File size: 182,136 Bytes
a144c29 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 | program(1.0)
[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}, {"coremltools-component-torch", "2.5.1"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "9.0"}})]
{
func main<ios17>(tensor<fp32, [1, 960000]> waveform) {
tensor<int32, [1]> input_1_axes_0 = const()[name = tensor<string, []>("input_1_axes_0"), val = tensor<int32, [1]>([1])];
tensor<string, []> waveform_to_fp16_dtype_0 = const()[name = tensor<string, []>("waveform_to_fp16_dtype_0"), val = tensor<string, []>("fp16")];
tensor<fp16, [1, 960000]> waveform_to_fp16 = cast(dtype = waveform_to_fp16_dtype_0, x = waveform)[name = tensor<string, []>("cast_49")];
tensor<fp16, [1, 1, 960000]> input_1_cast_fp16 = expand_dims(axes = input_1_axes_0, x = waveform_to_fp16)[name = tensor<string, []>("input_1_cast_fp16")];
tensor<string, []> input_3_pad_type_0 = const()[name = tensor<string, []>("input_3_pad_type_0"), val = tensor<string, []>("valid")];
tensor<int32, [1]> input_3_strides_0 = const()[name = tensor<string, []>("input_3_strides_0"), val = tensor<int32, [1]>([5])];
tensor<int32, [2]> input_3_pad_0 = const()[name = tensor<string, []>("input_3_pad_0"), val = tensor<int32, [2]>([0, 0])];
tensor<int32, [1]> input_3_dilations_0 = const()[name = tensor<string, []>("input_3_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<int32, []> input_3_groups_0 = const()[name = tensor<string, []>("input_3_groups_0"), val = tensor<int32, []>(1)];
tensor<fp16, [512, 1, 10]> model_feature_extractor_conv_layers_0_conv_weight_to_fp16 = const()[name = tensor<string, []>("model_feature_extractor_conv_layers_0_conv_weight_to_fp16"), val = tensor<fp16, [512, 1, 10]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
tensor<fp16, [1, 512, 191999]> input_3_cast_fp16 = conv(dilations = input_3_dilations_0, groups = input_3_groups_0, pad = input_3_pad_0, pad_type = input_3_pad_type_0, strides = input_3_strides_0, weight = model_feature_extractor_conv_layers_0_conv_weight_to_fp16, x = input_1_cast_fp16)[name = tensor<string, []>("input_3_cast_fp16")];
tensor<int32, [4]> reshape_0_shape_0 = const()[name = tensor<string, []>("reshape_0_shape_0"), val = tensor<int32, [4]>([1, 512, 1, 191999])];
tensor<fp16, [1, 512, 1, 191999]> reshape_0_cast_fp16 = reshape(shape = reshape_0_shape_0, x = input_3_cast_fp16)[name = tensor<string, []>("reshape_0_cast_fp16")];
tensor<int32, [2]> reduce_mean_0_axes_0 = const()[name = tensor<string, []>("reduce_mean_0_axes_0"), val = tensor<int32, [2]>([2, 3])];
tensor<bool, []> reduce_mean_0_keep_dims_0 = const()[name = tensor<string, []>("reduce_mean_0_keep_dims_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 512, 1, 1]> reduce_mean_0_cast_fp16 = reduce_mean(axes = reduce_mean_0_axes_0, keep_dims = reduce_mean_0_keep_dims_0, x = reshape_0_cast_fp16)[name = tensor<string, []>("reduce_mean_0_cast_fp16")];
tensor<fp16, [1, 512, 1, 191999]> sub_0_cast_fp16 = sub(x = reshape_0_cast_fp16, y = reduce_mean_0_cast_fp16)[name = tensor<string, []>("sub_0_cast_fp16")];
tensor<fp16, [1, 512, 1, 191999]> square_0_cast_fp16 = square(x = sub_0_cast_fp16)[name = tensor<string, []>("square_0_cast_fp16")];
tensor<int32, [2]> reduce_mean_2_axes_0 = const()[name = tensor<string, []>("reduce_mean_2_axes_0"), val = tensor<int32, [2]>([2, 3])];
tensor<bool, []> reduce_mean_2_keep_dims_0 = const()[name = tensor<string, []>("reduce_mean_2_keep_dims_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 512, 1, 1]> reduce_mean_2_cast_fp16 = reduce_mean(axes = reduce_mean_2_axes_0, keep_dims = reduce_mean_2_keep_dims_0, x = square_0_cast_fp16)[name = tensor<string, []>("reduce_mean_2_cast_fp16")];
tensor<fp16, []> add_0_y_0_to_fp16 = const()[name = tensor<string, []>("add_0_y_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 512, 1, 1]> add_0_cast_fp16 = add(x = reduce_mean_2_cast_fp16, y = add_0_y_0_to_fp16)[name = tensor<string, []>("add_0_cast_fp16")];
tensor<fp16, [1, 512, 1, 1]> sqrt_0_cast_fp16 = sqrt(x = add_0_cast_fp16)[name = tensor<string, []>("sqrt_0_cast_fp16")];
tensor<fp16, [1, 512, 1, 191999]> real_div_0_cast_fp16 = real_div(x = sub_0_cast_fp16, y = sqrt_0_cast_fp16)[name = tensor<string, []>("real_div_0_cast_fp16")];
tensor<int32, [3]> reshape_1_shape_0 = const()[name = tensor<string, []>("reshape_1_shape_0"), val = tensor<int32, [3]>([1, 512, 191999])];
tensor<fp16, [1, 512, 191999]> reshape_1_cast_fp16 = reshape(shape = reshape_1_shape_0, x = real_div_0_cast_fp16)[name = tensor<string, []>("reshape_1_cast_fp16")];
tensor<fp16, [1, 512, 1]> reshape_2_to_fp16 = const()[name = tensor<string, []>("reshape_2_to_fp16"), val = tensor<fp16, [1, 512, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(10368)))];
tensor<fp16, [1, 512, 191999]> mul_0_cast_fp16 = mul(x = reshape_1_cast_fp16, y = reshape_2_to_fp16)[name = tensor<string, []>("mul_0_cast_fp16")];
tensor<fp16, [1, 512, 1]> reshape_3_to_fp16 = const()[name = tensor<string, []>("reshape_3_to_fp16"), val = tensor<fp16, [1, 512, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(11456)))];
tensor<fp16, [1, 512, 191999]> add_1_cast_fp16 = add(x = mul_0_cast_fp16, y = reshape_3_to_fp16)[name = tensor<string, []>("add_1_cast_fp16")];
tensor<string, []> input_5_mode_0 = const()[name = tensor<string, []>("input_5_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 512, 191999]> input_5_cast_fp16 = gelu(mode = input_5_mode_0, x = add_1_cast_fp16)[name = tensor<string, []>("input_5_cast_fp16")];
tensor<string, []> x_3_pad_type_0 = const()[name = tensor<string, []>("x_3_pad_type_0"), val = tensor<string, []>("valid")];
tensor<int32, [1]> x_3_strides_0 = const()[name = tensor<string, []>("x_3_strides_0"), val = tensor<int32, [1]>([2])];
tensor<int32, [2]> x_3_pad_0 = const()[name = tensor<string, []>("x_3_pad_0"), val = tensor<int32, [2]>([0, 0])];
tensor<int32, [1]> x_3_dilations_0 = const()[name = tensor<string, []>("x_3_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<int32, []> x_3_groups_0 = const()[name = tensor<string, []>("x_3_groups_0"), val = tensor<int32, []>(1)];
tensor<fp16, [512, 512, 3]> model_feature_extractor_conv_layers_1_conv_weight_to_fp16 = const()[name = tensor<string, []>("model_feature_extractor_conv_layers_1_conv_weight_to_fp16"), val = tensor<fp16, [512, 512, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(12544)))];
tensor<fp16, [1, 512, 95999]> x_3_cast_fp16 = conv(dilations = x_3_dilations_0, groups = x_3_groups_0, pad = x_3_pad_0, pad_type = x_3_pad_type_0, strides = x_3_strides_0, weight = model_feature_extractor_conv_layers_1_conv_weight_to_fp16, x = input_5_cast_fp16)[name = tensor<string, []>("x_3_cast_fp16")];
tensor<string, []> input_7_mode_0 = const()[name = tensor<string, []>("input_7_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 512, 95999]> input_7_cast_fp16 = gelu(mode = input_7_mode_0, x = x_3_cast_fp16)[name = tensor<string, []>("input_7_cast_fp16")];
tensor<string, []> x_5_pad_type_0 = const()[name = tensor<string, []>("x_5_pad_type_0"), val = tensor<string, []>("valid")];
tensor<int32, [1]> x_5_strides_0 = const()[name = tensor<string, []>("x_5_strides_0"), val = tensor<int32, [1]>([2])];
tensor<int32, [2]> x_5_pad_0 = const()[name = tensor<string, []>("x_5_pad_0"), val = tensor<int32, [2]>([0, 0])];
tensor<int32, [1]> x_5_dilations_0 = const()[name = tensor<string, []>("x_5_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<int32, []> x_5_groups_0 = const()[name = tensor<string, []>("x_5_groups_0"), val = tensor<int32, []>(1)];
tensor<fp16, [512, 512, 3]> model_feature_extractor_conv_layers_2_conv_weight_to_fp16 = const()[name = tensor<string, []>("model_feature_extractor_conv_layers_2_conv_weight_to_fp16"), val = tensor<fp16, [512, 512, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(1585472)))];
tensor<fp16, [1, 512, 47999]> x_5_cast_fp16 = conv(dilations = x_5_dilations_0, groups = x_5_groups_0, pad = x_5_pad_0, pad_type = x_5_pad_type_0, strides = x_5_strides_0, weight = model_feature_extractor_conv_layers_2_conv_weight_to_fp16, x = input_7_cast_fp16)[name = tensor<string, []>("x_5_cast_fp16")];
tensor<string, []> input_9_mode_0 = const()[name = tensor<string, []>("input_9_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 512, 47999]> input_9_cast_fp16 = gelu(mode = input_9_mode_0, x = x_5_cast_fp16)[name = tensor<string, []>("input_9_cast_fp16")];
tensor<string, []> x_7_pad_type_0 = const()[name = tensor<string, []>("x_7_pad_type_0"), val = tensor<string, []>("valid")];
tensor<int32, [1]> x_7_strides_0 = const()[name = tensor<string, []>("x_7_strides_0"), val = tensor<int32, [1]>([2])];
tensor<int32, [2]> x_7_pad_0 = const()[name = tensor<string, []>("x_7_pad_0"), val = tensor<int32, [2]>([0, 0])];
tensor<int32, [1]> x_7_dilations_0 = const()[name = tensor<string, []>("x_7_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<int32, []> x_7_groups_0 = const()[name = tensor<string, []>("x_7_groups_0"), val = tensor<int32, []>(1)];
tensor<fp16, [512, 512, 3]> model_feature_extractor_conv_layers_3_conv_weight_to_fp16 = const()[name = tensor<string, []>("model_feature_extractor_conv_layers_3_conv_weight_to_fp16"), val = tensor<fp16, [512, 512, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(3158400)))];
tensor<fp16, [1, 512, 23999]> x_7_cast_fp16 = conv(dilations = x_7_dilations_0, groups = x_7_groups_0, pad = x_7_pad_0, pad_type = x_7_pad_type_0, strides = x_7_strides_0, weight = model_feature_extractor_conv_layers_3_conv_weight_to_fp16, x = input_9_cast_fp16)[name = tensor<string, []>("x_7_cast_fp16")];
tensor<string, []> input_11_mode_0 = const()[name = tensor<string, []>("input_11_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 512, 23999]> input_11_cast_fp16 = gelu(mode = input_11_mode_0, x = x_7_cast_fp16)[name = tensor<string, []>("input_11_cast_fp16")];
tensor<string, []> x_9_pad_type_0 = const()[name = tensor<string, []>("x_9_pad_type_0"), val = tensor<string, []>("valid")];
tensor<int32, [1]> x_9_strides_0 = const()[name = tensor<string, []>("x_9_strides_0"), val = tensor<int32, [1]>([2])];
tensor<int32, [2]> x_9_pad_0 = const()[name = tensor<string, []>("x_9_pad_0"), val = tensor<int32, [2]>([0, 0])];
tensor<int32, [1]> x_9_dilations_0 = const()[name = tensor<string, []>("x_9_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<int32, []> x_9_groups_0 = const()[name = tensor<string, []>("x_9_groups_0"), val = tensor<int32, []>(1)];
tensor<fp16, [512, 512, 3]> model_feature_extractor_conv_layers_4_conv_weight_to_fp16 = const()[name = tensor<string, []>("model_feature_extractor_conv_layers_4_conv_weight_to_fp16"), val = tensor<fp16, [512, 512, 3]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(4731328)))];
tensor<fp16, [1, 512, 11999]> x_9_cast_fp16 = conv(dilations = x_9_dilations_0, groups = x_9_groups_0, pad = x_9_pad_0, pad_type = x_9_pad_type_0, strides = x_9_strides_0, weight = model_feature_extractor_conv_layers_4_conv_weight_to_fp16, x = input_11_cast_fp16)[name = tensor<string, []>("x_9_cast_fp16")];
tensor<string, []> input_13_mode_0 = const()[name = tensor<string, []>("input_13_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 512, 11999]> input_13_cast_fp16 = gelu(mode = input_13_mode_0, x = x_9_cast_fp16)[name = tensor<string, []>("input_13_cast_fp16")];
tensor<string, []> x_11_pad_type_0 = const()[name = tensor<string, []>("x_11_pad_type_0"), val = tensor<string, []>("valid")];
tensor<int32, [1]> x_11_strides_0 = const()[name = tensor<string, []>("x_11_strides_0"), val = tensor<int32, [1]>([2])];
tensor<int32, [2]> x_11_pad_0 = const()[name = tensor<string, []>("x_11_pad_0"), val = tensor<int32, [2]>([0, 0])];
tensor<int32, [1]> x_11_dilations_0 = const()[name = tensor<string, []>("x_11_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<int32, []> x_11_groups_0 = const()[name = tensor<string, []>("x_11_groups_0"), val = tensor<int32, []>(1)];
tensor<fp16, [512, 512, 2]> model_feature_extractor_conv_layers_5_conv_weight_to_fp16 = const()[name = tensor<string, []>("model_feature_extractor_conv_layers_5_conv_weight_to_fp16"), val = tensor<fp16, [512, 512, 2]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(6304256)))];
tensor<fp16, [1, 512, 5999]> x_11_cast_fp16 = conv(dilations = x_11_dilations_0, groups = x_11_groups_0, pad = x_11_pad_0, pad_type = x_11_pad_type_0, strides = x_11_strides_0, weight = model_feature_extractor_conv_layers_5_conv_weight_to_fp16, x = input_13_cast_fp16)[name = tensor<string, []>("x_11_cast_fp16")];
tensor<string, []> input_15_mode_0 = const()[name = tensor<string, []>("input_15_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 512, 5999]> input_15_cast_fp16 = gelu(mode = input_15_mode_0, x = x_11_cast_fp16)[name = tensor<string, []>("input_15_cast_fp16")];
tensor<string, []> x_13_pad_type_0 = const()[name = tensor<string, []>("x_13_pad_type_0"), val = tensor<string, []>("valid")];
tensor<int32, [1]> x_13_strides_0 = const()[name = tensor<string, []>("x_13_strides_0"), val = tensor<int32, [1]>([2])];
tensor<int32, [2]> x_13_pad_0 = const()[name = tensor<string, []>("x_13_pad_0"), val = tensor<int32, [2]>([0, 0])];
tensor<int32, [1]> x_13_dilations_0 = const()[name = tensor<string, []>("x_13_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<int32, []> x_13_groups_0 = const()[name = tensor<string, []>("x_13_groups_0"), val = tensor<int32, []>(1)];
tensor<fp16, [512, 512, 2]> model_feature_extractor_conv_layers_6_conv_weight_to_fp16 = const()[name = tensor<string, []>("model_feature_extractor_conv_layers_6_conv_weight_to_fp16"), val = tensor<fp16, [512, 512, 2]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(7352896)))];
tensor<fp16, [1, 512, 2999]> x_13_cast_fp16 = conv(dilations = x_13_dilations_0, groups = x_13_groups_0, pad = x_13_pad_0, pad_type = x_13_pad_type_0, strides = x_13_strides_0, weight = model_feature_extractor_conv_layers_6_conv_weight_to_fp16, x = input_15_cast_fp16)[name = tensor<string, []>("x_13_cast_fp16")];
tensor<string, []> x_15_mode_0 = const()[name = tensor<string, []>("x_15_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 512, 2999]> x_15_cast_fp16 = gelu(mode = x_15_mode_0, x = x_13_cast_fp16)[name = tensor<string, []>("x_15_cast_fp16")];
tensor<int32, [3]> input_17_perm_0 = const()[name = tensor<string, []>("input_17_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
tensor<int32, [1]> input_19_axes_0 = const()[name = tensor<string, []>("input_19_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [512]> model_encoder_feature_projection_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_feature_projection_layer_norm_weight_to_fp16"), val = tensor<fp16, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8401536)))];
tensor<fp16, [512]> model_encoder_feature_projection_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_feature_projection_layer_norm_bias_to_fp16"), val = tensor<fp16, [512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8402624)))];
tensor<fp16, []> var_14_to_fp16 = const()[name = tensor<string, []>("op_14_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 2999, 512]> input_17_cast_fp16 = transpose(perm = input_17_perm_0, x = x_15_cast_fp16)[name = tensor<string, []>("transpose_122")];
