amy-en-1p46m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "amy-en-1p46m", |
| 4 | "language": "en_US", |
| 5 | "voice": "en_US-amy-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": 2908568, |
| 15 | "weights_sha256": "8a5e3665640cb4df561af2c297e0a9df12ea100555e380916dcc1064687e94f0", |
| 16 | "total_parameters": 1454284, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/en_US-amy-medium-retrain-20260712/a5-duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "db83e3ffc5c8488d9b26b5c6f8eb44308cb698f5d6140ecc2c854f58b65a9db6", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 129, |
| 24 | "hidden": 64, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 265, |
| 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": 135940, |
| 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": "ff79af0ca014dd21b29b097196d2486740ae0d6ebb82be5fe425d5221e6f8421" |
| 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": "76068934f28b60d5a203aecb48f9ee14b726b66a04a17795f3b471308c6ab40c" |
| 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": "10ae2cb1b3f2e5a79361eeb0bf2215c61d40c019b261d90cd9e6d4e82b46ae51" |
| 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": "0cdf00f9691d7c5a37dfb588bd7f5ef483330820056dfbf73400cfcde2f4522b" |
| 88 | }, |
| 89 | { |
| 90 | "name": "blocks.0.scale", |
| 91 | "shape": [ |
| 92 | 1 |
| 93 | ], |
| 94 | "dtype": "float16", |
| 95 | "offset_bytes": 82176, |
| 96 | "nbytes": 2, |
| 97 | "sha256": "4751ec2f62d6c6007f1741f49c1e2b073ef2ac41d2c1ff3450cecab5ff9f27c1" |
| 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": "c40ada89aba8af8afaaaab90ec0a5b273c7add905c5e492ead9d1296aedbe9bc" |
| 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": "9f3ee54cb142a95e0bc5beb2a202e8b3893f1918479add87dc24ced1d867ec64" |
| 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": "c5354edad42e0baf6d17da956bdb8725cd04e24780ece478905192495c2cafc1" |
| 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": "b1be8482444f952a6642c32f7edd8fb5cc3ffaad6a1d87e7b44979f7d0c0da31" |
| 142 | }, |
| 143 | { |
| 144 | "name": "blocks.1.scale", |
| 145 | "shape": [ |
| 146 | 1 |
| 147 | ], |
| 148 | "dtype": "float16", |
| 149 | "offset_bytes": 164354, |
| 150 | "nbytes": 2, |
| 151 | "sha256": "eabaacdc93772676a83ef893fd872a672754084792c633b85ec1209a7fdee5a2" |
| 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": "9380efaece0ff828baed3d9563a7f60098f556f77b98bdac8b61f9d49c407e18" |
| 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": "62784a9c6548ae633e136a6ecb84306460179ded933c1254783d460c507c04ca" |
| 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": "9d82d5995bf8271e00c55bd90389ed9601fb26ed0648403f9d1f880bb4bc7473" |
| 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": "e8957059d17d55fbb3166b1a7ab25794dddd148050d729bd7a77bfb8004ea40a" |
| 196 | }, |
| 197 | { |
| 198 | "name": "blocks.2.scale", |
| 199 | "shape": [ |
| 200 | 1 |
| 201 | ], |
| 202 | "dtype": "float16", |
| 203 | "offset_bytes": 246532, |
| 204 | "nbytes": 2, |
| 205 | "sha256": "27ca009eecb1ff93ea34c65ee7c372a15c01bf59c40c6736906c2d158cd64025" |
| 206 | }, |
| 207 | { |
| 208 | "name": "embedding.weight", |
| 209 | "shape": [ |
| 210 | 129, |
| 211 | 64 |
| 212 | ], |
| 213 | "dtype": "float16", |
| 214 | "offset_bytes": 246534, |
| 215 | "nbytes": 16512, |
| 216 | "sha256": "594dd084be8c32d2f86b0f5b94a73c388d4cc28b5a214d475e1cbe1b7e91a07b" |
| 217 | }, |
| 218 | { |
| 219 | "name": "input_proj.bias", |
| 220 | "shape": [ |
| 221 | 64 |
| 222 | ], |
| 223 | "dtype": "float16", |
| 224 | "offset_bytes": 263046, |
| 225 | "nbytes": 128, |
| 226 | "sha256": "cb18b942b1b1bc9c367b756f8bf21863e851db279f828ff57a52f70189a52b87" |
| 227 | }, |
| 228 | { |
| 229 | "name": "input_proj.weight", |
| 230 | "shape": [ |
| 231 | 64, |
| 232 | 67, |
| 233 | 1 |
| 234 | ], |
| 235 | "dtype": "float16", |
| 236 | "offset_bytes": 263174, |
| 237 | "nbytes": 8576, |
| 238 | "sha256": "68013ef208034c6cd870b31671708c85197f30cb5aae38f8fd7c4422863834af" |
| 239 | }, |
| 240 | { |
| 241 | "name": "output.bias", |
| 242 | "shape": [ |
| 243 | 1 |
| 244 | ], |
| 245 | "dtype": "float16", |
| 246 | "offset_bytes": 271750, |
| 247 | "nbytes": 2, |
