fr-siwis-1p57m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "fr_FR-flores-20260908", |
| 4 | "language": "fr_FR", |
| 5 | "voice": "fr_FR-siwis-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": 3130648, |
| 15 | "weights_sha256": "1adb251d94de69dfd15e12260122c7648b1af0b48a06cc9c69db8ea7b196aa9c", |
| 16 | "total_parameters": 1565324, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/fr_FR/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "e4d381c5a7b5a566d033a48938a096ee803c7d2cb9698da70200951dfa70e2d8", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 142, |
| 24 | "hidden": 64, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 525, |
| 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": 136772, |
| 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": "41328c49d55a7c09fea3884d20c74119abb0cb6b9e9f7d030505fc7f4382600b" |
| 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": "56abcd445706b317d883db577ea2a98f588a40ad28db06e359bb761894269e5d" |
| 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": "e4b8df8f3d46961fbe4ce99e0c29c654ba7b5eae144223c268c97ea11fe57621" |
| 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": "05bd9383c4fb1745cf50ae06a478759bd9ade50f41f796e29d89ca92f5c5c64a" |
| 88 | }, |
| 89 | { |
| 90 | "name": "blocks.0.scale", |
| 91 | "shape": [ |
| 92 | 1 |
| 93 | ], |
| 94 | "dtype": "float16", |
| 95 | "offset_bytes": 82176, |
| 96 | "nbytes": 2, |
| 97 | "sha256": "45e45fccd08f20ebc7559a184335da94b909922bb375c3312c4e5515e43f0ece" |
| 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": "881df82295fb6ff9f248db8ac7e59f448e15d1ce0ad892328946909dd65a2d83" |
| 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": "6707ec00775cb7b2e05e17c86a5816644a559f25de3784e91a8e930e65c42978" |
| 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": "3a9d277280f5da7bfdca53914833b67e824718093d6fea24ffdd47c889ffda0e" |
| 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": "0a1a62435258ed3731d55dcbdba3a19e701119ac9000dda63b07378e7793ffb9" |
| 142 | }, |
| 143 | { |
| 144 | "name": "blocks.1.scale", |
| 145 | "shape": [ |
| 146 | 1 |
| 147 | ], |
| 148 | "dtype": "float16", |
| 149 | "offset_bytes": 164354, |
| 150 | "nbytes": 2, |
| 151 | "sha256": "40fb6c5c5b1f52128437e41120b2b5b27e456718723e632d6f970920b28ee0ba" |
| 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": "0e08b28d13a88d86d47bccd498f5e0e95ae4cf705e3ccc6eb925d48356ef5b10" |
| 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": "68d9596ea86a283fbe49b318135500cb4d0ddf6130d722fcf0d654e09d8dc140" |
| 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": "6b8ae2b7ae30be5a9b1aeb56a4f19dd82ba40f2332c3ae4a68e509810e5c9615" |
| 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": "a1d07b86a11bd7f992a86635b4aa2feca2db2cb011cc95c2909c07495d311d1d" |
| 196 | }, |
| 197 | { |
| 198 | "name": "blocks.2.scale", |
| 199 | "shape": [ |
| 200 | 1 |
| 201 | ], |
| 202 | "dtype": "float16", |
| 203 | "offset_bytes": 246532, |
| 204 | "nbytes": 2, |
| 205 | "sha256": "7f2253d7e228b22a08bda1f09c516f6fead81df6536eb02fa991a34bb38d9be8" |
| 206 | }, |
| 207 | { |
| 208 | "name": "embedding.weight", |
| 209 | "shape": [ |
| 210 | 142, |
| 211 | 64 |
| 212 | ], |
| 213 | "dtype": "float16", |
| 214 | "offset_bytes": 246534, |
| 215 | "nbytes": 18176, |
| 216 | "sha256": "7c4335f8eafa43769a3322b9323be97656229ea0c064a549982d15f3949c5a8e" |
| 217 | }, |
| 218 | { |
| 219 | "name": "input_proj.bias", |
| 220 | "shape": [ |
| 221 | 64 |
| 222 | ], |
| 223 | "dtype": "float16", |
| 224 | "offset_bytes": 264710, |
| 225 | "nbytes": 128, |
| 226 | "sha256": "ea37a209f2b7e4a349d6075909bfac0c86d7772e347895e3ac625b6d0080b582" |
| 227 | }, |
| 228 | { |
| 229 | "name": "input_proj.weight", |
| 230 | "shape": [ |
| 231 | 64, |
| 232 | 67, |
| 233 | 1 |
| 234 | ], |
| 235 | "dtype": "float16", |
| 236 | "offset_bytes": 264838, |
| 237 | "nbytes": 8576, |
| 238 | "sha256": "f6cadfd2fa057fc3cf57ddc7073d7696a38b11cf784022432d5b572fa1bddfbe" |
| 239 | }, |
| 240 | { |
| 241 | "name": "output.bias", |
| 242 | "shape": [ |
| 243 | 1 |
| 244 | ], |
| 245 | "dtype": "float16", |
| 246 | "offset_bytes": 273414, |
| 247 | "nbytes": 2, |
| 248 | "sha256": "bfa0451ac7a2c62d6b50a06711493763ae36e5bbe81d2964db842fbdd10148ff" |
| 249 | }, |
| 250 | { |
| 251 | "name": "output.weight", |
| 252 | "shape": [ |
| 253 | 1, |
| 254 | 64, |
| 255 | 1 |
| 256 | ], |
| 257 | "dtype": "float16", |
| 258 | "offset_bytes": 273416, |
| 259 | "nbytes": 128, |
| 260 | "sha256": "2dfc72e412ca766fbbf2882843291c7e0734c0e81434cb7a8332d80d93f26a27" |
| 261 | } |
| 262 | ] |
| 263 | }, |
| 264 | "acoustic": { |
| 265 | "checkpoint": "artifacts/voices/fr_FR/joint/latent-student.pt", |
| 266 | "checkpoint_sha256": "c6f182939a6f0d7910441ef63983e33b7c9059a453d01e3226756630c133763c", |
| 267 | "config": { |
| 268 | "architecture": "token_context", |
| 269 | "vocab_size": 142, |