tensor<fp16, [1, 2999, 512]> input_19_cast_fp16 = layer_norm(axes = input_19_axes_0, beta = model_encoder_feature_projection_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_feature_projection_layer_norm_weight_to_fp16, x = input_17_cast_fp16)[name = tensor<string, []>("input_19_cast_fp16")];
tensor<fp16, [768, 512]> model_encoder_feature_projection_projection_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_feature_projection_projection_weight_to_fp16"), val = tensor<fp16, [768, 512]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(8403712)))];
tensor<fp16, [768]> model_encoder_feature_projection_projection_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_feature_projection_projection_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(9190208)))];
tensor<fp16, [1, 2999, 768]> linear_0_cast_fp16 = linear(bias = model_encoder_feature_projection_projection_bias_to_fp16, weight = model_encoder_feature_projection_projection_weight_to_fp16, x = input_19_cast_fp16)[name = tensor<string, []>("linear_0_cast_fp16")];
tensor<int32, [3]> input_23_perm_0 = const()[name = tensor<string, []>("input_23_perm_0"), val = tensor<int32, [3]>([0, -1, -2])];
tensor<string, []> x_19_pad_type_0 = const()[name = tensor<string, []>("x_19_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [2]> x_19_pad_0 = const()[name = tensor<string, []>("x_19_pad_0"), val = tensor<int32, [2]>([64, 64])];
tensor<int32, []> x_19_groups_0 = const()[name = tensor<string, []>("x_19_groups_0"), val = tensor<int32, []>(16)];
tensor<int32, [1]> x_19_strides_0 = const()[name = tensor<string, []>("x_19_strides_0"), val = tensor<int32, [1]>([1])];
tensor<int32, [1]> x_19_dilations_0 = const()[name = tensor<string, []>("x_19_dilations_0"), val = tensor<int32, [1]>([1])];
tensor<fp16, [768, 48, 128]> weight_19_to_fp16 = const()[name = tensor<string, []>("weight_19_to_fp16"), val = tensor<fp16, [768, 48, 128]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(9191808)))];
tensor<fp16, [768]> model_encoder_transformer_pos_conv_embed_conv_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_pos_conv_embed_conv_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(18629056)))];
tensor<fp16, [1, 768, 2999]> input_23_cast_fp16 = transpose(perm = input_23_perm_0, x = linear_0_cast_fp16)[name = tensor<string, []>("transpose_121")];
tensor<fp16, [1, 768, 3000]> x_19_cast_fp16 = conv(bias = model_encoder_transformer_pos_conv_embed_conv_bias_to_fp16, dilations = x_19_dilations_0, groups = x_19_groups_0, pad = x_19_pad_0, pad_type = x_19_pad_type_0, strides = x_19_strides_0, weight = weight_19_to_fp16, x = input_23_cast_fp16)[name = tensor<string, []>("x_19_cast_fp16")];
tensor<int32, [3]> x_21_begin_0 = const()[name = tensor<string, []>("x_21_begin_0"), val = tensor<int32, [3]>([0, 0, 0])];
tensor<int32, [3]> x_21_end_0 = const()[name = tensor<string, []>("x_21_end_0"), val = tensor<int32, [3]>([1, 768, 2999])];
tensor<bool, [3]> x_21_end_mask_0 = const()[name = tensor<string, []>("x_21_end_mask_0"), val = tensor<bool, [3]>([true, true, false])];
tensor<fp16, [1, 768, 2999]> x_21_cast_fp16 = slice_by_index(begin = x_21_begin_0, end = x_21_end_0, end_mask = x_21_end_mask_0, x = x_19_cast_fp16)[name = tensor<string, []>("x_21_cast_fp16")];
tensor<string, []> x_23_mode_0 = const()[name = tensor<string, []>("x_23_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 768, 2999]> x_23_cast_fp16 = gelu(mode = x_23_mode_0, x = x_21_cast_fp16)[name = tensor<string, []>("x_23_cast_fp16")];
tensor<int32, [3]> var_153_perm_0 = const()[name = tensor<string, []>("op_153_perm_0"), val = tensor<int32, [3]>([0, -1, -2])];
tensor<fp16, [1, 2999, 768]> var_153_cast_fp16 = transpose(perm = var_153_perm_0, x = x_23_cast_fp16)[name = tensor<string, []>("transpose_120")];
tensor<fp16, [1, 2999, 768]> input_25_cast_fp16 = add(x = linear_0_cast_fp16, y = var_153_cast_fp16)[name = tensor<string, []>("input_25_cast_fp16")];
tensor<int32, [1]> input_27_axes_0 = const()[name = tensor<string, []>("input_27_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(18630656)))];
tensor<fp16, [768]> model_encoder_transformer_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(18632256)))];
tensor<fp16, [1, 2999, 768]> input_27_cast_fp16 = layer_norm(axes = input_27_axes_0, beta = model_encoder_transformer_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layer_norm_weight_to_fp16, x = input_25_cast_fp16)[name = tensor<string, []>("input_27_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_0_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(18633856)))];
tensor<fp16, [768]> model_encoder_transformer_layers_0_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(19813568)))];
tensor<fp16, [1, 2999, 768]> linear_1_cast_fp16 = linear(bias = model_encoder_transformer_layers_0_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_0_attention_q_proj_weight_to_fp16, x = input_27_cast_fp16)[name = tensor<string, []>("linear_1_cast_fp16")];
tensor<int32, [4]> var_173 = const()[name = tensor<string, []>("op_173"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_174_cast_fp16 = reshape(shape = var_173, x = linear_1_cast_fp16)[name = tensor<string, []>("op_174_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_0_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(19815168)))];
tensor<fp16, [768]> model_encoder_transformer_layers_0_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(20994880)))];
tensor<fp16, [1, 2999, 768]> linear_2_cast_fp16 = linear(bias = model_encoder_transformer_layers_0_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_0_attention_k_proj_weight_to_fp16, x = input_27_cast_fp16)[name = tensor<string, []>("linear_2_cast_fp16")];
tensor<int32, [4]> var_179 = const()[name = tensor<string, []>("op_179"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_180_cast_fp16 = reshape(shape = var_179, x = linear_2_cast_fp16)[name = tensor<string, []>("op_180_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_0_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(20996480)))];
tensor<fp16, [768]> model_encoder_transformer_layers_0_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(22176192)))];
tensor<fp16, [1, 2999, 768]> linear_3_cast_fp16 = linear(bias = model_encoder_transformer_layers_0_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_0_attention_v_proj_weight_to_fp16, x = input_27_cast_fp16)[name = tensor<string, []>("linear_3_cast_fp16")];
tensor<int32, [4]> var_185 = const()[name = tensor<string, []>("op_185"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_186_cast_fp16 = reshape(shape = var_185, x = linear_3_cast_fp16)[name = tensor<string, []>("op_186_cast_fp16")];
tensor<int32, [4]> v_1_perm_0 = const()[name = tensor<string, []>("v_1_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_1_y_0_to_fp16 = const()[name = tensor<string, []>("mul_1_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_1_cast_fp16 = mul(x = var_174_cast_fp16, y = mul_1_y_0_to_fp16)[name = tensor<string, []>("mul_1_cast_fp16")];
tensor<bool, []> matmul_0_transpose_y_0 = const()[name = tensor<string, []>("matmul_0_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_0_transpose_x_0 = const()[name = tensor<string, []>("matmul_0_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_48_perm_0 = const()[name = tensor<string, []>("transpose_48_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_49_perm_0 = const()[name = tensor<string, []>("transpose_49_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_49 = transpose(perm = transpose_49_perm_0, x = var_180_cast_fp16)[name = tensor<string, []>("transpose_117")];
tensor<fp16, [1, 12, 2999, 64]> transpose_48 = transpose(perm = transpose_48_perm_0, x = mul_1_cast_fp16)[name = tensor<string, []>("transpose_118")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_0_cast_fp16 = matmul(transpose_x = matmul_0_transpose_x_0, transpose_y = matmul_0_transpose_y_0, x = transpose_48, y = transpose_49)[name = tensor<string, []>("matmul_0_cast_fp16")];
tensor<int32, []> softmax_0_axis_0 = const()[name = tensor<string, []>("softmax_0_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_0_cast_fp16 = softmax(axis = softmax_0_axis_0, x = matmul_0_cast_fp16)[name = tensor<string, []>("softmax_0_cast_fp16")];
tensor<bool, []> attn_output_1_transpose_x_0 = const()[name = tensor<string, []>("attn_output_1_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_1_transpose_y_0 = const()[name = tensor<string, []>("attn_output_1_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_1_cast_fp16 = transpose(perm = v_1_perm_0, x = var_186_cast_fp16)[name = tensor<string, []>("transpose_119")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_1_cast_fp16 = matmul(transpose_x = attn_output_1_transpose_x_0, transpose_y = attn_output_1_transpose_y_0, x = softmax_0_cast_fp16, y = v_1_cast_fp16)[name = tensor<string, []>("attn_output_1_cast_fp16")];
tensor<int32, [4]> var_189_perm_0 = const()[name = tensor<string, []>("op_189_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_190 = const()[name = tensor<string, []>("op_190"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_189_cast_fp16 = transpose(perm = var_189_perm_0, x = attn_output_1_cast_fp16)[name = tensor<string, []>("transpose_116")];
tensor<fp16, [1, 2999, 768]> input_29_cast_fp16 = reshape(shape = var_190, x = var_189_cast_fp16)[name = tensor<string, []>("input_29_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_0_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(22177792)))];
tensor<fp16, [768]> model_encoder_transformer_layers_0_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(23357504)))];
tensor<fp16, [1, 2999, 768]> linear_4_cast_fp16 = linear(bias = model_encoder_transformer_layers_0_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_0_attention_out_proj_weight_to_fp16, x = input_29_cast_fp16)[name = tensor<string, []>("linear_4_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_33_cast_fp16 = add(x = input_27_cast_fp16, y = linear_4_cast_fp16)[name = tensor<string, []>("input_33_cast_fp16")];
tensor<int32, [1]> input_35_axes_0 = const()[name = tensor<string, []>("input_35_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_0_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(23359104)))];
tensor<fp16, [768]> model_encoder_transformer_layers_0_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(23360704)))];
tensor<fp16, [1, 2999, 768]> input_35_cast_fp16 = layer_norm(axes = input_35_axes_0, beta = model_encoder_transformer_layers_0_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_0_layer_norm_weight_to_fp16, x = input_33_cast_fp16)[name = tensor<string, []>("input_35_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_0_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(23362304)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_0_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(28080960)))];
tensor<fp16, [1, 2999, 3072]> linear_5_cast_fp16 = linear(bias = model_encoder_transformer_layers_0_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_0_feed_forward_intermediate_dense_weight_to_fp16, x = input_35_cast_fp16)[name = tensor<string, []>("linear_5_cast_fp16")];
tensor<string, []> input_37_mode_0 = const()[name = tensor<string, []>("input_37_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_37_cast_fp16 = gelu(mode = input_37_mode_0, x = linear_5_cast_fp16)[name = tensor<string, []>("input_37_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_0_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(28087168)))];
tensor<fp16, [768]> model_encoder_transformer_layers_0_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(32805824)))];
tensor<fp16, [1, 2999, 768]> linear_6_cast_fp16 = linear(bias = model_encoder_transformer_layers_0_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_0_feed_forward_output_dense_weight_to_fp16, x = input_37_cast_fp16)[name = tensor<string, []>("linear_6_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_43_cast_fp16 = add(x = input_35_cast_fp16, y = linear_6_cast_fp16)[name = tensor<string, []>("input_43_cast_fp16")];
tensor<int32, [1]> x_31_axes_0 = const()[name = tensor<string, []>("x_31_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_0_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(32807424)))];
tensor<fp16, [768]> model_encoder_transformer_layers_0_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_0_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(32809024)))];
tensor<fp16, [1, 2999, 768]> x_31_cast_fp16 = layer_norm(axes = x_31_axes_0, beta = model_encoder_transformer_layers_0_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_0_final_layer_norm_weight_to_fp16, x = input_43_cast_fp16)[name = tensor<string, []>("x_31_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_1_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(32810624)))];
tensor<fp16, [768]> model_encoder_transformer_layers_1_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(33990336)))];
tensor<fp16, [1, 2999, 768]> linear_7_cast_fp16 = linear(bias = model_encoder_transformer_layers_1_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_1_attention_q_proj_weight_to_fp16, x = x_31_cast_fp16)[name = tensor<string, []>("linear_7_cast_fp16")];
tensor<int32, [4]> var_230 = const()[name = tensor<string, []>("op_230"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_231_cast_fp16 = reshape(shape = var_230, x = linear_7_cast_fp16)[name = tensor<string, []>("op_231_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_1_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(33991936)))];
tensor<fp16, [768]> model_encoder_transformer_layers_1_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(35171648)))];
tensor<fp16, [1, 2999, 768]> linear_8_cast_fp16 = linear(bias = model_encoder_transformer_layers_1_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_1_attention_k_proj_weight_to_fp16, x = x_31_cast_fp16)[name = tensor<string, []>("linear_8_cast_fp16")];
tensor<int32, [4]> var_236 = const()[name = tensor<string, []>("op_236"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_237_cast_fp16 = reshape(shape = var_236, x = linear_8_cast_fp16)[name = tensor<string, []>("op_237_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_1_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(35173248)))];
tensor<fp16, [768]> model_encoder_transformer_layers_1_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(36352960)))];
tensor<fp16, [1, 2999, 768]> linear_9_cast_fp16 = linear(bias = model_encoder_transformer_layers_1_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_1_attention_v_proj_weight_to_fp16, x = x_31_cast_fp16)[name = tensor<string, []>("linear_9_cast_fp16")];
tensor<int32, [4]> var_242 = const()[name = tensor<string, []>("op_242"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_243_cast_fp16 = reshape(shape = var_242, x = linear_9_cast_fp16)[name = tensor<string, []>("op_243_cast_fp16")];
tensor<int32, [4]> v_3_perm_0 = const()[name = tensor<string, []>("v_3_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_2_y_0_to_fp16 = const()[name = tensor<string, []>("mul_2_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_2_cast_fp16 = mul(x = var_231_cast_fp16, y = mul_2_y_0_to_fp16)[name = tensor<string, []>("mul_2_cast_fp16")];