| 248 | "sha256": "871a61a0998f1e9bad057b42a9122fef714022b56902e635c0cf154989b2fcb0" |
| 249 | }, |
| 250 | { |
| 251 | "name": "output.weight", |
| 252 | "shape": [ |
| 253 | 1, |
| 254 | 64, |
| 255 | 1 |
| 256 | ], |
| 257 | "dtype": "float16", |
| 258 | "offset_bytes": 271752, |
| 259 | "nbytes": 128, |
| 260 | "sha256": "9654e674867a895f86ee7a5d34efac219aa033f14e96d14f18ccb75782022601" |
| 261 | } |
| 262 | ] |
| 263 | }, |
| 264 | "acoustic": { |
| 265 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/en_US-amy-medium-retrain-20260712/e2-wide192-ac64-joint/latent-student.pt", |
| 266 | "checkpoint_sha256": "8f6bb5f39bce0189b949bd07287db3b58eb545e5fec909008613b814fc3eb413", |
| 267 | "config": { |
| 268 | "architecture": "token_context", |
| 269 | "vocab_size": 145, |
| 270 | "hidden": 64, |
| 271 | "depth": 4, |
| 272 | "kernel_size": 5, |
| 273 | "out_channels": 192, |
| 274 | "token_depth": 3 |
| 275 | }, |
| 276 | "parameters": 318023, |
| 277 | "tensors": [ |
| 278 | { |
| 279 | "name": "embedding.weight", |
| 280 | "shape": [ |
| 281 | 145, |
| 282 | 64 |
| 283 | ], |
| 284 | "dtype": "float16", |
| 285 | "offset_bytes": 271880, |
| 286 | "nbytes": 18560, |
| 287 | "sha256": "ed54bb0f8cbe23e82710c27a6fd1c377950d6d87ba30e7e5f479be96e3177866" |
| 288 | }, |
| 289 | { |
| 290 | "name": "frame_blocks.0.net.0.bias", |
| 291 | "shape": [ |
| 292 | 64 |
| 293 | ], |
| 294 | "dtype": "float16", |
| 295 | "offset_bytes": 290440, |
| 296 | "nbytes": 128, |
| 297 | "sha256": "f3facdb9652ea0c71f2e34ae44c533eb71f28e315b35a4a64ff322be98a6e9ed" |
| 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": 290568, |
| 308 | "nbytes": 40960, |
| 309 | "sha256": "8726dde1b8296d01d65a5c4578121349e29eb4376ee5a5427e29141813e9e3af" |
| 310 | }, |
| 311 | { |
| 312 | "name": "frame_blocks.0.net.2.bias", |
| 313 | "shape": [ |
| 314 | 64 |
| 315 | ], |
| 316 | "dtype": "float16", |
| 317 | "offset_bytes": 331528, |
| 318 | "nbytes": 128, |
| 319 | "sha256": "8709814d951c851bf6546cac26b27aa47a2f473d3a3a15316672e0943903449e" |
| 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": 331656, |
| 330 | "nbytes": 40960, |
| 331 | "sha256": "52850fbc77cf44b3db529d2e2606c4af31761d87e917521e9786ceb9d844cf21" |
| 332 | }, |
| 333 | { |
| 334 | "name": "frame_blocks.0.scale", |
| 335 | "shape": [ |
| 336 | 1 |
| 337 | ], |
| 338 | "dtype": "float16", |
| 339 | "offset_bytes": 372616, |
| 340 | "nbytes": 2, |
| 341 | "sha256": "4b73f90f711849b51375230dcb000cbe768a31a16ef54bd63007e44d3bde5ea1" |
| 342 | }, |
| 343 | { |
| 344 | "name": "frame_blocks.1.net.0.bias", |
| 345 | "shape": [ |
| 346 | 64 |
| 347 | ], |
| 348 | "dtype": "float16", |
| 349 | "offset_bytes": 372618, |
| 350 | "nbytes": 128, |
| 351 | "sha256": "b467f0c9257eb9f6a7eef951bf0d5c1a100d71a045c6adc55630a11e0f9bf5a3" |
| 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": 372746, |
| 362 | "nbytes": 40960, |
| 363 | "sha256": "7be78e0b4951bb5590d97a0a9a4d64336892e14ff8e96056491a12dde38d8a9b" |
| 364 | }, |
| 365 | { |
| 366 | "name": "frame_blocks.1.net.2.bias", |
| 367 | "shape": [ |
| 368 | 64 |
| 369 | ], |
| 370 | "dtype": "float16", |
| 371 | "offset_bytes": 413706, |
| 372 | "nbytes": 128, |
| 373 | "sha256": "bf330904774e71496ee564411799ec2770785eecbca75dfc60fdb06cfc8162fb" |
| 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": 413834, |
| 384 | "nbytes": 40960, |
| 385 | "sha256": "95358b12fae29c04ca44837c63a1c40bd25c60a1135c318a1b981c9962096d54" |
| 386 | }, |
| 387 | { |
| 388 | "name": "frame_blocks.1.scale", |
| 389 | "shape": [ |
| 390 | 1 |
| 391 | ], |
| 392 | "dtype": "float16", |
| 393 | "offset_bytes": 454794, |
| 394 | "nbytes": 2, |
| 395 | "sha256": "5e3d75a2758a973feedc73f54b4267314aa349d34331ae9ab8dbc37b3418db16" |
| 396 | }, |
| 397 | { |
| 398 | "name": "frame_blocks.2.net.0.bias", |
| 399 | "shape": [ |
| 400 | 64 |
| 401 | ], |
| 402 | "dtype": "float16", |
| 403 | "offset_bytes": 454796, |
| 404 | "nbytes": 128, |
| 405 | "sha256": "44c10b3440f7709284ed447df5956f9e07dda4f79c850e1296eff66a91b2b93a" |
| 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": 454924, |
| 416 | "nbytes": 40960, |
| 417 | "sha256": "a01e1b55160b87b847a53fc689ce7f4f148b6ab32a9d510d88e3d277d958ef5d" |
