tr-dfki-0p51m/manifest.json
| 1 | { |
| 2 | "format": "roota.raw-fp16.v1", |
| 3 | "package_name": "tr_TR-flores-20260908", |
| 4 | "language": "tr_TR", |
| 5 | "voice": "tr_TR-dfki-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": 1020182, |
| 15 | "weights_sha256": "a6ea40fd831c77f62e8056a3bc65d92a5da193754dfdcd8dbf14be9a895c88b2", |
| 16 | "total_parameters": 510091, |
| 17 | "components": { |
| 18 | "duration": { |
| 19 | "checkpoint": "artifacts/voices/tr_TR-tiny/duration/duration-student.pt", |
| 20 | "checkpoint_sha256": "3cfc59237846d5b0d2e36dda5514c50abb2b86ce7b56b2298a4910e5ff5051a3", |
| 21 | "config": { |
| 22 | "architecture": "duration_conv", |
| 23 | "vocab_size": 123, |
| 24 | "hidden": 32, |
| 25 | "depth": 3, |
| 26 | "kernel_size": 5, |
| 27 | "max_tokens": 639, |
| 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": 36036, |
| 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": "b83f052fc4b71c7f676d02224b6abea0349549565b10200f51f0a5dbd3375c5e" |
| 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": "0da4f4adf36e78139efd3b4c226752bbb103889dccac65532a858be9efe5ccaf" |
| 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": "ed96ec4b90fde98a103a93a47b1af3bb47a9fc26de30db85e7a70f9d4cce69d2" |
| 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": "18ec9c1f525dcb75a39799193812a9b02cf79d4eead740931e0b72b350526ec2" |
| 90 | }, |
| 91 | { |
| 92 | "name": "blocks.0.scale", |
| 93 | "shape": [ |
| 94 | 1 |
| 95 | ], |
| 96 | "dtype": "float16", |
| 97 | "offset_bytes": 20608, |
| 98 | "nbytes": 2, |
| 99 | "sha256": "147e82faca64a021f4af180c2acbeaa8be6a43478105110702869e353cda8c46" |
| 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": "b745d953d0737b02b245aee6086f7e2fe7f3254f36e640995e7f5d6f75d53536" |
| 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": "457614aa14e574bfc225e64b43bebd50fab84c51e22fd2340455a15b28cdfb89" |
| 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": "032702ce4bfdfe6bf477b3e13b22df50204dc7c57731d7bee876cd17654eb932" |
| 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": "893d310a63ba474e198dbc7b1aea5c4d2f785b7f6b8a905d32725e35d3d9302c" |
| 144 | }, |
| 145 | { |
| 146 | "name": "blocks.1.scale", |
| 147 | "shape": [ |
| 148 | 1 |
| 149 | ], |
| 150 | "dtype": "float16", |
| 151 | "offset_bytes": 41218, |
| 152 | "nbytes": 2, |
| 153 | "sha256": "f06e15ebb9473c729e94f2f809b2b6e9f35758f8f7ff9bd9a9c2393d9987b592" |
| 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": "166fa992509c0e353d4867ac9efc0aed38788ac26876bff44ff24bce0dbf741f" |
| 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": "a49603aa78874b077b12d9aa661bfb5a6f2a7fea293ff3538e598fcdc6eb0dc3" |
| 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": "c047661c803654ef6e748ca759c72d8c144cf2655f3d18df2e13877dbac90aca" |
| 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": "2b1bed22be38af3f6ba440f226ddbbec73bd3ca1536ce68dcd8411fc87cefb74" |
| 198 | }, |
| 199 | { |
| 200 | "name": "blocks.2.scale", |
| 201 | "shape": [ |
| 202 | 1 |
| 203 | ], |
| 204 | "dtype": "float16", |
| 205 | "offset_bytes": 61828, |
| 206 | "nbytes": 2, |
| 207 | "sha256": "4fa11afd6b5331ae25bfb1346bf61d7e878f2986c22292d266d22ec4ba0d9f08" |
| 208 | }, |
| 209 | { |
| 210 | "name": "embedding.weight", |
| 211 | "shape": [ |
| 212 | 123, |
| 213 | 32 |
| 214 | ], |
| 215 | "dtype": "float16", |