tensor<bool, []> matmul_1_transpose_y_0 = const()[name = tensor<string, []>("matmul_1_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_1_transpose_x_0 = const()[name = tensor<string, []>("matmul_1_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_50_perm_0 = const()[name = tensor<string, []>("transpose_50_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_51_perm_0 = const()[name = tensor<string, []>("transpose_51_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_51 = transpose(perm = transpose_51_perm_0, x = var_237_cast_fp16)[name = tensor<string, []>("transpose_113")];
tensor<fp16, [1, 12, 2999, 64]> transpose_50 = transpose(perm = transpose_50_perm_0, x = mul_2_cast_fp16)[name = tensor<string, []>("transpose_114")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_1_cast_fp16 = matmul(transpose_x = matmul_1_transpose_x_0, transpose_y = matmul_1_transpose_y_0, x = transpose_50, y = transpose_51)[name = tensor<string, []>("matmul_1_cast_fp16")];
tensor<int32, []> softmax_1_axis_0 = const()[name = tensor<string, []>("softmax_1_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_1_cast_fp16 = softmax(axis = softmax_1_axis_0, x = matmul_1_cast_fp16)[name = tensor<string, []>("softmax_1_cast_fp16")];
tensor<bool, []> attn_output_3_transpose_x_0 = const()[name = tensor<string, []>("attn_output_3_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_3_transpose_y_0 = const()[name = tensor<string, []>("attn_output_3_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_3_cast_fp16 = transpose(perm = v_3_perm_0, x = var_243_cast_fp16)[name = tensor<string, []>("transpose_115")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_3_cast_fp16 = matmul(transpose_x = attn_output_3_transpose_x_0, transpose_y = attn_output_3_transpose_y_0, x = softmax_1_cast_fp16, y = v_3_cast_fp16)[name = tensor<string, []>("attn_output_3_cast_fp16")];
tensor<int32, [4]> var_246_perm_0 = const()[name = tensor<string, []>("op_246_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_247 = const()[name = tensor<string, []>("op_247"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_246_cast_fp16 = transpose(perm = var_246_perm_0, x = attn_output_3_cast_fp16)[name = tensor<string, []>("transpose_112")];
tensor<fp16, [1, 2999, 768]> input_45_cast_fp16 = reshape(shape = var_247, x = var_246_cast_fp16)[name = tensor<string, []>("input_45_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_1_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(36354560)))];
tensor<fp16, [768]> model_encoder_transformer_layers_1_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(37534272)))];
tensor<fp16, [1, 2999, 768]> linear_10_cast_fp16 = linear(bias = model_encoder_transformer_layers_1_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_1_attention_out_proj_weight_to_fp16, x = input_45_cast_fp16)[name = tensor<string, []>("linear_10_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_49_cast_fp16 = add(x = x_31_cast_fp16, y = linear_10_cast_fp16)[name = tensor<string, []>("input_49_cast_fp16")];
tensor<int32, [1]> input_51_axes_0 = const()[name = tensor<string, []>("input_51_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_1_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(37535872)))];
tensor<fp16, [768]> model_encoder_transformer_layers_1_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(37537472)))];
tensor<fp16, [1, 2999, 768]> input_51_cast_fp16 = layer_norm(axes = input_51_axes_0, beta = model_encoder_transformer_layers_1_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_1_layer_norm_weight_to_fp16, x = input_49_cast_fp16)[name = tensor<string, []>("input_51_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_1_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(37539072)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_1_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42257728)))];
tensor<fp16, [1, 2999, 3072]> linear_11_cast_fp16 = linear(bias = model_encoder_transformer_layers_1_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_1_feed_forward_intermediate_dense_weight_to_fp16, x = input_51_cast_fp16)[name = tensor<string, []>("linear_11_cast_fp16")];
tensor<string, []> input_53_mode_0 = const()[name = tensor<string, []>("input_53_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_53_cast_fp16 = gelu(mode = input_53_mode_0, x = linear_11_cast_fp16)[name = tensor<string, []>("input_53_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_1_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42263936)))];
tensor<fp16, [768]> model_encoder_transformer_layers_1_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46982592)))];
tensor<fp16, [1, 2999, 768]> linear_12_cast_fp16 = linear(bias = model_encoder_transformer_layers_1_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_1_feed_forward_output_dense_weight_to_fp16, x = input_53_cast_fp16)[name = tensor<string, []>("linear_12_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_59_cast_fp16 = add(x = input_51_cast_fp16, y = linear_12_cast_fp16)[name = tensor<string, []>("input_59_cast_fp16")];
tensor<int32, [1]> x_37_axes_0 = const()[name = tensor<string, []>("x_37_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_1_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46984192)))];
tensor<fp16, [768]> model_encoder_transformer_layers_1_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_1_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46985792)))];
tensor<fp16, [1, 2999, 768]> x_37_cast_fp16 = layer_norm(axes = x_37_axes_0, beta = model_encoder_transformer_layers_1_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_1_final_layer_norm_weight_to_fp16, x = input_59_cast_fp16)[name = tensor<string, []>("x_37_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_2_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46987392)))];
tensor<fp16, [768]> model_encoder_transformer_layers_2_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(48167104)))];
tensor<fp16, [1, 2999, 768]> linear_13_cast_fp16 = linear(bias = model_encoder_transformer_layers_2_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_2_attention_q_proj_weight_to_fp16, x = x_37_cast_fp16)[name = tensor<string, []>("linear_13_cast_fp16")];
tensor<int32, [4]> var_287 = const()[name = tensor<string, []>("op_287"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_288_cast_fp16 = reshape(shape = var_287, x = linear_13_cast_fp16)[name = tensor<string, []>("op_288_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_2_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(48168704)))];
tensor<fp16, [768]> model_encoder_transformer_layers_2_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49348416)))];
tensor<fp16, [1, 2999, 768]> linear_14_cast_fp16 = linear(bias = model_encoder_transformer_layers_2_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_2_attention_k_proj_weight_to_fp16, x = x_37_cast_fp16)[name = tensor<string, []>("linear_14_cast_fp16")];
tensor<int32, [4]> var_293 = const()[name = tensor<string, []>("op_293"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_294_cast_fp16 = reshape(shape = var_293, x = linear_14_cast_fp16)[name = tensor<string, []>("op_294_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_2_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49350016)))];
tensor<fp16, [768]> model_encoder_transformer_layers_2_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50529728)))];
tensor<fp16, [1, 2999, 768]> linear_15_cast_fp16 = linear(bias = model_encoder_transformer_layers_2_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_2_attention_v_proj_weight_to_fp16, x = x_37_cast_fp16)[name = tensor<string, []>("linear_15_cast_fp16")];
tensor<int32, [4]> var_299 = const()[name = tensor<string, []>("op_299"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_300_cast_fp16 = reshape(shape = var_299, x = linear_15_cast_fp16)[name = tensor<string, []>("op_300_cast_fp16")];
tensor<int32, [4]> v_5_perm_0 = const()[name = tensor<string, []>("v_5_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_3_y_0_to_fp16 = const()[name = tensor<string, []>("mul_3_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_3_cast_fp16 = mul(x = var_288_cast_fp16, y = mul_3_y_0_to_fp16)[name = tensor<string, []>("mul_3_cast_fp16")];
tensor<bool, []> matmul_2_transpose_y_0 = const()[name = tensor<string, []>("matmul_2_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_2_transpose_x_0 = const()[name = tensor<string, []>("matmul_2_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_52_perm_0 = const()[name = tensor<string, []>("transpose_52_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_53_perm_0 = const()[name = tensor<string, []>("transpose_53_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_53 = transpose(perm = transpose_53_perm_0, x = var_294_cast_fp16)[name = tensor<string, []>("transpose_109")];
tensor<fp16, [1, 12, 2999, 64]> transpose_52 = transpose(perm = transpose_52_perm_0, x = mul_3_cast_fp16)[name = tensor<string, []>("transpose_110")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_2_cast_fp16 = matmul(transpose_x = matmul_2_transpose_x_0, transpose_y = matmul_2_transpose_y_0, x = transpose_52, y = transpose_53)[name = tensor<string, []>("matmul_2_cast_fp16")];
tensor<int32, []> softmax_2_axis_0 = const()[name = tensor<string, []>("softmax_2_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_2_cast_fp16 = softmax(axis = softmax_2_axis_0, x = matmul_2_cast_fp16)[name = tensor<string, []>("softmax_2_cast_fp16")];
tensor<bool, []> attn_output_5_transpose_x_0 = const()[name = tensor<string, []>("attn_output_5_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_5_transpose_y_0 = const()[name = tensor<string, []>("attn_output_5_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_5_cast_fp16 = transpose(perm = v_5_perm_0, x = var_300_cast_fp16)[name = tensor<string, []>("transpose_111")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_5_cast_fp16 = matmul(transpose_x = attn_output_5_transpose_x_0, transpose_y = attn_output_5_transpose_y_0, x = softmax_2_cast_fp16, y = v_5_cast_fp16)[name = tensor<string, []>("attn_output_5_cast_fp16")];
tensor<int32, [4]> var_303_perm_0 = const()[name = tensor<string, []>("op_303_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_304 = const()[name = tensor<string, []>("op_304"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_303_cast_fp16 = transpose(perm = var_303_perm_0, x = attn_output_5_cast_fp16)[name = tensor<string, []>("transpose_108")];
tensor<fp16, [1, 2999, 768]> input_61_cast_fp16 = reshape(shape = var_304, x = var_303_cast_fp16)[name = tensor<string, []>("input_61_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_2_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50531328)))];
tensor<fp16, [768]> model_encoder_transformer_layers_2_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51711040)))];
tensor<fp16, [1, 2999, 768]> linear_16_cast_fp16 = linear(bias = model_encoder_transformer_layers_2_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_2_attention_out_proj_weight_to_fp16, x = input_61_cast_fp16)[name = tensor<string, []>("linear_16_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_65_cast_fp16 = add(x = x_37_cast_fp16, y = linear_16_cast_fp16)[name = tensor<string, []>("input_65_cast_fp16")];
tensor<int32, [1]> input_67_axes_0 = const()[name = tensor<string, []>("input_67_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_2_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51712640)))];
tensor<fp16, [768]> model_encoder_transformer_layers_2_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51714240)))];
tensor<fp16, [1, 2999, 768]> input_67_cast_fp16 = layer_norm(axes = input_67_axes_0, beta = model_encoder_transformer_layers_2_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_2_layer_norm_weight_to_fp16, x = input_65_cast_fp16)[name = tensor<string, []>("input_67_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_2_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51715840)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_2_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56434496)))];
tensor<fp16, [1, 2999, 3072]> linear_17_cast_fp16 = linear(bias = model_encoder_transformer_layers_2_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_2_feed_forward_intermediate_dense_weight_to_fp16, x = input_67_cast_fp16)[name = tensor<string, []>("linear_17_cast_fp16")];
tensor<string, []> input_69_mode_0 = const()[name = tensor<string, []>("input_69_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_69_cast_fp16 = gelu(mode = input_69_mode_0, x = linear_17_cast_fp16)[name = tensor<string, []>("input_69_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_2_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56440704)))];
tensor<fp16, [768]> model_encoder_transformer_layers_2_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(61159360)))];
tensor<fp16, [1, 2999, 768]> linear_18_cast_fp16 = linear(bias = model_encoder_transformer_layers_2_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_2_feed_forward_output_dense_weight_to_fp16, x = input_69_cast_fp16)[name = tensor<string, []>("linear_18_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_75_cast_fp16 = add(x = input_67_cast_fp16, y = linear_18_cast_fp16)[name = tensor<string, []>("input_75_cast_fp16")];
tensor<int32, [1]> x_43_axes_0 = const()[name = tensor<string, []>("x_43_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_2_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(61160960)))];
tensor<fp16, [768]> model_encoder_transformer_layers_2_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_2_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(61162560)))];
tensor<fp16, [1, 2999, 768]> x_43_cast_fp16 = layer_norm(axes = x_43_axes_0, beta = model_encoder_transformer_layers_2_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_2_final_layer_norm_weight_to_fp16, x = input_75_cast_fp16)[name = tensor<string, []>("x_43_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_3_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(61164160)))];
tensor<fp16, [768]> model_encoder_transformer_layers_3_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(62343872)))];
tensor<fp16, [1, 2999, 768]> linear_19_cast_fp16 = linear(bias = model_encoder_transformer_layers_3_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_3_attention_q_proj_weight_to_fp16, x = x_43_cast_fp16)[name = tensor<string, []>("linear_19_cast_fp16")];
tensor<int32, [4]> var_344 = const()[name = tensor<string, []>("op_344"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_345_cast_fp16 = reshape(shape = var_344, x = linear_19_cast_fp16)[name = tensor<string, []>("op_345_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_3_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(62345472)))];
tensor<fp16, [768]> model_encoder_transformer_layers_3_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(63525184)))];
tensor<fp16, [1, 2999, 768]> linear_20_cast_fp16 = linear(bias = model_encoder_transformer_layers_3_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_3_attention_k_proj_weight_to_fp16, x = x_43_cast_fp16)[name = tensor<string, []>("linear_20_cast_fp16")];
tensor<int32, [4]> var_350 = const()[name = tensor<string, []>("op_350"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_351_cast_fp16 = reshape(shape = var_350, x = linear_20_cast_fp16)[name = tensor<string, []>("op_351_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_3_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(63526784)))];