| 418 | }, |
| 419 | { |
| 420 | "name": "frame_blocks.2.net.2.bias", |
| 421 | "shape": [ |
| 422 | 64 |
| 423 | ], |
| 424 | "dtype": "float16", |
| 425 | "offset_bytes": 495884, |
| 426 | "nbytes": 128, |
| 427 | "sha256": "863d3160caa15b687ad54572ee42866506f16cbff55d71fca93be254d1d83ee9" |
| 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": 496012, |
| 438 | "nbytes": 40960, |
| 439 | "sha256": "2f96e47d9c54ca8ccc11e60bcc6ffe6a21d82b8e301db8e1d08cb3c9fcf37215" |
| 440 | }, |
| 441 | { |
| 442 | "name": "frame_blocks.2.scale", |
| 443 | "shape": [ |
| 444 | 1 |
| 445 | ], |
| 446 | "dtype": "float16", |
| 447 | "offset_bytes": 536972, |
| 448 | "nbytes": 2, |
| 449 | "sha256": "903212996fa1b93e35a7ef3876f32f17068470b3b9de74445b5d62012fd66af0" |
| 450 | }, |
| 451 | { |
| 452 | "name": "frame_blocks.3.net.0.bias", |
| 453 | "shape": [ |
| 454 | 64 |
| 455 | ], |
| 456 | "dtype": "float16", |
| 457 | "offset_bytes": 536974, |
| 458 | "nbytes": 128, |
| 459 | "sha256": "f4c4633ea8c2fd206abb72cb07939ec705b855239d243050dc27cf2379518577" |
| 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": 537102, |
| 470 | "nbytes": 40960, |
| 471 | "sha256": "4dcc7c4d50d558a17c59ad748b2121eb6b8819ce0ef0ed801ea005cabc1f9ec7" |
| 472 | }, |
| 473 | { |
| 474 | "name": "frame_blocks.3.net.2.bias", |
| 475 | "shape": [ |
| 476 | 64 |
| 477 | ], |
| 478 | "dtype": "float16", |
| 479 | "offset_bytes": 578062, |
| 480 | "nbytes": 128, |
| 481 | "sha256": "bf3ba31d2603f3fcc9aad4aa4b0f918a9917046e8475aedeb9f3626a01cc5a3f" |
| 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": 578190, |
| 492 | "nbytes": 40960, |
| 493 | "sha256": "51ad010e3b4a43aa82a4871681c53ec69d0da63fa5e64ae5cc568f2ee8939596" |
| 494 | }, |
| 495 | { |
| 496 | "name": "frame_blocks.3.scale", |
| 497 | "shape": [ |
| 498 | 1 |
| 499 | ], |
| 500 | "dtype": "float16", |
| 501 | "offset_bytes": 619150, |
| 502 | "nbytes": 2, |
| 503 | "sha256": "7b7143fd05b65fbdf36f7228ac6ab69136bbc478a5efeb38b2518e505a1721fa" |
| 504 | }, |
| 505 | { |
| 506 | "name": "frame_input_proj.bias", |
| 507 | "shape": [ |
| 508 | 64 |
| 509 | ], |
| 510 | "dtype": "float16", |
| 511 | "offset_bytes": 619152, |
| 512 | "nbytes": 128, |
| 513 | "sha256": "185109013b74eb372552305eaf839ada90c0323e1c8ec82fe2811ca53332a1cd" |
| 514 | }, |
| 515 | { |
| 516 | "name": "frame_input_proj.weight", |
| 517 | "shape": [ |
| 518 | 64, |
| 519 | 67, |
| 520 | 1 |
| 521 | ], |
| 522 | "dtype": "float16", |
| 523 | "offset_bytes": 619280, |
| 524 | "nbytes": 8576, |
| 525 | "sha256": "1f3cfed56ab1df58ccae5c4c33ade533fb8cd3618ea17cda7b614b42ef96fbde" |
| 526 | }, |
| 527 | { |
| 528 | "name": "output.bias", |
| 529 | "shape": [ |
| 530 | 192 |
| 531 | ], |
| 532 | "dtype": "float16", |
| 533 | "offset_bytes": 627856, |
| 534 | "nbytes": 384, |
| 535 | "sha256": "150ad914e52a3d10f595f143d5f4e1efeed0db3b0cd27d1db4ba7f21617d32cb" |
| 536 | }, |
| 537 | { |
| 538 | "name": "output.weight", |
| 539 | "shape": [ |
| 540 | 192, |
| 541 | 64, |
| 542 | 1 |
| 543 | ], |
| 544 | "dtype": "float16", |
| 545 | "offset_bytes": 628240, |
| 546 | "nbytes": 24576, |
| 547 | "sha256": "dd32a2f1c25dfe96f9791ef6060351d9c7279edb16a7a3e386911729325da1bf" |
| 548 | }, |
| 549 | { |
| 550 | "name": "token_blocks.0.net.0.bias", |
| 551 | "shape": [ |
| 552 | 64 |
| 553 | ], |
| 554 | "dtype": "float16", |
| 555 | "offset_bytes": 652816, |
| 556 | "nbytes": 128, |
| 557 | "sha256": "3adfba3261dc707947e784319f7d58deb2f5b2a07eb9e194cb98d024c537b113" |
| 558 | }, |
| 559 | { |
| 560 | "name": "token_blocks.0.net.0.weight", |
| 561 | "shape": [ |
| 562 | 64, |
| 563 | 64, |
| 564 | 5 |
| 565 | ], |
| 566 | "dtype": "float16", |
| 567 | "offset_bytes": 652944, |
| 568 | "nbytes": 40960, |
| 569 | "sha256": "81fcffd7f42c4ee74a387c97c07ba7ab409e87a9e586da448d86bcb029ab5a50" |
| 570 | }, |
| 571 | { |
| 572 | "name": "token_blocks.0.net.2.bias", |
| 573 | "shape": [ |
| 574 | 64 |
| 575 | ], |
| 576 | "dtype": "float16", |
| 577 | "offset_bytes": 693904, |
| 578 | "nbytes": 128, |
| 579 | "sha256": "407b0e1c451139b67c93eff8609ea16db149ca92c10759967af703d9c5a34e1e" |
| 580 | }, |
| 581 | { |
| 582 | "name": "token_blocks.0.net.2.weight", |
| 583 | "shape": [ |
| 584 | 64, |
