | "masked_conv1d", |
| 1573 | "depthwise_conv1d", |
| 1574 | "low_rank_linear_or_conv1d" |
| 1575 | ], |
| 1576 | "runtime_boundary": { |
| 1577 | "neural_stack_self_contained": true, |
| 1578 | "arbitrary_text_self_contained": false, |
| 1579 | "accepted_runtime_input": "Piper phoneme ID sequence matching piper-phoneme-config.json", |
| 1580 | "output": "mono float waveform at manifest.sample_rate" |
| 1581 | }, |
| 1582 | "frontend": { |
| 1583 | "status": "not fully self-contained for arbitrary text yet", |
| 1584 | "input_contract": "runtime may accept Piper phoneme ID sequences directly", |
| 1585 | "included_config": "piper-phoneme-config.json", |
| 1586 | "piper_phoneme_type": "espeak", |
| 1587 | "piper_espeak_voice": "en", |
| 1588 | "phoneme_id_map_entries": 157, |
| 1589 | "next_work": "embed or rewrite the Piper/eSpeak-compatible frontend, or keep a strict phoneme-ID API" |
| 1590 | }, |
| 1591 | "size_budget": { |
| 1592 | "target_max_bytes": 4000000, |
| 1593 | "target_max_mib": 3.814697265625, |
| 1594 | "actual_model_asset_bytes": 2797270, |
| 1595 | "target_max_parameters": 1500000, |
| 1596 | "under_parameter_budget": true, |
| 1597 | "under_4mb_asset_budget": true |
| 1598 | }, |
| 1599 | "limitations": [ |
| 1600 | "This is a raw neural asset package, not a compiled standalone runtime binary.", |
| 1601 | "No PyTorch checkpoint, ONNX graph, or training state is included.", |
| 1602 | "Arbitrary text still requires a Piper/eSpeak-compatible frontend outside this package." |
| 1603 | ] |
| 1604 | } |
| 1605 | |