tensor<fp16, [768]> model_encoder_transformer_layers_3_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64706496)))];
tensor<fp16, [1, 2999, 768]> linear_21_cast_fp16 = linear(bias = model_encoder_transformer_layers_3_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_3_attention_v_proj_weight_to_fp16, x = x_43_cast_fp16)[name = tensor<string, []>("linear_21_cast_fp16")];
tensor<int32, [4]> var_356 = const()[name = tensor<string, []>("op_356"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_357_cast_fp16 = reshape(shape = var_356, x = linear_21_cast_fp16)[name = tensor<string, []>("op_357_cast_fp16")];
tensor<int32, [4]> v_7_perm_0 = const()[name = tensor<string, []>("v_7_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_4_y_0_to_fp16 = const()[name = tensor<string, []>("mul_4_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_4_cast_fp16 = mul(x = var_345_cast_fp16, y = mul_4_y_0_to_fp16)[name = tensor<string, []>("mul_4_cast_fp16")];
tensor<bool, []> matmul_3_transpose_y_0 = const()[name = tensor<string, []>("matmul_3_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_3_transpose_x_0 = const()[name = tensor<string, []>("matmul_3_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_54_perm_0 = const()[name = tensor<string, []>("transpose_54_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_55_perm_0 = const()[name = tensor<string, []>("transpose_55_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_55 = transpose(perm = transpose_55_perm_0, x = var_351_cast_fp16)[name = tensor<string, []>("transpose_105")];
tensor<fp16, [1, 12, 2999, 64]> transpose_54 = transpose(perm = transpose_54_perm_0, x = mul_4_cast_fp16)[name = tensor<string, []>("transpose_106")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_3_cast_fp16 = matmul(transpose_x = matmul_3_transpose_x_0, transpose_y = matmul_3_transpose_y_0, x = transpose_54, y = transpose_55)[name = tensor<string, []>("matmul_3_cast_fp16")];
tensor<int32, []> softmax_3_axis_0 = const()[name = tensor<string, []>("softmax_3_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_3_cast_fp16 = softmax(axis = softmax_3_axis_0, x = matmul_3_cast_fp16)[name = tensor<string, []>("softmax_3_cast_fp16")];
tensor<bool, []> attn_output_7_transpose_x_0 = const()[name = tensor<string, []>("attn_output_7_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_7_transpose_y_0 = const()[name = tensor<string, []>("attn_output_7_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_7_cast_fp16 = transpose(perm = v_7_perm_0, x = var_357_cast_fp16)[name = tensor<string, []>("transpose_107")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_7_cast_fp16 = matmul(transpose_x = attn_output_7_transpose_x_0, transpose_y = attn_output_7_transpose_y_0, x = softmax_3_cast_fp16, y = v_7_cast_fp16)[name = tensor<string, []>("attn_output_7_cast_fp16")];
tensor<int32, [4]> var_360_perm_0 = const()[name = tensor<string, []>("op_360_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_361 = const()[name = tensor<string, []>("op_361"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_360_cast_fp16 = transpose(perm = var_360_perm_0, x = attn_output_7_cast_fp16)[name = tensor<string, []>("transpose_104")];
tensor<fp16, [1, 2999, 768]> input_77_cast_fp16 = reshape(shape = var_361, x = var_360_cast_fp16)[name = tensor<string, []>("input_77_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_3_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64708096)))];
tensor<fp16, [768]> model_encoder_transformer_layers_3_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(65887808)))];
tensor<fp16, [1, 2999, 768]> linear_22_cast_fp16 = linear(bias = model_encoder_transformer_layers_3_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_3_attention_out_proj_weight_to_fp16, x = input_77_cast_fp16)[name = tensor<string, []>("linear_22_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_81_cast_fp16 = add(x = x_43_cast_fp16, y = linear_22_cast_fp16)[name = tensor<string, []>("input_81_cast_fp16")];
tensor<int32, [1]> input_83_axes_0 = const()[name = tensor<string, []>("input_83_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_3_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(65889408)))];
tensor<fp16, [768]> model_encoder_transformer_layers_3_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(65891008)))];
tensor<fp16, [1, 2999, 768]> input_83_cast_fp16 = layer_norm(axes = input_83_axes_0, beta = model_encoder_transformer_layers_3_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_3_layer_norm_weight_to_fp16, x = input_81_cast_fp16)[name = tensor<string, []>("input_83_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_3_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(65892608)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_3_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(70611264)))];
tensor<fp16, [1, 2999, 3072]> linear_23_cast_fp16 = linear(bias = model_encoder_transformer_layers_3_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_3_feed_forward_intermediate_dense_weight_to_fp16, x = input_83_cast_fp16)[name = tensor<string, []>("linear_23_cast_fp16")];
tensor<string, []> input_85_mode_0 = const()[name = tensor<string, []>("input_85_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_85_cast_fp16 = gelu(mode = input_85_mode_0, x = linear_23_cast_fp16)[name = tensor<string, []>("input_85_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_3_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(70617472)))];
tensor<fp16, [768]> model_encoder_transformer_layers_3_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(75336128)))];
tensor<fp16, [1, 2999, 768]> linear_24_cast_fp16 = linear(bias = model_encoder_transformer_layers_3_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_3_feed_forward_output_dense_weight_to_fp16, x = input_85_cast_fp16)[name = tensor<string, []>("linear_24_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_91_cast_fp16 = add(x = input_83_cast_fp16, y = linear_24_cast_fp16)[name = tensor<string, []>("input_91_cast_fp16")];
tensor<int32, [1]> x_49_axes_0 = const()[name = tensor<string, []>("x_49_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_3_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(75337728)))];
tensor<fp16, [768]> model_encoder_transformer_layers_3_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_3_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(75339328)))];
tensor<fp16, [1, 2999, 768]> x_49_cast_fp16 = layer_norm(axes = x_49_axes_0, beta = model_encoder_transformer_layers_3_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_3_final_layer_norm_weight_to_fp16, x = input_91_cast_fp16)[name = tensor<string, []>("x_49_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_4_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(75340928)))];
tensor<fp16, [768]> model_encoder_transformer_layers_4_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(76520640)))];
tensor<fp16, [1, 2999, 768]> linear_25_cast_fp16 = linear(bias = model_encoder_transformer_layers_4_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_4_attention_q_proj_weight_to_fp16, x = x_49_cast_fp16)[name = tensor<string, []>("linear_25_cast_fp16")];
tensor<int32, [4]> var_401 = const()[name = tensor<string, []>("op_401"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_402_cast_fp16 = reshape(shape = var_401, x = linear_25_cast_fp16)[name = tensor<string, []>("op_402_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_4_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(76522240)))];
tensor<fp16, [768]> model_encoder_transformer_layers_4_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(77701952)))];
tensor<fp16, [1, 2999, 768]> linear_26_cast_fp16 = linear(bias = model_encoder_transformer_layers_4_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_4_attention_k_proj_weight_to_fp16, x = x_49_cast_fp16)[name = tensor<string, []>("linear_26_cast_fp16")];
tensor<int32, [4]> var_407 = const()[name = tensor<string, []>("op_407"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_408_cast_fp16 = reshape(shape = var_407, x = linear_26_cast_fp16)[name = tensor<string, []>("op_408_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_4_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(77703552)))];
tensor<fp16, [768]> model_encoder_transformer_layers_4_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(78883264)))];
tensor<fp16, [1, 2999, 768]> linear_27_cast_fp16 = linear(bias = model_encoder_transformer_layers_4_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_4_attention_v_proj_weight_to_fp16, x = x_49_cast_fp16)[name = tensor<string, []>("linear_27_cast_fp16")];
tensor<int32, [4]> var_413 = const()[name = tensor<string, []>("op_413"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_414_cast_fp16 = reshape(shape = var_413, x = linear_27_cast_fp16)[name = tensor<string, []>("op_414_cast_fp16")];
tensor<int32, [4]> v_9_perm_0 = const()[name = tensor<string, []>("v_9_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_5_y_0_to_fp16 = const()[name = tensor<string, []>("mul_5_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_5_cast_fp16 = mul(x = var_402_cast_fp16, y = mul_5_y_0_to_fp16)[name = tensor<string, []>("mul_5_cast_fp16")];
tensor<bool, []> matmul_4_transpose_y_0 = const()[name = tensor<string, []>("matmul_4_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_4_transpose_x_0 = const()[name = tensor<string, []>("matmul_4_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_56_perm_0 = const()[name = tensor<string, []>("transpose_56_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_57_perm_0 = const()[name = tensor<string, []>("transpose_57_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_57 = transpose(perm = transpose_57_perm_0, x = var_408_cast_fp16)[name = tensor<string, []>("transpose_101")];
tensor<fp16, [1, 12, 2999, 64]> transpose_56 = transpose(perm = transpose_56_perm_0, x = mul_5_cast_fp16)[name = tensor<string, []>("transpose_102")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_4_cast_fp16 = matmul(transpose_x = matmul_4_transpose_x_0, transpose_y = matmul_4_transpose_y_0, x = transpose_56, y = transpose_57)[name = tensor<string, []>("matmul_4_cast_fp16")];
tensor<int32, []> softmax_4_axis_0 = const()[name = tensor<string, []>("softmax_4_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_4_cast_fp16 = softmax(axis = softmax_4_axis_0, x = matmul_4_cast_fp16)[name = tensor<string, []>("softmax_4_cast_fp16")];
tensor<bool, []> attn_output_9_transpose_x_0 = const()[name = tensor<string, []>("attn_output_9_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_9_transpose_y_0 = const()[name = tensor<string, []>("attn_output_9_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_9_cast_fp16 = transpose(perm = v_9_perm_0, x = var_414_cast_fp16)[name = tensor<string, []>("transpose_103")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_9_cast_fp16 = matmul(transpose_x = attn_output_9_transpose_x_0, transpose_y = attn_output_9_transpose_y_0, x = softmax_4_cast_fp16, y = v_9_cast_fp16)[name = tensor<string, []>("attn_output_9_cast_fp16")];
tensor<int32, [4]> var_417_perm_0 = const()[name = tensor<string, []>("op_417_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_418 = const()[name = tensor<string, []>("op_418"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_417_cast_fp16 = transpose(perm = var_417_perm_0, x = attn_output_9_cast_fp16)[name = tensor<string, []>("transpose_100")];
tensor<fp16, [1, 2999, 768]> input_93_cast_fp16 = reshape(shape = var_418, x = var_417_cast_fp16)[name = tensor<string, []>("input_93_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_4_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(78884864)))];
tensor<fp16, [768]> model_encoder_transformer_layers_4_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(80064576)))];
tensor<fp16, [1, 2999, 768]> linear_28_cast_fp16 = linear(bias = model_encoder_transformer_layers_4_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_4_attention_out_proj_weight_to_fp16, x = input_93_cast_fp16)[name = tensor<string, []>("linear_28_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_97_cast_fp16 = add(x = x_49_cast_fp16, y = linear_28_cast_fp16)[name = tensor<string, []>("input_97_cast_fp16")];
tensor<int32, [1]> input_99_axes_0 = const()[name = tensor<string, []>("input_99_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_4_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(80066176)))];
tensor<fp16, [768]> model_encoder_transformer_layers_4_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(80067776)))];
tensor<fp16, [1, 2999, 768]> input_99_cast_fp16 = layer_norm(axes = input_99_axes_0, beta = model_encoder_transformer_layers_4_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_4_layer_norm_weight_to_fp16, x = input_97_cast_fp16)[name = tensor<string, []>("input_99_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_4_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(80069376)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_4_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(84788032)))];
tensor<fp16, [1, 2999, 3072]> linear_29_cast_fp16 = linear(bias = model_encoder_transformer_layers_4_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_4_feed_forward_intermediate_dense_weight_to_fp16, x = input_99_cast_fp16)[name = tensor<string, []>("linear_29_cast_fp16")];
tensor<string, []> input_101_mode_0 = const()[name = tensor<string, []>("input_101_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_101_cast_fp16 = gelu(mode = input_101_mode_0, x = linear_29_cast_fp16)[name = tensor<string, []>("input_101_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_4_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(84794240)))];
tensor<fp16, [768]> model_encoder_transformer_layers_4_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(89512896)))];
tensor<fp16, [1, 2999, 768]> linear_30_cast_fp16 = linear(bias = model_encoder_transformer_layers_4_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_4_feed_forward_output_dense_weight_to_fp16, x = input_101_cast_fp16)[name = tensor<string, []>("linear_30_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_107_cast_fp16 = add(x = input_99_cast_fp16, y = linear_30_cast_fp16)[name = tensor<string, []>("input_107_cast_fp16")];
tensor<int32, [1]> x_55_axes_0 = const()[name = tensor<string, []>("x_55_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_4_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(89514496)))];
tensor<fp16, [768]> model_encoder_transformer_layers_4_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_4_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(89516096)))];
tensor<fp16, [1, 2999, 768]> x_55_cast_fp16 = layer_norm(axes = x_55_axes_0, beta = model_encoder_transformer_layers_4_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_4_final_layer_norm_weight_to_fp16, x = input_107_cast_fp16)[name = tensor<string, []>("x_55_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_5_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(89517696)))];
