de-thorsten-0p51m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "de_DE-flores-20260908", |
| 4 | "language": "de_DE", |
| 5 | "voice": "de_DE-thorsten-medium", |
| 6 | "created_from": "tools/export_roota_self_contained_package.py", |
| 7 | "sample_rate": 22050, |
| 8 | "hop_length": 256, |
| 9 | "inference": { |
| 10 | "duration_length_scale": 1.0, |
| 11 | "duration_policy": "round(exp(log_duration) * duration_length_scale)" |
| 12 | }, |
| 13 | "weights_file": "weights.fp16.bin", |
| 14 | "weights_size_bytes": 1023222, |
| 15 | "weights_sha256": "c1d58bca79bc34e5f24c9652eb8c984d9eaf01b22c86868d25291eba9c871b2b", |
| 16 | "total_parameters": 511611, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/de_DE-tiny/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "45c9353a1ed174eff04589f9e2ef57e574cf78303ba4df54f511d61fea359585", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 142, |
| 24 | "hidden": 32, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 759, |
| 28 | "max_duration": 80, |
| 29 | "target_duration_scale": 1.0, |
| 30 | "pause_preserve_scale": 0.0, |
| 31 | "pause_preserve_window": 0, |
| 32 | "pause_token_ids": [ |
| 33 | 4, |
| 34 | 8, |
| 35 | 10, |
| 36 | 11, |
| 37 | 12, |
| 38 | 13 |
| 39 | ], |
| 40 | "long_preserve_threshold": 0.0, |
| 41 | "long_preserve_scale": 0.0, |
| 42 | "duration_vocab_size_forced": 0, |
| 43 | "duration_oov_id": 59 |
| 44 | }, |
| 45 | "parameters": 36644, |
| 46 | "tensors": [ |
| 47 | { |
| 48 | "name": "blocks.0.net.0.bias", |
| 49 | "shape": [ |
| 50 | 32 |
| 51 | ], |
| 52 | "dtype": "float16", |
| 53 | "offset_bytes": 0, |
| 54 | "nbytes": 64, |
| 55 | "sha256": "36b5b8a4fd6ffeb2243e5d9b793d6fd1b0971cea981e2f0517b7e10b9d19af1e" |
| 56 | }, |
| 57 | { |
| 58 | "name": "blocks.0.net.0.weight", |
| 59 | "shape": [ |
| 60 | 32, |
| 61 | 32, |
| 62 | 5 |
| 63 | ], |
| 64 | "dtype": "float16", |
| 65 | "offset_bytes": 64, |
| 66 | "nbytes": 10240, |
| 67 | "sha256": "9e2581b46f88f75e358fb91302c10adec3459011a21e5ba87800f1ce5d236dee" |
| 68 | }, |
| 69 | { |
| 70 | "name": "blocks.0.net.2.bias", |
| 71 | "shape": [ |
| 72 | 32 |
| 73 | ], |
| 74 | "dtype": "float16", |
| 75 | "offset_bytes": 10304, |
| 76 | "nbytes": 64, |
| 77 | "sha256": "0b8110c32482441226212bb7fdc57b1c10f76730ca00dadf78c73edad90ce74b" |
| 78 | }, |
| 79 | { |
| 80 | "name": "blocks.0.net.2.weight", |
| 81 | "shape": [ |
| 82 | 32, |
| 83 | 32, |
| 84 | 5 |
| 85 | ], |
| 86 | "dtype": "float16", |
| 87 | "offset_bytes": 10368, |
| 88 | "nbytes": 10240, |
| 89 | "sha256": "94a1998866c4825dafbd71d632b7809bf2a759285e073239579f98979ec2dda3" |
| 90 | }, |
| 91 | { |
| 92 | "name": "blocks.0.scale", |
| 93 | "shape": [ |
| 94 | 1 |
| 95 | ], |
| 96 | "dtype": "float16", |
| 97 | "offset_bytes": 20608, |
| 98 | "nbytes": 2, |
| 99 | "sha256": "195f58bc6d6b7b36335c95e08343825a7ae6f30437b4a7e6fa7b89d76907570a" |
| 100 | }, |
| 101 | { |
| 102 | "name": "blocks.1.net.0.bias", |
| 103 | "shape": [ |
| 104 | 32 |
| 105 | ], |
| 106 | "dtype": "float16", |
| 107 | "offset_bytes": 20610, |
| 108 | "nbytes": 64, |
| 109 | "sha256": "a888599f9f575f09aee92b7f4b7deb0619d36fcc07728093094aa51de55be1d4" |
| 110 | }, |
| 111 | { |
| 112 | "name": "blocks.1.net.0.weight", |
| 113 | "shape": [ |
| 114 | 32, |
| 115 | 32, |
| 116 | 5 |
| 117 | ], |
| 118 | "dtype": "float16", |
| 119 | "offset_bytes": 20674, |
| 120 | "nbytes": 10240, |
| 121 | "sha256": "f01e0b02b3f148423302ac7239b4834a20a8f8ba4b84200b3157c09476ecf615" |
| 122 | }, |
| 123 | { |
| 124 | "name": "blocks.1.net.2.bias", |
| 125 | "shape": [ |
| 126 | 32 |
| 127 | ], |
| 128 | "dtype": "float16", |
| 129 | "offset_bytes": 30914, |
| 130 | "nbytes": 64, |
| 131 | "sha256": "41a94422b8a876aac9b86274b3118ff12e2f489d7a2accf46a3b87a019422d9e" |
| 132 | }, |
| 133 | { |
| 134 | "name": "blocks.1.net.2.weight", |
| 135 | "shape": [ |
| 136 | 32, |
| 137 | 32, |
| 138 | 5 |
| 139 | ], |
| 140 | "dtype": "float16", |
| 141 | "offset_bytes": 30978, |
| 142 | "nbytes": 10240, |
| 143 | "sha256": "70d3aa33a902ac9eb4a0a32ff9e409813478e7178efec614ec7cc7661c362e91" |
| 144 | }, |
| 145 | { |
| 146 | "name": "blocks.1.scale", |
| 147 | "shape": [ |
| 148 | 1 |
| 149 | ], |
| 150 | "dtype": "float16", |
| 151 | "offset_bytes": 41218, |
| 152 | "nbytes": 2, |
| 153 | "sha256": "cd68af92df5e5341dbf259860c321ef5d8aed0f730540dc8947cd7d862be3838" |
| 154 | }, |
| 155 | { |
| 156 | "name": "blocks.2.net.0.bias", |
| 157 | "shape": [ |
| 158 | 32 |
| 159 | ], |
| 160 | "dtype": "float16", |
| 161 | "offset_bytes": 41220, |
| 162 | "nbytes": 64, |
| 163 | "sha256": "b3bf8ca3be3359454ffbe5c3c25800d51c18ff6bf40b5dd5518601e92033353f" |
| 164 | }, |
| 165 | { |
| 166 | "name": "blocks.2.net.0.weight", |
| 167 | "shape": [ |
| 168 | 32, |
| 169 | 32, |
| 170 | 5 |
| 171 | ], |
| 172 | "dtype": "float16", |
| 173 | "offset_bytes": 41284, |
| 174 | "nbytes": 10240, |
| 175 | "sha256": "598b18481c70f06d000c4c8afcb4a9c6f40b3c917e760bf795ff9e20df575a03" |
| 176 | }, |
| 177 | { |
| 178 | "name": "blocks.2.net.2.bias", |
| 179 | "shape": [ |
| 180 | 32 |
| 181 | ], |
| 182 | "dtype": "float16", |
| 183 | "offset_bytes": 51524, |
| 184 | "nbytes": 64, |
| 185 | "sha256": "fe362c0d9fe4a03b8fe1eecb9ce3bd5c8be3c0f9034245ee871c7ca7581c8dcd" |
| 186 | }, |
| 187 | { |