| 216 | "offset_bytes": 61830, |
| 217 | "nbytes": 7872, |
| 218 | "sha256": "62e8100192a185ce4f3266ed6b8967aacfa37c65c5281b3122afa59f45764742" |
| 219 | }, |
| 220 | { |
| 221 | "name": "input_proj.bias", |
| 222 | "shape": [ |
| 223 | 32 |
| 224 | ], |
| 225 | "dtype": "float16", |
| 226 | "offset_bytes": 69702, |
| 227 | "nbytes": 64, |
| 228 | "sha256": "64ebff37475e9cc917a561766bb1ad981ab52d3f877d7fcb3cac48d5778656f2" |
| 229 | }, |
| 230 | { |
| 231 | "name": "input_proj.weight", |
| 232 | "shape": [ |
| 233 | 32, |
| 234 | 35, |
| 235 | 1 |
| 236 | ], |
| 237 | "dtype": "float16", |
| 238 | "offset_bytes": 69766, |
| 239 | "nbytes": 2240, |
| 240 | "sha256": "c651fc7598918b3426aa29a442110d2a01e0bda3f8633c99c2aabd6f6278a4fb" |
| 241 | }, |
| 242 | { |
| 243 | "name": "output.bias", |
| 244 | "shape": [ |
| 245 | 1 |
| 246 | ], |
| 247 | "dtype": "float16", |
| 248 | "offset_bytes": 72006, |
| 249 | "nbytes": 2, |
| 250 | "sha256": "d693cc2578faf3ab81141851866a7c6e08e8d2b0247458af6920d35939bd142b" |
| 251 | }, |
| 252 | { |
| 253 | "name": "output.weight", |
| 254 | "shape": [ |
| 255 | 1, |
| 256 | 32, |
| 257 | 1 |
| 258 | ], |
| 259 | "dtype": "float16", |
| 260 | "offset_bytes": 72008, |
| 261 | "nbytes": 64, |
| 262 | "sha256": "2c700e485cc7e65cd328efdaa75e9048c3e2d0f12cc6217e770681c1e5b2c839" |
| 263 | } |
| 264 | ] |
| 265 | }, |
| 266 | "acoustic": { |
| 267 | "checkpoint": "artifacts/voices/tr_TR-tiny/joint/latent-student.pt", |
| 268 | "checkpoint_sha256": "de5cec8f3933b380220f9263fd6cb0153edf64aed438cce5594fd23335c86569", |
| 269 | "config": { |
| 270 | "architecture": "token_context", |
| 271 | "vocab_size": 123, |
| 272 | "hidden": 48, |
| 273 | "depth": 3, |
| 274 | "kernel_size": 5, |
| 275 | "out_channels": 192, |
| 276 | "token_depth": 3 |
| 277 | }, |
| 278 | "parameters": 159078, |
| 279 | "tensors": [ |
| 280 | { |
| 281 | "name": "embedding.weight", |
| 282 | "shape": [ |
| 283 | 123, |
| 284 | 48 |
| 285 | ], |
| 286 | "dtype": "float16", |
| 287 | "offset_bytes": 72072, |
| 288 | "nbytes": 11808, |
| 289 | "sha256": "c5f6f28a9633b1108a47f3270c718c2c132d671fc3dde17207500f644dd67a6d" |
| 290 | }, |
| 291 | { |
| 292 | "name": "frame_blocks.0.net.0.bias", |
| 293 | "shape": [ |
| 294 | 48 |
| 295 | ], |
| 296 | "dtype": "float16", |
| 297 | "offset_bytes": 83880, |
| 298 | "nbytes": 96, |
| 299 | "sha256": "13b5fdb8da5ae14971f2a32ce814d0876b3b9ce12e9926b3c21a81619c0ebf68" |
| 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": 83976, |
| 310 | "nbytes": 23040, |
| 311 | "sha256": "2c5f05df4b19aacac1cdb4a53cd3ae6ea64d54091c9896b8fbb487213fbefa87" |
| 312 | }, |
| 313 | { |
| 314 | "name": "frame_blocks.0.net.2.bias", |
| 315 | "shape": [ |
| 316 | 48 |
| 317 | ], |
| 318 | "dtype": "float16", |
| 319 | "offset_bytes": 107016, |
| 320 | "nbytes": 96, |
| 321 | "sha256": "c534465d0e7606c42c8ba60f9e8817bd1eeb8a738a65510437b7325bca128d56" |
| 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": 107112, |
| 332 | "nbytes": 23040, |
| 333 | "sha256": "a117a7244ebe18ea47052364d27a1362bc6ab98fb009027758ee68907049b5a1" |
| 334 | }, |
| 335 | { |
| 336 | "name": "frame_blocks.0.scale", |
| 337 | "shape": [ |
| 338 | 1 |
| 339 | ], |
| 340 | "dtype": "float16", |
| 341 | "offset_bytes": 130152, |
| 342 | "nbytes": 2, |
| 343 | "sha256": "7c358cc7ba794345c407ade4475096eecbfe6d06f443596710913af68452b7b8" |
| 344 | }, |
| 345 | { |