tensor<fp16, [768]> model_encoder_transformer_layers_5_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(90697408)))];
tensor<fp16, [1, 2999, 768]> linear_31_cast_fp16 = linear(bias = model_encoder_transformer_layers_5_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_5_attention_q_proj_weight_to_fp16, x = x_55_cast_fp16)[name = tensor<string, []>("linear_31_cast_fp16")];
tensor<int32, [4]> var_458 = const()[name = tensor<string, []>("op_458"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_459_cast_fp16 = reshape(shape = var_458, x = linear_31_cast_fp16)[name = tensor<string, []>("op_459_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_5_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(90699008)))];
tensor<fp16, [768]> model_encoder_transformer_layers_5_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(91878720)))];
tensor<fp16, [1, 2999, 768]> linear_32_cast_fp16 = linear(bias = model_encoder_transformer_layers_5_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_5_attention_k_proj_weight_to_fp16, x = x_55_cast_fp16)[name = tensor<string, []>("linear_32_cast_fp16")];
tensor<int32, [4]> var_464 = const()[name = tensor<string, []>("op_464"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_465_cast_fp16 = reshape(shape = var_464, x = linear_32_cast_fp16)[name = tensor<string, []>("op_465_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_5_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(91880320)))];
tensor<fp16, [768]> model_encoder_transformer_layers_5_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(93060032)))];
tensor<fp16, [1, 2999, 768]> linear_33_cast_fp16 = linear(bias = model_encoder_transformer_layers_5_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_5_attention_v_proj_weight_to_fp16, x = x_55_cast_fp16)[name = tensor<string, []>("linear_33_cast_fp16")];
tensor<int32, [4]> var_470 = const()[name = tensor<string, []>("op_470"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_471_cast_fp16 = reshape(shape = var_470, x = linear_33_cast_fp16)[name = tensor<string, []>("op_471_cast_fp16")];
tensor<int32, [4]> v_11_perm_0 = const()[name = tensor<string, []>("v_11_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_6_y_0_to_fp16 = const()[name = tensor<string, []>("mul_6_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_6_cast_fp16 = mul(x = var_459_cast_fp16, y = mul_6_y_0_to_fp16)[name = tensor<string, []>("mul_6_cast_fp16")];
tensor<bool, []> matmul_5_transpose_y_0 = const()[name = tensor<string, []>("matmul_5_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_5_transpose_x_0 = const()[name = tensor<string, []>("matmul_5_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_58_perm_0 = const()[name = tensor<string, []>("transpose_58_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_59_perm_0 = const()[name = tensor<string, []>("transpose_59_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_59 = transpose(perm = transpose_59_perm_0, x = var_465_cast_fp16)[name = tensor<string, []>("transpose_97")];
tensor<fp16, [1, 12, 2999, 64]> transpose_58 = transpose(perm = transpose_58_perm_0, x = mul_6_cast_fp16)[name = tensor<string, []>("transpose_98")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_5_cast_fp16 = matmul(transpose_x = matmul_5_transpose_x_0, transpose_y = matmul_5_transpose_y_0, x = transpose_58, y = transpose_59)[name = tensor<string, []>("matmul_5_cast_fp16")];
tensor<int32, []> softmax_5_axis_0 = const()[name = tensor<string, []>("softmax_5_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_5_cast_fp16 = softmax(axis = softmax_5_axis_0, x = matmul_5_cast_fp16)[name = tensor<string, []>("softmax_5_cast_fp16")];
tensor<bool, []> attn_output_11_transpose_x_0 = const()[name = tensor<string, []>("attn_output_11_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_11_transpose_y_0 = const()[name = tensor<string, []>("attn_output_11_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_11_cast_fp16 = transpose(perm = v_11_perm_0, x = var_471_cast_fp16)[name = tensor<string, []>("transpose_99")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_11_cast_fp16 = matmul(transpose_x = attn_output_11_transpose_x_0, transpose_y = attn_output_11_transpose_y_0, x = softmax_5_cast_fp16, y = v_11_cast_fp16)[name = tensor<string, []>("attn_output_11_cast_fp16")];
tensor<int32, [4]> var_474_perm_0 = const()[name = tensor<string, []>("op_474_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_475 = const()[name = tensor<string, []>("op_475"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_474_cast_fp16 = transpose(perm = var_474_perm_0, x = attn_output_11_cast_fp16)[name = tensor<string, []>("transpose_96")];
tensor<fp16, [1, 2999, 768]> input_109_cast_fp16 = reshape(shape = var_475, x = var_474_cast_fp16)[name = tensor<string, []>("input_109_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_5_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(93061632)))];
tensor<fp16, [768]> model_encoder_transformer_layers_5_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(94241344)))];
tensor<fp16, [1, 2999, 768]> linear_34_cast_fp16 = linear(bias = model_encoder_transformer_layers_5_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_5_attention_out_proj_weight_to_fp16, x = input_109_cast_fp16)[name = tensor<string, []>("linear_34_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_113_cast_fp16 = add(x = x_55_cast_fp16, y = linear_34_cast_fp16)[name = tensor<string, []>("input_113_cast_fp16")];
tensor<int32, [1]> input_115_axes_0 = const()[name = tensor<string, []>("input_115_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_5_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(94242944)))];
tensor<fp16, [768]> model_encoder_transformer_layers_5_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(94244544)))];
tensor<fp16, [1, 2999, 768]> input_115_cast_fp16 = layer_norm(axes = input_115_axes_0, beta = model_encoder_transformer_layers_5_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_5_layer_norm_weight_to_fp16, x = input_113_cast_fp16)[name = tensor<string, []>("input_115_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_5_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(94246144)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_5_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(98964800)))];
tensor<fp16, [1, 2999, 3072]> linear_35_cast_fp16 = linear(bias = model_encoder_transformer_layers_5_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_5_feed_forward_intermediate_dense_weight_to_fp16, x = input_115_cast_fp16)[name = tensor<string, []>("linear_35_cast_fp16")];
tensor<string, []> input_117_mode_0 = const()[name = tensor<string, []>("input_117_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_117_cast_fp16 = gelu(mode = input_117_mode_0, x = linear_35_cast_fp16)[name = tensor<string, []>("input_117_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_5_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(98971008)))];
tensor<fp16, [768]> model_encoder_transformer_layers_5_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(103689664)))];
tensor<fp16, [1, 2999, 768]> linear_36_cast_fp16 = linear(bias = model_encoder_transformer_layers_5_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_5_feed_forward_output_dense_weight_to_fp16, x = input_117_cast_fp16)[name = tensor<string, []>("linear_36_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_123_cast_fp16 = add(x = input_115_cast_fp16, y = linear_36_cast_fp16)[name = tensor<string, []>("input_123_cast_fp16")];
tensor<int32, [1]> x_61_axes_0 = const()[name = tensor<string, []>("x_61_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_5_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(103691264)))];
tensor<fp16, [768]> model_encoder_transformer_layers_5_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_5_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(103692864)))];
tensor<fp16, [1, 2999, 768]> x_61_cast_fp16 = layer_norm(axes = x_61_axes_0, beta = model_encoder_transformer_layers_5_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_5_final_layer_norm_weight_to_fp16, x = input_123_cast_fp16)[name = tensor<string, []>("x_61_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_6_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(103694464)))];
tensor<fp16, [768]> model_encoder_transformer_layers_6_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(104874176)))];
tensor<fp16, [1, 2999, 768]> linear_37_cast_fp16 = linear(bias = model_encoder_transformer_layers_6_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_6_attention_q_proj_weight_to_fp16, x = x_61_cast_fp16)[name = tensor<string, []>("linear_37_cast_fp16")];
tensor<int32, [4]> var_515 = const()[name = tensor<string, []>("op_515"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_516_cast_fp16 = reshape(shape = var_515, x = linear_37_cast_fp16)[name = tensor<string, []>("op_516_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_6_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(104875776)))];
tensor<fp16, [768]> model_encoder_transformer_layers_6_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(106055488)))];
tensor<fp16, [1, 2999, 768]> linear_38_cast_fp16 = linear(bias = model_encoder_transformer_layers_6_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_6_attention_k_proj_weight_to_fp16, x = x_61_cast_fp16)[name = tensor<string, []>("linear_38_cast_fp16")];
tensor<int32, [4]> var_521 = const()[name = tensor<string, []>("op_521"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_522_cast_fp16 = reshape(shape = var_521, x = linear_38_cast_fp16)[name = tensor<string, []>("op_522_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_6_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(106057088)))];
tensor<fp16, [768]> model_encoder_transformer_layers_6_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(107236800)))];
tensor<fp16, [1, 2999, 768]> linear_39_cast_fp16 = linear(bias = model_encoder_transformer_layers_6_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_6_attention_v_proj_weight_to_fp16, x = x_61_cast_fp16)[name = tensor<string, []>("linear_39_cast_fp16")];
tensor<int32, [4]> var_527 = const()[name = tensor<string, []>("op_527"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_528_cast_fp16 = reshape(shape = var_527, x = linear_39_cast_fp16)[name = tensor<string, []>("op_528_cast_fp16")];
tensor<int32, [4]> v_13_perm_0 = const()[name = tensor<string, []>("v_13_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_7_y_0_to_fp16 = const()[name = tensor<string, []>("mul_7_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_7_cast_fp16 = mul(x = var_516_cast_fp16, y = mul_7_y_0_to_fp16)[name = tensor<string, []>("mul_7_cast_fp16")];
tensor<bool, []> matmul_6_transpose_y_0 = const()[name = tensor<string, []>("matmul_6_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_6_transpose_x_0 = const()[name = tensor<string, []>("matmul_6_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_60_perm_0 = const()[name = tensor<string, []>("transpose_60_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_61_perm_0 = const()[name = tensor<string, []>("transpose_61_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_61 = transpose(perm = transpose_61_perm_0, x = var_522_cast_fp16)[name = tensor<string, []>("transpose_93")];
tensor<fp16, [1, 12, 2999, 64]> transpose_60 = transpose(perm = transpose_60_perm_0, x = mul_7_cast_fp16)[name = tensor<string, []>("transpose_94")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_6_cast_fp16 = matmul(transpose_x = matmul_6_transpose_x_0, transpose_y = matmul_6_transpose_y_0, x = transpose_60, y = transpose_61)[name = tensor<string, []>("matmul_6_cast_fp16")];
tensor<int32, []> softmax_6_axis_0 = const()[name = tensor<string, []>("softmax_6_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_6_cast_fp16 = softmax(axis = softmax_6_axis_0, x = matmul_6_cast_fp16)[name = tensor<string, []>("softmax_6_cast_fp16")];
tensor<bool, []> attn_output_13_transpose_x_0 = const()[name = tensor<string, []>("attn_output_13_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_13_transpose_y_0 = const()[name = tensor<string, []>("attn_output_13_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_13_cast_fp16 = transpose(perm = v_13_perm_0, x = var_528_cast_fp16)[name = tensor<string, []>("transpose_95")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_13_cast_fp16 = matmul(transpose_x = attn_output_13_transpose_x_0, transpose_y = attn_output_13_transpose_y_0, x = softmax_6_cast_fp16, y = v_13_cast_fp16)[name = tensor<string, []>("attn_output_13_cast_fp16")];
tensor<int32, [4]> var_531_perm_0 = const()[name = tensor<string, []>("op_531_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_532 = const()[name = tensor<string, []>("op_532"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_531_cast_fp16 = transpose(perm = var_531_perm_0, x = attn_output_13_cast_fp16)[name = tensor<string, []>("transpose_92")];
tensor<fp16, [1, 2999, 768]> input_125_cast_fp16 = reshape(shape = var_532, x = var_531_cast_fp16)[name = tensor<string, []>("input_125_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_6_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(107238400)))];
tensor<fp16, [768]> model_encoder_transformer_layers_6_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(108418112)))];
tensor<fp16, [1, 2999, 768]> linear_40_cast_fp16 = linear(bias = model_encoder_transformer_layers_6_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_6_attention_out_proj_weight_to_fp16, x = input_125_cast_fp16)[name = tensor<string, []>("linear_40_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_129_cast_fp16 = add(x = x_61_cast_fp16, y = linear_40_cast_fp16)[name = tensor<string, []>("input_129_cast_fp16")];
tensor<int32, [1]> input_131_axes_0 = const()[name = tensor<string, []>("input_131_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_6_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(108419712)))];
tensor<fp16, [768]> model_encoder_transformer_layers_6_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(108421312)))];
tensor<fp16, [1, 2999, 768]> input_131_cast_fp16 = layer_norm(axes = input_131_axes_0, beta = model_encoder_transformer_layers_6_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_6_layer_norm_weight_to_fp16, x = input_129_cast_fp16)[name = tensor<string, []>("input_131_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_6_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(108422912)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_6_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(113141568)))];
tensor<fp16, [1, 2999, 3072]> linear_41_cast_fp16 = linear(bias = model_encoder_transformer_layers_6_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_6_feed_forward_intermediate_dense_weight_to_fp16, x = input_131_cast_fp16)[name = tensor<string, []>("linear_41_cast_fp16")];
tensor<string, []> input_133_mode_0 = const()[name = tensor<string, []>("input_133_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_133_cast_fp16 = gelu(mode = input_133_mode_0, x = linear_41_cast_fp16)[name = tensor<string, []>("input_133_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_6_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(113147776)))];