| 188 | "name": "blocks.2.net.2.weight", |
| 189 | "shape": [ |
| 190 | 32, |
| 191 | 32, |
| 192 | 5 |
| 193 | ], |
| 194 | "dtype": "float16", |
| 195 | "offset_bytes": 51588, |
| 196 | "nbytes": 10240, |
| 197 | "sha256": "9aa8c06aa3a4350e613f96e7634a082325fc9b8c726749b38f942f2b7e8f8083" |
| 198 | }, |
| 199 | { |
| 200 | "name": "blocks.2.scale", |
| 201 | "shape": [ |
| 202 | 1 |
| 203 | ], |
| 204 | "dtype": "float16", |
| 205 | "offset_bytes": 61828, |
| 206 | "nbytes": 2, |
| 207 | "sha256": "7c780e8d1180b5c258bdc08506855bca08c111a40cb4b38313b153a712504948" |
| 208 | }, |
| 209 | { |
| 210 | "name": "embedding.weight", |
| 211 | "shape": [ |
| 212 | 142, |
| 213 | 32 |
| 214 | ], |
| 215 | "dtype": "float16", |
| 216 | "offset_bytes": 61830, |
| 217 | "nbytes": 9088, |
| 218 | "sha256": "987010766fc56029aa6b1c10105b8b03d33ec7293810a6fa18e169c18d2f1f1a" |
| 219 | }, |
| 220 | { |
| 221 | "name": "input_proj.bias", |
| 222 | "shape": [ |
| 223 | 32 |
| 224 | ], |
| 225 | "dtype": "float16", |
| 226 | "offset_bytes": 70918, |
| 227 | "nbytes": 64, |
| 228 | "sha256": "0d30223ffc491ae02450919be65ae3f36527240ab9343d2f9a31f322b61d1f32" |
| 229 | }, |
| 230 | { |
| 231 | "name": "input_proj.weight", |
| 232 | "shape": [ |
| 233 | 32, |
| 234 | 35, |
| 235 | 1 |
| 236 | ], |
| 237 | "dtype": "float16", |
| 238 | "offset_bytes": 70982, |
| 239 | "nbytes": 2240, |
| 240 | "sha256": "72fae90d8895e15980ae418697f70b751eb5ada00cb6c35d95a0fb173e41e4ff" |
| 241 | }, |
| 242 | { |
| 243 | "name": "output.bias", |
| 244 | "shape": [ |
| 245 | 1 |
| 246 | ], |
| 247 | "dtype": "float16", |
| 248 | "offset_bytes": 73222, |
| 249 | "nbytes": 2, |
| 250 | "sha256": "4138c6ae42992f3ca53610b0c858e0f5a5a8d7e3424a81feb73b89f0f1877ba9" |
| 251 | }, |
| 252 | { |
| 253 | "name": "output.weight", |
| 254 | "shape": [ |
| 255 | 1, |
| 256 | 32, |
| 257 | 1 |
| 258 | ], |
| 259 | "dtype": "float16", |
| 260 | "offset_bytes": 73224, |
| 261 | "nbytes": 64, |
| 262 | "sha256": "114a0f2b87f181424e4397c752609f38e947e9d1908cc6eaa2aa406c74b3f906" |
| 263 | } |
| 264 | ] |
| 265 | }, |
| 266 | "acoustic": { |
| 267 | "checkpoint": "artifacts/voices/de_DE-tiny/joint/latent-student.pt", |
| 268 | "checkpoint_sha256": "cb8ba4f54dde827961ba48b737df69e56905cfb320250e6f67696fed6615503d", |
| 269 | "config": { |
| 270 | "architecture": "token_context", |
| 271 | "vocab_size": 142, |
| 272 | "hidden": 48, |
| 273 | "depth": 3, |
| 274 | "kernel_size": 5, |
| 275 | "out_channels": 192, |
| 276 | "token_depth": 3 |
| 277 | }, |
| 278 | "parameters": 159990, |
| 279 | "tensors": [ |
| 280 | { |
| 281 | "name": "embedding.weight", |
| 282 | "shape": [ |
| 283 | 142, |
| 284 | 48 |
| 285 | ], |
| 286 | "dtype": "float16", |
| 287 | "offset_bytes": 73288, |
| 288 | "nbytes": 13632, |
| 289 | "sha256": "d785988b1513d7daa18f4291da348e692ff9fcbda56bac7974b33c73f08d1ec5" |
| 290 | }, |
| 291 | { |
| 292 | "name": "frame_blocks.0.net.0.bias", |
| 293 | "shape": [ |
| 294 | 48 |
| 295 | ], |
| 296 | "dtype": "float16", |
| 297 | "offset_bytes": 86920, |
| 298 | "nbytes": 96, |
| 299 | "sha256": "fe310ede189c7d7f04fa5a9d6cd649ba0a8ee2958f757c38e4782ad566fcebde" |
| 300 | }, |
| 301 | { |
| 302 | "name": "frame_blocks.0.net.0.weight", |
| 303 | "shape": [ |
| 304 | 48, |
| 305 | 48, |
| 306 | 5 |
| 307 | ], |
| 308 | "dtype": "float16", |
| 309 | "offset_bytes": 87016, |
| 310 | "nbytes": 23040, |
| 311 | "sha256": "6ea9f41d05c32ae17b24a271f7c0fbbc28786430c457ac85e7f83a09071a48ae" |
| 312 | }, |
| 313 | { |
| 314 | "name": "frame_blocks.0.net.2.bias", |
| 315 | "shape": [ |
| 316 | 48 |
| 317 | ], |
| 318 | "dtype": "float16", |
| 319 | "offset_bytes": 110056, |
| 320 | "nbytes": 96, |
| 321 | "sha256": "1312bfc1de81247a25c59ffef372f0f571b690907169d19cb09b75d726cd7bd4" |
| 322 | }, |
| 323 | { |
| 324 | "name": "frame_blocks.0.net.2.weight", |
| 325 | "shape": [ |
| 326 | 48, |
| 327 | 48, |
| 328 | 5 |
| 329 | ], |
| 330 | "dtype": "float16", |
| 331 | "offset_bytes": 110152, |
| 332 | "nbytes": 23040, |
| 333 | "sha256": "2e45d5673eaeba8631fb4902939581ff46eb928f4725f98736911e277f223de6" |
| 334 | }, |
| 335 | { |
| 336 | "name": "frame_blocks.0.scale", |
| 337 | "shape": [ |
| 338 | 1 |
| 339 | ], |
| 340 | "dtype": "float16", |
| 341 | "offset_bytes": 133192, |
| 342 | "nbytes": 2, |
| 343 | "sha256": "aa479a8b11e401613c3f4bfc6982ed5f444f9d4019a3bd6e58dd7d5078da1a7c" |
| 344 | }, |
| 345 | { |
| 346 | "name": "frame_blocks.1.net.0.bias", |
| 347 | "shape": [ |
| 348 | 48 |
| 349 | ], |
| 350 | "dtype": "float16", |
| 351 | "offset_bytes": 133194, |
| 352 | "nbytes": 96, |
| 353 | "sha256": "b29444329eb04b2fc04888b4aef5ed8c4f50ab8020ee4f1473ab78795ce315ef" |
| 354 | }, |
| 355 | { |
| 356 | "name": "frame_blocks.1.net.0.weight", |
| 357 | "shape": [ |
| 358 | 48, |
| 359 | 48, |
| 360 | 5 |
| 361 | ], |
| 362 | "dtype": "float16", |
| 363 | "offset_bytes": 133290, |
| 364 | "nbytes": 23040, |
| 365 | "sha256": "45a153fba9e2285c17e5a316569c2dcc01fccb7de191235a0c2a439527b7bfd7" |
| 366 | }, |
| 367 | { |
| 368 | "name": "frame_blocks.1.net.2.bias", |
| 369 | "shape": [ |
| 370 | 48 |
| 371 | ], |
| 372 | "dtype": "float16", |
| 373 | "offset_bytes": 156330, |
| 374 | "nbytes": 96, |
| 375 | "sha256": "46f0179eda0d3ff74345f1952a5bdad6f3457bbd4fc52e968a89862756dbe97a" |
