amy-en-1p8m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "amy-en-1p8m", |
| 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": 3668760, |
| 15 | "weights_sha256": "bf2daa62fc65a1cb96ef0967fa4f28c45aae8528e62f957e8944370cabbe99b2", |
| 16 | "total_parameters": 1834380, |
| 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/w3-wide192-joint/latent-student.pt", |
| 266 | "checkpoint_sha256": "138cb83654ff3f1b8bc6fa5e3870c49d9330b9ab50f62c8a623b57282f5a4c62", |
| 267 | "config": { |
| 268 | "architecture": "token_context", |
| 269 | "vocab_size": 145, |
| 270 | "hidden": 96, |
| 271 | "depth": 4, |
| 272 | "kernel_size": 5, |
| 273 | "out_channels": 192, |
| 274 | "token_depth": 3 |
| 275 | }, |
| 276 | "parameters": 698119, |
| 277 | "tensors": [ |
| 278 | { |
| 279 | "name": "embedding.weight", |
| 280 | "shape": [ |
| 281 | 145, |
| 282 | 96 |
| 283 | ], |
| 284 | "dtype": "float16", |
| 285 | "offset_bytes": 271880, |
| 286 | "nbytes": 27840, |
| 287 | "sha256": "8fe01f580d82e00fc2c4b5791bd37c59b7a9f743b8cc53033a7c421a3497c87a" |
| 288 | }, |
| 289 | { |
| 290 | "name": "frame_blocks.0.net.0.bias", |
| 291 | "shape": [ |
| 292 | 96 |
| 293 | ], |
| 294 | "dtype": "float16", |
| 295 | "offset_bytes": 299720, |
| 296 | "nbytes": 192, |
| 297 | "sha256": "3895d7ea3b66e7412f7f4346b49c44b87ca3dc17fc73e7c5548d1ee8a3b324b4" |
| 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": 299912, |
| 308 | "nbytes": 92160, |
| 309 | "sha256": "4db2ed1cc92f88c96a78e74eca68ff86a443ae8208ef30e26e0cadf30309550f" |
| 310 | }, |
| 311 | { |
| 312 | "name": "frame_blocks.0.net.2.bias", |
| 313 | "shape": [ |
| 314 | 96 |
| 315 | ], |
| 316 | "dtype": "float16", |
| 317 | "offset_bytes": 392072, |
| 318 | "nbytes": 192, |
| 319 | "sha256": "41099c1d9075fcf080daad2430314c24ac36ea1cc85a929f14ed827355175abf" |
| 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": 392264, |
| 330 | "nbytes": 92160, |
| 331 | "sha256": "33e9ae4bcb0199a6a5a78b4a2cb7580a7ff5668c6824f67a807d6fa1463e6b19" |
| 332 | }, |
| 333 | { |
| 334 | "name": "frame_blocks.0.scale", |
| 335 | "shape": [ |
| 336 | 1 |
| 337 | ], |
| 338 | "dtype": "float16", |
| 339 | "offset_bytes": 484424, |
| 340 | "nbytes": 2, |
| 341 | "sha256": "668d511224022d12d6542cebba55d7d3a66676927f2f71879cee9116c9925007" |
| 342 | }, |
| 343 | { |
| 344 | "name": "frame_blocks.1.net.0.bias", |
| 345 | "shape": [ |
| 346 | 96 |
| 347 | ], |
| 348 | "dtype": "float16", |
| 349 | "offset_bytes": 484426, |
| 350 | "nbytes": 192, |
| 351 | "sha256": "5eb42163b30b5f88c2e5d81b17dacc7731274e2663836c255207448912b92262" |
| 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": 484618, |
| 362 | "nbytes": 92160, |
| 363 | "sha256": "21d34fccf21b9cfdef4301359b98350b0d47abf815f217def4230650bbda3842" |
| 364 | }, |
| 365 | { |
| 366 | "name": "frame_blocks.1.net.2.bias", |
| 367 | "shape": [ |
| 368 | 96 |
| 369 | ], |
| 370 | "dtype": "float16", |
| 371 | "offset_bytes": 576778, |