tensor<fp16, [768]> model_encoder_transformer_layers_6_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(117866432)))];
tensor<fp16, [1, 2999, 768]> linear_42_cast_fp16 = linear(bias = model_encoder_transformer_layers_6_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_6_feed_forward_output_dense_weight_to_fp16, x = input_133_cast_fp16)[name = tensor<string, []>("linear_42_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_139_cast_fp16 = add(x = input_131_cast_fp16, y = linear_42_cast_fp16)[name = tensor<string, []>("input_139_cast_fp16")];
tensor<int32, [1]> x_67_axes_0 = const()[name = tensor<string, []>("x_67_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_6_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(117868032)))];
tensor<fp16, [768]> model_encoder_transformer_layers_6_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_6_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(117869632)))];
tensor<fp16, [1, 2999, 768]> x_67_cast_fp16 = layer_norm(axes = x_67_axes_0, beta = model_encoder_transformer_layers_6_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_6_final_layer_norm_weight_to_fp16, x = input_139_cast_fp16)[name = tensor<string, []>("x_67_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_7_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(117871232)))];
tensor<fp16, [768]> model_encoder_transformer_layers_7_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(119050944)))];
tensor<fp16, [1, 2999, 768]> linear_43_cast_fp16 = linear(bias = model_encoder_transformer_layers_7_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_7_attention_q_proj_weight_to_fp16, x = x_67_cast_fp16)[name = tensor<string, []>("linear_43_cast_fp16")];
tensor<int32, [4]> var_572 = const()[name = tensor<string, []>("op_572"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_573_cast_fp16 = reshape(shape = var_572, x = linear_43_cast_fp16)[name = tensor<string, []>("op_573_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_7_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(119052544)))];
tensor<fp16, [768]> model_encoder_transformer_layers_7_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(120232256)))];
tensor<fp16, [1, 2999, 768]> linear_44_cast_fp16 = linear(bias = model_encoder_transformer_layers_7_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_7_attention_k_proj_weight_to_fp16, x = x_67_cast_fp16)[name = tensor<string, []>("linear_44_cast_fp16")];
tensor<int32, [4]> var_578 = const()[name = tensor<string, []>("op_578"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_579_cast_fp16 = reshape(shape = var_578, x = linear_44_cast_fp16)[name = tensor<string, []>("op_579_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_7_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(120233856)))];
tensor<fp16, [768]> model_encoder_transformer_layers_7_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(121413568)))];
tensor<fp16, [1, 2999, 768]> linear_45_cast_fp16 = linear(bias = model_encoder_transformer_layers_7_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_7_attention_v_proj_weight_to_fp16, x = x_67_cast_fp16)[name = tensor<string, []>("linear_45_cast_fp16")];
tensor<int32, [4]> var_584 = const()[name = tensor<string, []>("op_584"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_585_cast_fp16 = reshape(shape = var_584, x = linear_45_cast_fp16)[name = tensor<string, []>("op_585_cast_fp16")];
tensor<int32, [4]> v_15_perm_0 = const()[name = tensor<string, []>("v_15_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_8_y_0_to_fp16 = const()[name = tensor<string, []>("mul_8_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_8_cast_fp16 = mul(x = var_573_cast_fp16, y = mul_8_y_0_to_fp16)[name = tensor<string, []>("mul_8_cast_fp16")];
tensor<bool, []> matmul_7_transpose_y_0 = const()[name = tensor<string, []>("matmul_7_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_7_transpose_x_0 = const()[name = tensor<string, []>("matmul_7_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_62_perm_0 = const()[name = tensor<string, []>("transpose_62_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_63_perm_0 = const()[name = tensor<string, []>("transpose_63_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_63 = transpose(perm = transpose_63_perm_0, x = var_579_cast_fp16)[name = tensor<string, []>("transpose_89")];
tensor<fp16, [1, 12, 2999, 64]> transpose_62 = transpose(perm = transpose_62_perm_0, x = mul_8_cast_fp16)[name = tensor<string, []>("transpose_90")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_7_cast_fp16 = matmul(transpose_x = matmul_7_transpose_x_0, transpose_y = matmul_7_transpose_y_0, x = transpose_62, y = transpose_63)[name = tensor<string, []>("matmul_7_cast_fp16")];
tensor<int32, []> softmax_7_axis_0 = const()[name = tensor<string, []>("softmax_7_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_7_cast_fp16 = softmax(axis = softmax_7_axis_0, x = matmul_7_cast_fp16)[name = tensor<string, []>("softmax_7_cast_fp16")];
tensor<bool, []> attn_output_15_transpose_x_0 = const()[name = tensor<string, []>("attn_output_15_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_15_transpose_y_0 = const()[name = tensor<string, []>("attn_output_15_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_15_cast_fp16 = transpose(perm = v_15_perm_0, x = var_585_cast_fp16)[name = tensor<string, []>("transpose_91")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_15_cast_fp16 = matmul(transpose_x = attn_output_15_transpose_x_0, transpose_y = attn_output_15_transpose_y_0, x = softmax_7_cast_fp16, y = v_15_cast_fp16)[name = tensor<string, []>("attn_output_15_cast_fp16")];
tensor<int32, [4]> var_588_perm_0 = const()[name = tensor<string, []>("op_588_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_589 = const()[name = tensor<string, []>("op_589"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_588_cast_fp16 = transpose(perm = var_588_perm_0, x = attn_output_15_cast_fp16)[name = tensor<string, []>("transpose_88")];
tensor<fp16, [1, 2999, 768]> input_141_cast_fp16 = reshape(shape = var_589, x = var_588_cast_fp16)[name = tensor<string, []>("input_141_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_7_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(121415168)))];
tensor<fp16, [768]> model_encoder_transformer_layers_7_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(122594880)))];
tensor<fp16, [1, 2999, 768]> linear_46_cast_fp16 = linear(bias = model_encoder_transformer_layers_7_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_7_attention_out_proj_weight_to_fp16, x = input_141_cast_fp16)[name = tensor<string, []>("linear_46_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_145_cast_fp16 = add(x = x_67_cast_fp16, y = linear_46_cast_fp16)[name = tensor<string, []>("input_145_cast_fp16")];
tensor<int32, [1]> input_147_axes_0 = const()[name = tensor<string, []>("input_147_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_7_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(122596480)))];
tensor<fp16, [768]> model_encoder_transformer_layers_7_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(122598080)))];
tensor<fp16, [1, 2999, 768]> input_147_cast_fp16 = layer_norm(axes = input_147_axes_0, beta = model_encoder_transformer_layers_7_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_7_layer_norm_weight_to_fp16, x = input_145_cast_fp16)[name = tensor<string, []>("input_147_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_7_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(122599680)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_7_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(127318336)))];
tensor<fp16, [1, 2999, 3072]> linear_47_cast_fp16 = linear(bias = model_encoder_transformer_layers_7_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_7_feed_forward_intermediate_dense_weight_to_fp16, x = input_147_cast_fp16)[name = tensor<string, []>("linear_47_cast_fp16")];
tensor<string, []> input_149_mode_0 = const()[name = tensor<string, []>("input_149_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_149_cast_fp16 = gelu(mode = input_149_mode_0, x = linear_47_cast_fp16)[name = tensor<string, []>("input_149_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_7_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(127324544)))];
tensor<fp16, [768]> model_encoder_transformer_layers_7_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(132043200)))];
tensor<fp16, [1, 2999, 768]> linear_48_cast_fp16 = linear(bias = model_encoder_transformer_layers_7_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_7_feed_forward_output_dense_weight_to_fp16, x = input_149_cast_fp16)[name = tensor<string, []>("linear_48_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_155_cast_fp16 = add(x = input_147_cast_fp16, y = linear_48_cast_fp16)[name = tensor<string, []>("input_155_cast_fp16")];
tensor<int32, [1]> x_73_axes_0 = const()[name = tensor<string, []>("x_73_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_7_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(132044800)))];
tensor<fp16, [768]> model_encoder_transformer_layers_7_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_7_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(132046400)))];
tensor<fp16, [1, 2999, 768]> x_73_cast_fp16 = layer_norm(axes = x_73_axes_0, beta = model_encoder_transformer_layers_7_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_7_final_layer_norm_weight_to_fp16, x = input_155_cast_fp16)[name = tensor<string, []>("x_73_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_8_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(132048000)))];
tensor<fp16, [768]> model_encoder_transformer_layers_8_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(133227712)))];
tensor<fp16, [1, 2999, 768]> linear_49_cast_fp16 = linear(bias = model_encoder_transformer_layers_8_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_8_attention_q_proj_weight_to_fp16, x = x_73_cast_fp16)[name = tensor<string, []>("linear_49_cast_fp16")];
tensor<int32, [4]> var_629 = const()[name = tensor<string, []>("op_629"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_630_cast_fp16 = reshape(shape = var_629, x = linear_49_cast_fp16)[name = tensor<string, []>("op_630_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_8_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(133229312)))];
tensor<fp16, [768]> model_encoder_transformer_layers_8_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(134409024)))];
tensor<fp16, [1, 2999, 768]> linear_50_cast_fp16 = linear(bias = model_encoder_transformer_layers_8_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_8_attention_k_proj_weight_to_fp16, x = x_73_cast_fp16)[name = tensor<string, []>("linear_50_cast_fp16")];
tensor<int32, [4]> var_635 = const()[name = tensor<string, []>("op_635"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_636_cast_fp16 = reshape(shape = var_635, x = linear_50_cast_fp16)[name = tensor<string, []>("op_636_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_8_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(134410624)))];
tensor<fp16, [768]> model_encoder_transformer_layers_8_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(135590336)))];
tensor<fp16, [1, 2999, 768]> linear_51_cast_fp16 = linear(bias = model_encoder_transformer_layers_8_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_8_attention_v_proj_weight_to_fp16, x = x_73_cast_fp16)[name = tensor<string, []>("linear_51_cast_fp16")];
tensor<int32, [4]> var_641 = const()[name = tensor<string, []>("op_641"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_642_cast_fp16 = reshape(shape = var_641, x = linear_51_cast_fp16)[name = tensor<string, []>("op_642_cast_fp16")];
tensor<int32, [4]> v_17_perm_0 = const()[name = tensor<string, []>("v_17_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_9_y_0_to_fp16 = const()[name = tensor<string, []>("mul_9_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_9_cast_fp16 = mul(x = var_630_cast_fp16, y = mul_9_y_0_to_fp16)[name = tensor<string, []>("mul_9_cast_fp16")];
tensor<bool, []> matmul_8_transpose_y_0 = const()[name = tensor<string, []>("matmul_8_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_8_transpose_x_0 = const()[name = tensor<string, []>("matmul_8_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_64_perm_0 = const()[name = tensor<string, []>("transpose_64_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_65_perm_0 = const()[name = tensor<string, []>("transpose_65_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_65 = transpose(perm = transpose_65_perm_0, x = var_636_cast_fp16)[name = tensor<string, []>("transpose_85")];
tensor<fp16, [1, 12, 2999, 64]> transpose_64 = transpose(perm = transpose_64_perm_0, x = mul_9_cast_fp16)[name = tensor<string, []>("transpose_86")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_8_cast_fp16 = matmul(transpose_x = matmul_8_transpose_x_0, transpose_y = matmul_8_transpose_y_0, x = transpose_64, y = transpose_65)[name = tensor<string, []>("matmul_8_cast_fp16")];
tensor<int32, []> softmax_8_axis_0 = const()[name = tensor<string, []>("softmax_8_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_8_cast_fp16 = softmax(axis = softmax_8_axis_0, x = matmul_8_cast_fp16)[name = tensor<string, []>("softmax_8_cast_fp16")];
tensor<bool, []> attn_output_17_transpose_x_0 = const()[name = tensor<string, []>("attn_output_17_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_17_transpose_y_0 = const()[name = tensor<string, []>("attn_output_17_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_17_cast_fp16 = transpose(perm = v_17_perm_0, x = var_642_cast_fp16)[name = tensor<string, []>("transpose_87")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_17_cast_fp16 = matmul(transpose_x = attn_output_17_transpose_x_0, transpose_y = attn_output_17_transpose_y_0, x = softmax_8_cast_fp16, y = v_17_cast_fp16)[name = tensor<string, []>("attn_output_17_cast_fp16")];
tensor<int32, [4]> var_645_perm_0 = const()[name = tensor<string, []>("op_645_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_646 = const()[name = tensor<string, []>("op_646"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_645_cast_fp16 = transpose(perm = var_645_perm_0, x = attn_output_17_cast_fp16)[name = tensor<string, []>("transpose_84")];
tensor<fp16, [1, 2999, 768]> input_157_cast_fp16 = reshape(shape = var_646, x = var_645_cast_fp16)[name = tensor<string, []>("input_157_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_8_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(135591936)))];
tensor<fp16, [768]> model_encoder_transformer_layers_8_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(136771648)))];
tensor<fp16, [1, 2999, 768]> linear_52_cast_fp16 = linear(bias = model_encoder_transformer_layers_8_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_8_attention_out_proj_weight_to_fp16, x = input_157_cast_fp16)[name = tensor<string, []>("linear_52_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_161_cast_fp16 = add(x = x_73_cast_fp16, y = linear_52_cast_fp16)[name = tensor<string, []>("input_161_cast_fp16")];
tensor<int32, [1]> input_163_axes_0 = const()[name = tensor<string, []>("input_163_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_8_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(136773248)))];