| 346 | "name": "frame_blocks.1.net.0.bias", |
| 347 | "shape": [ |
| 348 | 48 |
| 349 | ], |
| 350 | "dtype": "float16", |
| 351 | "offset_bytes": 130154, |
| 352 | "nbytes": 96, |
| 353 | "sha256": "7e380c416e5e6750379d83589c615d83ee0bfd50a19e533a9765643380ee7cbe" |
| 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": 130250, |
| 364 | "nbytes": 23040, |
| 365 | "sha256": "c0706d07c8d2c9d8d59b9dea01c83e6d54996c3dfd1b1f2ac038ba8ea5682364" |
| 366 | }, |
| 367 | { |
| 368 | "name": "frame_blocks.1.net.2.bias", |
| 369 | "shape": [ |
| 370 | 48 |
| 371 | ], |
| 372 | "dtype": "float16", |
| 373 | "offset_bytes": 153290, |
| 374 | "nbytes": 96, |
| 375 | "sha256": "c7052a6a327b64c7e73b41fd42501b44336a4c4726e0b5e70f63cb755fca011b" |
| 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": 153386, |
| 386 | "nbytes": 23040, |
| 387 | "sha256": "62fd639984e83e87e0ce0280e3cbe4af96433695c1feca75576e73bea0a71a1f" |
| 388 | }, |
| 389 | { |
| 390 | "name": "frame_blocks.1.scale", |
| 391 | "shape": [ |
| 392 | 1 |
| 393 | ], |
| 394 | "dtype": "float16", |
| 395 | "offset_bytes": 176426, |
| 396 | "nbytes": 2, |
| 397 | "sha256": "09a348f60040bcab4bbf5087dad11311035c50b540293213714fda38c8c2ee34" |
| 398 | }, |
| 399 | { |
| 400 | "name": "frame_blocks.2.net.0.bias", |
| 401 | "shape": [ |
| 402 | 48 |
| 403 | ], |
| 404 | "dtype": "float16", |
| 405 | "offset_bytes": 176428, |
| 406 | "nbytes": 96, |
| 407 | "sha256": "1e6910def7cb84fb666b9434e146681285291d336b7aa405134cdc53066d7e54" |
| 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": 176524, |
| 418 | "nbytes": 23040, |
| 419 | "sha256": "8e6248ae1a287f416a172d541a324073d106973c1262e79091ba2be6f40c20fb" |
| 420 | }, |
| 421 | { |
| 422 | "name": "frame_blocks.2.net.2.bias", |
| 423 | "shape": [ |
| 424 | 48 |
| 425 | ], |
| 426 | "dtype": "float16", |
| 427 | "offset_bytes": 199564, |
| 428 | "nbytes": 96, |
| 429 | "sha256": "fecca09a1733ad63e7f7b5f66335633c2e6682b4ce421d2fba730122b2d86856" |
| 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": 199660, |
| 440 | "nbytes": 23040, |
| 441 | "sha256": "251b0470780eee706c8bba3204e7259af78b0e13d2cbabe98fb225035aade48a" |
| 442 | }, |
| 443 | { |
| 444 | "name": "frame_blocks.2.scale", |
| 445 | "shape": [ |
| 446 | 1 |
| 447 | ], |
| 448 | "dtype": "float16", |
| 449 | "offset_bytes": 222700, |
| 450 | "nbytes": 2, |
| 451 | "sha256": "b14d4e75db3a4082e9e11b88dd8dbfe1a7485c33c4a6150bdffeee577a0965a8" |
| 452 | }, |
| 453 | { |
| 454 | "name": "frame_input_proj.bias", |
| 455 | "shape": [ |
| 456 | 48 |
| 457 | ], |
| 458 | "dtype": "float16", |
| 459 | "offset_bytes": 222702, |
| 460 | "nbytes": 96, |
| 461 | "sha256": "03831d6e0486a902fc61e04b8fef7c56522171f2ec2683855e5c7130a5e6abb4" |
| 462 | }, |
| 463 | { |
| 464 | "name": "frame_input_proj.weight", |
| 465 | "shape": [ |
| 466 | 48, |
| 467 | 51, |
| 468 | 1 |
| 469 | ], |
| 470 | "dtype": "float16", |
| 471 | "offset_bytes": 222798, |
| 472 | "nbytes": 4896, |
| 473 | "sha256": "4443ccb54e73e536f4a70a2dd0040cff36ebb7131de5f8dd665be21d57ba4dc0" |
| 474 | }, |
| 475 | { |
| 476 | "name": "output.bias", |
| 477 | "shape": [ |
| 478 | 192 |
| 479 | ], |
| 480 | "dtype": "float16", |
| 481 | "offset_bytes": 227694, |
| 482 | "nbytes": 384, |
| 483 | "sha256": "96eaeed1bbaa04d3ffdc73b55bb583fe2bb3801f1d87f4e1ee3302343a0b5cdd" |
| 484 | }, |
| 485 | { |
| 486 | "name": "output.weight", |