| 376 | }, |
| 377 | { |
| 378 | "name": "frame_blocks.1.net.2.weight", |
| 379 | "shape": [ |
| 380 | 48, |
| 381 | 48, |
| 382 | 5 |
| 383 | ], |
| 384 | "dtype": "float16", |
| 385 | "offset_bytes": 156426, |
| 386 | "nbytes": 23040, |
| 387 | "sha256": "a9eba82214f830f7281cd9e95f1df7269b346370525a1c06a3835436d7f9ac4f" |
| 388 | }, |
| 389 | { |
| 390 | "name": "frame_blocks.1.scale", |
| 391 | "shape": [ |
| 392 | 1 |
| 393 | ], |
| 394 | "dtype": "float16", |
| 395 | "offset_bytes": 179466, |
| 396 | "nbytes": 2, |
| 397 | "sha256": "bf37c3988f6d7c368fa7111364f5b65ca037fac7db1c4b8fd730bf58265083ea" |
| 398 | }, |
| 399 | { |
| 400 | "name": "frame_blocks.2.net.0.bias", |
| 401 | "shape": [ |
| 402 | 48 |
| 403 | ], |
| 404 | "dtype": "float16", |
| 405 | "offset_bytes": 179468, |
| 406 | "nbytes": 96, |
| 407 | "sha256": "5f8887ba7202f906b47cbba86b5762642cd1f59fa621509b03cb0a4eca488e60" |
| 408 | }, |
| 409 | { |
| 410 | "name": "frame_blocks.2.net.0.weight", |
| 411 | "shape": [ |
| 412 | 48, |
| 413 | 48, |
| 414 | 5 |
| 415 | ], |
| 416 | "dtype": "float16", |
| 417 | "offset_bytes": 179564, |
| 418 | "nbytes": 23040, |
| 419 | "sha256": "cb26db022dc7319036c54bf23057f38c94106dd46950e09ec5cf2934cb137cd9" |
| 420 | }, |
| 421 | { |
| 422 | "name": "frame_blocks.2.net.2.bias", |
| 423 | "shape": [ |
| 424 | 48 |
| 425 | ], |
| 426 | "dtype": "float16", |
| 427 | "offset_bytes": 202604, |
| 428 | "nbytes": 96, |
| 429 | "sha256": "76308b506c0bdd071152170fc0f880b3ba957c886e1ae616a7db1554f90c9778" |
| 430 | }, |
| 431 | { |
| 432 | "name": "frame_blocks.2.net.2.weight", |
| 433 | "shape": [ |
| 434 | 48, |
| 435 | 48, |
| 436 | 5 |
| 437 | ], |
| 438 | "dtype": "float16", |
| 439 | "offset_bytes": 202700, |
| 440 | "nbytes": 23040, |
| 441 | "sha256": "9ba05fd2fe346e51bfbe5acd19c795da0e060a6f07e9701cde243808737b8df1" |
| 442 | }, |
| 443 | { |
| 444 | "name": "frame_blocks.2.scale", |
| 445 | "shape": [ |
| 446 | 1 |
| 447 | ], |
| 448 | "dtype": "float16", |
| 449 | "offset_bytes": 225740, |
| 450 | "nbytes": 2, |
| 451 | "sha256": "d867dfd39cdf8fd12daa69604d09460eb337f895d98e2e3aa136ab2bcbfdbe32" |
| 452 | }, |
| 453 | { |
| 454 | "name": "frame_input_proj.bias", |
| 455 | "shape": [ |
| 456 | 48 |
| 457 | ], |
| 458 | "dtype": "float16", |
| 459 | "offset_bytes": 225742, |
| 460 | "nbytes": 96, |
| 461 | "sha256": "3508383f928d4a285c7742f3b92b057b499edbb804cd35912d1c3f38232800de" |
| 462 | }, |
| 463 | { |
| 464 | "name": "frame_input_proj.weight", |
| 465 | "shape": [ |
| 466 | 48, |
| 467 | 51, |
| 468 | 1 |
| 469 | ], |
| 470 | "dtype": "float16", |
| 471 | "offset_bytes": 225838, |
| 472 | "nbytes": 4896, |
| 473 | "sha256": "4eb5ceda2a9a7d21da7f239d0bdd202b67e010dbbd5b551aca2ded484a6f3e79" |
| 474 | }, |
| 475 | { |
| 476 | "name": "output.bias", |
| 477 | "shape": [ |
| 478 | 192 |
| 479 | ], |
| 480 | "dtype": "float16", |
| 481 | "offset_bytes": 230734, |
| 482 | "nbytes": 384, |
| 483 | "sha256": "ee94c4d4f5d139a780b0784f4fd99b91a093f5a45467fa2a61bf62f15146d082" |
| 484 | }, |
| 485 | { |
| 486 | "name": "output.weight", |
| 487 | "shape": [ |
| 488 | 192, |
| 489 | 48, |
| 490 | 1 |
| 491 | ], |
| 492 | "dtype": "float16", |
| 493 | "offset_bytes": 231118, |
| 494 | "nbytes": 18432, |
| 495 | "sha256": "0a3c6ca1dfc28706d50fce880d838dfee49032a746ef9bb21a326c6ba0ce5521" |
| 496 | }, |
| 497 | { |
| 498 | "name": "token_blocks.0.net.0.bias", |
| 499 | "shape": [ |
| 500 | 48 |
| 501 | ], |
| 502 | "dtype": "float16", |
| 503 | "offset_bytes": 249550, |
| 504 | "nbytes": 96, |
| 505 | "sha256": "5703171d793952315814243fcbe0f7a9461df888141c96092729370062446ca1" |
| 506 | }, |
| 507 | { |
| 508 | "name": "token_blocks.0.net.0.weight", |
| 509 | "shape": [ |
| 510 | 48, |
| 511 | 48, |
| 512 | 5 |
| 513 | ], |
| 514 | "dtype": "float16", |
| 515 | "offset_bytes": 249646, |
| 516 | "nbytes": 23040, |
| 517 | "sha256": "7b6671caff9fcc5b10716e3db12c42536f323740003ecc016f1f48377aea3eec" |
| 518 | }, |
| 519 | { |
| 520 | "name": "token_blocks.0.net.2.bias", |
| 521 | "shape": [ |
| 522 | 48 |
| 523 | ], |
| 524 | "dtype": "float16", |
| 525 | "offset_bytes": 272686, |
| 526 | "nbytes": 96, |
| 527 | "sha256": "45882f07654fce411186966107ad82975fd8e99fe2c9fc9f0307cce2aa35d437" |
| 528 | }, |
| 529 | { |
| 530 | "name": "token_blocks.0.net.2.weight", |
| 531 | "shape": [ |
| 532 | 48, |
| 533 | 48, |
| 534 | 5 |
| 535 | ], |
| 536 | "dtype": "float16", |
| 537 | "offset_bytes": 272782, |
| 538 | "nbytes": 23040, |
| 539 | "sha256": "12f1a184fcfb18ca0081fd0127085c7a96af16d6246ada0be85ed0290782e2c4" |
| 540 | }, |
| 541 | { |
| 542 | "name": "token_blocks.0.scale", |
| 543 | "shape": [ |
| 544 | 1 |
| 545 | ], |
| 546 | "dtype": "float16", |
| 547 | "offset_bytes": 295822, |
| 548 | "nbytes": 2, |
| 549 | "sha256": "dc1f4e30338b75850742878446b5eb4d10e8e3f5160af405b010c0395206986a" |
| 550 | }, |
| 551 | { |
| 552 | "name": "token_blocks.1.net.0.bias", |
| 553 | "shape": [ |
| 554 | 48 |
| 555 | ], |
| 556 | "dtype": "float16", |
| 557 | "offset_bytes": 295824, |
| 558 | "nbytes": 96, |
| 559 | "sha256": "027a36b7deae5dada86e9c82b3f38ceb1fad55a42d4862e9645dee41eb986643" |
| 560 | }, |
| 561 | { |
| 562 | "name": "token_blocks.1.net.0.weight", |
| 563 | "shape": [ |
| 564 | 48, |