tensor<fp16, [768]> model_encoder_transformer_layers_8_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(136774848)))];
tensor<fp16, [1, 2999, 768]> input_163_cast_fp16 = layer_norm(axes = input_163_axes_0, beta = model_encoder_transformer_layers_8_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_8_layer_norm_weight_to_fp16, x = input_161_cast_fp16)[name = tensor<string, []>("input_163_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_8_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(136776448)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_8_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(141495104)))];
tensor<fp16, [1, 2999, 3072]> linear_53_cast_fp16 = linear(bias = model_encoder_transformer_layers_8_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_8_feed_forward_intermediate_dense_weight_to_fp16, x = input_163_cast_fp16)[name = tensor<string, []>("linear_53_cast_fp16")];
tensor<string, []> input_165_mode_0 = const()[name = tensor<string, []>("input_165_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_165_cast_fp16 = gelu(mode = input_165_mode_0, x = linear_53_cast_fp16)[name = tensor<string, []>("input_165_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_8_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(141501312)))];
tensor<fp16, [768]> model_encoder_transformer_layers_8_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(146219968)))];
tensor<fp16, [1, 2999, 768]> linear_54_cast_fp16 = linear(bias = model_encoder_transformer_layers_8_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_8_feed_forward_output_dense_weight_to_fp16, x = input_165_cast_fp16)[name = tensor<string, []>("linear_54_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_171_cast_fp16 = add(x = input_163_cast_fp16, y = linear_54_cast_fp16)[name = tensor<string, []>("input_171_cast_fp16")];
tensor<int32, [1]> x_79_axes_0 = const()[name = tensor<string, []>("x_79_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_8_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(146221568)))];
tensor<fp16, [768]> model_encoder_transformer_layers_8_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_8_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(146223168)))];
tensor<fp16, [1, 2999, 768]> x_79_cast_fp16 = layer_norm(axes = x_79_axes_0, beta = model_encoder_transformer_layers_8_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_8_final_layer_norm_weight_to_fp16, x = input_171_cast_fp16)[name = tensor<string, []>("x_79_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_9_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(146224768)))];
tensor<fp16, [768]> model_encoder_transformer_layers_9_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(147404480)))];
tensor<fp16, [1, 2999, 768]> linear_55_cast_fp16 = linear(bias = model_encoder_transformer_layers_9_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_9_attention_q_proj_weight_to_fp16, x = x_79_cast_fp16)[name = tensor<string, []>("linear_55_cast_fp16")];
tensor<int32, [4]> var_686 = const()[name = tensor<string, []>("op_686"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_687_cast_fp16 = reshape(shape = var_686, x = linear_55_cast_fp16)[name = tensor<string, []>("op_687_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_9_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(147406080)))];
tensor<fp16, [768]> model_encoder_transformer_layers_9_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(148585792)))];
tensor<fp16, [1, 2999, 768]> linear_56_cast_fp16 = linear(bias = model_encoder_transformer_layers_9_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_9_attention_k_proj_weight_to_fp16, x = x_79_cast_fp16)[name = tensor<string, []>("linear_56_cast_fp16")];
tensor<int32, [4]> var_692 = const()[name = tensor<string, []>("op_692"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_693_cast_fp16 = reshape(shape = var_692, x = linear_56_cast_fp16)[name = tensor<string, []>("op_693_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_9_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(148587392)))];
tensor<fp16, [768]> model_encoder_transformer_layers_9_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(149767104)))];
tensor<fp16, [1, 2999, 768]> linear_57_cast_fp16 = linear(bias = model_encoder_transformer_layers_9_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_9_attention_v_proj_weight_to_fp16, x = x_79_cast_fp16)[name = tensor<string, []>("linear_57_cast_fp16")];
tensor<int32, [4]> var_698 = const()[name = tensor<string, []>("op_698"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_699_cast_fp16 = reshape(shape = var_698, x = linear_57_cast_fp16)[name = tensor<string, []>("op_699_cast_fp16")];
tensor<int32, [4]> v_19_perm_0 = const()[name = tensor<string, []>("v_19_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_10_y_0_to_fp16 = const()[name = tensor<string, []>("mul_10_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_10_cast_fp16 = mul(x = var_687_cast_fp16, y = mul_10_y_0_to_fp16)[name = tensor<string, []>("mul_10_cast_fp16")];
tensor<bool, []> matmul_9_transpose_y_0 = const()[name = tensor<string, []>("matmul_9_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_9_transpose_x_0 = const()[name = tensor<string, []>("matmul_9_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_66_perm_0 = const()[name = tensor<string, []>("transpose_66_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_67_perm_0 = const()[name = tensor<string, []>("transpose_67_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_67 = transpose(perm = transpose_67_perm_0, x = var_693_cast_fp16)[name = tensor<string, []>("transpose_81")];
tensor<fp16, [1, 12, 2999, 64]> transpose_66 = transpose(perm = transpose_66_perm_0, x = mul_10_cast_fp16)[name = tensor<string, []>("transpose_82")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_9_cast_fp16 = matmul(transpose_x = matmul_9_transpose_x_0, transpose_y = matmul_9_transpose_y_0, x = transpose_66, y = transpose_67)[name = tensor<string, []>("matmul_9_cast_fp16")];
tensor<int32, []> softmax_9_axis_0 = const()[name = tensor<string, []>("softmax_9_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_9_cast_fp16 = softmax(axis = softmax_9_axis_0, x = matmul_9_cast_fp16)[name = tensor<string, []>("softmax_9_cast_fp16")];
tensor<bool, []> attn_output_19_transpose_x_0 = const()[name = tensor<string, []>("attn_output_19_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_19_transpose_y_0 = const()[name = tensor<string, []>("attn_output_19_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_19_cast_fp16 = transpose(perm = v_19_perm_0, x = var_699_cast_fp16)[name = tensor<string, []>("transpose_83")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_19_cast_fp16 = matmul(transpose_x = attn_output_19_transpose_x_0, transpose_y = attn_output_19_transpose_y_0, x = softmax_9_cast_fp16, y = v_19_cast_fp16)[name = tensor<string, []>("attn_output_19_cast_fp16")];
tensor<int32, [4]> var_702_perm_0 = const()[name = tensor<string, []>("op_702_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_703 = const()[name = tensor<string, []>("op_703"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_702_cast_fp16 = transpose(perm = var_702_perm_0, x = attn_output_19_cast_fp16)[name = tensor<string, []>("transpose_80")];
tensor<fp16, [1, 2999, 768]> input_173_cast_fp16 = reshape(shape = var_703, x = var_702_cast_fp16)[name = tensor<string, []>("input_173_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_9_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(149768704)))];
tensor<fp16, [768]> model_encoder_transformer_layers_9_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(150948416)))];
tensor<fp16, [1, 2999, 768]> linear_58_cast_fp16 = linear(bias = model_encoder_transformer_layers_9_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_9_attention_out_proj_weight_to_fp16, x = input_173_cast_fp16)[name = tensor<string, []>("linear_58_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_177_cast_fp16 = add(x = x_79_cast_fp16, y = linear_58_cast_fp16)[name = tensor<string, []>("input_177_cast_fp16")];
tensor<int32, [1]> input_179_axes_0 = const()[name = tensor<string, []>("input_179_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_9_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(150950016)))];
tensor<fp16, [768]> model_encoder_transformer_layers_9_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(150951616)))];
tensor<fp16, [1, 2999, 768]> input_179_cast_fp16 = layer_norm(axes = input_179_axes_0, beta = model_encoder_transformer_layers_9_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_9_layer_norm_weight_to_fp16, x = input_177_cast_fp16)[name = tensor<string, []>("input_179_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_9_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(150953216)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_9_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(155671872)))];
tensor<fp16, [1, 2999, 3072]> linear_59_cast_fp16 = linear(bias = model_encoder_transformer_layers_9_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_9_feed_forward_intermediate_dense_weight_to_fp16, x = input_179_cast_fp16)[name = tensor<string, []>("linear_59_cast_fp16")];
tensor<string, []> input_181_mode_0 = const()[name = tensor<string, []>("input_181_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_181_cast_fp16 = gelu(mode = input_181_mode_0, x = linear_59_cast_fp16)[name = tensor<string, []>("input_181_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_9_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(155678080)))];
tensor<fp16, [768]> model_encoder_transformer_layers_9_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(160396736)))];
tensor<fp16, [1, 2999, 768]> linear_60_cast_fp16 = linear(bias = model_encoder_transformer_layers_9_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_9_feed_forward_output_dense_weight_to_fp16, x = input_181_cast_fp16)[name = tensor<string, []>("linear_60_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_187_cast_fp16 = add(x = input_179_cast_fp16, y = linear_60_cast_fp16)[name = tensor<string, []>("input_187_cast_fp16")];
tensor<int32, [1]> x_85_axes_0 = const()[name = tensor<string, []>("x_85_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_9_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(160398336)))];
tensor<fp16, [768]> model_encoder_transformer_layers_9_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_9_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(160399936)))];
tensor<fp16, [1, 2999, 768]> x_85_cast_fp16 = layer_norm(axes = x_85_axes_0, beta = model_encoder_transformer_layers_9_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_9_final_layer_norm_weight_to_fp16, x = input_187_cast_fp16)[name = tensor<string, []>("x_85_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_10_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(160401536)))];
tensor<fp16, [768]> model_encoder_transformer_layers_10_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(161581248)))];
tensor<fp16, [1, 2999, 768]> linear_61_cast_fp16 = linear(bias = model_encoder_transformer_layers_10_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_10_attention_q_proj_weight_to_fp16, x = x_85_cast_fp16)[name = tensor<string, []>("linear_61_cast_fp16")];
tensor<int32, [4]> var_743 = const()[name = tensor<string, []>("op_743"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_744_cast_fp16 = reshape(shape = var_743, x = linear_61_cast_fp16)[name = tensor<string, []>("op_744_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_10_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(161582848)))];
tensor<fp16, [768]> model_encoder_transformer_layers_10_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(162762560)))];
tensor<fp16, [1, 2999, 768]> linear_62_cast_fp16 = linear(bias = model_encoder_transformer_layers_10_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_10_attention_k_proj_weight_to_fp16, x = x_85_cast_fp16)[name = tensor<string, []>("linear_62_cast_fp16")];
tensor<int32, [4]> var_749 = const()[name = tensor<string, []>("op_749"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_750_cast_fp16 = reshape(shape = var_749, x = linear_62_cast_fp16)[name = tensor<string, []>("op_750_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_10_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(162764160)))];
tensor<fp16, [768]> model_encoder_transformer_layers_10_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(163943872)))];
tensor<fp16, [1, 2999, 768]> linear_63_cast_fp16 = linear(bias = model_encoder_transformer_layers_10_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_10_attention_v_proj_weight_to_fp16, x = x_85_cast_fp16)[name = tensor<string, []>("linear_63_cast_fp16")];
tensor<int32, [4]> var_755 = const()[name = tensor<string, []>("op_755"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_756_cast_fp16 = reshape(shape = var_755, x = linear_63_cast_fp16)[name = tensor<string, []>("op_756_cast_fp16")];
tensor<int32, [4]> v_21_perm_0 = const()[name = tensor<string, []>("v_21_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_11_y_0_to_fp16 = const()[name = tensor<string, []>("mul_11_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_11_cast_fp16 = mul(x = var_744_cast_fp16, y = mul_11_y_0_to_fp16)[name = tensor<string, []>("mul_11_cast_fp16")];
tensor<bool, []> matmul_10_transpose_y_0 = const()[name = tensor<string, []>("matmul_10_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_10_transpose_x_0 = const()[name = tensor<string, []>("matmul_10_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_68_perm_0 = const()[name = tensor<string, []>("transpose_68_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_69_perm_0 = const()[name = tensor<string, []>("transpose_69_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_69 = transpose(perm = transpose_69_perm_0, x = var_750_cast_fp16)[name = tensor<string, []>("transpose_77")];
tensor<fp16, [1, 12, 2999, 64]> transpose_68 = transpose(perm = transpose_68_perm_0, x = mul_11_cast_fp16)[name = tensor<string, []>("transpose_78")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_10_cast_fp16 = matmul(transpose_x = matmul_10_transpose_x_0, transpose_y = matmul_10_transpose_y_0, x = transpose_68, y = transpose_69)[name = tensor<string, []>("matmul_10_cast_fp16")];
tensor<int32, []> softmax_10_axis_0 = const()[name = tensor<string, []>("softmax_10_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_10_cast_fp16 = softmax(axis = softmax_10_axis_0, x = matmul_10_cast_fp16)[name = tensor<string, []>("softmax_10_cast_fp16")];
tensor<bool, []> attn_output_21_transpose_x_0 = const()[name = tensor<string, []>("attn_output_21_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_21_transpose_y_0 = const()[name = tensor<string, []>("attn_output_21_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_21_cast_fp16 = transpose(perm = v_21_perm_0, x = var_756_cast_fp16)[name = tensor<string, []>("transpose_79")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_21_cast_fp16 = matmul(transpose_x = attn_output_21_transpose_x_0, transpose_y = attn_output_21_transpose_y_0, x = softmax_10_cast_fp16, y = v_21_cast_fp16)[name = tensor<string, []>("attn_output_21_cast_fp16")];