| 270 | "hidden": 96, |
| 271 | "depth": 4, |
| 272 | "kernel_size": 5, |
| 273 | "out_channels": 192, |
| 274 | "token_depth": 3 |
| 275 | }, |
| 276 | "parameters": 697831, |
| 277 | "tensors": [ |
| 278 | { |
| 279 | "name": "embedding.weight", |
| 280 | "shape": [ |
| 281 | 142, |
| 282 | 96 |
| 283 | ], |
| 284 | "dtype": "float16", |
| 285 | "offset_bytes": 273544, |
| 286 | "nbytes": 27264, |
| 287 | "sha256": "db64c38e03558daa03ddabb5600908bee54d421db78097b552c5199d16d2fce0" |
| 288 | }, |
| 289 | { |
| 290 | "name": "frame_blocks.0.net.0.bias", |
| 291 | "shape": [ |
| 292 | 96 |
| 293 | ], |
| 294 | "dtype": "float16", |
| 295 | "offset_bytes": 300808, |
| 296 | "nbytes": 192, |
| 297 | "sha256": "723fdb1db12340af8d49309b42ef89294667e00f32a60f2d7beeb97fa2a7a912" |
| 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": 301000, |
| 308 | "nbytes": 92160, |
| 309 | "sha256": "51a9e030754bc3cc4617c706106d1b390d307b26647db2fb65814797fb6de29a" |
| 310 | }, |
| 311 | { |
| 312 | "name": "frame_blocks.0.net.2.bias", |
| 313 | "shape": [ |
| 314 | 96 |
| 315 | ], |
| 316 | "dtype": "float16", |
| 317 | "offset_bytes": 393160, |
| 318 | "nbytes": 192, |
| 319 | "sha256": "40ad9cd6d44d9e3a48c53869dfd48c257c934d92ced9a2d3e0ac2610f118cc67" |
| 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": 393352, |
| 330 | "nbytes": 92160, |
| 331 | "sha256": "f0cfa0a154319bea4603c38fe73be68e1a746264f273987c2ff5038c850d143f" |
| 332 | }, |
| 333 | { |
| 334 | "name": "frame_blocks.0.scale", |
| 335 | "shape": [ |
| 336 | 1 |
| 337 | ], |
| 338 | "dtype": "float16", |
| 339 | "offset_bytes": 485512, |
| 340 | "nbytes": 2, |
| 341 | "sha256": "169abef7b18fd2b398c7aed29140e38087287b8cd76df76e15868005485e6c8a" |
| 342 | }, |
| 343 | { |
| 344 | "name": "frame_blocks.1.net.0.bias", |
| 345 | "shape": [ |
| 346 | 96 |
| 347 | ], |
| 348 | "dtype": "float16", |
| 349 | "offset_bytes": 485514, |
| 350 | "nbytes": 192, |
| 351 | "sha256": "cd86b6fcec6fc92954a0b396d700ed3679a17456169fd7386783848f6700dcae" |
| 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": 485706, |
| 362 | "nbytes": 92160, |
| 363 | "sha256": "f6ceda0a38168389009abcccd564acb5dab156e464e6d393786f4e1d64fa56d3" |
| 364 | }, |
| 365 | { |
| 366 | "name": "frame_blocks.1.net.2.bias", |
| 367 | "shape": [ |
| 368 | 96 |
| 369 | ], |
| 370 | "dtype": "float16", |
| 371 | "offset_bytes": 577866, |
| 372 | "nbytes": 192, |
| 373 | "sha256": "55ca5221b3c2c416a918c1ba531fbe0a6311ed1eae73bb62524c95e049a3b026" |
| 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": 578058, |
| 384 | "nbytes": 92160, |
| 385 | "sha256": "b5dbcf6ebac270e779e6e99c2ed3c379d86d71fcef440343311a9435db1d4f65" |
| 386 | }, |
| 387 | { |
| 388 | "name": "frame_blocks.1.scale", |
| 389 | "shape": [ |
| 390 | 1 |
| 391 | ], |
| 392 | "dtype": "float16", |
| 393 | "offset_bytes": 670218, |
| 394 | "nbytes": 2, |
| 395 | "sha256": "a7634dfd197f2abd47af5b59e7e2fb8f5abf6eae6b2f55432ae5055b4dbc4063" |
| 396 | }, |
| 397 | { |
| 398 | "name": "frame_blocks.2.net.0.bias", |
| 399 | "shape": [ |
| 400 | 96 |
| 401 | ], |
| 402 | "dtype": "float16", |
| 403 | "offset_bytes": 670220, |
| 404 | "nbytes": 192, |
| 405 | "sha256": "160e0720badbad855eac211cc7218a5b6a045c38220d01d1435a80d7a7bb378e" |
| 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": 670412, |
| 416 | "nbytes": 92160, |
| 417 | "sha256": "e89516c81de95b667f36469c3bf07b835b422e06160dfaf3ee5aec089ec63b92" |
| 418 | }, |
| 419 | { |
| 420 | "name": "frame_blocks.2.net.2.bias", |
| 421 | "shape": [ |
| 422 | 96 |
| 423 | ], |
| 424 | "dtype": "float16", |
| 425 | "offset_bytes": 762572, |
| 426 | "nbytes": 192, |
| 427 | "sha256": "ded614ebb314fef2fe33416e0f048e840b3f9ed6844c7cdf24cf3a8affaf108f" |
| 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": 762764, |
| 438 | "nbytes": 92160, |
| 439 | "sha256": "27dda1e43c17cae55d3307b9099f633d987933ad3a461e7dd33462ca33e55442" |
| 440 | }, |
| 441 | { |
| 442 | "name": "frame_blocks.2.scale", |
| 443 | "shape": [ |
| 444 | 1 |
| 445 | ], |
| 446 | "dtype": "float16", |
| 447 | "offset_bytes": 854924, |
| 448 | "nbytes": 2, |
| 449 | "sha256": "bd02df9754ad0556dd2933e4e9aa41f818350220d54eafd6756812c61b74bb38" |
| 450 | }, |
| 451 | { |
| 452 | "name": "frame_blocks.3.net.0.bias", |
| 453 | "shape": [ |
| 454 | 96 |
| 455 | ], |
| 456 | "dtype": "float16", |
| 457 | "offset_bytes": 854926, |
| 458 | "nbytes": 192, |
| 459 | "sha256": "490ae90cb2add22e395442524732515a2b462b9836b7e1061e0f5ccc70d46de1" |
| 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": 855118, |
| 470 | "nbytes": 92160, |
| 471 | "sha256": "d61bdf6b5364ab6899024f7a195f10a9f9097c3c29a651e28704571dd3f88eb3" |
| 472 | }, |
| 473 | { |
| 474 | "name": "frame_blocks.3.net.2.bias", |
| 475 | "shape": [ |
| 476 | 96 |
| 477 | ], |
| 478 | "dtype": "float16", |
| 479 | "offset_bytes": 947278, |
| 480 | "nbytes": 192, |