| 565 | 48, |
| 566 | 5 |
| 567 | ], |
| 568 | "dtype": "float16", |
| 569 | "offset_bytes": 295920, |
| 570 | "nbytes": 23040, |
| 571 | "sha256": "32fab8e48d69b160037fef71b806af50bc37cc960f618a85ee42c3d392bc93df" |
| 572 | }, |
| 573 | { |
| 574 | "name": "token_blocks.1.net.2.bias", |
| 575 | "shape": [ |
| 576 | 48 |
| 577 | ], |
| 578 | "dtype": "float16", |
| 579 | "offset_bytes": 318960, |
| 580 | "nbytes": 96, |
| 581 | "sha256": "658c904f34f1c667b0b04d3274dfe6d662cbc169d4803784dcb311f5349cfb43" |
| 582 | }, |
| 583 | { |
| 584 | "name": "token_blocks.1.net.2.weight", |
| 585 | "shape": [ |
| 586 | 48, |
| 587 | 48, |
| 588 | 5 |
| 589 | ], |
| 590 | "dtype": "float16", |
| 591 | "offset_bytes": 319056, |
| 592 | "nbytes": 23040, |
| 593 | "sha256": "fbbdb00402ee30b90330a6451bc2342d1ad868067587279fb574a385b158a5ca" |
| 594 | }, |
| 595 | { |
| 596 | "name": "token_blocks.1.scale", |
| 597 | "shape": [ |
| 598 | 1 |
| 599 | ], |
| 600 | "dtype": "float16", |
| 601 | "offset_bytes": 342096, |
| 602 | "nbytes": 2, |
| 603 | "sha256": "f21b0ef89d8ee5daf6ba6cd441652a647229941e85b1537b5c35ede6b4d519a8" |
| 604 | }, |
| 605 | { |
| 606 | "name": "token_blocks.2.net.0.bias", |
| 607 | "shape": [ |
| 608 | 48 |
| 609 | ], |
| 610 | "dtype": "float16", |
| 611 | "offset_bytes": 342098, |
| 612 | "nbytes": 96, |
| 613 | "sha256": "7d5152e3560ef94f6005eb4b7faafe4667933d064eab81b7cff0286dafbdde75" |
| 614 | }, |
| 615 | { |
| 616 | "name": "token_blocks.2.net.0.weight", |
| 617 | "shape": [ |
| 618 | 48, |
| 619 | 48, |
| 620 | 5 |
| 621 | ], |
| 622 | "dtype": "float16", |
| 623 | "offset_bytes": 342194, |
| 624 | "nbytes": 23040, |
| 625 | "sha256": "57f51444447a17c9c00f29385a29e488075132694ae660db1bb11ee75ed4e837" |
| 626 | }, |
| 627 | { |
| 628 | "name": "token_blocks.2.net.2.bias", |
| 629 | "shape": [ |
| 630 | 48 |
| 631 | ], |
| 632 | "dtype": "float16", |
| 633 | "offset_bytes": 365234, |
| 634 | "nbytes": 96, |
| 635 | "sha256": "19e9b8bd472781afc88742149feb05156d1a0e38eb4192bff1e5153d0c650293" |
| 636 | }, |
| 637 | { |
| 638 | "name": "token_blocks.2.net.2.weight", |
| 639 | "shape": [ |
| 640 | 48, |
| 641 | 48, |
| 642 | 5 |
| 643 | ], |
| 644 | "dtype": "float16", |
| 645 | "offset_bytes": 365330, |
| 646 | "nbytes": 23040, |
| 647 | "sha256": "24159eedb8777eabdc6785ff04022b656c5d6cc42452daa7539170ff84b73c5d" |
| 648 | }, |
| 649 | { |
| 650 | "name": "token_blocks.2.scale", |
| 651 | "shape": [ |
| 652 | 1 |
| 653 | ], |
| 654 | "dtype": "float16", |
| 655 | "offset_bytes": 388370, |
| 656 | "nbytes": 2, |
| 657 | "sha256": "a1beadd91861530c43c74be436443c9cd3e063910eb1e698137592038f758398" |
| 658 | }, |
| 659 | { |
| 660 | "name": "token_input_proj.bias", |
| 661 | "shape": [ |
| 662 | 48 |
| 663 | ], |
| 664 | "dtype": "float16", |
| 665 | "offset_bytes": 388372, |
| 666 | "nbytes": 96, |
| 667 | "sha256": "1608c81a9e38fd9e0647150add8c1e12b149239d1c601ce2d705764e55f06f71" |
| 668 | }, |
| 669 | { |
| 670 | "name": "token_input_proj.weight", |
| 671 | "shape": [ |
| 672 | 48, |
| 673 | 50, |
| 674 | 1 |
| 675 | ], |
| 676 | "dtype": "float16", |
| 677 | "offset_bytes": 388468, |
| 678 | "nbytes": 4800, |
| 679 | "sha256": "42229ad332f7367864238967d64c0a1c6fb822d58bd8e959212c5a6f99eb5c47" |
| 680 | } |
| 681 | ] |
| 682 | }, |
| 683 | "decoder": { |
| 684 | "checkpoint": "artifacts/voices/de_DE-tiny/joint/decoder-student.pt", |
| 685 | "checkpoint_sha256": "40919a6a1cf797643e5dcba70af149fddf9dcb71f75b0dbc2dbea8867e1606ac", |
| 686 | "config": { |
| 687 | "in_channels": 192, |
| 688 | "source_in_channels": 192, |
| 689 | "channels": [ |
| 690 | 96, |
| 691 | 48, |
| 692 | 24, |
| 693 | 12 |
| 694 | ], |
| 695 | "res_layers": 1, |
| 696 | "variant": "piperlite", |
| 697 | "lrc_code_dim": 40, |
| 698 | "lrc_encoder_hidden": 64, |
| 699 | "rank_ratio": 0.5, |
| 700 | "activation": "leaky_relu", |
| 701 | "stage_affine": false, |
| 702 | "factorized_pre_rank": 0, |
| 703 | "piper_res_factor_rank_ratio": 0.0, |
| 704 | "res_bank_scale_mode": "kept", |
| 705 | "stage0_branches": [ |
| 706 | 0, |
| 707 | 1, |
| 708 | 2 |
| 709 | ], |
| 710 | "stage1_branches": [ |
| 711 | 0, |
| 712 | 1, |
| 713 | 2 |
| 714 | ], |
| 715 | "stage2_branches": [ |
| 716 | 0, |
| 717 | 1, |
| 718 | 2 |
| 719 | ], |
| 720 | "stage3_branches": [ |
| 721 | 0, |
| 722 | 1, |
| 723 | 2 |
| 724 | ], |
| 725 | "fsd_dim": 72, |
| 726 | "fsd_blocks": 5, |
| 727 | "fsd_film_rank": 12, |
| 728 | "fsd_head_rank": 48, |
| 729 | "wavehax_channels": 32, |
| 730 | "wavehax_blocks": 8, |
| 731 | "wavehax_mult_channels": 2, |
| 732 | "wavehax_kernel_freq": 7, |
| 733 | "wavehax_kernel_time": 7, |
| 734 | "wavehax_f0_channel": 80, |
| 735 | "wavehax_voiced_channel": 81, |
| 736 | "wavehax_sample_rate": 24000, |
| 737 | "wavehax_prior_power": 0.1, |
| 738 | "wavehax_prior_noise": 0.01, |
| 739 | "whx_channels": 32, |
| 740 | "whx_hidden": 64, |
| 741 | "whx_blocks": 8, |
| 742 | "whx_n_fft": 512, |
| 743 | "whx_f0_channel": 80, |
| 744 | "whx_voiced_channel": 81, |
| 745 | "whx_sample_rate": 24000, |
| 746 | "hift_width": 224, |
| 747 | "hift_gen_channels": 192, |
| 748 | "hift_style_dim": 64, |
| 749 | "hift_body_blocks": 3, |
| 750 | "hift_f0_channel": 80, |
| 751 | "hift_voiced_channel": 81, |
| 752 | "hift_sample_rate": 24000, |