tensor<int32, [4]> var_759_perm_0 = const()[name = tensor<string, []>("op_759_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_760 = const()[name = tensor<string, []>("op_760"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_759_cast_fp16 = transpose(perm = var_759_perm_0, x = attn_output_21_cast_fp16)[name = tensor<string, []>("transpose_76")];
tensor<fp16, [1, 2999, 768]> input_189_cast_fp16 = reshape(shape = var_760, x = var_759_cast_fp16)[name = tensor<string, []>("input_189_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_10_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(163945472)))];
tensor<fp16, [768]> model_encoder_transformer_layers_10_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(165125184)))];
tensor<fp16, [1, 2999, 768]> linear_64_cast_fp16 = linear(bias = model_encoder_transformer_layers_10_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_10_attention_out_proj_weight_to_fp16, x = input_189_cast_fp16)[name = tensor<string, []>("linear_64_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_193_cast_fp16 = add(x = x_85_cast_fp16, y = linear_64_cast_fp16)[name = tensor<string, []>("input_193_cast_fp16")];
tensor<int32, [1]> input_195_axes_0 = const()[name = tensor<string, []>("input_195_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_10_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(165126784)))];
tensor<fp16, [768]> model_encoder_transformer_layers_10_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(165128384)))];
tensor<fp16, [1, 2999, 768]> input_195_cast_fp16 = layer_norm(axes = input_195_axes_0, beta = model_encoder_transformer_layers_10_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_10_layer_norm_weight_to_fp16, x = input_193_cast_fp16)[name = tensor<string, []>("input_195_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_10_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(165129984)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_10_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(169848640)))];
tensor<fp16, [1, 2999, 3072]> linear_65_cast_fp16 = linear(bias = model_encoder_transformer_layers_10_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_10_feed_forward_intermediate_dense_weight_to_fp16, x = input_195_cast_fp16)[name = tensor<string, []>("linear_65_cast_fp16")];
tensor<string, []> input_197_mode_0 = const()[name = tensor<string, []>("input_197_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_197_cast_fp16 = gelu(mode = input_197_mode_0, x = linear_65_cast_fp16)[name = tensor<string, []>("input_197_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_10_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(169854848)))];
tensor<fp16, [768]> model_encoder_transformer_layers_10_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(174573504)))];
tensor<fp16, [1, 2999, 768]> linear_66_cast_fp16 = linear(bias = model_encoder_transformer_layers_10_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_10_feed_forward_output_dense_weight_to_fp16, x = input_197_cast_fp16)[name = tensor<string, []>("linear_66_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_203_cast_fp16 = add(x = input_195_cast_fp16, y = linear_66_cast_fp16)[name = tensor<string, []>("input_203_cast_fp16")];
tensor<int32, [1]> x_91_axes_0 = const()[name = tensor<string, []>("x_91_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_10_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(174575104)))];
tensor<fp16, [768]> model_encoder_transformer_layers_10_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_10_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(174576704)))];
tensor<fp16, [1, 2999, 768]> x_91_cast_fp16 = layer_norm(axes = x_91_axes_0, beta = model_encoder_transformer_layers_10_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_10_final_layer_norm_weight_to_fp16, x = input_203_cast_fp16)[name = tensor<string, []>("x_91_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_11_attention_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_q_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(174578304)))];
tensor<fp16, [768]> model_encoder_transformer_layers_11_attention_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(175758016)))];
tensor<fp16, [1, 2999, 768]> linear_67_cast_fp16 = linear(bias = model_encoder_transformer_layers_11_attention_q_proj_bias_to_fp16, weight = model_encoder_transformer_layers_11_attention_q_proj_weight_to_fp16, x = x_91_cast_fp16)[name = tensor<string, []>("linear_67_cast_fp16")];
tensor<int32, [4]> var_800 = const()[name = tensor<string, []>("op_800"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_801_cast_fp16 = reshape(shape = var_800, x = linear_67_cast_fp16)[name = tensor<string, []>("op_801_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_11_attention_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_k_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(175759616)))];
tensor<fp16, [768]> model_encoder_transformer_layers_11_attention_k_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(176939328)))];
tensor<fp16, [1, 2999, 768]> linear_68_cast_fp16 = linear(bias = model_encoder_transformer_layers_11_attention_k_proj_bias_to_fp16, weight = model_encoder_transformer_layers_11_attention_k_proj_weight_to_fp16, x = x_91_cast_fp16)[name = tensor<string, []>("linear_68_cast_fp16")];
tensor<int32, [4]> var_806 = const()[name = tensor<string, []>("op_806"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_807_cast_fp16 = reshape(shape = var_806, x = linear_68_cast_fp16)[name = tensor<string, []>("op_807_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_11_attention_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_v_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(176940928)))];
tensor<fp16, [768]> model_encoder_transformer_layers_11_attention_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(178120640)))];
tensor<fp16, [1, 2999, 768]> linear_69_cast_fp16 = linear(bias = model_encoder_transformer_layers_11_attention_v_proj_bias_to_fp16, weight = model_encoder_transformer_layers_11_attention_v_proj_weight_to_fp16, x = x_91_cast_fp16)[name = tensor<string, []>("linear_69_cast_fp16")];
tensor<int32, [4]> var_812 = const()[name = tensor<string, []>("op_812"), val = tensor<int32, [4]>([1, 2999, 12, 64])];
tensor<fp16, [1, 2999, 12, 64]> var_813_cast_fp16 = reshape(shape = var_812, x = linear_69_cast_fp16)[name = tensor<string, []>("op_813_cast_fp16")];
tensor<int32, [4]> v_perm_0 = const()[name = tensor<string, []>("v_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<fp16, []> mul_12_y_0_to_fp16 = const()[name = tensor<string, []>("mul_12_y_0_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 2999, 12, 64]> mul_12_cast_fp16 = mul(x = var_801_cast_fp16, y = mul_12_y_0_to_fp16)[name = tensor<string, []>("mul_12_cast_fp16")];
tensor<bool, []> matmul_11_transpose_y_0 = const()[name = tensor<string, []>("matmul_11_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<bool, []> matmul_11_transpose_x_0 = const()[name = tensor<string, []>("matmul_11_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<int32, [4]> transpose_70_perm_0 = const()[name = tensor<string, []>("transpose_70_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_71_perm_0 = const()[name = tensor<string, []>("transpose_71_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 2999, 64]> transpose_71 = transpose(perm = transpose_71_perm_0, x = var_807_cast_fp16)[name = tensor<string, []>("transpose_73")];
tensor<fp16, [1, 12, 2999, 64]> transpose_70 = transpose(perm = transpose_70_perm_0, x = mul_12_cast_fp16)[name = tensor<string, []>("transpose_74")];
tensor<fp16, [1, 12, 2999, 2999]> matmul_11_cast_fp16 = matmul(transpose_x = matmul_11_transpose_x_0, transpose_y = matmul_11_transpose_y_0, x = transpose_70, y = transpose_71)[name = tensor<string, []>("matmul_11_cast_fp16")];
tensor<int32, []> softmax_11_axis_0 = const()[name = tensor<string, []>("softmax_11_axis_0"), val = tensor<int32, []>(-1)];
tensor<fp16, [1, 12, 2999, 2999]> softmax_11_cast_fp16 = softmax(axis = softmax_11_axis_0, x = matmul_11_cast_fp16)[name = tensor<string, []>("softmax_11_cast_fp16")];
tensor<bool, []> attn_output_transpose_x_0 = const()[name = tensor<string, []>("attn_output_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_output_transpose_y_0 = const()[name = tensor<string, []>("attn_output_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 12, 2999, 64]> v_cast_fp16 = transpose(perm = v_perm_0, x = var_813_cast_fp16)[name = tensor<string, []>("transpose_75")];
tensor<fp16, [1, 12, 2999, 64]> attn_output_cast_fp16 = matmul(transpose_x = attn_output_transpose_x_0, transpose_y = attn_output_transpose_y_0, x = softmax_11_cast_fp16, y = v_cast_fp16)[name = tensor<string, []>("attn_output_cast_fp16")];
tensor<int32, [4]> var_816_perm_0 = const()[name = tensor<string, []>("op_816_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_817 = const()[name = tensor<string, []>("op_817"), val = tensor<int32, [3]>([1, -1, 768])];
tensor<fp16, [1, 2999, 12, 64]> var_816_cast_fp16 = transpose(perm = var_816_perm_0, x = attn_output_cast_fp16)[name = tensor<string, []>("transpose_72")];
tensor<fp16, [1, 2999, 768]> input_205_cast_fp16 = reshape(shape = var_817, x = var_816_cast_fp16)[name = tensor<string, []>("input_205_cast_fp16")];
tensor<fp16, [768, 768]> model_encoder_transformer_layers_11_attention_out_proj_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_out_proj_weight_to_fp16"), val = tensor<fp16, [768, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(178122240)))];
tensor<fp16, [768]> model_encoder_transformer_layers_11_attention_out_proj_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_attention_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(179301952)))];
tensor<fp16, [1, 2999, 768]> linear_70_cast_fp16 = linear(bias = model_encoder_transformer_layers_11_attention_out_proj_bias_to_fp16, weight = model_encoder_transformer_layers_11_attention_out_proj_weight_to_fp16, x = input_205_cast_fp16)[name = tensor<string, []>("linear_70_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_209_cast_fp16 = add(x = x_91_cast_fp16, y = linear_70_cast_fp16)[name = tensor<string, []>("input_209_cast_fp16")];
tensor<int32, [1]> input_211_axes_0 = const()[name = tensor<string, []>("input_211_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_11_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(179303552)))];
tensor<fp16, [768]> model_encoder_transformer_layers_11_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(179305152)))];
tensor<fp16, [1, 2999, 768]> input_211_cast_fp16 = layer_norm(axes = input_211_axes_0, beta = model_encoder_transformer_layers_11_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_11_layer_norm_weight_to_fp16, x = input_209_cast_fp16)[name = tensor<string, []>("input_211_cast_fp16")];
tensor<fp16, [3072, 768]> model_encoder_transformer_layers_11_feed_forward_intermediate_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_feed_forward_intermediate_dense_weight_to_fp16"), val = tensor<fp16, [3072, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(179306752)))];
tensor<fp16, [3072]> model_encoder_transformer_layers_11_feed_forward_intermediate_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_feed_forward_intermediate_dense_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(184025408)))];
tensor<fp16, [1, 2999, 3072]> linear_71_cast_fp16 = linear(bias = model_encoder_transformer_layers_11_feed_forward_intermediate_dense_bias_to_fp16, weight = model_encoder_transformer_layers_11_feed_forward_intermediate_dense_weight_to_fp16, x = input_211_cast_fp16)[name = tensor<string, []>("linear_71_cast_fp16")];
tensor<string, []> input_213_mode_0 = const()[name = tensor<string, []>("input_213_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 2999, 3072]> input_213_cast_fp16 = gelu(mode = input_213_mode_0, x = linear_71_cast_fp16)[name = tensor<string, []>("input_213_cast_fp16")];
tensor<fp16, [768, 3072]> model_encoder_transformer_layers_11_feed_forward_output_dense_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_feed_forward_output_dense_weight_to_fp16"), val = tensor<fp16, [768, 3072]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(184031616)))];
tensor<fp16, [768]> model_encoder_transformer_layers_11_feed_forward_output_dense_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_feed_forward_output_dense_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(188750272)))];
tensor<fp16, [1, 2999, 768]> linear_72_cast_fp16 = linear(bias = model_encoder_transformer_layers_11_feed_forward_output_dense_bias_to_fp16, weight = model_encoder_transformer_layers_11_feed_forward_output_dense_weight_to_fp16, x = input_213_cast_fp16)[name = tensor<string, []>("linear_72_cast_fp16")];
tensor<fp16, [1, 2999, 768]> input_219_cast_fp16 = add(x = input_211_cast_fp16, y = linear_72_cast_fp16)[name = tensor<string, []>("input_219_cast_fp16")];
tensor<int32, [1]> input_axes_0 = const()[name = tensor<string, []>("input_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> model_encoder_transformer_layers_11_final_layer_norm_weight_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(188751872)))];
tensor<fp16, [768]> model_encoder_transformer_layers_11_final_layer_norm_bias_to_fp16 = const()[name = tensor<string, []>("model_encoder_transformer_layers_11_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(188753472)))];
tensor<fp16, [1, 2999, 768]> input_cast_fp16 = layer_norm(axes = input_axes_0, beta = model_encoder_transformer_layers_11_final_layer_norm_bias_to_fp16, epsilon = var_14_to_fp16, gamma = model_encoder_transformer_layers_11_final_layer_norm_weight_to_fp16, x = input_219_cast_fp16)[name = tensor<string, []>("input_cast_fp16")];
tensor<fp16, [29, 768]> model_aux_weight_to_fp16 = const()[name = tensor<string, []>("model_aux_weight_to_fp16"), val = tensor<fp16, [29, 768]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(188755072)))];
tensor<fp16, [29]> model_aux_bias_to_fp16 = const()[name = tensor<string, []>("model_aux_bias_to_fp16"), val = tensor<fp16, [29]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(188799680)))];
tensor<fp16, [1, 2999, 29]> linear_73_cast_fp16 = linear(bias = model_aux_bias_to_fp16, weight = model_aux_weight_to_fp16, x = input_cast_fp16)[name = tensor<string, []>("linear_73_cast_fp16")];
tensor<int32, [1]> var_850_axes_0 = const()[name = tensor<string, []>("op_850_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<bool, []> var_850_keep_dims_0 = const()[name = tensor<string, []>("op_850_keep_dims_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 2999, 1]> var_850_cast_fp16 = reduce_log_sum_exp(axes = var_850_axes_0, keep_dims = var_850_keep_dims_0, x = linear_73_cast_fp16)[name = tensor<string, []>("op_850_cast_fp16")];
tensor<fp16, [1, 2999, 29]> var_852_cast_fp16 = sub(x = linear_73_cast_fp16, y = var_850_cast_fp16)[name = tensor<string, []>("op_852_cast_fp16")];
tensor<string, []> var_852_cast_fp16_to_fp32_dtype_0 = const()[name = tensor<string, []>("op_852_cast_fp16_to_fp32_dtype_0"), val = tensor<string, []>("fp32")];
tensor<fp32, [1, 2999, 29]> emissions = cast(dtype = var_852_cast_fp16_to_fp32_dtype_0, x = var_852_cast_fp16)[name = tensor<string, []>("cast_48")];
} -> (emissions);
} |