| 372 | "nbytes": 192, |
| 373 | "sha256": "1c1d8ec36189be43f8e5727d21c90df69ec9dd25ae3afa656f7ab1990dea9f03" |
| 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": 576970, |
| 384 | "nbytes": 92160, |
| 385 | "sha256": "1c0ab17a205834e4d185bb4a2193e844493bd98cfbf4cb47774f554d2e669995" |
| 386 | }, |
| 387 | { |
| 388 | "name": "frame_blocks.1.scale", |
| 389 | "shape": [ |
| 390 | 1 |
| 391 | ], |
| 392 | "dtype": "float16", |
| 393 | "offset_bytes": 669130, |
| 394 | "nbytes": 2, |
| 395 | "sha256": "c3a9ce8326f05f8639235470b036becc19f9a5bb601b94bba44d9fe3ea812787" |
| 396 | }, |
| 397 | { |
| 398 | "name": "frame_blocks.2.net.0.bias", |
| 399 | "shape": [ |
| 400 | 96 |
| 401 | ], |
| 402 | "dtype": "float16", |
| 403 | "offset_bytes": 669132, |
| 404 | "nbytes": 192, |
| 405 | "sha256": "75430352d8be75ec8d6aae0cba9532f30a018fcd15e551b8f5021015707b7662" |
| 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": 669324, |
| 416 | "nbytes": 92160, |
| 417 | "sha256": "38e2a52784f94d8e7fd843d7a31e3d9aefffb7653158fe6248d33bc2a69cde7d" |
| 418 | }, |
| 419 | { |
| 420 | "name": "frame_blocks.2.net.2.bias", |
| 421 | "shape": [ |
| 422 | 96 |
| 423 | ], |
| 424 | "dtype": "float16", |
| 425 | "offset_bytes": 761484, |
| 426 | "nbytes": 192, |
| 427 | "sha256": "f03157a21079a20148f8efade1b790b7dac294dac899a36c4c0e4c267aa4435e" |
| 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": 761676, |
| 438 | "nbytes": 92160, |
| 439 | "sha256": "fce444c46c7fc7b087a6c8df29a5f097d0be595074c3d2cae29712a14c22563e" |
| 440 | }, |
| 441 | { |
| 442 | "name": "frame_blocks.2.scale", |
| 443 | "shape": [ |
| 444 | 1 |
| 445 | ], |
| 446 | "dtype": "float16", |
| 447 | "offset_bytes": 853836, |
| 448 | "nbytes": 2, |
| 449 | "sha256": "3e37d4b36203bc7aae8db24a6a8fccad35d20599e295d8c3654b8a9b344f73da" |
| 450 | }, |
| 451 | { |
| 452 | "name": "frame_blocks.3.net.0.bias", |
| 453 | "shape": [ |
| 454 | 96 |
| 455 | ], |
| 456 | "dtype": "float16", |
| 457 | "offset_bytes": 853838, |
| 458 | "nbytes": 192, |
| 459 | "sha256": "0b9bdd1105b3738f47179efcff8d5008864da82696426a246e33de5b9ea86e36" |
| 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": 854030, |
| 470 | "nbytes": 92160, |
| 471 | "sha256": "b760e41321d3660668d10278daa29d42a33bc1e67ce9ccdacf3b4764d19a5e14" |
| 472 | }, |
| 473 | { |
| 474 | "name": "frame_blocks.3.net.2.bias", |
| 475 | "shape": [ |
| 476 | 96 |
| 477 | ], |
| 478 | "dtype": "float16", |
| 479 | "offset_bytes": 946190, |
| 480 | "nbytes": 192, |
| 481 | "sha256": "3745ac86ea566cf804533f2f7d9334a2bdb6f075d5de7fb6c16532671f05f804" |
| 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": 946382, |
| 492 | "nbytes": 92160, |
| 493 | "sha256": "fcce8b71e1b89633378c19684beb04e918f1c508338dae570015d70bad9863c8" |
| 494 | }, |
| 495 | { |
| 496 | "name": "frame_blocks.3.scale", |
| 497 | "shape": [ |
| 498 | 1 |
| 499 | ], |
| 500 | "dtype": "float16", |
| 501 | "offset_bytes": 1038542, |
| 502 | "nbytes": 2, |