| 481 | "sha256": "a09af629fb4bb24567b038f9021636fc8cd54300f84cf796748d8541878ed5a3" |
| 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": 947470, |
| 492 | "nbytes": 92160, |
| 493 | "sha256": "372fef8312c4d383a0a3b010fb178aa69a3b58f763989f3ef325f70dbabfe10e" |
| 494 | }, |
| 495 | { |
| 496 | "name": "frame_blocks.3.scale", |
| 497 | "shape": [ |
| 498 | 1 |
| 499 | ], |
| 500 | "dtype": "float16", |
| 501 | "offset_bytes": 1039630, |
| 502 | "nbytes": 2, |
| 503 | "sha256": "45de83b3c61da569ccf94d022360ef4ee292c4e0ea3bf0ab6359964f8ee8eedb" |
| 504 | }, |
| 505 | { |
| 506 | "name": "frame_input_proj.bias", |
| 507 | "shape": [ |
| 508 | 96 |
| 509 | ], |
| 510 | "dtype": "float16", |
| 511 | "offset_bytes": 1039632, |
| 512 | "nbytes": 192, |
| 513 | "sha256": "92c0672fd217385a2c9ddfb28d84e53fe75071adf1389ba205cbb2dfb8064855" |
| 514 | }, |
| 515 | { |
| 516 | "name": "frame_input_proj.weight", |
| 517 | "shape": [ |
| 518 | 96, |
| 519 | 99, |
| 520 | 1 |
| 521 | ], |
| 522 | "dtype": "float16", |
| 523 | "offset_bytes": 1039824, |
| 524 | "nbytes": 19008, |
| 525 | "sha256": "d43b0faaa0bc8bd01f9f6995165e448fac0d3656b9144ad8885ad08bce533e88" |
| 526 | }, |
| 527 | { |
| 528 | "name": "output.bias", |
| 529 | "shape": [ |
| 530 | 192 |
| 531 | ], |
| 532 | "dtype": "float16", |
| 533 | "offset_bytes": 1058832, |
| 534 | "nbytes": 384, |
| 535 | "sha256": "fa9ab1f34aa4671609fef9be48c44c140773998193ffe4382af8f5e21fbcb2cb" |
| 536 | }, |
| 537 | { |
| 538 | "name": "output.weight", |
| 539 | "shape": [ |
| 540 | 192, |
| 541 | 96, |
| 542 | 1 |
| 543 | ], |
| 544 | "dtype": "float16", |
| 545 | "offset_bytes": 1059216, |
| 546 | "nbytes": 36864, |
| 547 | "sha256": "cef710b1bd55c16bee54929d8c2035ec273e25eff0f4a9c8eba85621fbf60f7f" |
| 548 | }, |
| 549 | { |
| 550 | "name": "token_blocks.0.net.0.bias", |
| 551 | "shape": [ |
| 552 | 96 |
| 553 | ], |
| 554 | "dtype": "float16", |
| 555 | "offset_bytes": 1096080, |
| 556 | "nbytes": 192, |
| 557 | "sha256": "8622e4be5d32861ea5228a44280db684fdc05e59f0503bc22c89212c65851ef0" |
| 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": 1096272, |
| 568 | "nbytes": 92160, |
| 569 | "sha256": "2ba3389c6a1c6ec5f44ef54af4532e70059d622fe76c054c2aa4cf4b6e165f53" |
| 570 | }, |
| 571 | { |
| 572 | "name": "token_blocks.0.net.2.bias", |
| 573 | "shape": [ |
| 574 | 96 |
| 575 | ], |
| 576 | "dtype": "float16", |
| 577 | "offset_bytes": 1188432, |
| 578 | "nbytes": 192, |
| 579 | "sha256": "097182976cdb65b8e300f43c014da03be79e463927bbd971e94fc62b9abf772e" |
| 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": 1188624, |
| 590 | "nbytes": 92160, |
| 591 | "sha256": "a62810ad25bee086f47f23e1f8feadde63b647031aae4a004c1d506b2d2ab8f2" |
| 592 | }, |
| 593 | { |
| 594 | "name": "token_blocks.0.scale", |
| 595 | "shape": [ |
| 596 | 1 |
| 597 | ], |
| 598 | "dtype": "float16", |
| 599 | "offset_bytes": 1280784, |
| 600 | "nbytes": 2, |
| 601 | "sha256": "fc32610d482804adbc6115d0e01cbbe03d51514b62057bb8b037341f8080a5e3" |
| 602 | }, |
| 603 | { |
| 604 | "name": "token_blocks.1.net.0.bias", |
| 605 | "shape": [ |
| 606 | 96 |
| 607 | ], |
| 608 | "dtype": "float16", |
| 609 | "offset_bytes": 1280786, |
| 610 | "nbytes": 192, |
| 611 | "sha256": "c3c318d92862e4b5711a77682374780178f9d30586a16c373f66a6a1da99f374" |
| 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": 1280978, |
| 622 | "nbytes": 92160, |
| 623 | "sha256": "10d4ff12dda8e76cb9f607f821093f17c1882638ef7552620042425e9e587590" |
| 624 | }, |
| 625 | { |
| 626 | "name": "token_blocks.1.net.2.bias", |
| 627 | "shape": [ |
| 628 | 96 |
| 629 | ], |
| 630 | "dtype": "float16", |
| 631 | "offset_bytes": 1373138, |
| 632 | "nbytes": 192, |
| 633 | "sha256": "d8394bbac128086e782c8e1b6b60314e90583533d6705c693b21be586b30fcb4" |
| 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": 1373330, |
| 644 | "nbytes": 92160, |
| 645 | "sha256": "f1c6069ea5338ce9df103f6872b934048a19fc998a77e4e8c97ef6a1664fa35c" |
| 646 | }, |
| 647 | { |
| 648 | "name": "token_blocks.1.scale", |
| 649 | "shape": [ |
| 650 | 1 |
| 651 | ], |
| 652 | "dtype": "float16", |
| 653 | "offset_bytes": 1465490, |
| 654 | "nbytes": 2, |
| 655 | "sha256": "38764d5d876dc6708f84ade34b0e7b4e615b18eef2631ade31eb1883875d845e" |
| 656 | }, |
| 657 | { |
| 658 | "name": "token_blocks.2.net.0.bias", |
| 659 | "shape": [ |
| 660 | 96 |
| 661 | ], |
| 662 | "dtype": "float16", |
| 663 | "offset_bytes": 1465492, |
| 664 | "nbytes": 192, |
| 665 | "sha256": "57fd6336c4f5136d9838b22f1eda680572316892a8193ca2064aef12cb847839" |
| 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": 1465684, |
| 676 | "nbytes": 92160, |
| 677 | "sha256": "0af51ae34c6c148eb66bdd427003fafe2714cf540c26e9e5475ab8ce2c3a6806" |
| 678 | }, |
| 679 | { |
| 680 | "name": "token_blocks.2.net.2.bias", |
| 681 | "shape": [ |
| 682 | 96 |
| 683 | ], |
| 684 | "dtype": "float16", |
| 685 | "offset_bytes": 1557844, |