| 753 | "stage_projection_bottlenecks": [], |
| 754 | "teacher_init_checkpoint": null, |
| 755 | "teacher_init_method": "first", |
| 756 | "teacher_init": null, |
| 757 | "init_decoder_checkpoint": "artifacts/voices/de_DE-tiny/decoder/decoder-student.pt", |
| 758 | "decoder_init": { |
| 759 | "checkpoint": "artifacts/voices/de_DE-tiny/decoder/decoder-student.pt", |
| 760 | "decoder_parameters": 314977, |
| 761 | "matched_config": { |
| 762 | "in_channels": 192, |
| 763 | "channels": [ |
| 764 | 96, |
| 765 | 48, |
| 766 | 24, |
| 767 | 12 |
| 768 | ], |
| 769 | "res_layers": 1, |
| 770 | "variant": "piperlite", |
| 771 | "rank_ratio": 0.5, |
| 772 | "activation": "leaky_relu", |
| 773 | "stage_affine": false, |
| 774 | "factorized_pre_rank": 0, |
| 775 | "piper_res_factor_rank_ratio": 0.0, |
| 776 | "res_bank_scale_mode": "kept", |
| 777 | "stage0_branches": [ |
| 778 | 0, |
| 779 | 1, |
| 780 | 2 |
| 781 | ], |
| 782 | "stage1_branches": [ |
| 783 | 0, |
| 784 | 1, |
| 785 | 2 |
| 786 | ], |
| 787 | "stage2_branches": [ |
| 788 | 0, |
| 789 | 1, |
| 790 | 2 |
| 791 | ], |
| 792 | "stage3_branches": [ |
| 793 | 0, |
| 794 | 1, |
| 795 | 2 |
| 796 | ], |
| 797 | "fsd_dim": 72, |
| 798 | "fsd_blocks": 5, |
| 799 | "fsd_film_rank": 12, |
| 800 | "fsd_head_rank": 48, |
| 801 | "wavehax_channels": 32, |
| 802 | "wavehax_blocks": 8, |
| 803 | "wavehax_mult_channels": 2, |
| 804 | "wavehax_kernel_freq": 7, |
| 805 | "wavehax_kernel_time": 7, |
| 806 | "wavehax_f0_channel": 80, |
| 807 | "wavehax_voiced_channel": 81, |
| 808 | "wavehax_sample_rate": 24000, |
| 809 | "wavehax_prior_power": 0.1, |
| 810 | "wavehax_prior_noise": 0.01, |
| 811 | "whx_channels": 32, |
| 812 | "whx_hidden": 64, |
| 813 | "whx_blocks": 8, |
| 814 | "whx_n_fft": 512, |
| 815 | "whx_f0_channel": 80, |
| 816 | "whx_voiced_channel": 81, |
| 817 | "whx_sample_rate": 24000, |
| 818 | "hift_width": 224, |
| 819 | "hift_gen_channels": 192, |
| 820 | "hift_style_dim": 64, |
| 821 | "hift_body_blocks": 3, |
| 822 | "hift_f0_channel": 80, |
| 823 | "hift_voiced_channel": 81, |
| 824 | "hift_sample_rate": 24000, |
| 825 | "stage_projection_bottlenecks": [], |
| 826 | "post_filter_channels": 0, |
| 827 | "post_filter_layers": 0, |
| 828 | "post_filter_kernel": 9, |
| 829 | "post_filter_scale": 0.25, |
| 830 | "pre_tanh_repair_channels": 0, |
| 831 | "pre_tanh_repair_layers": 0, |
| 832 | "pre_tanh_repair_kernel": 7, |
| 833 | "pre_tanh_repair_scale": 0.15, |
| 834 | "istft_n_fft": 512, |
| 835 | "istft_tail_head_rank": 20 |
| 836 | }, |
| 837 | "allowed_post_filter_mismatches": [], |
| 838 | "allowed_pre_tanh_repair_mismatches": [], |
| 839 | "allowed_activation_mismatches": [], |
| 840 | "missing_keys": [], |
| 841 | "unexpected_keys": [] |
| 842 | }, |
| 843 | "post_filter_channels": 0, |
| 844 | "post_filter_layers": 0, |
| 845 | "post_filter_kernel": 9, |
| 846 | "post_filter_scale": 0.25, |
| 847 | "pre_tanh_repair_channels": 0, |
| 848 | "pre_tanh_repair_layers": 0, |
| 849 | "pre_tanh_repair_kernel": 7, |
| 850 | "pre_tanh_repair_scale": 0.15, |
| 851 | "istft_n_fft": 512, |
| 852 | "istft_tail_head_rank": 20, |
| 853 | "ap_amplitude_weight": 0.0, |
| 854 | "ap_phase_weight": 0.0, |
| 855 | "ap_complex_weight": 0.0, |
| 856 | "spectral_head_init": "default", |
| 857 | "spectral_head_init_scale": 0.001, |
| 858 | "ap_amp_init_bias": -6.0, |
| 859 | "ap_phase_real_init_bias": 1.0, |
| 860 | "spectral_head_init_summary": null, |
| 861 | "feature_exact_weight": 0.0, |
| 862 | "feature_exact_keys": [ |
| 863 | "pre", |
| 864 | "up0" |
| 865 | ], |
| 866 | "feature_exact_teacher_init": null, |
| 867 | "signature_pack_dir": "artifacts/voices/de_DE-tiny/signatures", |
| 868 | "exact_feature_pack_dir": null, |
| 869 | "signature_hint_weight": 0.05, |
| 870 | "signature_temporal_weight": 0.4, |
| 871 | "signature_phase_weight": 0.0, |
| 872 | "signature_phase_bins": 0, |
| 873 | "signature_keys": [ |
| 874 | "stage0_mix", |
| 875 | "stage1_mix", |
| 876 | "stage2_mix", |
| 877 | "pre_tanh", |
| 878 | "audio" |
| 879 | ], |
| 880 | "bottleneck_code_checkpoint": null, |
| 881 | "bottleneck_code_weight": 0.0, |
| 882 | "bottleneck_code_keys": [], |
| 883 | "bottleneck_code_training_only_parameters": 0, |
| 884 | "echo_tail_weight": 0.0, |
| 885 | "echo_tail_min_ms": 8.0, |
| 886 | "echo_tail_max_ms": 40.0, |
| 887 | "echo_tail_lags": 9, |
| 888 | "echo_tail_margin": 0.02, |
| 889 | "paired_acoustic_residual_weight": 0.0, |
| 890 | "paired_acoustic_residual_max_scale": 0.25, |
| 891 | "lrc_pred_code_mix_prob": 0.0, |
| 892 | "lrc_pred_code_checkpoint": null, |
| 893 | "lrc_pred_code_residual_prob": 0.0, |
| 894 | "lrc_pred_code_residual_max_scale": 0.25, |
| 895 | "lrc_pred_code_cache": null |
| 896 | }, |
| 897 | "parameters": 314977, |
| 898 | "tensors": [ |
| 899 | { |
| 900 | "name": "post.bias", |
| 901 | "shape": [ |
| 902 | 1 |
| 903 | ], |
| 904 | "dtype": "float16", |
| 905 | "offset_bytes": 393268, |
| 906 | "nbytes": 2, |
| 907 | "sha256": "c2853e3a26c6801cdcc379f629a95a17fd4f0de6de28550298cfbaa300e68764" |
| 908 | }, |
| 909 | { |
| 910 | "name": "post.weight", |
| 911 | "shape": [ |
| 912 | 1, |
| 913 | 12, |
| 914 | 7 |
| 915 | ], |
| 916 | "dtype": "float16", |
| 917 | "offset_bytes": 393270, |
| 918 | "nbytes": 168, |