| 503 | "sha256": "dd71c3764f19ce33270861e90f525062de5d7b969b023777da03b8a1628fa389" |
| 504 | }, |
| 505 | { |
| 506 | "name": "frame_input_proj.bias", |
| 507 | "shape": [ |
| 508 | 96 |
| 509 | ], |
| 510 | "dtype": "float16", |
| 511 | "offset_bytes": 1038544, |
| 512 | "nbytes": 192, |
| 513 | "sha256": "ec3676f400fb1d86a89e087ebd151b94380cde7f34400fb45b4139290373434b" |
| 514 | }, |
| 515 | { |
| 516 | "name": "frame_input_proj.weight", |
| 517 | "shape": [ |
| 518 | 96, |
| 519 | 99, |
| 520 | 1 |
| 521 | ], |
| 522 | "dtype": "float16", |
| 523 | "offset_bytes": 1038736, |
| 524 | "nbytes": 19008, |
| 525 | "sha256": "a2abadb0edba4169cfefe1f877fb0b7ad989da75c4018df0833aaad0665f55d5" |
| 526 | }, |
| 527 | { |
| 528 | "name": "output.bias", |
| 529 | "shape": [ |
| 530 | 192 |
| 531 | ], |
| 532 | "dtype": "float16", |
| 533 | "offset_bytes": 1057744, |
| 534 | "nbytes": 384, |
| 535 | "sha256": "11f8a725fa1d2547245fe76fdd793bb39ea1ed8ce6662335af88d8b39d30e8e5" |
| 536 | }, |
| 537 | { |
| 538 | "name": "output.weight", |
| 539 | "shape": [ |
| 540 | 192, |
| 541 | 96, |
| 542 | 1 |
| 543 | ], |
| 544 | "dtype": "float16", |
| 545 | "offset_bytes": 1058128, |
| 546 | "nbytes": 36864, |
| 547 | "sha256": "273aa6b97f8b6be69e7789b3a96ed7f2fc87732b4155cd7b7862a3c0e75f1d1f" |
| 548 | }, |
| 549 | { |
| 550 | "name": "token_blocks.0.net.0.bias", |
| 551 | "shape": [ |
| 552 | 96 |
| 553 | ], |
| 554 | "dtype": "float16", |
| 555 | "offset_bytes": 1094992, |
| 556 | "nbytes": 192, |
| 557 | "sha256": "f7ee665f3efbb813d79734805608877911ee14149bbf3bf9bb3764fd7822f10a" |
| 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": 1095184, |
| 568 | "nbytes": 92160, |
| 569 | "sha256": "85f83d8a3f66cbe19a3ba8927aef672eb5bded036bcd4436e03c37f4babc57d4" |
| 570 | }, |
| 571 | { |
| 572 | "name": "token_blocks.0.net.2.bias", |
| 573 | "shape": [ |
| 574 | 96 |
| 575 | ], |
| 576 | "dtype": "float16", |
| 577 | "offset_bytes": 1187344, |
| 578 | "nbytes": 192, |
| 579 | "sha256": "6ab420ddd933e98b12cbe0b2d3cb1d9df146be85bf18ef98a5348b77ca52acc5" |
| 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": 1187536, |
| 590 | "nbytes": 92160, |
| 591 | "sha256": "7b1ce0d8d96a8f0ed8a95db77e9f6d0cb5a0043968b0b26eb6e8a9649b822e25" |
| 592 | }, |
| 593 | { |
| 594 | "name": "token_blocks.0.scale", |
| 595 | "shape": [ |
| 596 | 1 |
| 597 | ], |
| 598 | "dtype": "float16", |
| 599 | "offset_bytes": 1279696, |
| 600 | "nbytes": 2, |
| 601 | "sha256": "20747991457f50df38a30fbdac5b3c730cbb3e0d311f3365babdf8fec38cd23a" |
| 602 | }, |
| 603 | { |
| 604 | "name": "token_blocks.1.net.0.bias", |
| 605 | "shape": [ |
| 606 | 96 |
| 607 | ], |
| 608 | "dtype": "float16", |
| 609 | "offset_bytes": 1279698, |
| 610 | "nbytes": 192, |
| 611 | "sha256": "460a92181d854e24ae56c56e1a16c0e56d984ca5dea1c868fd34a05f52f76780" |
| 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": 1279890, |
| 622 | "nbytes": 92160, |