| 686 | "nbytes": 192, |
| 687 | "sha256": "1bdc0b3193a54af22d037bc9b7da3d55d2d084b059b907949be61a3409a08ec1" |
| 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": 1558036, |
| 698 | "nbytes": 92160, |
| 699 | "sha256": "7b01bc5fd16f1817979db30c735d1fc2f6dee195aa21b1b885d3f44871bdf7ea" |
| 700 | }, |
| 701 | { |
| 702 | "name": "token_blocks.2.scale", |
| 703 | "shape": [ |
| 704 | 1 |
| 705 | ], |
| 706 | "dtype": "float16", |
| 707 | "offset_bytes": 1650196, |
| 708 | "nbytes": 2, |
| 709 | "sha256": "c174ae89692397984cec80c60ff335caee3d5b1328f8353abf1d8f1a85260c52" |
| 710 | }, |
| 711 | { |
| 712 | "name": "token_input_proj.bias", |
| 713 | "shape": [ |
| 714 | 96 |
| 715 | ], |
| 716 | "dtype": "float16", |
| 717 | "offset_bytes": 1650198, |
| 718 | "nbytes": 192, |
| 719 | "sha256": "4bdc4475f86e226d56781b84e69bdb1c4063bdc8e9dbb4b7af4f4cf5577b6b13" |
| 720 | }, |
| 721 | { |
| 722 | "name": "token_input_proj.weight", |
| 723 | "shape": [ |
| 724 | 96, |
| 725 | 98, |
| 726 | 1 |
| 727 | ], |
| 728 | "dtype": "float16", |
| 729 | "offset_bytes": 1650390, |
| 730 | "nbytes": 18816, |
| 731 | "sha256": "b1d0b375b5a52b2121f846ac5dbb28e18f59a71c84ce07ffb84d0eb90f9d48b8" |
| 732 | } |
| 733 | ] |
| 734 | }, |
| 735 | "decoder": { |
| 736 | "checkpoint": "artifacts/voices/fr_FR/joint/decoder-student.pt", |
| 737 | "checkpoint_sha256": "c016c608ab2c64d71077b93f98b2da0e56468b5c050eccb882731dba8b046ed3", |
| 738 | "config": { |
| 739 | "in_channels": 192, |
| 740 | "source_in_channels": 192, |
| 741 | "channels": [ |
| 742 | 160, |
| 743 | 80, |
| 744 | 40, |
| 745 | 20 |
| 746 | ], |
| 747 | "res_layers": 1, |
| 748 | "variant": "piperlite", |
| 749 | "lrc_code_dim": 40, |
| 750 | "lrc_encoder_hidden": 64, |
| 751 | "rank_ratio": 0.5, |
| 752 | "activation": "leaky_relu", |
| 753 | "stage_affine": false, |
| 754 | "factorized_pre_rank": 0, |
| 755 | "piper_res_factor_rank_ratio": 0.0, |
| 756 | "res_bank_scale_mode": "kept", |
| 757 | "stage0_branches": [ |
| 758 | 0, |
| 759 | 1, |
| 760 | 2 |
| 761 | ], |
| 762 | "stage1_branches": [ |
| 763 | 0, |
| 764 | 1, |
| 765 | 2 |
| 766 | ], |
| 767 | "stage2_branches": [ |
| 768 | 0, |
| 769 | 1, |
| 770 | 2 |
| 771 | ], |
| 772 | "stage3_branches": [ |
| 773 | 0, |
| 774 | 1, |
| 775 | 2 |
| 776 | ], |
| 777 | "fsd_dim": 72, |
| 778 | "fsd_blocks": 5, |
| 779 | "fsd_film_rank": 12, |
| 780 | "fsd_head_rank": 48, |
| 781 | "wavehax_channels": 32, |
| 782 | "wavehax_blocks": 8, |
| 783 | "wavehax_mult_channels": 2, |
| 784 | "wavehax_kernel_freq": 7, |
| 785 | "wavehax_kernel_time": 7, |
| 786 | "wavehax_f0_channel": 80, |
| 787 | "wavehax_voiced_channel": 81, |
| 788 | "wavehax_sample_rate": 24000, |
| 789 | "wavehax_prior_power": 0.1, |
| 790 | "wavehax_prior_noise": 0.01, |
| 791 | "whx_channels": 32, |
| 792 | "whx_hidden": 64, |
| 793 | "whx_blocks": 8, |
| 794 | "whx_n_fft": 512, |
| 795 | "whx_f0_channel": 80, |
| 796 | "whx_voiced_channel": 81, |
| 797 | "whx_sample_rate": 24000, |
| 798 | "hift_width": 224, |
| 799 | "hift_gen_channels": 192, |
| 800 | "hift_style_dim": 64, |
| 801 | "hift_body_blocks": 3, |
| 802 | "hift_f0_channel": 80, |
| 803 | "hift_voiced_channel": 81, |
| 804 | "hift_sample_rate": 24000, |
| 805 | "stage_projection_bottlenecks": [], |
| 806 | "teacher_init_checkpoint": null, |
| 807 | "teacher_init_method": "first", |
| 808 | "teacher_init": null, |
| 809 | "init_decoder_checkpoint": "artifacts/voices/fr_FR/decoder/decoder-student.pt", |
| 810 | "decoder_init": { |
| 811 | "checkpoint": "artifacts/voices/fr_FR/decoder/decoder-student.pt", |
| 812 | "decoder_parameters": 730721, |
| 813 | "matched_config": { |
| 814 | "in_channels": 192, |
| 815 | "channels": [ |
| 816 | 160, |
| 817 | 80, |
| 818 | 40, |
| 819 | 20 |
| 820 | ], |
| 821 | "res_layers": 1, |
| 822 | "variant": "piperlite", |
| 823 | "rank_ratio": 0.5, |
| 824 | "activation": "leaky_relu", |
| 825 | "stage_affine": false, |
| 826 | "factorized_pre_rank": 0, |
| 827 | "piper_res_factor_rank_ratio": 0.0, |
| 828 | "res_bank_scale_mode": "kept", |
| 829 | "stage0_branches": [ |
| 830 | 0, |
| 831 | 1, |
| 832 | 2 |
| 833 | ], |
| 834 | "stage1_branches": [ |
| 835 | 0, |
| 836 | 1, |
| 837 | 2 |
| 838 | ], |
| 839 | "stage2_branches": [ |
| 840 | 0, |
| 841 | 1, |
| 842 | 2 |
| 843 | ], |
| 844 | "stage3_branches": [ |
| 845 | 0, |
| 846 | 1, |
| 847 | 2 |
| 848 | ], |
| 849 | "fsd_dim": 72, |
| 850 | "fsd_blocks": 5, |
| 851 | "fsd_film_rank": 12, |
| 852 | "fsd_head_rank": 48, |
| 853 | "wavehax_channels": 32, |
| 854 | "wavehax_blocks": 8, |
| 855 | "wavehax_mult_channels": 2, |
| 856 | "wavehax_kernel_freq": 7, |
| 857 | "wavehax_kernel_time": 7, |
| 858 | "wavehax_f0_channel": 80, |
| 859 | "wavehax_voiced_channel": 81, |
| 860 | "wavehax_sample_rate": 24000, |
| 861 | "wavehax_prior_power": 0.1, |
| 862 | "wavehax_prior_noise": 0.01, |
| 863 | "whx_channels": 32, |