| 585 | 64, |
| 586 | 5 |
| 587 | ], |
| 588 | "dtype": "float16", |
| 589 | "offset_bytes": 694032, |
| 590 | "nbytes": 40960, |
| 591 | "sha256": "54765629e5add5343879de7737c606d6f59073bfdc79f9ceb71850cff0752abc" |
| 592 | }, |
| 593 | { |
| 594 | "name": "token_blocks.0.scale", |
| 595 | "shape": [ |
| 596 | 1 |
| 597 | ], |
| 598 | "dtype": "float16", |
| 599 | "offset_bytes": 734992, |
| 600 | "nbytes": 2, |
| 601 | "sha256": "fb9dab57470045611754ea57105f2fe2607dcb22e0ca4923331496190d5e980d" |
| 602 | }, |
| 603 | { |
| 604 | "name": "token_blocks.1.net.0.bias", |
| 605 | "shape": [ |
| 606 | 64 |
| 607 | ], |
| 608 | "dtype": "float16", |
| 609 | "offset_bytes": 734994, |
| 610 | "nbytes": 128, |
| 611 | "sha256": "0269c879665521f30043c8b06ed347bad484b6281741c0fd3c4d51c50bf0e6ea" |
| 612 | }, |
| 613 | { |
| 614 | "name": "token_blocks.1.net.0.weight", |
| 615 | "shape": [ |
| 616 | 64, |
| 617 | 64, |
| 618 | 5 |
| 619 | ], |
| 620 | "dtype": "float16", |
| 621 | "offset_bytes": 735122, |
| 622 | "nbytes": 40960, |
| 623 | "sha256": "d54e851206dab1b20de6f6c1be3e2a563f2a27a99c0d2dc35389ef1bf3dc153a" |
| 624 | }, |
| 625 | { |
| 626 | "name": "token_blocks.1.net.2.bias", |
| 627 | "shape": [ |
| 628 | 64 |
| 629 | ], |
| 630 | "dtype": "float16", |
| 631 | "offset_bytes": 776082, |
| 632 | "nbytes": 128, |
| 633 | "sha256": "c12562c3e2ed4cfc284466e6f065a6f586de47355ef903180bf2bc08afba2e33" |
| 634 | }, |
| 635 | { |
| 636 | "name": "token_blocks.1.net.2.weight", |
| 637 | "shape": [ |
| 638 | 64, |
| 639 | 64, |
| 640 | 5 |
| 641 | ], |
| 642 | "dtype": "float16", |
| 643 | "offset_bytes": 776210, |
| 644 | "nbytes": 40960, |
| 645 | "sha256": "589665b5a5cd5a7cc94c5eb2bf42447176b29ed6f97ce094e489e3c7da015171" |
| 646 | }, |
| 647 | { |
| 648 | "name": "token_blocks.1.scale", |
| 649 | "shape": [ |
| 650 | 1 |
| 651 | ], |
| 652 | "dtype": "float16", |
| 653 | "offset_bytes": 817170, |
| 654 | "nbytes": 2, |
| 655 | "sha256": "b89f6c155ea697da28d2033ff939c76048c6d8d281b32b9e077a252c3a0251e3" |
| 656 | }, |
| 657 | { |
| 658 | "name": "token_blocks.2.net.0.bias", |
| 659 | "shape": [ |
| 660 | 64 |
| 661 | ], |
| 662 | "dtype": "float16", |
| 663 | "offset_bytes": 817172, |
| 664 | "nbytes": 128, |
| 665 | "sha256": "847e4b2220b1e5dc1b4ec2b3d75bdeacdb1c9f42c895da7d7848071239dc191f" |
| 666 | }, |
| 667 | { |
| 668 | "name": "token_blocks.2.net.0.weight", |
| 669 | "shape": [ |
| 670 | 64, |
| 671 | 64, |
| 672 | 5 |
| 673 | ], |
| 674 | "dtype": "float16", |
| 675 | "offset_bytes": 817300, |
| 676 | "nbytes": 40960, |
| 677 | "sha256": "fda281b8a749f55cb839aec37017a5243a9a36dc73240f13bc73086cdfa9c85e" |
| 678 | }, |
| 679 | { |
| 680 | "name": "token_blocks.2.net.2.bias", |
| 681 | "shape": [ |
| 682 | 64 |
| 683 | ], |
| 684 | "dtype": "float16", |
| 685 | "offset_bytes": 858260, |
| 686 | "nbytes": 128, |
| 687 | "sha256": "a7450bb57306d56e7ec549f32a751f5c71cdd39683d9b5a6dd356531392f2d65" |
| 688 | }, |
| 689 | { |
| 690 | "name": "token_blocks.2.net.2.weight", |
| 691 | "shape": [ |
| 692 | 64, |
| 693 | 64, |
| 694 | 5 |
| 695 | ], |
| 696 | "dtype": "float16", |
| 697 | "offset_bytes": 858388, |
| 698 | "nbytes": 40960, |
| 699 | "sha256": "bbc2e600335b756323d2c4cb078c59284475cccbfd0771cb04ba02970f2e3e18" |
| 700 | }, |
| 701 | { |
| 702 | "name": "token_blocks.2.scale", |
| 703 | "shape": [ |
| 704 | 1 |
| 705 | ], |
| 706 | "dtype": "float16", |
| 707 | "offset_bytes": 899348, |
| 708 | "nbytes": 2, |
| 709 | "sha256": "9837985ab5c01cce23af80fb7d09b7b4e4e16abf34d342ebc5ff6001c19b2805" |
| 710 | }, |
| 711 | { |
| 712 | "name": "token_input_proj.bias", |
| 713 | "shape": [ |
| 714 | 64 |
| 715 | ], |
| 716 | "dtype": "float16", |
| 717 | "offset_bytes": 899350, |
| 718 | "nbytes": 128, |
| 719 | "sha256": "fc0100bc3b17bd5e8ef7420e63d631e87a4e59c8fefcdb674ebc0c5b86755aa2" |
| 720 | }, |
| 721 | { |
| 722 | "name": "token_input_proj.weight", |
| 723 | "shape": [ |
| 724 | 64, |
| 725 | 66, |
| 726 | 1 |
| 727 | ], |
| 728 | "dtype": "float16", |
| 729 | "offset_bytes": 899478, |
| 730 | "nbytes": 8448, |
| 731 | "sha256": "c1dc5393ece500cae9fe22cdd7d049ffbc0a685be621356029000450fc0e3a91" |
| 732 | } |
| 733 | ] |
| 734 | }, |