| 623 | "sha256": "5fe91763cf56bc4aadc0beb69720106a4cb9692b613afc4879211fd7766f168d" |
| 624 | }, |
| 625 | { |
| 626 | "name": "token_blocks.1.net.2.bias", |
| 627 | "shape": [ |
| 628 | 96 |
| 629 | ], |
| 630 | "dtype": "float16", |
| 631 | "offset_bytes": 1372050, |
| 632 | "nbytes": 192, |
| 633 | "sha256": "641d6f69631dbc3682989882d34b4225bc063a35a6ea8a1ac3cfc064d9b49f49" |
| 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": 1372242, |
| 644 | "nbytes": 92160, |
| 645 | "sha256": "f86c80035540e51a4f1d25a3b7993891aea15849da8b7816029b44e12e148f4c" |
| 646 | }, |
| 647 | { |
| 648 | "name": "token_blocks.1.scale", |
| 649 | "shape": [ |
| 650 | 1 |
| 651 | ], |
| 652 | "dtype": "float16", |
| 653 | "offset_bytes": 1464402, |
| 654 | "nbytes": 2, |
| 655 | "sha256": "9d9ddb12085922901cc30b377271aa4874b530bb3b6120afa8a378a7714badfc" |
| 656 | }, |
| 657 | { |
| 658 | "name": "token_blocks.2.net.0.bias", |
| 659 | "shape": [ |
| 660 | 96 |
| 661 | ], |
| 662 | "dtype": "float16", |
| 663 | "offset_bytes": 1464404, |
| 664 | "nbytes": 192, |
| 665 | "sha256": "b3a70823c0164d778348fe425b407b3f3ccdddb8b0db2d80ccb10fa6ec36a426" |
| 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": 1464596, |
| 676 | "nbytes": 92160, |
| 677 | "sha256": "2d81dcfbadfa282e14a2aab8477795ef097f4c41830fbb0f9bdf5e45d30f0a6e" |
| 678 | }, |
| 679 | { |
| 680 | "name": "token_blocks.2.net.2.bias", |
| 681 | "shape": [ |
| 682 | 96 |
| 683 | ], |
| 684 | "dtype": "float16", |
| 685 | "offset_bytes": 1556756, |
| 686 | "nbytes": 192, |
| 687 | "sha256": "82797aa6ff459c204e80d726cc5bc1ee754e4812c25573b5bfe02c94506041e7" |
| 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": 1556948, |
| 698 | "nbytes": 92160, |
| 699 | "sha256": "3d67578f16eab1ece2f6214b0725fefffb9264253cdb8377e2671243cc0319d0" |
| 700 | }, |
| 701 | { |
| 702 | "name": "token_blocks.2.scale", |
| 703 | "shape": [ |
| 704 | 1 |
| 705 | ], |
| 706 | "dtype": "float16", |
| 707 | "offset_bytes": 1649108, |
| 708 | "nbytes": 2, |
| 709 | "sha256": "913ff7834221bb7056374f12c0971fa66884577980ee4b4aae2215f32e46b354" |
| 710 | }, |
| 711 | { |
| 712 | "name": "token_input_proj.bias", |
| 713 | "shape": [ |
| 714 | 96 |
| 715 | ], |
| 716 | "dtype": "float16", |
| 717 | "offset_bytes": 1649110, |
| 718 | "nbytes": 192, |
| 719 | "sha256": "e2a3db4fe24f7f0d227725c41670e174f2e67fa5153e998cd7fbffb48b6122be" |
| 720 | }, |
| 721 | { |
| 722 | "name": "token_input_proj.weight", |
| 723 | "shape": [ |
| 724 | 96, |
| 725 | 98, |
| 726 | 1 |
| 727 | ], |
| 728 | "dtype": "float16", |
| 729 | "offset_bytes": 1649302, |
| 730 | "nbytes": 18816, |
| 731 | "sha256": "01b0c7ec6e8b907f6fd70de40ace79fdedcd13a0c4a1207fab68dc1a6f13346a" |
| 732 | } |
| 733 | ] |
| 734 | }, |
| 735 | "decoder": { |
| 736 | "checkpoint": "artifacts/sub10m-search/root-a-piper-vits/en_US-amy-medium-retrain-20260712/w3-wide192-joint/decoder-student.pt", |
| 737 | "checkpoint_sha256": "7ae32b637170996fb73ec16c7a9b319d3d2a1b8106ceb7f596ea6064db9f9f41", |