| 864 | "whx_hidden": 64, |
| 865 | "whx_blocks": 8, |
| 866 | "whx_n_fft": 512, |
| 867 | "whx_f0_channel": 80, |
| 868 | "whx_voiced_channel": 81, |
| 869 | "whx_sample_rate": 24000, |
| 870 | "hift_width": 224, |
| 871 | "hift_gen_channels": 192, |
| 872 | "hift_style_dim": 64, |
| 873 | "hift_body_blocks": 3, |
| 874 | "hift_f0_channel": 80, |
| 875 | "hift_voiced_channel": 81, |
| 876 | "hift_sample_rate": 24000, |
| 877 | "stage_projection_bottlenecks": [], |
| 878 | "post_filter_channels": 0, |
| 879 | "post_filter_layers": 0, |
| 880 | "post_filter_kernel": 9, |
| 881 | "post_filter_scale": 0.25, |
| 882 | "pre_tanh_repair_channels": 0, |
| 883 | "pre_tanh_repair_layers": 0, |
| 884 | "pre_tanh_repair_kernel": 7, |
| 885 | "pre_tanh_repair_scale": 0.15, |
| 886 | "istft_n_fft": 512, |
| 887 | "istft_tail_head_rank": 20 |
| 888 | }, |
| 889 | "allowed_post_filter_mismatches": [], |
| 890 | "allowed_pre_tanh_repair_mismatches": [], |
| 891 | "allowed_activation_mismatches": [], |
| 892 | "missing_keys": [], |
| 893 | "unexpected_keys": [] |
| 894 | }, |
| 895 | "post_filter_channels": 0, |
| 896 | "post_filter_layers": 0, |
| 897 | "post_filter_kernel": 9, |
| 898 | "post_filter_scale": 0.25, |
| 899 | "pre_tanh_repair_channels": 0, |
| 900 | "pre_tanh_repair_layers": 0, |
| 901 | "pre_tanh_repair_kernel": 7, |
| 902 | "pre_tanh_repair_scale": 0.15, |
| 903 | "istft_n_fft": 512, |
| 904 | "istft_tail_head_rank": 20, |
| 905 | "ap_amplitude_weight": 0.0, |
| 906 | "ap_phase_weight": 0.0, |
| 907 | "ap_complex_weight": 0.0, |
| 908 | "spectral_head_init": "default", |
| 909 | "spectral_head_init_scale": 0.001, |
| 910 | "ap_amp_init_bias": -6.0, |
| 911 | "ap_phase_real_init_bias": 1.0, |
| 912 | "spectral_head_init_summary": null, |
| 913 | "feature_exact_weight": 0.0, |
| 914 | "feature_exact_keys": [ |
| 915 | "pre", |
| 916 | "up0" |
| 917 | ], |
| 918 | "feature_exact_teacher_init": null, |
| 919 | "signature_pack_dir": "artifacts/voices/fr_FR/signatures", |
| 920 | "exact_feature_pack_dir": null, |
| 921 | "signature_hint_weight": 0.05, |
| 922 | "signature_temporal_weight": 0.4, |
| 923 | "signature_phase_weight": 0.0, |
| 924 | "signature_phase_bins": 0, |
| 925 | "signature_keys": [ |
| 926 | "stage0_mix", |
| 927 | "stage1_mix", |
| 928 | "stage2_mix", |
| 929 | "pre_tanh", |
| 930 | "audio" |
| 931 | ], |
| 932 | "bottleneck_code_checkpoint": null, |
| 933 | "bottleneck_code_weight": 0.0, |
| 934 | "bottleneck_code_keys": [], |
| 935 | "bottleneck_code_training_only_parameters": 0, |
| 936 | "echo_tail_weight": 0.0, |
| 937 | "echo_tail_min_ms": 8.0, |
| 938 | "echo_tail_max_ms": 40.0, |
| 939 | "echo_tail_lags": 9, |
| 940 | "echo_tail_margin": 0.02, |
| 941 | "paired_acoustic_residual_weight": 0.0, |
| 942 | "paired_acoustic_residual_max_scale": 0.25, |
| 943 | "lrc_pred_code_mix_prob": 0.0, |
| 944 | "lrc_pred_code_checkpoint": null, |
| 945 | "lrc_pred_code_residual_prob": 0.0, |
| 946 | "lrc_pred_code_residual_max_scale": 0.25, |
| 947 | "lrc_pred_code_cache": null |
| 948 | }, |
| 949 | "parameters": 730721, |
| 950 | "tensors": [ |
| 951 | { |
| 952 | "name": "post.bias", |
| 953 | "shape": [ |
| 954 | 1 |
| 955 | ], |
| 956 | "dtype": "float16", |
| 957 | "offset_bytes": 1669206, |
| 958 | "nbytes": 2, |
| 959 | "sha256": "ea698a850acd9a300e0b315487849e53c370b15b21fbfce29244cc1c0ef6a22e" |
| 960 | }, |
| 961 | { |
| 962 | "name": "post.weight", |
| 963 | "shape": [ |
| 964 | 1, |
| 965 | 20, |
| 966 | 7 |
| 967 | ], |
| 968 | "dtype": "float16", |
| 969 | "offset_bytes": 1669208, |
| 970 | "nbytes": 280, |
| 971 | "sha256": "5d9cd6b92c2572e55684de627196491b8de6ec139e69a753f29df20562d43ade" |
| 972 | }, |
| 973 | { |
| 974 | "name": "pre.bias", |
| 975 | "shape": [ |
| 976 | 160 |
| 977 | ], |
| 978 | "dtype": "float16", |
| 979 | "offset_bytes": 1669488, |
| 980 | "nbytes": 320, |
| 981 | "sha256": "77ebaa50ce96489fdb4b5864814b5d77dd8c486dc3b24af127fde3e961abd4fe" |
| 982 | }, |
| 983 | { |
| 984 | "name": "pre.weight", |
| 985 | "shape": [ |
| 986 | 160, |
| 987 | 192, |
| 988 | 7 |
| 989 | ], |
| 990 | "dtype": "float16", |
| 991 | "offset_bytes": 1669808, |
| 992 | "nbytes": 430080, |
| 993 | "sha256": "1018f99689f3d5cc7775bf1de0cd66c2fa65115057e99e9c2dd9aa69fcaae284" |
| 994 | }, |
| 995 | { |
| 996 | "name": "res0.0.blocks.0.conv1.bias", |
| 997 | "shape": [ |
| 998 | 80 |
| 999 | ], |
| 1000 | "dtype": "float16", |
| 1001 | "offset_bytes": 2099888, |
| 1002 | "nbytes": 160, |
| 1003 | "sha256": "a866471c45d02eb32c77cc046b585e3d456a8225e2c1f2c6fa97e813d1573d6f" |
| 1004 | }, |
| 1005 | { |
| 1006 | "name": "res0.0.blocks.0.conv1.weight", |
| 1007 | "shape": [ |
| 1008 | 80, |
| 1009 | 80, |
| 1010 | 3 |
| 1011 | ], |
| 1012 | "dtype": "float16", |
| 1013 | "offset_bytes": 2100048, |
| 1014 | "nbytes": 38400, |
| 1015 | "sha256": "723d31067ab5e0aefacb1e38a3e5514e9ce4716fc8d46ebd2a9d67f2f464278f" |