| 919 | "sha256": "a346c1caeae0aa2b327ecee56a29869100935af4ee65130251eb177e1f78f26f" |
| 920 | }, |
| 921 | { |
| 922 | "name": "pre.bias", |
| 923 | "shape": [ |
| 924 | 96 |
| 925 | ], |
| 926 | "dtype": "float16", |
| 927 | "offset_bytes": 393438, |
| 928 | "nbytes": 192, |
| 929 | "sha256": "3ccccc982ea68d0daecc8bf509166e83ad7b5d8b80bf467b26ec64dde68a77a9" |
| 930 | }, |
| 931 | { |
| 932 | "name": "pre.weight", |
| 933 | "shape": [ |
| 934 | 96, |
| 935 | 192, |
| 936 | 7 |
| 937 | ], |
| 938 | "dtype": "float16", |
| 939 | "offset_bytes": 393630, |
| 940 | "nbytes": 258048, |
| 941 | "sha256": "46c91f7fb91cdff8f647a6551537715c17826f11d7c9d9a9969e57c4ee07ee47" |
| 942 | }, |
| 943 | { |
| 944 | "name": "res0.0.blocks.0.conv1.bias", |
| 945 | "shape": [ |
| 946 | 48 |
| 947 | ], |
| 948 | "dtype": "float16", |
| 949 | "offset_bytes": 651678, |
| 950 | "nbytes": 96, |
| 951 | "sha256": "dc081d68a60e5e3d01c20377d31aeb04a138c2da244ffbb7416e6851caa37d98" |
| 952 | }, |
| 953 | { |
| 954 | "name": "res0.0.blocks.0.conv1.weight", |
| 955 | "shape": [ |
| 956 | 48, |
| 957 | 48, |
| 958 | 3 |
| 959 | ], |
| 960 | "dtype": "float16", |
| 961 | "offset_bytes": 651774, |
| 962 | "nbytes": 13824, |
| 963 | "sha256": "53c9bc53a1330ebce9c323291ddef34008b99eeba4992356391f84eed67aa4df" |
| 964 | }, |
| 965 | { |
| 966 | "name": "res0.0.blocks.0.conv2.bias", |
| 967 | "shape": [ |
| 968 | 48 |
| 969 | ], |
| 970 | "dtype": "float16", |
| 971 | "offset_bytes": 665598, |
| 972 | "nbytes": 96, |
| 973 | "sha256": "5eaa00b618d0dfd00fe5ddb4ecb8a35e47b5631cc6d2edc825e8f747f9355bda" |
| 974 | }, |
| 975 | { |
| 976 | "name": "res0.0.blocks.0.conv2.weight", |
| 977 | "shape": [ |
| 978 | 48, |
| 979 | 48, |
| 980 | 3 |
| 981 | ], |
| 982 | "dtype": "float16", |
| 983 | "offset_bytes": 665694, |
| 984 | "nbytes": 13824, |
| 985 | "sha256": "3bcfc92da4d941016cdd081c976a92105597a035a11e57fc2d5b0e354447f2e8" |
| 986 | }, |
| 987 | { |
| 988 | "name": "res0.0.blocks.1.conv1.bias", |
| 989 | "shape": [ |
| 990 | 48 |
| 991 | ], |
| 992 | "dtype": "float16", |
| 993 | "offset_bytes": 679518, |
| 994 | "nbytes": 96, |
| 995 | "sha256": "ca3cd210bef9eef39bfc6269d76b5c96931c1dff1c1e8c061742b28c5933f8ca" |
| 996 | }, |
| 997 | { |
| 998 | "name": "res0.0.blocks.1.conv1.weight", |
| 999 | "shape": [ |
| 1000 | 48, |
| 1001 | 48, |
| 1002 | 5 |
| 1003 | ], |
| 1004 | "dtype": "float16", |
| 1005 | "offset_bytes": 679614, |
| 1006 | "nbytes": 23040, |
| 1007 | "sha256": "8a4d550283b0b6a5b5166ef6bdc33fec62f42ffdc1beef3e2a6db56b3f025198" |
| 1008 | }, |
| 1009 | { |
| 1010 | "name": "res0.0.blocks.1.conv2.bias", |
| 1011 | "shape": [ |
| 1012 | 48 |
| 1013 | ], |
| 1014 | "dtype": "float16", |
| 1015 | "offset_bytes": 702654, |
| 1016 | "nbytes": 96, |
| 1017 | "sha256": "c19c4083008e4bc5e16c24e6a82384dbc02983db0e58aac1ea013c53a5fced78" |
| 1018 | }, |
| 1019 | { |
| 1020 | "name": "res0.0.blocks.1.conv2.weight", |
| 1021 | "shape": [ |
| 1022 | 48, |
| 1023 | 48, |
| 1024 | 5 |
| 1025 | ], |
| 1026 | "dtype": "float16", |
| 1027 | "offset_bytes": 702750, |
| 1028 | "nbytes": 23040, |
| 1029 | "sha256": "e4743dcbda7362fe27f9ae27c1dac9c37fdd220e40ebbf278d539d4f1a423b92" |
| 1030 | }, |
| 1031 | { |
| 1032 | "name": "res0.0.blocks.2.conv1.bias", |
| 1033 | "shape": [ |
| 1034 | 48 |
| 1035 | ], |
| 1036 | "dtype": "float16", |
| 1037 | "offset_bytes": 725790, |
| 1038 | "nbytes": 96, |
| 1039 | "sha256": "ccf5ed3af5b64e7169506ea1e4f1c22531d64990e393ceec480410b500d12b16" |
| 1040 | }, |
| 1041 | { |
| 1042 | "name": "res0.0.blocks.2.conv1.weight", |
| 1043 | "shape": [ |
| 1044 | 48, |
| 1045 | 48, |
| 1046 | 7 |
| 1047 | ], |
| 1048 | "dtype": "float16", |
| 1049 | "offset_bytes": 725886, |
| 1050 | "nbytes": 32256, |
| 1051 | "sha256": "5271e92d66b57e7f257b47fb78a53f92548241061feb7806dcc4affc526b867d" |
| 1052 | }, |
| 1053 | { |
| 1054 | "name": "res0.0.blocks.2.conv2.bias", |
| 1055 | "shape": [ |
| 1056 | 48 |
| 1057 | ], |
| 1058 | "dtype": "float16", |
| 1059 | "offset_bytes": 758142, |
| 1060 | "nbytes": 96, |
| 1061 | "sha256": "0894d5afae6af7a1334cf06dd870a5588f9f5c89bd4307d25312316c723fa9bc" |
| 1062 | }, |
| 1063 | { |
| 1064 | "name": "res0.0.blocks.2.conv2.weight", |
| 1065 | "shape": [ |
| 1066 | 48, |
| 1067 | 48, |
| 1068 | 7 |
| 1069 | ], |
| 1070 | "dtype": "float16", |
| 1071 | "offset_bytes": 758238, |
| 1072 | "nbytes": 32256, |
| 1073 | "sha256": "04f1f8fec29f3c617c05e6a5ef22cc1188eed7e87100e62e5964c2b0012449cf" |
| 1074 | }, |
| 1075 | { |
| 1076 | "name": "res1.0.blocks.0.conv1.bias", |
| 1077 | "shape": [ |
| 1078 | 24 |
| 1079 | ], |
| 1080 | "dtype": "float16", |
| 1081 | "offset_bytes": 790494, |
| 1082 | "nbytes": 48, |
| 1083 | "sha256": "f8860087cdc229cf2fe275769e6a0ce2814ed1355f6f719fd5d5ffb6d66421ee" |
| 1084 | }, |
| 1085 | { |
| 1086 | "name": "res1.0.blocks.0.conv1.weight", |
| 1087 | "shape": [ |
| 1088 | 24, |
| 1089 | 24, |
| 1090 | 3 |
| 1091 | ], |
| 1092 | "dtype": "float16", |
| 1093 | "offset_bytes": 790542, |
| 1094 | "nbytes": 3456, |
| 1095 | "sha256": "f2f17ac3a8c363477ea5650dafd4776f817ba7349d88e0ef4b7d59c33120f574" |
| 1096 | }, |
| 1097 | { |
| 1098 | "name": "res1.0.blocks.0.conv2.bias", |
| 1099 | "shape": [ |
| 1100 | 24 |
| 1101 | ], |
| 1102 | "dtype": "float16", |
| 1103 | "offset_bytes": 793998, |