| 487 | "shape": [ |
| 488 | 192, |
| 489 | 48, |
| 490 | 1 |
| 491 | ], |
| 492 | "dtype": "float16", |
| 493 | "offset_bytes": 228078, |
| 494 | "nbytes": 18432, |
| 495 | "sha256": "1e088640a0449e6ff4ae5d4eecbe2ecd4506e4e1d70e3f6f51974aff383c2bc7" |
| 496 | }, |
| 497 | { |
| 498 | "name": "token_blocks.0.net.0.bias", |
| 499 | "shape": [ |
| 500 | 48 |
| 501 | ], |
| 502 | "dtype": "float16", |
| 503 | "offset_bytes": 246510, |
| 504 | "nbytes": 96, |
| 505 | "sha256": "c8b2570ce9270d7a3c05e362f43309a1324b8576b65b9bbb641a09b57ffefe29" |
| 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": 246606, |
| 516 | "nbytes": 23040, |
| 517 | "sha256": "5c8b886de85a9a3b6a3341fee22945a17e9c0d9331425696bc5c418aab824b6c" |
| 518 | }, |
| 519 | { |
| 520 | "name": "token_blocks.0.net.2.bias", |
| 521 | "shape": [ |
| 522 | 48 |
| 523 | ], |
| 524 | "dtype": "float16", |
| 525 | "offset_bytes": 269646, |
| 526 | "nbytes": 96, |
| 527 | "sha256": "e40bf2737776460b3baf4aba70bd21d18b0cadb386e1aae586fe970d0d1e04c3" |
| 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": 269742, |
| 538 | "nbytes": 23040, |
| 539 | "sha256": "982a50fd4ba6727276d5ff8823731baa2d4ed89192dc8d765b82101005a13c3f" |
| 540 | }, |
| 541 | { |
| 542 | "name": "token_blocks.0.scale", |
| 543 | "shape": [ |
| 544 | 1 |
| 545 | ], |
| 546 | "dtype": "float16", |
| 547 | "offset_bytes": 292782, |
| 548 | "nbytes": 2, |
| 549 | "sha256": "613e4a4e3ca53afc93a63c9c2bc636b38e4b2ecfeba6100980bd736f2c921f55" |
| 550 | }, |
| 551 | { |
| 552 | "name": "token_blocks.1.net.0.bias", |
| 553 | "shape": [ |
| 554 | 48 |
| 555 | ], |
| 556 | "dtype": "float16", |
| 557 | "offset_bytes": 292784, |
| 558 | "nbytes": 96, |
| 559 | "sha256": "f9c8212ba5bcc43fee8f9ebcdb663acb66bd4b37efd70dfe97174aedfc124d31" |
| 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": 292880, |
| 570 | "nbytes": 23040, |
| 571 | "sha256": "9dbf0c2e45805a015139bf67d5767f5e54ae0ca83fc410c4b4e381721efdcd66" |
| 572 | }, |
| 573 | { |
| 574 | "name": "token_blocks.1.net.2.bias", |
| 575 | "shape": [ |
| 576 | 48 |
| 577 | ], |
| 578 | "dtype": "float16", |
| 579 | "offset_bytes": 315920, |
| 580 | "nbytes": 96, |
| 581 | "sha256": "7f5404c31b193e223e9c8901de65809c73d8f79241d801c38e82151068fff4c6" |
| 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": 316016, |
| 592 | "nbytes": 23040, |
| 593 | "sha256": "b22931e53a9514c8e4a3aceaa056ee51ae9aed486d0bc45797df4cf1edf02b29" |
| 594 | }, |
| 595 | { |
| 596 | "name": "token_blocks.1.scale", |
| 597 | "shape": [ |
| 598 | 1 |
| 599 | ], |
| 600 | "dtype": "float16", |
| 601 | "offset_bytes": 339056, |
| 602 | "nbytes": 2, |
| 603 | "sha256": "44c808fd166dbb8961fb795587e5ee0082f8a2dc781ff06a3f02223dcc06a7da" |
| 604 | }, |
| 605 | { |
| 606 | "name": "token_blocks.2.net.0.bias", |
| 607 | "shape": [ |
| 608 | 48 |
| 609 | ], |
| 610 | "dtype": "float16", |
| 611 | "offset_bytes": 339058, |
| 612 | "nbytes": 96, |
| 613 | "sha256": "d0720fd6480496505960f75ba2c4026241897c5cf95679a7a8c56d54e8d64c35" |
| 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": 339154, |
| 624 | "nbytes": 23040, |
| 625 | "sha256": "919bcaff9aaa4b6ec46790f508f2e75f76436e4a9a10a7e3a35a075cf4b9ef44" |
| 626 | }, |
| 627 | { |
| 628 | "name": "token_blocks.2.net.2.bias", |
| 629 | "shape": [ |
| 630 | 48 |
| 631 | ], |
| 632 | "dtype": "float16", |
| 633 | "offset_bytes": 362194, |