| 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": 1668118, |
| 908 | "nbytes": 2, |
| 909 | "sha256": "b428b8fc8c77c79f4ad5966585e1f543396cc6a5432057616e07fc9304f830ea" |
| 910 | }, |
| 911 | { |
| 912 | "name": "post.weight", |
| 913 | "shape": [ |
| 914 | 1, |
| 915 | 24, |
| 916 | 7 |
| 917 | ], |
| 918 | "dtype": "float16", |
| 919 | "offset_bytes": 1668120, |
| 920 | "nbytes": 336, |
| 921 | "sha256": "cc4bd0e71a6d786e4ccec40e09a56e9e460c002feacdeb4c4a303d9dd51b2555" |
| 922 | }, |
| 923 | { |
| 924 | "name": "pre.bias", |
| 925 | "shape": [ |
| 926 | 192 |
| 927 | ], |
| 928 | "dtype": "float16", |
| 929 | "offset_bytes": 1668456, |
| 930 | "nbytes": 384, |
| 931 | "sha256": "b9f3ce8837fc92b13153d9cdddfd551a017fdb03339ca78162cbd7bae2146db5" |
| 932 | }, |
| 933 | { |
| 934 | "name": "pre.weight", |
| 935 | "shape": [ |
| 936 | 192, |
| 937 | 192, |
| 938 | 7 |
| 939 | ], |
| 940 | "dtype": "float16", |
| 941 | "offset_bytes": 1668840, |
| 942 | "nbytes": 516096, |
| 943 | "sha256": "c13d479ffa828ff79a03a3eb45c33a41c4b5645cab4bf3034e523e4b7f2701f0" |
| 944 | }, |
| 945 | { |
| 946 | "name": "res0.0.blocks.0.conv1.bias", |
| 947 | "shape": [ |
| 948 | 96 |
| 949 | ], |
| 950 | "dtype": "float16", |
| 951 | "offset_bytes": 2184936, |
| 952 | "nbytes": 192, |
| 953 | "sha256": "3e985ec0a740ccfe94bb32c956bc3259a4759a610ad894f8a7ad1e8ea6ce1566" |
| 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": 2185128, |
| 964 | "nbytes": 55296, |
| 965 | "sha256": "3e6e159b18bf11307ce7ec3782662b8425a6e2d1654bf0b68f8540ff47b687e8" |
| 966 | }, |
| 967 | { |
| 968 | "name": "res0.0.blocks.0.conv2.bias", |
| 969 | "shape": [ |
| 970 | 96 |
| 971 | ], |
| 972 | "dtype": "float16", |
| 973 | "offset_bytes": 2240424, |
| 974 | "nbytes": 192, |
| 975 | "sha256": "3f0fe001a57b7a441432ee0c10291f6c0913f949e3c905b60c60d876c4fbb719" |
| 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": 2240616, |
| 986 | "nbytes": 55296, |
| 987 | "sha256": "b43e4e3b2c6b982e4817eec2eb6a73e8586b1f5f6786681780d0983afbeb6bd9" |
| 988 | }, |
| 989 | { |
| 990 | "name": "res0.0.blocks.1.conv1.bias", |
| 991 | "shape": [ |
| 992 | 96 |
| 993 | ], |
| 994 | "dtype": "float16", |
| 995 | "offset_bytes": 2295912, |
| 996 | "nbytes": 192, |
| 997 | "sha256": "e6802d769fad666b2a25666bb2bada248903d6bd13e4fd222bf1d864b8a7b78c" |
| 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": 2296104, |
| 1008 | "nbytes": 92160, |
| 1009 | "sha256": "a1f0f7b72cb6eca83d63bea9121270a001bef2641608037e0ca5a1831f699563" |
| 1010 | }, |
| 1011 | { |
| 1012 | "name": "res0.0.blocks.1.conv2.bias", |
| 1013 | "shape": [ |
| 1014 | 96 |
| 1015 | ], |
| 1016 | "dtype": "float16", |
| 1017 | "offset_bytes": 2388264, |
| 1018 | "nbytes": 192, |
| 1019 | "sha256": "d6adc9dae02e91ca48303f176106a1fddb02031690b8a5aa21bea67818eb3394" |
| 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": 2388456, |
| 1030 | "nbytes": 92160, |