| 1104 | "nbytes": 48, |
| 1105 | "sha256": "ecad7a6b1d1c6978e1f62945ca98a140e538f5b01dc15cd5fea85eb105b74f89" |
| 1106 | }, |
| 1107 | { |
| 1108 | "name": "res1.0.blocks.0.conv2.weight", |
| 1109 | "shape": [ |
| 1110 | 24, |
| 1111 | 24, |
| 1112 | 3 |
| 1113 | ], |
| 1114 | "dtype": "float16", |
| 1115 | "offset_bytes": 794046, |
| 1116 | "nbytes": 3456, |
| 1117 | "sha256": "ccd37f143be018a8fd72cde2de68c28da5e074950fb3c5ebb8229ca5f479a861" |
| 1118 | }, |
| 1119 | { |
| 1120 | "name": "res1.0.blocks.1.conv1.bias", |
| 1121 | "shape": [ |
| 1122 | 24 |
| 1123 | ], |
| 1124 | "dtype": "float16", |
| 1125 | "offset_bytes": 797502, |
| 1126 | "nbytes": 48, |
| 1127 | "sha256": "7e7ebd357cace0eb908d827d2e1d79e5880e2b96a62e63e93b815afafd31ebb0" |
| 1128 | }, |
| 1129 | { |
| 1130 | "name": "res1.0.blocks.1.conv1.weight", |
| 1131 | "shape": [ |
| 1132 | 24, |
| 1133 | 24, |
| 1134 | 5 |
| 1135 | ], |
| 1136 | "dtype": "float16", |
| 1137 | "offset_bytes": 797550, |
| 1138 | "nbytes": 5760, |
| 1139 | "sha256": "97edf5246c1187c16ec920ccfd4bfa46813af5bd532c17bd48bb2983072c1f20" |
| 1140 | }, |
| 1141 | { |
| 1142 | "name": "res1.0.blocks.1.conv2.bias", |
| 1143 | "shape": [ |
| 1144 | 24 |
| 1145 | ], |
| 1146 | "dtype": "float16", |
| 1147 | "offset_bytes": 803310, |
| 1148 | "nbytes": 48, |
| 1149 | "sha256": "2a915ff350da9f649b6a885337ddfb27b4ea95beaa653d5cfdb70452cd2041ca" |
| 1150 | }, |
| 1151 | { |
| 1152 | "name": "res1.0.blocks.1.conv2.weight", |
| 1153 | "shape": [ |
| 1154 | 24, |
| 1155 | 24, |
| 1156 | 5 |
| 1157 | ], |
| 1158 | "dtype": "float16", |
| 1159 | "offset_bytes": 803358, |
| 1160 | "nbytes": 5760, |
| 1161 | "sha256": "e3ac307d067778c68a41b0c3d221f3833e548cb96ce4f37061e102ab48704f82" |
| 1162 | }, |
| 1163 | { |
| 1164 | "name": "res1.0.blocks.2.conv1.bias", |
| 1165 | "shape": [ |
| 1166 | 24 |
| 1167 | ], |
| 1168 | "dtype": "float16", |
| 1169 | "offset_bytes": 809118, |
| 1170 | "nbytes": 48, |
| 1171 | "sha256": "aae5fc730b1439ae8177daf0084d5928e3a84a7e16f929f25e3840805aa867e4" |
| 1172 | }, |
| 1173 | { |
| 1174 | "name": "res1.0.blocks.2.conv1.weight", |
| 1175 | "shape": [ |
| 1176 | 24, |
| 1177 | 24, |
| 1178 | 7 |
| 1179 | ], |
| 1180 | "dtype": "float16", |
| 1181 | "offset_bytes": 809166, |
| 1182 | "nbytes": 8064, |
| 1183 | "sha256": "88f2a7e2822505536a0df4dd51d0b6da2f6e975f7861ada30b64178cb1e6f75f" |
| 1184 | }, |
| 1185 | { |
| 1186 | "name": "res1.0.blocks.2.conv2.bias", |
| 1187 | "shape": [ |
| 1188 | 24 |
| 1189 | ], |
| 1190 | "dtype": "float16", |
| 1191 | "offset_bytes": 817230, |
| 1192 | "nbytes": 48, |
| 1193 | "sha256": "f29d71c911ee00ac7ce62ebfa851fc59300eb3949592c72b435f643804ff20f2" |
| 1194 | }, |
| 1195 | { |
| 1196 | "name": "res1.0.blocks.2.conv2.weight", |
| 1197 | "shape": [ |
| 1198 | 24, |
| 1199 | 24, |
| 1200 | 7 |
| 1201 | ], |
| 1202 | "dtype": "float16", |
| 1203 | "offset_bytes": 817278, |
| 1204 | "nbytes": 8064, |
| 1205 | "sha256": "b3ae8992d627efa18cbbd314438b59928057899c0b3cf344eb8489754ff4a1af" |
| 1206 | }, |
| 1207 | { |
| 1208 | "name": "res2.0.blocks.0.conv1.bias", |
| 1209 | "shape": [ |
| 1210 | 12 |
| 1211 | ], |
| 1212 | "dtype": "float16", |
| 1213 | "offset_bytes": 825342, |
| 1214 | "nbytes": 24, |
| 1215 | "sha256": "45019f4077d077fc1bec5109d65265fb5fe30e5c38e3424425a05d8109615c86" |
| 1216 | }, |
| 1217 | { |
| 1218 | "name": "res2.0.blocks.0.conv1.weight", |
| 1219 | "shape": [ |
| 1220 | 12, |
| 1221 | 12, |
| 1222 | 3 |
| 1223 | ], |
| 1224 | "dtype": "float16", |
| 1225 | "offset_bytes": 825366, |
| 1226 | "nbytes": 864, |
| 1227 | "sha256": "5dc53202cf286d44b19653ef8ad5dd31bdf0a5cf9e0a9fe1dd7de009a680c2c3" |
| 1228 | }, |
| 1229 | { |
| 1230 | "name": "res2.0.blocks.0.conv2.bias", |
| 1231 | "shape": [ |
| 1232 | 12 |
| 1233 | ], |
| 1234 | "dtype": "float16", |
| 1235 | "offset_bytes": 826230, |
| 1236 | "nbytes": 24, |
| 1237 | "sha256": "53886c5ab43aac14f7fe2eb342fcd38bde60519fbdd7f86a35a09e79fc2f44ee" |
| 1238 | }, |
| 1239 | { |
| 1240 | "name": "res2.0.blocks.0.conv2.weight", |
| 1241 | "shape": [ |
| 1242 | 12, |
| 1243 | 12, |
| 1244 | 3 |
| 1245 | ], |
| 1246 | "dtype": "float16", |
| 1247 | "offset_bytes": 826254, |
| 1248 | "nbytes": 864, |
| 1249 | "sha256": "cb5e88cafe5683d41274f3e6ac9c2f83a27bf2f0830532a5ae4675bd75fc43fb" |
| 1250 | }, |
| 1251 | { |
| 1252 | "name": "res2.0.blocks.1.conv1.bias", |
| 1253 | "shape": [ |
| 1254 | 12 |
| 1255 | ], |
| 1256 | "dtype": "float16", |
| 1257 | "offset_bytes": 827118, |
| 1258 | "nbytes": 24, |
| 1259 | "sha256": "c492ad6ef8e6e0e95538ee75317497dc1d156e45be2dbcf927755b866a5a0a74" |
| 1260 | }, |
| 1261 | { |
| 1262 | "name": "res2.0.blocks.1.conv1.weight", |
| 1263 | "shape": [ |
| 1264 | 12, |
| 1265 | 12, |
| 1266 | 5 |
| 1267 | ], |
| 1268 | "dtype": "float16", |
| 1269 | "offset_bytes": 827142, |
| 1270 | "nbytes": 1440, |
| 1271 | "sha256": "7bc6a71f9b1858ea51525683f9660ee347b3273171296c1ff763dbbc61242722" |
| 1272 | }, |
| 1273 | { |
| 1274 | "name": "res2.0.blocks.1.conv2.bias", |
| 1275 | "shape": [ |
| 1276 | 12 |
| 1277 | ], |
| 1278 | "dtype": "float16", |
| 1279 | "offset_bytes": 828582, |
| 1280 | "nbytes": 24, |
| 1281 | "sha256": "fad68527d1da296c6665ab71d67661851aa6d4d1c9bf9b722d2d48d44b65188c" |
| 1282 | }, |
| 1283 | { |