| 634 | "nbytes": 96, |
| 635 | "sha256": "66cba93dade203d3f65b9997ffbcf54a0886593c87de85556a95f78432c85971" |
| 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": 362290, |
| 646 | "nbytes": 23040, |
| 647 | "sha256": "6854b9055957950f38c84fbf22c1968d4358b21eecd2d5c7eb4c2b94b07c3953" |
| 648 | }, |
| 649 | { |
| 650 | "name": "token_blocks.2.scale", |
| 651 | "shape": [ |
| 652 | 1 |
| 653 | ], |
| 654 | "dtype": "float16", |
| 655 | "offset_bytes": 385330, |
| 656 | "nbytes": 2, |
| 657 | "sha256": "641b8a45edb49f782f6596103cbb74058b8a77a9cf8aa1bf39c005573e21fd5c" |
| 658 | }, |
| 659 | { |
| 660 | "name": "token_input_proj.bias", |
| 661 | "shape": [ |
| 662 | 48 |
| 663 | ], |
| 664 | "dtype": "float16", |
| 665 | "offset_bytes": 385332, |
| 666 | "nbytes": 96, |
| 667 | "sha256": "03e6660169ee870edce85ca4a7974a48e6efcd7505ef6fda37382a36edc994c7" |
| 668 | }, |
| 669 | { |
| 670 | "name": "token_input_proj.weight", |
| 671 | "shape": [ |
| 672 | 48, |
| 673 | 50, |
| 674 | 1 |
| 675 | ], |
| 676 | "dtype": "float16", |
| 677 | "offset_bytes": 385428, |
| 678 | "nbytes": 4800, |
| 679 | "sha256": "fc4d7091b5e2227d61ec74ba0c347ecf397e05740a13661421767c0be3ca342e" |
| 680 | } |
| 681 | ] |
| 682 | }, |
| 683 | "decoder": { |
| 684 | "checkpoint": "artifacts/voices/tr_TR-tiny/joint/decoder-student.pt", |
| 685 | "checkpoint_sha256": "6f69e5c35596ae99b4acf27a90665d0fcadc5977b728bff436ca16e228b1da2d", |
| 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/tr_TR-tiny/decoder/decoder-student.pt", |
| 758 | "decoder_init": { |
| 759 | "checkpoint": "artifacts/voices/tr_TR-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/tr_TR-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": 390228, |
| 906 | "nbytes": 2, |
| 907 | "sha256": "8c14d1cd7cfa8ad23622959e3acfb2d75da7f6e802cdfd7743569cae21365e9a" |
| 908 | }, |
| 909 | { |
| 910 | "name": "post.weight", |
| 911 | "shape": [ |
| 912 | 1, |
| 913 | 12, |
| 914 | 7 |
| 915 | ], |
| 916 | "dtype": "float16", |
| 917 | "offset_bytes": 390230, |
| 918 | "nbytes": 168, |
| 919 | "sha256": "f1691864ec542475ed12cc81e234a113f6e3b22de93f0792e43d1622065d67ff" |
| 920 | }, |
| 921 | { |
| 922 | "name": "pre.bias", |
| 923 | "shape": [ |
| 924 | 96 |
| 925 | ], |
| 926 | "dtype": "float16", |
| 927 | "offset_bytes": 390398, |
| 928 | "nbytes": 192, |
| 929 | "sha256": "44a17063b6545217a7d19445ff2e75004ad2f0156ce0dfcdc594cf6218e6b40e" |
| 930 | }, |
| 931 | { |
| 932 | "name": "pre.weight", |
| 933 | "shape": [ |
| 934 | 96, |
| 935 | 192, |
| 936 | 7 |
| 937 | ], |
| 938 | "dtype": "float16", |
| 939 | "offset_bytes": 390590, |
| 940 | "nbytes": 258048, |
| 941 | "sha256": "eb7945730f94965147c8945940db85f813d421610f922edd865a2d147e1f4f69" |
| 942 | }, |
| 943 | { |
| 944 | "name": "res0.0.blocks.0.conv1.bias", |
| 945 | "shape": [ |
| 946 | 48 |
| 947 | ], |
| 948 | "dtype": "float16", |
| 949 | "offset_bytes": 648638, |
| 950 | "nbytes": 96, |
| 951 | "sha256": "97cf532e074c5f71570a3b604fc20cab6680d82256f7cebe31939453cefbdc82" |
| 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": 648734, |
| 962 | "nbytes": 13824, |
| 963 | "sha256": "88177c35f891d97b3eb5cec62710ecc1788de9aec7cbfd1acbdd17d558854a52" |
| 964 | }, |
| 965 | { |
| 966 | "name": "res0.0.blocks.0.conv2.bias", |
| 967 | "shape": [ |
| 968 | 48 |
| 969 | ], |
| 970 | "dtype": "float16", |