| 1031 | "sha256": "ee93c8153f0d4e3895578b4562ded08e919b1c2e5e389baa8ee9f8a772162331" |
| 1032 | }, |
| 1033 | { |
| 1034 | "name": "res0.0.blocks.2.conv1.bias", |
| 1035 | "shape": [ |
| 1036 | 96 |
| 1037 | ], |
| 1038 | "dtype": "float16", |
| 1039 | "offset_bytes": 2480616, |
| 1040 | "nbytes": 192, |
| 1041 | "sha256": "67c7d43e6608fb63947d50a7d09f89ec8d55d8b7124e607e820287f12fc83b4d" |
| 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": 2480808, |
| 1052 | "nbytes": 129024, |
| 1053 | "sha256": "fb548de036af0a53802e2d09bf6c3e6433baed55f97f201d453f7248efc8238e" |
| 1054 | }, |
| 1055 | { |
| 1056 | "name": "res0.0.blocks.2.conv2.bias", |
| 1057 | "shape": [ |
| 1058 | 96 |
| 1059 | ], |
| 1060 | "dtype": "float16", |
| 1061 | "offset_bytes": 2609832, |
| 1062 | "nbytes": 192, |
| 1063 | "sha256": "53a48693c2c8c423a95f5f7dbd302512df58b1eb7ace4ec5c8de9f980150297a" |
| 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": 2610024, |
| 1074 | "nbytes": 129024, |
| 1075 | "sha256": "ae860c8b6f54063653d2bae406fbb8a2cd9e606b384ebf8003bce41b2a8e14f3" |
| 1076 | }, |
| 1077 | { |
| 1078 | "name": "res1.0.blocks.0.conv1.bias", |
| 1079 | "shape": [ |
| 1080 | 48 |
| 1081 | ], |
| 1082 | "dtype": "float16", |
| 1083 | "offset_bytes": 2739048, |
| 1084 | "nbytes": 96, |
| 1085 | "sha256": "cdca7d7367eace48d3af51cf0b4845e44c53fc171e00d9b0cf8b5e8797399687" |
| 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": 2739144, |
| 1096 | "nbytes": 13824, |
| 1097 | "sha256": "d62afe225f80cd5ef452890c55b694a9089c495b6ea2a6eabe3d1f13cc9a5c92" |
| 1098 | }, |
| 1099 | { |
| 1100 | "name": "res1.0.blocks.0.conv2.bias", |
| 1101 | "shape": [ |
| 1102 | 48 |
| 1103 | ], |
| 1104 | "dtype": "float16", |
| 1105 | "offset_bytes": 2752968, |
| 1106 | "nbytes": 96, |
| 1107 | "sha256": "c4654264ec3d07a092b15888e978990c8c11f002a75612122b1b3d13833721d1" |
| 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": 2753064, |
| 1118 | "nbytes": 13824, |
| 1119 | "sha256": "b494d043be209a1fecc008f05a65a66974a37659e0d4c217ccb9056d11a3090f" |
| 1120 | }, |
| 1121 | { |
| 1122 | "name": "res1.0.blocks.1.conv1.bias", |
| 1123 | "shape": [ |
| 1124 | 48 |
| 1125 | ], |
| 1126 | "dtype": "float16", |
| 1127 | "offset_bytes": 2766888, |
| 1128 | "nbytes": 96, |
| 1129 | "sha256": "41e463eca93590a0023e76c0c991b705b315a50c15258c2af53eb03b5eb4d5cf" |
| 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": 2766984, |
| 1140 | "nbytes": 23040, |
| 1141 | "sha256": "ec711a10429274a25f22335c0a0a83ef6274693850438936ad9d5442fda5fe65" |
| 1142 | }, |
| 1143 | { |
| 1144 | "name": "res1.0.blocks.1.conv2.bias", |
| 1145 | "shape": [ |
| 1146 | 48 |
| 1147 | ], |
| 1148 | "dtype": "float16", |
| 1149 | "offset_bytes": 2790024, |
| 1150 | "nbytes": 96, |
| 1151 | "sha256": "67545836c28654996a611ce6fc2f74d2aafed4ce04aa13a425f6f855dbb1c73e" |
| 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": 2790120, |
| 1162 | "nbytes": 23040, |