| 1016 | }, |
| 1017 | { |
| 1018 | "name": "res0.0.blocks.0.conv2.bias", |
| 1019 | "shape": [ |
| 1020 | 80 |
| 1021 | ], |
| 1022 | "dtype": "float16", |
| 1023 | "offset_bytes": 2138448, |
| 1024 | "nbytes": 160, |
| 1025 | "sha256": "21620822c4577f94be04fa1eb9febab18fb23644634bb39fe8a7d5d10bdf810c" |
| 1026 | }, |
| 1027 | { |
| 1028 | "name": "res0.0.blocks.0.conv2.weight", |
| 1029 | "shape": [ |
| 1030 | 80, |
| 1031 | 80, |
| 1032 | 3 |
| 1033 | ], |
| 1034 | "dtype": "float16", |
| 1035 | "offset_bytes": 2138608, |
| 1036 | "nbytes": 38400, |
| 1037 | "sha256": "fd22721dfd24519e4e169a496064228d730789e254fbe438c00acca06d9d7f75" |
| 1038 | }, |
| 1039 | { |
| 1040 | "name": "res0.0.blocks.1.conv1.bias", |
| 1041 | "shape": [ |
| 1042 | 80 |
| 1043 | ], |
| 1044 | "dtype": "float16", |
| 1045 | "offset_bytes": 2177008, |
| 1046 | "nbytes": 160, |
| 1047 | "sha256": "b865370eaf27ecbb9b97f414df0d9fe83cfa6e69b9b55bb285c83359ce920b78" |
| 1048 | }, |
| 1049 | { |
| 1050 | "name": "res0.0.blocks.1.conv1.weight", |
| 1051 | "shape": [ |
| 1052 | 80, |
| 1053 | 80, |
| 1054 | 5 |
| 1055 | ], |
| 1056 | "dtype": "float16", |
| 1057 | "offset_bytes": 2177168, |
| 1058 | "nbytes": 64000, |
| 1059 | "sha256": "2bba81a7d36a112acd1c6eabb7ae514e55008b5e1a719c492951ca3e10d625e5" |
| 1060 | }, |
| 1061 | { |
| 1062 | "name": "res0.0.blocks.1.conv2.bias", |
| 1063 | "shape": [ |
| 1064 | 80 |
| 1065 | ], |
| 1066 | "dtype": "float16", |
| 1067 | "offset_bytes": 2241168, |
| 1068 | "nbytes": 160, |
| 1069 | "sha256": "c8540bf6fa4e0340b5a6b739745d5ecd42dcdf70f482f416e068c3af89c1bf68" |
| 1070 | }, |
| 1071 | { |
| 1072 | "name": "res0.0.blocks.1.conv2.weight", |
| 1073 | "shape": [ |
| 1074 | 80, |
| 1075 | 80, |
| 1076 | 5 |
| 1077 | ], |
| 1078 | "dtype": "float16", |
| 1079 | "offset_bytes": 2241328, |
| 1080 | "nbytes": 64000, |
| 1081 | "sha256": "7995927c3287aee35607c2075f75fd5b716dc0fa89674463ec1685655994e363" |
| 1082 | }, |
| 1083 | { |
| 1084 | "name": "res0.0.blocks.2.conv1.bias", |
| 1085 | "shape": [ |
| 1086 | 80 |
| 1087 | ], |
| 1088 | "dtype": "float16", |
| 1089 | "offset_bytes": 2305328, |
| 1090 | "nbytes": 160, |
| 1091 | "sha256": "d23d12be274ced2c88494b54085cf32328498880ccc9c1d169e9b4e41f0b3547" |
| 1092 | }, |
| 1093 | { |
| 1094 | "name": "res0.0.blocks.2.conv1.weight", |
| 1095 | "shape": [ |
| 1096 | 80, |
| 1097 | 80, |
| 1098 | 7 |
| 1099 | ], |
| 1100 | "dtype": "float16", |
| 1101 | "offset_bytes": 2305488, |
| 1102 | "nbytes": 89600, |
| 1103 | "sha256": "71137aac2d0eaa6f43b28214883ca99b6019fe3e96bf3a0a2a662884cdef2f0d" |
| 1104 | }, |
| 1105 | { |
| 1106 | "name": "res0.0.blocks.2.conv2.bias", |
| 1107 | "shape": [ |
| 1108 | 80 |
| 1109 | ], |
| 1110 | "dtype": "float16", |
| 1111 | "offset_bytes": 2395088, |
| 1112 | "nbytes": 160, |
| 1113 | "sha256": "117ffa17b13e40caf9427f1308da011f72276bb44e4d80561296ff467d00c5d5" |
| 1114 | }, |
| 1115 | { |
| 1116 | "name": "res0.0.blocks.2.conv2.weight", |
| 1117 | "shape": [ |
| 1118 | 80, |
| 1119 | 80, |
| 1120 | 7 |
| 1121 | ], |
| 1122 | "dtype": "float16", |
| 1123 | "offset_bytes": 2395248, |
| 1124 | "nbytes": 89600, |
| 1125 | "sha256": "52251c2d503644fe671e9af18127344748c952979f1a28f827c0255f546a66ca" |
| 1126 | }, |
| 1127 | { |
| 1128 | "name": "res1.0.blocks.0.conv1.bias", |
| 1129 | "shape": [ |
| 1130 | 40 |
| 1131 | ], |
| 1132 | "dtype": "float16", |
| 1133 | "offset_bytes": 2484848, |
| 1134 | "nbytes": 80, |
| 1135 | "sha256": "b58b46b010b357fafe8673bf608d431958987a519fdc5b603127dc004f81cafe" |
| 1136 | }, |
| 1137 | { |
| 1138 | "name": "res1.0.blocks.0.conv1.weight", |
| 1139 | "shape": [ |
| 1140 | 40, |
| 1141 | 40, |
| 1142 | 3 |
| 1143 | ], |
| 1144 | "dtype": "float16", |
| 1145 | "offset_bytes": 2484928, |
| 1146 | "nbytes": 9600, |
| 1147 | "sha256": "bfb216a404c3b6d89b85978bd24ee71dbd23d6c082a4225fb511c96444597e89" |
| 1148 | }, |
| 1149 | { |
| 1150 | "name": "res1.0.blocks.0.conv2.bias", |
| 1151 | "shape": [ |
| 1152 | 40 |
| 1153 | ], |
| 1154 | "dtype": "float16", |
| 1155 | "offset_bytes": 2494528, |
| 1156 | "nbytes": 80, |
| 1157 | "sha256": "10358adf8f206ea55889106404a1e5b7fbad14803578bcb11ab0bcceac4ab4bb" |
| 1158 | }, |
| 1159 | { |
| 1160 | "name": "res1.0.blocks.0.conv2.weight", |
| 1161 | "shape": [ |
| 1162 | 40, |
| 1163 | 40, |
| 1164 | 3 |
| 1165 | ], |
| 1166 | "dtype": "float16", |
| 1167 | "offset_bytes": 2494608, |
| 1168 | "nbytes": 9600, |
| 1169 | "sha256": "89e625a4137d534bae4b420988c3ba4625d1709415c7a46818ffdb4442d7fb86" |
| 1170 | }, |
| 1171 | { |
| 1172 | "name": "res1.0.blocks.1.conv1.bias", |
| 1173 | "shape": [ |
| 1174 | 40 |
| 1175 | ], |
| 1176 | "dtype": "float16", |
| 1177 | "offset_bytes": 2504208, |
| 1178 | "nbytes": 80, |
| 1179 | "sha256": "0118a0f686107c77ab02f180348ba1eb4ffe7a809d0551ad0071fe7edcb22af3" |
| 1180 | }, |
| 1181 | { |