| 1284 | "name": "res2.0.blocks.1.conv2.weight", |
| 1285 | "shape": [ |
| 1286 | 12, |
| 1287 | 12, |
| 1288 | 5 |
| 1289 | ], |
| 1290 | "dtype": "float16", |
| 1291 | "offset_bytes": 828606, |
| 1292 | "nbytes": 1440, |
| 1293 | "sha256": "70a6f37053dde5f67c090098b3b3baf0f27244b93b0a7cd12a34aef1ec7b7c98" |
| 1294 | }, |
| 1295 | { |
| 1296 | "name": "res2.0.blocks.2.conv1.bias", |
| 1297 | "shape": [ |
| 1298 | 12 |
| 1299 | ], |
| 1300 | "dtype": "float16", |
| 1301 | "offset_bytes": 830046, |
| 1302 | "nbytes": 24, |
| 1303 | "sha256": "6bba5c9d96d8cbb919fa24eb61fbeda90bea148423b560464375e544c394fddf" |
| 1304 | }, |
| 1305 | { |
| 1306 | "name": "res2.0.blocks.2.conv1.weight", |
| 1307 | "shape": [ |
| 1308 | 12, |
| 1309 | 12, |
| 1310 | 7 |
| 1311 | ], |
| 1312 | "dtype": "float16", |
| 1313 | "offset_bytes": 830070, |
| 1314 | "nbytes": 2016, |
| 1315 | "sha256": "07dabfc809718d51e24ec67c035804272ce2931d0715ba7ca26ab2bb1fef2085" |
| 1316 | }, |
| 1317 | { |
| 1318 | "name": "res2.0.blocks.2.conv2.bias", |
| 1319 | "shape": [ |
| 1320 | 12 |
| 1321 | ], |
| 1322 | "dtype": "float16", |
| 1323 | "offset_bytes": 832086, |
| 1324 | "nbytes": 24, |
| 1325 | "sha256": "349f831f36127427c464a712a73fab5b70e5e6917b83149535a0e88544d12dd2" |
| 1326 | }, |
| 1327 | { |
| 1328 | "name": "res2.0.blocks.2.conv2.weight", |
| 1329 | "shape": [ |
| 1330 | 12, |
| 1331 | 12, |
| 1332 | 7 |
| 1333 | ], |
| 1334 | "dtype": "float16", |
| 1335 | "offset_bytes": 832110, |
| 1336 | "nbytes": 2016, |
| 1337 | "sha256": "42bd9be12600fba763de7032060d1b988f2d1eb41ac30f60a02af23f73de557c" |
| 1338 | }, |
| 1339 | { |
| 1340 | "name": "up0.bias", |
| 1341 | "shape": [ |
| 1342 | 48 |
| 1343 | ], |
| 1344 | "dtype": "float16", |
| 1345 | "offset_bytes": 834126, |
| 1346 | "nbytes": 96, |
| 1347 | "sha256": "59f009dc031be31ac8aaa5a0ce95141063ad7443333fed40fc244db416e1e962" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "up0.weight", |
| 1351 | "shape": [ |
| 1352 | 96, |
| 1353 | 48, |
| 1354 | 16 |
| 1355 | ], |
| 1356 | "dtype": "float16", |
| 1357 | "offset_bytes": 834222, |
| 1358 | "nbytes": 147456, |
| 1359 | "sha256": "dc8397d5ac49ee9f683b769ca4447d17e78f4d08b4ec64abf573a7b2f5db264c" |
| 1360 | }, |
| 1361 | { |
| 1362 | "name": "up1.bias", |
| 1363 | "shape": [ |
| 1364 | 24 |
| 1365 | ], |
| 1366 | "dtype": "float16", |
| 1367 | "offset_bytes": 981678, |
| 1368 | "nbytes": 48, |
| 1369 | "sha256": "9678d8adc7b3a21c61ba88e7dc9476b588979eac646a93354ed686ab5a15b560" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "up1.weight", |
| 1373 | "shape": [ |
| 1374 | 48, |
| 1375 | 24, |
| 1376 | 16 |
| 1377 | ], |
| 1378 | "dtype": "float16", |
| 1379 | "offset_bytes": 981726, |
| 1380 | "nbytes": 36864, |
| 1381 | "sha256": "79b74297301e9af7214bc23d15a128d15dd06161483a1b436f4f6f0e4b6fb55f" |
| 1382 | }, |
| 1383 | { |
| 1384 | "name": "up2.bias", |
| 1385 | "shape": [ |
| 1386 | 12 |
| 1387 | ], |
| 1388 | "dtype": "float16", |
| 1389 | "offset_bytes": 1018590, |
| 1390 | "nbytes": 24, |
| 1391 | "sha256": "6e7c15ff7577108d412ac69670d9bc1d135c384495536fcdc0c5771ceaf31dfd" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "up2.weight", |
| 1395 | "shape": [ |
| 1396 | 24, |
| 1397 | 12, |
| 1398 | 8 |
| 1399 | ], |
| 1400 | "dtype": "float16", |
| 1401 | "offset_bytes": 1018614, |
| 1402 | "nbytes": 4608, |
| 1403 | "sha256": "67926a26b092267aad4f6c4af63394d15943d8d45a3da5b93967adda1f61b4d8" |
| 1404 | } |
| 1405 | ] |
| 1406 | } |
| 1407 | }, |
| 1408 | "runtime_ops": [ |
| 1409 | "embedding_lookup", |
| 1410 | "conv1d_same", |
| 1411 | "conv1d_1x1", |
| 1412 | "conv_transpose1d", |
| 1413 | "silu", |
| 1414 | "leaky_relu", |
| 1415 | "tanh", |
| 1416 | "residual_add", |
| 1417 | "repeat_interleave", |
| 1418 | "masked_conv1d", |
| 1419 | "depthwise_conv1d", |
| 1420 | "low_rank_linear_or_conv1d" |
| 1421 | ], |
| 1422 | "runtime_boundary": { |
| 1423 | "neural_stack_self_contained": true, |
| 1424 | "arbitrary_text_self_contained": false, |
| 1425 | "accepted_runtime_input": "Piper phoneme ID sequence matching piper-phoneme-config.json", |
| 1426 | "output": "mono float waveform at manifest.sample_rate" |
| 1427 | }, |
| 1428 | "frontend": { |
| 1429 | "status": "not fully self-contained for arbitrary text yet", |
| 1430 | "input_contract": "runtime may accept Piper phoneme ID sequences directly", |
| 1431 | "included_config": "piper-phoneme-config.json", |
| 1432 | "piper_phoneme_type": "espeak", |
| 1433 | "piper_espeak_voice": "de", |
| 1434 | "phoneme_id_map_entries": 152, |
| 1435 | "next_work": "embed or rewrite the Piper/eSpeak-compatible frontend, or keep a strict phoneme-ID API" |
| 1436 | }, |
| 1437 | "size_budget": { |
| 1438 | "target_max_bytes": 4000000, |
| 1439 | "target_max_mib": 3.814697265625, |
| 1440 | "actual_model_asset_bytes": 1028041, |
| 1441 | "target_max_parameters": 1500000, |
| 1442 | "under_parameter_budget": true, |
| 1443 | "under_4mb_asset_budget": true |
| 1444 | }, |
| 1445 | "limitations": [ |
| 1446 | "This is a raw neural asset package, not a compiled standalone runtime binary.", |
| 1447 | "No PyTorch checkpoint, ONNX graph, or training state is included.", |
| 1448 | "Arbitrary text still requires a Piper/eSpeak-compatible frontend outside this package." |
| 1449 | ] |
| 1450 | } |
| 1451 | |