| 735 | "decoder": { |
| 736 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/en_US-amy-medium-retrain-20260712/e2-wide192-ac64-joint/decoder-student.pt", |
| 737 | "checkpoint_sha256": "82cea77f4e01dd314c996203c927e50a5365685d938297fadbb691c1f873f2eb", |
| 738 | "config": { |
| 739 | "in_channels": 192, |
| 740 | "source_in_channels": 192, |
| 741 | "channels": [ |
| 742 | 192, |
| 743 | 96, |
| 744 | 48, |
| 745 | 24 |
| 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/en_US-amy-medium-retrain-20260712/w1-wide192-40k/decoder-student.pt", |
| 786 | "decoder_init": { |
| 787 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/en_US-amy-medium-retrain-20260712/w1-wide192-40k/decoder-student.pt", |
| 788 | "decoder_parameters": 1000321, |
| 789 | "matched_config": { |
| 790 | "in_channels": 192, |
| 791 | "channels": [ |
| 792 | 192, |
| 793 | 96, |
| 794 | 48, |
| 795 | 24 |
| 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/en_US-amy-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": 1000321, |
| 900 | "tensors": [ |
| 901 | { |
| 902 | "name": "post.bias", |
| 903 | "shape": [ |
| 904 | 1 |
| 905 | ], |
| 906 | "dtype": "float16", |
| 907 | "offset_bytes": 907926, |
| 908 | "nbytes": 2, |
| 909 | "sha256": "6f41899897d38eeef914a3586d78cdbd2a0e36f31decaa0a9283570fec36fff9" |
| 910 | }, |
| 911 | { |
| 912 | "name": "post.weight", |
| 913 | "shape": [ |
| 914 | 1, |
| 915 | 24, |
| 916 | 7 |
| 917 | ], |
| 918 | "dtype": "float16", |
| 919 | "offset_bytes": 907928, |
| 920 | "nbytes": 336, |
| 921 | "sha256": "7c7d52a497ae58395d5f2c8888a8317eefdfac746a3bc7468493f494a25b1530" |
| 922 | }, |
| 923 | { |
| 924 | "name": "pre.bias", |
| 925 | "shape": [ |
| 926 | 192 |
| 927 | ], |
| 928 | "dtype": "float16", |
| 929 | "offset_bytes": 908264, |
| 930 | "nbytes": 384, |
| 931 | "sha256": "41d1eb3a101c911d6010722f06da2e0ea136224e8e753db36e9cf43f85866174" |
| 932 | }, |
| 933 | { |
| 934 | "name": "pre.weight", |
| 935 | "shape": [ |
| 936 | 192, |
| 937 | 192, |
| 938 | 7 |
| 939 | ], |
| 940 | "dtype": "float16", |
| 941 | "offset_bytes": 908648, |
| 942 | "nbytes": 516096, |
| 943 | "sha256": "52669d68f57f2d20bd688f1036ae4ac2398df6fb73c490fe807b4f360d510f74" |
| 944 | }, |
| 945 | { |
| 946 | "name": "res0.0.blocks.0.conv1.bias", |
| 947 | "shape": [ |
| 948 | 96 |
| 949 | ], |
| 950 | "dtype": "float16", |
| 951 | "offset_bytes": 1424744, |
| 952 | "nbytes": 192, |
| 953 | "sha256": "594717236c93d7e38afac5637ee4340b27cb6d77e8c91087310abcf2a11a5a17" |
| 954 | }, |
| 955 | { |
| 956 | "name": "res0.0.blocks.0.conv1.weight", |
| 957 | "shape": [ |
| 958 | 96, |
| 959 | 96, |
| 960 | 3 |
| 961 | ], |
| 962 | "dtype": "float16", |
| 963 | "offset_bytes": 1424936, |
| 964 | "nbytes": 55296, |
| 965 | "sha256": "e86a3ce8c0ce9be6933b0ee4e52ff4cd9bdbcdab313e15ed34e64bd534de5ba7" |
| 966 | }, |
| 967 | { |
| 968 | "name": "res0.0.blocks.0.conv2.bias", |
| 969 | "shape": [ |
| 970 | 96 |
| 971 | ], |
| 972 | "dtype": "float16", |
| 973 | "offset_bytes": 1480232, |
| 974 | "nbytes": 192, |
| 975 | "sha256": "fadec52974aff8493a6c4699510fc3dae192467cbaaedf3a60cffa621638a0be" |
| 976 | }, |
| 977 | { |
| 978 | "name": "res0.0.blocks.0.conv2.weight", |
| 979 | "shape": [ |
| 980 | 96, |
| 981 | 96, |
| 982 | 3 |
| 983 | ], |
| 984 | "dtype": "float16", |
| 985 | "offset_bytes": 1480424, |
| 986 | "nbytes": 55296, |
| 987 | "sha256": "4c6db3b91fa8b35a1b7869d8bc432ae1bd22496e7353f9c5fd51724108c1f221" |
| 988 | }, |
| 989 | { |
| 990 | "name": "res0.0.blocks.1.conv1.bias", |
| 991 | "shape": [ |
| 992 | 96 |
| 993 | ], |
| 994 | "dtype": "float16", |
| 995 | "offset_bytes": 1535720, |
| 996 | "nbytes": 192, |
| 997 | "sha256": "c026831d7d8c5e8a1140ede24a0ce309ff9bab4a32845502737c5cb76e4d80eb" |
| 998 | }, |
| 999 | { |
| 1000 | "name": "res0.0.blocks.1.conv1.weight", |
| 1001 | "shape": [ |
| 1002 | 96, |
| 1003 | 96, |
| 1004 | 5 |
| 1005 | ], |
| 1006 | "dtype": "float16", |
| 1007 | "offset_bytes": 1535912, |
| 1008 | "nbytes": 92160, |
| 1009 | "sha256": "cf532dd5e5c3cbd0fca85179bc1aefa71b4e200e34692470e08b32391eef1041" |
| 1010 | }, |
| 1011 | { |
| 1012 | "name": "res0.0.blocks.1.conv2.bias", |
| 1013 | "shape": [ |
| 1014 | 96 |
| 1015 | ], |