| 1182 | "name": "res1.0.blocks.1.conv1.weight", |
| 1183 | "shape": [ |
| 1184 | 40, |
| 1185 | 40, |
| 1186 | 5 |
| 1187 | ], |
| 1188 | "dtype": "float16", |
| 1189 | "offset_bytes": 2504288, |
| 1190 | "nbytes": 16000, |
| 1191 | "sha256": "ad11a4e50cef13b5b76e0e367f00f18d8c0548a439548426209184d9c2ef55d7" |
| 1192 | }, |
| 1193 | { |
| 1194 | "name": "res1.0.blocks.1.conv2.bias", |
| 1195 | "shape": [ |
| 1196 | 40 |
| 1197 | ], |
| 1198 | "dtype": "float16", |
| 1199 | "offset_bytes": 2520288, |
| 1200 | "nbytes": 80, |
| 1201 | "sha256": "81b337c6f306362cc3ae8e5b5c0661ed908bbb51512720093663a354003568ad" |
| 1202 | }, |
| 1203 | { |
| 1204 | "name": "res1.0.blocks.1.conv2.weight", |
| 1205 | "shape": [ |
| 1206 | 40, |
| 1207 | 40, |
| 1208 | 5 |
| 1209 | ], |
| 1210 | "dtype": "float16", |
| 1211 | "offset_bytes": 2520368, |
| 1212 | "nbytes": 16000, |
| 1213 | "sha256": "cacc6da38780fc2c6ea017511179e46047d2ebae91c0b61d2aa8d53dfec2d53d" |
| 1214 | }, |
| 1215 | { |
| 1216 | "name": "res1.0.blocks.2.conv1.bias", |
| 1217 | "shape": [ |
| 1218 | 40 |
| 1219 | ], |
| 1220 | "dtype": "float16", |
| 1221 | "offset_bytes": 2536368, |
| 1222 | "nbytes": 80, |
| 1223 | "sha256": "2c5b3f0ea0813b1bfc3a578faa39f365badb5353ad636d3e19553392b8a167f4" |
| 1224 | }, |
| 1225 | { |
| 1226 | "name": "res1.0.blocks.2.conv1.weight", |
| 1227 | "shape": [ |
| 1228 | 40, |
| 1229 | 40, |
| 1230 | 7 |
| 1231 | ], |
| 1232 | "dtype": "float16", |
| 1233 | "offset_bytes": 2536448, |
| 1234 | "nbytes": 22400, |
| 1235 | "sha256": "b68423ac1551f59085cb3f0bddb77b543c0f36dccf50b7167ee95bf848d45e2e" |
| 1236 | }, |
| 1237 | { |
| 1238 | "name": "res1.0.blocks.2.conv2.bias", |
| 1239 | "shape": [ |
| 1240 | 40 |
| 1241 | ], |
| 1242 | "dtype": "float16", |
| 1243 | "offset_bytes": 2558848, |
| 1244 | "nbytes": 80, |
| 1245 | "sha256": "2c481747015149f9734337daa5a4bbe7049b6a76c65d1e616805255469e9601b" |
| 1246 | }, |
| 1247 | { |
| 1248 | "name": "res1.0.blocks.2.conv2.weight", |
| 1249 | "shape": [ |
| 1250 | 40, |
| 1251 | 40, |
| 1252 | 7 |
| 1253 | ], |
| 1254 | "dtype": "float16", |
| 1255 | "offset_bytes": 2558928, |
| 1256 | "nbytes": 22400, |
| 1257 | "sha256": "8dc3a397d9fde852e719b5fdf6d2f02f1ac47e942d3821f2bb6b384bbaff2e5f" |
| 1258 | }, |
| 1259 | { |
| 1260 | "name": "res2.0.blocks.0.conv1.bias", |
| 1261 | "shape": [ |
| 1262 | 20 |
| 1263 | ], |
| 1264 | "dtype": "float16", |
| 1265 | "offset_bytes": 2581328, |
| 1266 | "nbytes": 40, |
| 1267 | "sha256": "46e0b31b800cb92f1abe45771dff2f1240c560e3847580a378869b5337e7157f" |
| 1268 | }, |
| 1269 | { |
| 1270 | "name": "res2.0.blocks.0.conv1.weight", |
| 1271 | "shape": [ |
| 1272 | 20, |
| 1273 | 20, |
| 1274 | 3 |
| 1275 | ], |
| 1276 | "dtype": "float16", |
| 1277 | "offset_bytes": 2581368, |
| 1278 | "nbytes": 2400, |
| 1279 | "sha256": "f7df48e0a099020bd47647a0b66fb776924effc5c96f57f86d7af88d7fb11be4" |
| 1280 | }, |
| 1281 | { |
| 1282 | "name": "res2.0.blocks.0.conv2.bias", |
| 1283 | "shape": [ |
| 1284 | 20 |
| 1285 | ], |
| 1286 | "dtype": "float16", |
| 1287 | "offset_bytes": 2583768, |
| 1288 | "nbytes": 40, |
| 1289 | "sha256": "7851b6034c59193699a750cb231ca29823a426060613024aac946fc31fa86bc9" |
| 1290 | }, |
| 1291 | { |
| 1292 | "name": "res2.0.blocks.0.conv2.weight", |
| 1293 | "shape": [ |
| 1294 | 20, |
| 1295 | 20, |
| 1296 | 3 |
| 1297 | ], |
| 1298 | "dtype": "float16", |
| 1299 | "offset_bytes": 2583808, |
| 1300 | "nbytes": 2400, |
| 1301 | "sha256": "682addd64ee00d1067f31d88980ffa5da9a0b91018b674037349caeb1b3e1e98" |
| 1302 | }, |
| 1303 | { |
| 1304 | "name": "res2.0.blocks.1.conv1.bias", |
| 1305 | "shape": [ |
| 1306 | 20 |
| 1307 | ], |
| 1308 | "dtype": "float16", |
| 1309 | "offset_bytes": 2586208, |
| 1310 | "nbytes": 40, |
| 1311 | "sha256": "154cf731d65de42582099fbd74fee0177e1463b6bb39f0f04b6cd5240b0786cf" |
| 1312 | }, |
| 1313 | { |
| 1314 | "name": "res2.0.blocks.1.conv1.weight", |
| 1315 | "shape": [ |
| 1316 | 20, |
| 1317 | 20, |
| 1318 | 5 |
| 1319 | ], |
| 1320 | "dtype": "float16", |
| 1321 | "offset_bytes": 2586248, |
| 1322 | "nbytes": 4000, |
| 1323 | "sha256": "89bb17a69d59d99bbad5da4a2a7f36360da426830ed50a0b616d319cc594c8c3" |
| 1324 | }, |
| 1325 | { |
| 1326 | "name": "res2.0.blocks.1.conv2.bias", |
| 1327 | "shape": [ |
| 1328 | 20 |
| 1329 | ], |
| 1330 | "dtype": "float16", |
| 1331 | "offset_bytes": 2590248, |
| 1332 | "nbytes": 40, |
| 1333 | "sha256": "65875cba8401f3fd69f5d1ca127d3caad831ef6ba376d95a16bd9ac92596dc33" |
| 1334 | }, |
| 1335 | { |
| 1336 | "name": "res2.0.blocks.1.conv2.weight", |
| 1337 | "shape": [ |
| 1338 | 20, |
| 1339 | 20, |
| 1340 | 5 |
| 1341 | ], |
| 1342 | "dtype": "float16", |
| 1343 | "offset_bytes": 2590288, |
| 1344 | "nbytes": 4000, |
| 1345 | "sha256": "0aa7e6724ffc2884b67d3e963a5c2e4dc5f5e7a65fdbb50e3744f217b49048ef" |
| 1346 | }, |
| 1347 | { |
| 1348 | "name": "res2.0.blocks.2.conv1.bias", |
| 1349 | "shape": [ |