| 971 | "offset_bytes": 662558, |
| 972 | "nbytes": 96, |
| 973 | "sha256": "ec30af433565a97fd84007c863bf3ed77d098b2c17f5101152c0b876b01f32f9" |
| 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": 662654, |
| 984 | "nbytes": 13824, |
| 985 | "sha256": "c03f23bd46afa8b6bba17188cefd4f7a8dc088736fb1f9e85ced3de26ae5fd68" |
| 986 | }, |
| 987 | { |
| 988 | "name": "res0.0.blocks.1.conv1.bias", |
| 989 | "shape": [ |
| 990 | 48 |
| 991 | ], |
| 992 | "dtype": "float16", |
| 993 | "offset_bytes": 676478, |
| 994 | "nbytes": 96, |
| 995 | "sha256": "52bc78aa4e296ba039e98b922f047f27ff51ad1eb114a937552ce03db233e580" |
| 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": 676574, |
| 1006 | "nbytes": 23040, |
| 1007 | "sha256": "f81b06437595fabc9d4eab7045b6ebfad1b0ac287449d1260c1e864b30559742" |
| 1008 | }, |
| 1009 | { |
| 1010 | "name": "res0.0.blocks.1.conv2.bias", |
| 1011 | "shape": [ |
| 1012 | 48 |
| 1013 | ], |
| 1014 | "dtype": "float16", |
| 1015 | "offset_bytes": 699614, |
| 1016 | "nbytes": 96, |
| 1017 | "sha256": "f7c8371114d4d597483adcdd952ea07c1d9e09d9f8449b285a7cb349f3956489" |
| 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": 699710, |
| 1028 | "nbytes": 23040, |
| 1029 | "sha256": "345dcedb34a0d3ece9e65e398b2fdf827adbbed01d3d52d0e7dc2cbfa4822856" |
| 1030 | }, |
| 1031 | { |
| 1032 | "name": "res0.0.blocks.2.conv1.bias", |
| 1033 | "shape": [ |
| 1034 | 48 |
| 1035 | ], |
| 1036 | "dtype": "float16", |
| 1037 | "offset_bytes": 722750, |
| 1038 | "nbytes": 96, |
| 1039 | "sha256": "5267466869ad14253d444a3fa2f5b89042c8dce983f59b112d70a4290272f7f7" |
| 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": 722846, |
| 1050 | "nbytes": 32256, |
| 1051 | "sha256": "60eddbfde7d700768b37d3ff880a6c437e57b90d80f84f3d08cebaae301e2d0d" |
| 1052 | }, |
| 1053 | { |
| 1054 | "name": "res0.0.blocks.2.conv2.bias", |
| 1055 | "shape": [ |
| 1056 | 48 |
| 1057 | ], |
| 1058 | "dtype": "float16", |
| 1059 | "offset_bytes": 755102, |
| 1060 | "nbytes": 96, |
| 1061 | "sha256": "3059cc1ef80bdd04f9d8a960c917d48d064a80b364cae3a028113a35e2b36dd3" |
| 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": 755198, |
| 1072 | "nbytes": 32256, |
| 1073 | "sha256": "4a15a0a966352384e5bd231aea0eeb8788f394f3ce0123bae032edd3c0d569d7" |
| 1074 | }, |
| 1075 | { |
| 1076 | "name": "res1.0.blocks.0.conv1.bias", |
| 1077 | "shape": [ |
| 1078 | 24 |
| 1079 | ], |
| 1080 | "dtype": "float16", |
| 1081 | "offset_bytes": 787454, |
| 1082 | "nbytes": 48, |
| 1083 | "sha256": "4b5f7709e943d4994fdd09fe215a5180d65d41d8489fbc4268dc0a17d5d754d2" |
| 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": 787502, |
| 1094 | "nbytes": 3456, |
| 1095 | "sha256": "7e23e13fffb25c688ec896f186353025a3916e86a8db87d3e2b1dbb204a45f4e" |
| 1096 | }, |
| 1097 | { |
| 1098 | "name": "res1.0.blocks.0.conv2.bias", |
| 1099 | "shape": [ |
| 1100 | 24 |
| 1101 | ], |
| 1102 | "dtype": "float16", |
| 1103 | "offset_bytes": 790958, |
| 1104 | "nbytes": 48, |
| 1105 | "sha256": "a3b8cd54f9f7ef83a7d4909c89343d67b916b8bbcfd1e51d4d545b81cd4dd9d0" |
| 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": 791006, |
| 1116 | "nbytes": 3456, |
| 1117 | "sha256": "1d906ec8bc07ca8449e7a3acc893cff01dde5221fc0bbe78bfa0b3d08424628d" |
| 1118 | }, |
| 1119 | { |
| 1120 | "name": "res1.0.blocks.1.conv1.bias", |
| 1121 | "shape": [ |
| 1122 | 24 |
| 1123 | ], |
| 1124 | "dtype": "float16", |