| 1016 | "dtype": "float16", |
| 1017 | "offset_bytes": 1628072, |
| 1018 | "nbytes": 192, |
| 1019 | "sha256": "917c83785ae52d978f17c3ae1b1ad345372fdd4959f8cbb2347317ad5117791c" |
| 1020 | }, |
| 1021 | { |
| 1022 | "name": "res0.0.blocks.1.conv2.weight", |
| 1023 | "shape": [ |
| 1024 | 96, |
| 1025 | 96, |
| 1026 | 5 |
| 1027 | ], |
| 1028 | "dtype": "float16", |
| 1029 | "offset_bytes": 1628264, |
| 1030 | "nbytes": 92160, |
| 1031 | "sha256": "1bb07f750731f1bb1c6f617f56ed6d01a8907066e2b7bba736ba18a227ba3846" |
| 1032 | }, |
| 1033 | { |
| 1034 | "name": "res0.0.blocks.2.conv1.bias", |
| 1035 | "shape": [ |
| 1036 | 96 |
| 1037 | ], |
| 1038 | "dtype": "float16", |
| 1039 | "offset_bytes": 1720424, |
| 1040 | "nbytes": 192, |
| 1041 | "sha256": "6b0094ff1b5212ae7dae9e42bc22e6acd496e7c3f84e67affc07085e2898d2b0" |
| 1042 | }, |
| 1043 | { |
| 1044 | "name": "res0.0.blocks.2.conv1.weight", |
| 1045 | "shape": [ |
| 1046 | 96, |
| 1047 | 96, |
| 1048 | 7 |
| 1049 | ], |
| 1050 | "dtype": "float16", |
| 1051 | "offset_bytes": 1720616, |
| 1052 | "nbytes": 129024, |
| 1053 | "sha256": "0c5af9a7aceb34c2e2f61e8008a13a1e029c7ea03fb3bb28fb7f7d57b644447e" |
| 1054 | }, |
| 1055 | { |
| 1056 | "name": "res0.0.blocks.2.conv2.bias", |
| 1057 | "shape": [ |
| 1058 | 96 |
| 1059 | ], |
| 1060 | "dtype": "float16", |
| 1061 | "offset_bytes": 1849640, |
| 1062 | "nbytes": 192, |
| 1063 | "sha256": "a4abfb56a712e9cac96a297c2b1121d3de08594ef71d117506a932b0f4f2d79c" |
| 1064 | }, |
| 1065 | { |
| 1066 | "name": "res0.0.blocks.2.conv2.weight", |
| 1067 | "shape": [ |
| 1068 | 96, |
| 1069 | 96, |
| 1070 | 7 |
| 1071 | ], |
| 1072 | "dtype": "float16", |
| 1073 | "offset_bytes": 1849832, |
| 1074 | "nbytes": 129024, |
| 1075 | "sha256": "ec98f1406de5613ba7c523d4c3a1b39ac1b77d4561db8ec00a1ec8163be82186" |
| 1076 | }, |
| 1077 | { |
| 1078 | "name": "res1.0.blocks.0.conv1.bias", |
| 1079 | "shape": [ |
| 1080 | 48 |
| 1081 | ], |
| 1082 | "dtype": "float16", |
| 1083 | "offset_bytes": 1978856, |
| 1084 | "nbytes": 96, |
| 1085 | "sha256": "655c27ab895cf10a15f6a571883de758717162a2a0739b8c8d1ad3d8afeab50e" |
| 1086 | }, |
| 1087 | { |
| 1088 | "name": "res1.0.blocks.0.conv1.weight", |
| 1089 | "shape": [ |
| 1090 | 48, |
| 1091 | 48, |
| 1092 | 3 |
| 1093 | ], |
| 1094 | "dtype": "float16", |
| 1095 | "offset_bytes": 1978952, |
| 1096 | "nbytes": 13824, |
| 1097 | "sha256": "a4a9faaa1ac240cfa2e9db1dccde291970cd5ae7f447b64b809023f9480c45e6" |
| 1098 | }, |
| 1099 | { |
| 1100 | "name": "res1.0.blocks.0.conv2.bias", |
| 1101 | "shape": [ |
| 1102 | 48 |
| 1103 | ], |
| 1104 | "dtype": "float16", |
| 1105 | "offset_bytes": 1992776, |
| 1106 | "nbytes": 96, |
| 1107 | "sha256": "c9b38683186b6b8fb5372e09a3d8e3e921e27943622aa880bd8a0f3c43256af9" |
| 1108 | }, |
| 1109 | { |
| 1110 | "name": "res1.0.blocks.0.conv2.weight", |
| 1111 | "shape": [ |
| 1112 | 48, |
| 1113 | 48, |
| 1114 | 3 |
| 1115 | ], |
| 1116 | "dtype": "float16", |
| 1117 | "offset_bytes": 1992872, |
| 1118 | "nbytes": 13824, |
| 1119 | "sha256": "2ee5a04d1ed25a10fe41ece5c589bcb73a25f06de54f353ea6089dc26c1f320f" |
| 1120 | }, |
| 1121 | { |
| 1122 | "name": "res1.0.blocks.1.conv1.bias", |
| 1123 | "shape": [ |
| 1124 | 48 |
| 1125 | ], |
| 1126 | "dtype": "float16", |
| 1127 | "offset_bytes": 2006696, |
| 1128 | "nbytes": 96, |
| 1129 | "sha256": "86de76655e98e4919521f2333ed0d20b573aab2bfc0a3418decd464e47ac765a" |
| 1130 | }, |
| 1131 | { |
| 1132 | "name": "res1.0.blocks.1.conv1.weight", |
| 1133 | "shape": [ |
| 1134 | 48, |
| 1135 | 48, |
| 1136 | 5 |
| 1137 | ], |
| 1138 | "dtype": "float16", |
| 1139 | "offset_bytes": 2006792, |
| 1140 | "nbytes": 23040, |
| 1141 | "sha256": "690c8b85cb317be0ab8e4b939e12000c16efd58fe44f853118559084fafe5f79" |
| 1142 | }, |
| 1143 | { |
| 1144 | "name": "res1.0.blocks.1.conv2.bias", |
| 1145 | "shape": [ |
| 1146 | 48 |
| 1147 | ], |
| 1148 | "dtype": "float16", |
| 1149 | "offset_bytes": 2029832, |
| 1150 | "nbytes": 96, |
| 1151 | "sha256": "87fdc079e84469fd9d7c3548f4e5d67a8ea68d26a3bd48bf1dd2bb2fc1e902c5" |
| 1152 | }, |
| 1153 | { |
| 1154 | "name": "res1.0.blocks.1.conv2.weight", |
| 1155 | "shape": [ |
| 1156 | 48, |
| 1157 | 48, |
| 1158 | 5 |
| 1159 | ], |
| 1160 | "dtype": "float16", |