| 1350 | 20 |
| 1351 | ], |
| 1352 | "dtype": "float16", |
| 1353 | "offset_bytes": 2594288, |
| 1354 | "nbytes": 40, |
| 1355 | "sha256": "0077ca8a985b2c9e91a9e73b5882f12dfc8533a57dd9a076e9f6a64301d84917" |
| 1356 | }, |
| 1357 | { |
| 1358 | "name": "res2.0.blocks.2.conv1.weight", |
| 1359 | "shape": [ |
| 1360 | 20, |
| 1361 | 20, |
| 1362 | 7 |
| 1363 | ], |
| 1364 | "dtype": "float16", |
| 1365 | "offset_bytes": 2594328, |
| 1366 | "nbytes": 5600, |
| 1367 | "sha256": "5e81a196b49271e39e3bd4d15af07f22c3393f672f0a27a09fb626138f99b2c7" |
| 1368 | }, |
| 1369 | { |
| 1370 | "name": "res2.0.blocks.2.conv2.bias", |
| 1371 | "shape": [ |
| 1372 | 20 |
| 1373 | ], |
| 1374 | "dtype": "float16", |
| 1375 | "offset_bytes": 2599928, |
| 1376 | "nbytes": 40, |
| 1377 | "sha256": "f3d4ea2c9a4df00b50e29df10289639c60d75e4fea2413197d22aec8d2b72c1d" |
| 1378 | }, |
| 1379 | { |
| 1380 | "name": "res2.0.blocks.2.conv2.weight", |
| 1381 | "shape": [ |
| 1382 | 20, |
| 1383 | 20, |
| 1384 | 7 |
| 1385 | ], |
| 1386 | "dtype": "float16", |
| 1387 | "offset_bytes": 2599968, |
| 1388 | "nbytes": 5600, |
| 1389 | "sha256": "ce3da09a3cbdd972ddd06b12cc360c05609d4bcb243a4a210b3adda82752c562" |
| 1390 | }, |
| 1391 | { |
| 1392 | "name": "up0.bias", |
| 1393 | "shape": [ |
| 1394 | 80 |
| 1395 | ], |
| 1396 | "dtype": "float16", |
| 1397 | "offset_bytes": 2605568, |
| 1398 | "nbytes": 160, |
| 1399 | "sha256": "fcb4c8c13a99df0a3206f98c3ab0d8d62bb8474ecca725c4c79eaf9a6b3afaea" |
| 1400 | }, |
| 1401 | { |
| 1402 | "name": "up0.weight", |
| 1403 | "shape": [ |
| 1404 | 160, |
| 1405 | 80, |
| 1406 | 16 |
| 1407 | ], |
| 1408 | "dtype": "float16", |
| 1409 | "offset_bytes": 2605728, |
| 1410 | "nbytes": 409600, |
| 1411 | "sha256": "b917d4c1cc68518dc29a5857a7557b75fa1a6434d5468ad3c0cd40e8927b01f9" |
| 1412 | }, |
| 1413 | { |
| 1414 | "name": "up1.bias", |
| 1415 | "shape": [ |
| 1416 | 40 |
| 1417 | ], |
| 1418 | "dtype": "float16", |
| 1419 | "offset_bytes": 3015328, |
| 1420 | "nbytes": 80, |
| 1421 | "sha256": "b8896db5a2a917093e60fa7e73fd5543be12f1cfb5c8aa7c22c1000142549462" |
| 1422 | }, |
| 1423 | { |
| 1424 | "name": "up1.weight", |
| 1425 | "shape": [ |
| 1426 | 80, |
| 1427 | 40, |
| 1428 | 16 |
| 1429 | ], |
| 1430 | "dtype": "float16", |
| 1431 | "offset_bytes": 3015408, |
| 1432 | "nbytes": 102400, |
| 1433 | "sha256": "a4a7ad21e408aa58564490d226974fcbd3acddc8f69fdf3576cd80aa9f5002ae" |
| 1434 | }, |
| 1435 | { |
| 1436 | "name": "up2.bias", |
| 1437 | "shape": [ |
| 1438 | 20 |
| 1439 | ], |
| 1440 | "dtype": "float16", |
| 1441 | "offset_bytes": 3117808, |
| 1442 | "nbytes": 40, |
| 1443 | "sha256": "e3cb4377f656552db6379477292ab9db07d9a683446e803cc231ed7a11989db4" |
| 1444 | }, |
| 1445 | { |
| 1446 | "name": "up2.weight", |
| 1447 | "shape": [ |
| 1448 | 40, |
| 1449 | 20, |
| 1450 | 8 |
| 1451 | ], |
| 1452 | "dtype": "float16", |
| 1453 | "offset_bytes": 3117848, |
| 1454 | "nbytes": 12800, |
| 1455 | "sha256": "f0cefc0ca9ca29502757feb27da4cd478817d1f42a495ef96874b9fcad8f5163" |
| 1456 | } |
| 1457 | ] |
| 1458 | } |
| 1459 | }, |
| 1460 | "runtime_ops": [ |
| 1461 | "embedding_lookup", |
| 1462 | "conv1d_same", |
| 1463 | "conv1d_1x1", |
| 1464 | "conv_transpose1d", |
| 1465 | "silu", |
| 1466 | "leaky_relu", |
| 1467 | "tanh", |
| 1468 | "residual_add", |
| 1469 | "repeat_interleave", |
| 1470 | "masked_conv1d", |
| 1471 | "depthwise_conv1d", |
| 1472 | "low_rank_linear_or_conv1d" |
| 1473 | ], |
| 1474 | "runtime_boundary": { |
| 1475 | "neural_stack_self_contained": true, |
| 1476 | "arbitrary_text_self_contained": false, |
| 1477 | "accepted_runtime_input": "Piper phoneme ID sequence matching piper-phoneme-config.json", |
| 1478 | "output": "mono float waveform at manifest.sample_rate" |
| 1479 | }, |
| 1480 | "frontend": { |
| 1481 | "status": "not fully self-contained for arbitrary text yet", |
| 1482 | "input_contract": "runtime may accept Piper phoneme ID sequences directly", |
| 1483 | "included_config": "piper-phoneme-config.json", |
| 1484 | "piper_phoneme_type": "espeak", |
| 1485 | "piper_espeak_voice": "fr", |
| 1486 | "phoneme_id_map_entries": 154, |
| 1487 | "next_work": "embed or rewrite the Piper/eSpeak-compatible frontend, or keep a strict phoneme-ID API" |
| 1488 | }, |
| 1489 | "size_budget": { |
| 1490 | "target_max_bytes": 4000000, |
| 1491 | "target_max_mib": 3.814697265625, |
| 1492 | "actual_model_asset_bytes": 3135523, |
| 1493 | "target_max_parameters": 1500000, |
| 1494 | "under_parameter_budget": false, |
| 1495 | "under_4mb_asset_budget": true |
| 1496 | }, |
| 1497 | "limitations": [ |
| 1498 | "This is a raw neural asset package, not a compiled standalone runtime binary.", |
| 1499 | "No PyTorch checkpoint, ONNX graph, or training state is included.", |
| 1500 | "Arbitrary text still requires a Piper/eSpeak-compatible frontend outside this package." |
| 1501 | ] |
| 1502 | } |
| 1503 | |