| 1125 | "offset_bytes": 794462, |
| 1126 | "nbytes": 48, |
| 1127 | "sha256": "b2d74b63f4bb012a5eebe768c7689fc7b7cff49f20b671a7ce93bd053c8abfb4" |
| 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": 794510, |
| 1138 | "nbytes": 5760, |
| 1139 | "sha256": "7b66aef1dbd4e23cab0ad976156a1db386e291e202043aaaba0b8838bb4c4b38" |
| 1140 | }, |
| 1141 | { |
| 1142 | "name": "res1.0.blocks.1.conv2.bias", |
| 1143 | "shape": [ |
| 1144 | 24 |
| 1145 | ], |
| 1146 | "dtype": "float16", |
| 1147 | "offset_bytes": 800270, |
| 1148 | "nbytes": 48, |
| 1149 | "sha256": "d6da51bce3d7fd6f741aee1fcb1b00a6f661aeacba24d30913e151509da3c0a4" |
| 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": 800318, |
| 1160 | "nbytes": 5760, |
| 1161 | "sha256": "6621d293f57546724113adfbee37c14cfa2dc9009a2881652ef177cbed2b4a72" |
| 1162 | }, |
| 1163 | { |
| 1164 | "name": "res1.0.blocks.2.conv1.bias", |
| 1165 | "shape": [ |
| 1166 | 24 |
| 1167 | ], |
| 1168 | "dtype": "float16", |
| 1169 | "offset_bytes": 806078, |
| 1170 | "nbytes": 48, |
| 1171 | "sha256": "58891748e3ed716fe8a14e3b0a9ab2a925241ef8cce21815cb2fbed245105727" |
| 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": 806126, |
| 1182 | "nbytes": 8064, |
| 1183 | "sha256": "e8ac84b1be9ecfac2a117fc8c86b195cae5ba68fc69c58ad68138ce6ad4bd844" |
| 1184 | }, |
| 1185 | { |
| 1186 | "name": "res1.0.blocks.2.conv2.bias", |
| 1187 | "shape": [ |
| 1188 | 24 |
| 1189 | ], |
| 1190 | "dtype": "float16", |
| 1191 | "offset_bytes": 814190, |
| 1192 | "nbytes": 48, |
| 1193 | "sha256": "2f606be706ffc06c6675eac1b8db5d5b35c15d31422e7f7ecba798cb79d92ce9" |
| 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": 814238, |
| 1204 | "nbytes": 8064, |
| 1205 | "sha256": "a12194f058a96d57989ae799dea307bc1df410f089e4399b42afe350582821d6" |
| 1206 | }, |
| 1207 | { |
| 1208 | "name": "res2.0.blocks.0.conv1.bias", |
| 1209 | "shape": [ |
| 1210 | 12 |
| 1211 | ], |
| 1212 | "dtype": "float16", |
| 1213 | "offset_bytes": 822302, |
| 1214 | "nbytes": 24, |
| 1215 | "sha256": "9856280eabb8298344bf035ee2a276f60d29907365d12a2fa473d81971798835" |
| 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": 822326, |
| 1226 | "nbytes": 864, |
| 1227 | "sha256": "ad209752f6a5cb32bad8222027739161cb436c81037fdf62e64e17e85c083e57" |
| 1228 | }, |
| 1229 | { |
| 1230 | "name": "res2.0.blocks.0.conv2.bias", |
| 1231 | "shape": [ |
| 1232 | 12 |
| 1233 | ], |
| 1234 | "dtype": "float16", |
| 1235 | "offset_bytes": 823190, |
| 1236 | "nbytes": 24, |
| 1237 | "sha256": "d317fbfe9425324d79874cae8fb2763aea9e3d2d2f222ecb886b553e5b60147a" |
| 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": 823214, |
| 1248 | "nbytes": 864, |
| 1249 | "sha256": "1b506a117d3e173c9169ccc4a5373841ed1a55cab8d8ae9ee97ae3b25c52f69e" |
| 1250 | }, |
| 1251 | { |
| 1252 | "name": "res2.0.blocks.1.conv1.bias", |
| 1253 | "shape": [ |
| 1254 | 12 |
| 1255 | ], |
| 1256 | "dtype": "float16", |
| 1257 | "offset_bytes": 824078, |
| 1258 | "nbytes": 24, |
| 1259 | "sha256": "d7477fa2467dfceaf691104a5991b40aa8a7498454fee518cf194a257abb7b87" |
| 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": 824102, |
| 1270 | "nbytes": 1440, |
| 1271 | "sha256": "e7376f66778562e4eb9ab4fe9de01e85dac38d85329b5a6f5a01550799ea5a67" |
| 1272 | }, |
| 1273 | { |
| 1274 | "name": "res2.0.blocks.1.conv2.bias", |
| 1275 | "shape": [ |
| 1276 | 12 |
| 1277 | ], |