| 1161 | "offset_bytes": 2029928, |
| 1162 | "nbytes": 23040, |
| 1163 | "sha256": "7f1a0e4322a99c253878fbdcd4a5e9aa45f54dcc5b72fee2c458dd4fa3b4fdcc" |
| 1164 | }, |
| 1165 | { |
| 1166 | "name": "res1.0.blocks.2.conv1.bias", |
| 1167 | "shape": [ |
| 1168 | 48 |
| 1169 | ], |
| 1170 | "dtype": "float16", |
| 1171 | "offset_bytes": 2052968, |
| 1172 | "nbytes": 96, |
| 1173 | "sha256": "8e1734ebbd1b21cae8e1d06943b3c367c50742f1a5b09f15bbb39f885448c23a" |
| 1174 | }, |
| 1175 | { |
| 1176 | "name": "res1.0.blocks.2.conv1.weight", |
| 1177 | "shape": [ |
| 1178 | 48, |
| 1179 | 48, |
| 1180 | 7 |
| 1181 | ], |
| 1182 | "dtype": "float16", |
| 1183 | "offset_bytes": 2053064, |
| 1184 | "nbytes": 32256, |
| 1185 | "sha256": "b45d4c2f215c1bba99f4094c95942a172bab2dedb8969b0f13c0cd0881317028" |
| 1186 | }, |
| 1187 | { |
| 1188 | "name": "res1.0.blocks.2.conv2.bias", |
| 1189 | "shape": [ |
| 1190 | 48 |
| 1191 | ], |
| 1192 | "dtype": "float16", |
| 1193 | "offset_bytes": 2085320, |
| 1194 | "nbytes": 96, |
| 1195 | "sha256": "1ae08b011df299e2da8a4821889b3ccbf1dfed36caf689ff84f8d37e6622203b" |
| 1196 | }, |
| 1197 | { |
| 1198 | "name": "res1.0.blocks.2.conv2.weight", |
| 1199 | "shape": [ |
| 1200 | 48, |
| 1201 | 48, |
| 1202 | 7 |
| 1203 | ], |
| 1204 | "dtype": "float16", |
| 1205 | "offset_bytes": 2085416, |
| 1206 | "nbytes": 32256, |
| 1207 | "sha256": "ab6ed1571f456586a3d9c91e4bbbae1933506c76eb19da8c032e69a8a0de4391" |
| 1208 | }, |
| 1209 | { |
| 1210 | "name": "res2.0.blocks.0.conv1.bias", |
| 1211 | "shape": [ |
| 1212 | 24 |
| 1213 | ], |
| 1214 | "dtype": "float16", |
| 1215 | "offset_bytes": 2117672, |
| 1216 | "nbytes": 48, |
| 1217 | "sha256": "3c34eb7c16e1b136f2b7509b8707cc375d0a214823c3987cff7e29413f645455" |
| 1218 | }, |
| 1219 | { |
| 1220 | "name": "res2.0.blocks.0.conv1.weight", |
| 1221 | "shape": [ |
| 1222 | 24, |
| 1223 | 24, |
| 1224 | 3 |
| 1225 | ], |
| 1226 | "dtype": "float16", |
| 1227 | "offset_bytes": 2117720, |
| 1228 | "nbytes": 3456, |
| 1229 | "sha256": "ab5007553a96e546c19d8020beadc855e5cd18b58d14049c1786cf9a060db422" |
| 1230 | }, |
| 1231 | { |
| 1232 | "name": "res2.0.blocks.0.conv2.bias", |
| 1233 | "shape": [ |
| 1234 | 24 |
| 1235 | ], |
| 1236 | "dtype": "float16", |
| 1237 | "offset_bytes": 2121176, |
| 1238 | "nbytes": 48, |
| 1239 | "sha256": "39f258ec506712a35d7d5040f8586bf728ccba1b086122a34647ac3f079c382c" |
| 1240 | }, |
| 1241 | { |
| 1242 | "name": "res2.0.blocks.0.conv2.weight", |
| 1243 | "shape": [ |
| 1244 | 24, |
| 1245 | 24, |
| 1246 | 3 |
| 1247 | ], |
| 1248 | "dtype": "float16", |
| 1249 | "offset_bytes": 2121224, |
| 1250 | "nbytes": 3456, |
| 1251 | "sha256": "878bef8c1bb0a23cf6f7b6e4a191c52dc0825eab96a67c5cfd4c1b4ef3c9bbe0" |
| 1252 | }, |
| 1253 | { |
| 1254 | "name": "res2.0.blocks.1.conv1.bias", |
| 1255 | "shape": [ |
| 1256 | 24 |
| 1257 | ], |
| 1258 | "dtype": "float16", |
| 1259 | "offset_bytes": 2124680, |
| 1260 | "nbytes": 48, |
| 1261 | "sha256": "5d50c8858508ee31c8bb7430874a55859403655f9241fef9baf73d5c86a0d3af" |
| 1262 | }, |
| 1263 | { |
| 1264 | "name": "res2.0.blocks.1.conv1.weight", |
| 1265 | "shape": [ |
| 1266 | 24, |
| 1267 | 24, |
| 1268 | 5 |
| 1269 | ], |
| 1270 | "dtype": "float16", |
| 1271 | "offset_bytes": 2124728, |
| 1272 | "nbytes": 5760, |
| 1273 | "sha256": "78646fb483afa1a71c767db57e1ef2f33a451de98836bb736e505868db774e76" |
| 1274 | }, |
| 1275 | { |
| 1276 | "name": "res2.0.blocks.1.conv2.bias", |
| 1277 | "shape": [ |
| 1278 | 24 |
| 1279 | ], |
| 1280 | "dtype": "float16", |
| 1281 | "offset_bytes": 2130488, |
| 1282 | "nbytes": 48, |
| 1283 | "sha256": "660a5ee4fda2afad8c7ee36bb70462ec22867c4bc57743f3540284acc75cb1a8" |
| 1284 | }, |
| 1285 | { |
| 1286 | "name": "res2.0.blocks.1.conv2.weight", |
| 1287 | "shape": [ |
| 1288 | 24, |
| 1289 | 24, |
| 1290 | 5 |
| 1291 | ], |
| 1292 | "dtype": "float16", |
| 1293 | "offset_bytes": 2130536, |
| 1294 | "nbytes": 5760, |
| 1295 | "sha256": "b810c3c9deeef54c6f946b82b583321ac6ee3f2c1a2fac51828c71f53e1d6932" |
| 1296 | }, |
| 1297 | { |
| 1298 | "name": "res2.0.blocks.2.conv1.bias", |
| 1299 | "shape": [ |
| 1300 | 24 |
| 1301 | ], |
| 1302 | "dtype": "float16", |
| 1303 | "offset_bytes": 2136296, |