| 1163 | "sha256": "64193a166a08ace8d5121fb68a9dadb05eb9a2a2472c2cd23f3bf975d0398409" |
| 1164 | }, |
| 1165 | { |
| 1166 | "name": "res1.0.blocks.2.conv1.bias", |
| 1167 | "shape": [ |
| 1168 | 48 |
| 1169 | ], |
| 1170 | "dtype": "float16", |
| 1171 | "offset_bytes": 2813160, |
| 1172 | "nbytes": 96, |
| 1173 | "sha256": "6ac3f7edcca93f9b899ba7def2cb279c01f608b61a791239901402fc2bf6559d" |
| 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": 2813256, |
| 1184 | "nbytes": 32256, |
| 1185 | "sha256": "8aa60110eacc8365559e6cd227f8017d9f26bef73fa3f51c1d50a04b936674f8" |
| 1186 | }, |
| 1187 | { |
| 1188 | "name": "res1.0.blocks.2.conv2.bias", |
| 1189 | "shape": [ |
| 1190 | 48 |
| 1191 | ], |
| 1192 | "dtype": "float16", |
| 1193 | "offset_bytes": 2845512, |
| 1194 | "nbytes": 96, |
| 1195 | "sha256": "938ed03944f463cd0e5c6ca8b555136f0303f7c272807103c1309c26f9b95cf9" |
| 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": 2845608, |
| 1206 | "nbytes": 32256, |
| 1207 | "sha256": "d02e1c9d4b086cf8dacfb28f35b5562595810733ad5f5df28b97c3547f82264e" |
| 1208 | }, |
| 1209 | { |
| 1210 | "name": "res2.0.blocks.0.conv1.bias", |
| 1211 | "shape": [ |
| 1212 | 24 |
| 1213 | ], |
| 1214 | "dtype": "float16", |
| 1215 | "offset_bytes": 2877864, |
| 1216 | "nbytes": 48, |
| 1217 | "sha256": "f6e9ddf48e2b7086b21df0b2b86901cc2b22516a606aa3122531b6424a829538" |
| 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": 2877912, |
| 1228 | "nbytes": 3456, |
| 1229 | "sha256": "f374904c49225ad600be9b92b7922c20f729e488760bf4f074ea0b5929df8005" |
| 1230 | }, |
| 1231 | { |
| 1232 | "name": "res2.0.blocks.0.conv2.bias", |
| 1233 | "shape": [ |
| 1234 | 24 |
| 1235 | ], |
| 1236 | "dtype": "float16", |
| 1237 | "offset_bytes": 2881368, |
| 1238 | "nbytes": 48, |
| 1239 | "sha256": "bc259fd84aabeb698c10cb8d759820e75f3edd6be3b1197e8665d7c766c5415f" |
| 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": 2881416, |
| 1250 | "nbytes": 3456, |
| 1251 | "sha256": "797c214f890a7003ec2a577724fb207dc00909c3dde1274b0f48b725cdfe9f8b" |
| 1252 | }, |
| 1253 | { |
| 1254 | "name": "res2.0.blocks.1.conv1.bias", |
| 1255 | "shape": [ |
| 1256 | 24 |
| 1257 | ], |
| 1258 | "dtype": "float16", |
| 1259 | "offset_bytes": 2884872, |
| 1260 | "nbytes": 48, |
| 1261 | "sha256": "370ed210690aca9fb394e2285533d96639dd4dcb8af494b32cc95a156c301a8b" |
| 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": 2884920, |
| 1272 | "nbytes": 5760, |
| 1273 | "sha256": "8a81002998431b60bf18ebc0332763e1c7b58b74fbd4ca23190dd145484da24c" |
| 1274 | }, |
| 1275 | { |
| 1276 | "name": "res2.0.blocks.1.conv2.bias", |
| 1277 | "shape": [ |
| 1278 | 24 |
| 1279 | ], |
| 1280 | "dtype": "float16", |
| 1281 | "offset_bytes": 2890680, |
| 1282 | "nbytes": 48, |
| 1283 | "sha256": "87dee9b73256228cfec1669c950243ef6f976d689c275998ce98da210efb4f4f" |
| 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": 2890728, |
| 1294 | "nbytes": 5760, |
| 1295 | "sha256": "70f821765a00ffc647910908422ec87af0980a6aa3079eaf2503bad26f6277e6" |