| 1278 | "dtype": "float16", |
| 1279 | "offset_bytes": 825542, |
| 1280 | "nbytes": 24, |
| 1281 | "sha256": "f27ac380c53732c2cd578f0f377e371266e07261ae3dd3d2f4da3a0b3855f66f" |
| 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": 825566, |
| 1292 | "nbytes": 1440, |
| 1293 | "sha256": "4e9db4b03358adb30a5e19b476dcad40deacbeab453164ed554353c6477681d5" |
| 1294 | }, |
| 1295 | { |
| 1296 | "name": "res2.0.blocks.2.conv1.bias", |
| 1297 | "shape": [ |
| 1298 | 12 |
| 1299 | ], |
| 1300 | "dtype": "float16", |
| 1301 | "offset_bytes": 827006, |
| 1302 | "nbytes": 24, |
| 1303 | "sha256": "429898d0b0ce8ab36f532281bf64f486d8c4753efda84303c4b3cd89c3b2464d" |
| 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": 827030, |
| 1314 | "nbytes": 2016, |
| 1315 | "sha256": "8195c356df443fa9bd3748b52366c890c3454097ac77daf22c73bb9a92f65e7a" |
| 1316 | }, |
| 1317 | { |
| 1318 | "name": "res2.0.blocks.2.conv2.bias", |
| 1319 | "shape": [ |
| 1320 | 12 |
| 1321 | ], |
| 1322 | "dtype": "float16", |
| 1323 | "offset_bytes": 829046, |
| 1324 | "nbytes": 24, |
| 1325 | "sha256": "44ba0efcf9910e0584fe33be7f1a0e991ef5655db257684b5061421ececf1f53" |
| 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": 829070, |
| 1336 | "nbytes": 2016, |
| 1337 | "sha256": "9ed41b4425ebef68ef5f873d1560fdbe18edb99f6a37d5c5ca8d4ce302a54a33" |
| 1338 | }, |
| 1339 | { |
| 1340 | "name": "up0.bias", |
| 1341 | "shape": [ |
| 1342 | 48 |
| 1343 | ], |
| 1344 | "dtype": "float16", |
| 1345 | "offset_bytes": 831086, |
| 1346 | "nbytes": 96, |
| 1347 | "sha256": "530f6ee092177e7387186518cdeee5bf409f8f6ae875c56b90f5dab2ce683c6c" |
| 1348 | }, |
| 1349 | { |
| 1350 | "name": "up0.weight", |
| 1351 | "shape": [ |
| 1352 | 96, |
| 1353 | 48, |
| 1354 | 16 |
| 1355 | ], |
| 1356 | "dtype": "float16", |
| 1357 | "offset_bytes": 831182, |
| 1358 | "nbytes": 147456, |
| 1359 | "sha256": "a31c9f9a41f1f0768fde27da4544995055ee57e8afa94d2fa0baa72079d2886d" |
| 1360 | }, |
| 1361 | { |
| 1362 | "name": "up1.bias", |
| 1363 | "shape": [ |
| 1364 | 24 |
| 1365 | ], |
| 1366 | "dtype": "float16", |
| 1367 | "offset_bytes": 978638, |
| 1368 | "nbytes": 48, |
| 1369 | "sha256": "071a03f958ecebc57c894351070dcf5913bc3072015471cca4c8a44397e62d49" |
| 1370 | }, |
| 1371 | { |
| 1372 | "name": "up1.weight", |
| 1373 | "shape": [ |
| 1374 | 48, |
| 1375 | 24, |
| 1376 | 16 |
| 1377 | ], |
| 1378 | "dtype": "float16", |
| 1379 | "offset_bytes": 978686, |
| 1380 | "nbytes": 36864, |
| 1381 | "sha256": "cba94e0bee2a416add4ce03dd738d8dbfdabc1a250fc28a1ec7f7c6b0142f09d" |
| 1382 | }, |
| 1383 | { |
| 1384 | "name": "up2.bias", |
| 1385 | "shape": [ |
| 1386 | 12 |
| 1387 | ], |
| 1388 | "dtype": "float16", |
| 1389 | "offset_bytes": 1015550, |
| 1390 | "nbytes": 24, |
| 1391 | "sha256": "bf968b9cdc59c5e8e62fa180f7e101438d1f592a26c0c7240b09c9d05df37446" |
| 1392 | }, |
| 1393 | { |
| 1394 | "name": "up2.weight", |
| 1395 | "shape": [ |
| 1396 | 24, |
| 1397 | 12, |
| 1398 | 8 |
| 1399 | ], |
| 1400 | "dtype": "float16", |
| 1401 | "offset_bytes": 1015574, |
| 1402 | "nbytes": 4608, |
| 1403 | "sha256": "6499a56ee4f2360e7f91e35191a1387e0b8ee3da7a3e49e46d9b22af8cfa2b7f" |
| 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": "tr", |
| 1434 | "phoneme_id_map_entries": 154, |
| 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": 1025142, |
| 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 | |