| 1304 | "nbytes": 48, |
| 1305 | "sha256": "20598d2ee3aea8c9ca979866a84c3309a06e0edcc7d333b8a8e12893e9ae7207" |
| 1306 | }, |
| 1307 | { |
| 1308 | "name": "res2.0.blocks.2.conv1.weight", |
| 1309 | "shape": [ |
| 1310 | 24, |
| 1311 | 24, |
| 1312 | 7 |
| 1313 | ], |
| 1314 | "dtype": "float16", |
| 1315 | "offset_bytes": 2136344, |
| 1316 | "nbytes": 8064, |
| 1317 | "sha256": "14864a5893baaf35b5bc23760bd68aea7fc0cefc142a47a1aee52aaa34de154e" |
| 1318 | }, |
| 1319 | { |
| 1320 | "name": "res2.0.blocks.2.conv2.bias", |
| 1321 | "shape": [ |
| 1322 | 24 |
| 1323 | ], |
| 1324 | "dtype": "float16", |
| 1325 | "offset_bytes": 2144408, |
| 1326 | "nbytes": 48, |
| 1327 | "sha256": "3e0d1ac22046007a79180b414072777ffc48b31882362646e7a6ad15a9250306" |
| 1328 | }, |
| 1329 | { |
| 1330 | "name": "res2.0.blocks.2.conv2.weight", |
| 1331 | "shape": [ |
| 1332 | 24, |
| 1333 | 24, |
| 1334 | 7 |
| 1335 | ], |
| 1336 | "dtype": "float16", |
| 1337 | "offset_bytes": 2144456, |
| 1338 | "nbytes": 8064, |
| 1339 | "sha256": "67d2afec63873af779b7ba8e5c98912221310790d490baaa9d9ed2a3ef4439ab" |
| 1340 | }, |
| 1341 | { |
| 1342 | "name": "up0.bias", |
| 1343 | "shape": [ |
| 1344 | 96 |
| 1345 | ], |
| 1346 | "dtype": "float16", |
| 1347 | "offset_bytes": 2152520, |
| 1348 | "nbytes": 192, |
| 1349 | "sha256": "12e69cdad1f29f8028151af70b6822b2f782cf21eca2da649948d4044fc51378" |
| 1350 | }, |
| 1351 | { |
| 1352 | "name": "up0.weight", |
| 1353 | "shape": [ |
| 1354 | 192, |
| 1355 | 96, |
| 1356 | 16 |
| 1357 | ], |
| 1358 | "dtype": "float16", |
| 1359 | "offset_bytes": 2152712, |
| 1360 | "nbytes": 589824, |
| 1361 | "sha256": "7fb23727b70524489b1001a6d9f5e1152815401573c46ea7705d466fc3340514" |
| 1362 | }, |
| 1363 | { |
| 1364 | "name": "up1.bias", |
| 1365 | "shape": [ |
| 1366 | 48 |
| 1367 | ], |
| 1368 | "dtype": "float16", |
| 1369 | "offset_bytes": 2742536, |
| 1370 | "nbytes": 96, |
| 1371 | "sha256": "a8aff8cb66a65115ce140eaa665ccd8478f4afded76bdc0e261895977c78f7a5" |
| 1372 | }, |
| 1373 | { |
| 1374 | "name": "up1.weight", |
| 1375 | "shape": [ |
| 1376 | 96, |
| 1377 | 48, |
| 1378 | 16 |
| 1379 | ], |
| 1380 | "dtype": "float16", |
| 1381 | "offset_bytes": 2742632, |
| 1382 | "nbytes": 147456, |
| 1383 | "sha256": "60633639c6b0cf9e6b3efaec5dd73adea9b536978443acdb090dcc2c1598ecf8" |
| 1384 | }, |
| 1385 | { |
| 1386 | "name": "up2.bias", |
| 1387 | "shape": [ |
| 1388 | 24 |
| 1389 | ], |
| 1390 | "dtype": "float16", |
| 1391 | "offset_bytes": 2890088, |
| 1392 | "nbytes": 48, |
| 1393 | "sha256": "293c0f400e2812dc9d593771902044ec0f9e38010922010bf9f706892f1e516e" |
| 1394 | }, |
| 1395 | { |
| 1396 | "name": "up2.weight", |
| 1397 | "shape": [ |
| 1398 | 48, |
| 1399 | 24, |
| 1400 | 8 |
| 1401 | ], |
| 1402 | "dtype": "float16", |
| 1403 | "offset_bytes": 2890136, |
| 1404 | "nbytes": 18432, |
| 1405 | "sha256": "67d99f6f7f78fea65c20e1189f21d100e753194f907bcdc946d37de443fd1f59" |
| 1406 | } |
| 1407 | ] |
| 1408 | } |
| 1409 | }, |
| 1410 | "runtime_ops": [ |
| 1411 | "embedding_lookup", |
| 1412 | "conv1d_same", |
| 1413 | "conv1d_1x1", |
| 1414 | "conv_transpose1d", |
| 1415 | "silu", |
| 1416 | "leaky_relu", |
| 1417 | "tanh", |
| 1418 | "residual_add", |
| 1419 | "repeat_interleave", |
| 1420 | "masked_conv1d", |
| 1421 | "depthwise_conv1d", |
| 1422 | "low_rank_linear_or_conv1d" |
| 1423 | ], |
| 1424 | "runtime_boundary": { |
| 1425 | "neural_stack_self_contained": true, |
| 1426 | "arbitrary_text_self_contained": false, |
| 1427 | "accepted_runtime_input": "Piper phoneme ID sequence matching piper-phoneme-config.json", |
| 1428 | "output": "mono float waveform at manifest.sample_rate" |
| 1429 | }, |
| 1430 | "frontend": { |
| 1431 | "status": "not fully self-contained for arbitrary text yet", |
| 1432 | "input_contract": "runtime may accept Piper phoneme ID sequences directly", |
| 1433 | "included_config": "piper-phoneme-config.json", |
| 1434 | "piper_phoneme_type": "espeak", |
| 1435 | "piper_espeak_voice": "en-us", |
| 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": 2913450, |
| 1443 | "target_max_parameters": 1500000, |
| 1444 | "under_parameter_budget": true, |
| 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 | |