| 1296 | }, |
| 1297 | { |
| 1298 | "name": "res2.0.blocks.2.conv1.bias", |
| 1299 | "shape": [ |
| 1300 | 24 |
| 1301 | ], |
| 1302 | "dtype": "float16", |
| 1303 | "offset_bytes": 2896488, |
| 1304 | "nbytes": 48, |
| 1305 | "sha256": "27df0f4bc674c080ae3025d6c1df45ae626bb67d8226e95cc48d402fe4716be0" |
| 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": 2896536, |
| 1316 | "nbytes": 8064, |
| 1317 | "sha256": "4a544c28dbca960bc949bb9df0d656e4eaac38454d28a629e58edfe6e7f42437" |
| 1318 | }, |
| 1319 | { |
| 1320 | "name": "res2.0.blocks.2.conv2.bias", |
| 1321 | "shape": [ |
| 1322 | 24 |
| 1323 | ], |
| 1324 | "dtype": "float16", |
| 1325 | "offset_bytes": 2904600, |
| 1326 | "nbytes": 48, |
| 1327 | "sha256": "09ccbb9f40aedc46a27cc30d396150a23fc35f587f09d8859cfb49e92706410c" |
| 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": 2904648, |
| 1338 | "nbytes": 8064, |
| 1339 | "sha256": "513b00e273d3c75e7b520720783834298994407d104eee3207c6ac803c1b4da6" |
| 1340 | }, |
| 1341 | { |
| 1342 | "name": "up0.bias", |
| 1343 | "shape": [ |
| 1344 | 96 |
| 1345 | ], |
| 1346 | "dtype": "float16", |
| 1347 | "offset_bytes": 2912712, |
| 1348 | "nbytes": 192, |
| 1349 | "sha256": "ba934062cd2796a6be8f4b9a6adb375b9b5021d8349c273f78c0bb4c864c44c0" |
| 1350 | }, |
| 1351 | { |
| 1352 | "name": "up0.weight", |
| 1353 | "shape": [ |
| 1354 | 192, |
| 1355 | 96, |
| 1356 | 16 |
| 1357 | ], |
| 1358 | "dtype": "float16", |
| 1359 | "offset_bytes": 2912904, |
| 1360 | "nbytes": 589824, |
| 1361 | "sha256": "a39f13f9011ddd1101b4115cabd71070b23b73e124f9e1da052a36a990488948" |
| 1362 | }, |
| 1363 | { |
| 1364 | "name": "up1.bias", |
| 1365 | "shape": [ |
| 1366 | 48 |
| 1367 | ], |
| 1368 | "dtype": "float16", |
| 1369 | "offset_bytes": 3502728, |
| 1370 | "nbytes": 96, |
| 1371 | "sha256": "3ed6cc941e9d226532fd50cfc1caeb7618a7a1d83078e84f6f8d8118839248a3" |
| 1372 | }, |
| 1373 | { |
| 1374 | "name": "up1.weight", |
| 1375 | "shape": [ |
| 1376 | 96, |
| 1377 | 48, |
| 1378 | 16 |
| 1379 | ], |
| 1380 | "dtype": "float16", |
| 1381 | "offset_bytes": 3502824, |
| 1382 | "nbytes": 147456, |
| 1383 | "sha256": "e79563ef81572af67f41ea365cbb40ef7f9a4331d2d9b27020f39ea0c9459744" |
| 1384 | }, |
| 1385 | { |
| 1386 | "name": "up2.bias", |
| 1387 | "shape": [ |
| 1388 | 24 |
| 1389 | ], |
| 1390 | "dtype": "float16", |
| 1391 | "offset_bytes": 3650280, |
| 1392 | "nbytes": 48, |
| 1393 | "sha256": "94171e0469180b290fba94cb9add373c760e59066fa754b36793da25468c15a5" |
| 1394 | }, |
| 1395 | { |
| 1396 | "name": "up2.weight", |
| 1397 | "shape": [ |
| 1398 | 48, |
| 1399 | 24, |
| 1400 | 8 |
| 1401 | ], |
| 1402 | "dtype": "float16", |
| 1403 | "offset_bytes": 3650328, |
| 1404 | "nbytes": 18432, |
| 1405 | "sha256": "d51dc128395b19745bab36eb93b9f54d04de364b3d1ebbc3a60793c943b72937" |
| 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": 3673642, |
| 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 | |