File size: 144,323 Bytes
efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b a44e8b1 efa763b |
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 |
program(1.0)
[buildInfo = dict<tensor<string, []>, tensor<string, []>>({{"coremlc-component-MIL", "3304.5.2"}, {"coremlc-version", "3304.6.2"}, {"coremltools-component-torch", "2.2.1"}, {"coremltools-source-dialect", "TorchScript"}, {"coremltools-version", "7.1"}})]
{
func main<ios16>(tensor<int32, [1]> cache_length, tensor<fp16, [1, 224]> decoder_key_padding_mask, tensor<fp16, [1, 384, 1, 1500]> encoder_output_embeds, tensor<int32, [1]> input_ids, tensor<fp16, [1, 1536, 1, 224]> key_cache, tensor<fp16, [1, 224]> kv_cache_update_mask, tensor<fp16, [1, 1536, 1, 224]> value_cache) {
tensor<int32, []> var_24_axis_0 = const()[name = tensor<string, []>("op_24_axis_0"), val = tensor<int32, []>(0)];
tensor<int32, []> var_24_batch_dims_0 = const()[name = tensor<string, []>("op_24_batch_dims_0"), val = tensor<int32, []>(0)];
tensor<fp16, [51864, 384]> embed_tokens_weight_to_fp16 = const()[name = tensor<string, []>("embed_tokens_weight_to_fp16"), val = tensor<fp16, [51864, 384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(64)))];
tensor<fp16, [1, 384]> var_24_cast_fp16 = gather(axis = var_24_axis_0, batch_dims = var_24_batch_dims_0, indices = input_ids, x = embed_tokens_weight_to_fp16)[name = tensor<string, []>("op_24_cast_fp16")];
tensor<int32, []> var_28_axis_0 = const()[name = tensor<string, []>("op_28_axis_0"), val = tensor<int32, []>(0)];
tensor<int32, []> var_28_batch_dims_0 = const()[name = tensor<string, []>("op_28_batch_dims_0"), val = tensor<int32, []>(0)];
tensor<fp16, [448, 384]> embed_positions_weight_to_fp16 = const()[name = tensor<string, []>("embed_positions_weight_to_fp16"), val = tensor<fp16, [448, 384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(39831680)))];
tensor<fp16, [1, 384]> var_28_cast_fp16 = gather(axis = var_28_axis_0, batch_dims = var_28_batch_dims_0, indices = cache_length, x = embed_positions_weight_to_fp16)[name = tensor<string, []>("op_28_cast_fp16")];
tensor<fp16, [1, 384]> hidden_states_1_cast_fp16 = add(x = var_24_cast_fp16, y = var_28_cast_fp16)[name = tensor<string, []>("hidden_states_1_cast_fp16")];
tensor<int32, [1]> var_42_axes_0 = const()[name = tensor<string, []>("op_42_axes_0"), val = tensor<int32, [1]>([2])];
tensor<fp16, [1, 384, 1]> var_42_cast_fp16 = expand_dims(axes = var_42_axes_0, x = hidden_states_1_cast_fp16)[name = tensor<string, []>("op_42_cast_fp16")];
tensor<int32, [1]> inputs_1_axes_0 = const()[name = tensor<string, []>("inputs_1_axes_0"), val = tensor<int32, [1]>([3])];
tensor<fp16, [1, 384, 1, 1]> inputs_1_cast_fp16 = expand_dims(axes = inputs_1_axes_0, x = var_42_cast_fp16)[name = tensor<string, []>("inputs_1_cast_fp16")];
tensor<int32, [4]> tile_0 = const()[name = tensor<string, []>("tile_0"), val = tensor<int32, [4]>([384, 384, 384, 384])];
tensor<int32, []> var_47_axis_0 = const()[name = tensor<string, []>("op_47_axis_0"), val = tensor<int32, []>(1)];
tensor<fp16, [1, 384, 1, 224]> var_47_cast_fp16_0, tensor<fp16, [1, 384, 1, 224]> var_47_cast_fp16_1, tensor<fp16, [1, 384, 1, 224]> var_47_cast_fp16_2, tensor<fp16, [1, 384, 1, 224]> var_47_cast_fp16_3 = split(axis = var_47_axis_0, split_sizes = tile_0, x = key_cache)[name = tensor<string, []>("op_47_cast_fp16")];
tensor<int32, [4]> tile_1 = const()[name = tensor<string, []>("tile_1"), val = tensor<int32, [4]>([384, 384, 384, 384])];
tensor<int32, []> var_54_axis_0 = const()[name = tensor<string, []>("op_54_axis_0"), val = tensor<int32, []>(1)];
tensor<fp16, [1, 384, 1, 224]> var_54_cast_fp16_0, tensor<fp16, [1, 384, 1, 224]> var_54_cast_fp16_1, tensor<fp16, [1, 384, 1, 224]> var_54_cast_fp16_2, tensor<fp16, [1, 384, 1, 224]> var_54_cast_fp16_3 = split(axis = var_54_axis_0, split_sizes = tile_1, x = value_cache)[name = tensor<string, []>("op_54_cast_fp16")];
tensor<int32, []> var_64 = const()[name = tensor<string, []>("op_64"), val = tensor<int32, []>(3)];
tensor<int32, []> var_71 = const()[name = tensor<string, []>("op_71"), val = tensor<int32, []>(1)];
tensor<bool, []> var_72 = const()[name = tensor<string, []>("op_72"), val = tensor<bool, []>(true)];
tensor<int32, [1]> var_84 = const()[name = tensor<string, []>("op_84"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_1_cast_fp16 = reduce_mean(axes = var_84, keep_dims = var_72, x = inputs_1_cast_fp16)[name = tensor<string, []>("channels_mean_1_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_1_cast_fp16 = sub(x = inputs_1_cast_fp16, y = channels_mean_1_cast_fp16)[name = tensor<string, []>("zero_mean_1_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_1_cast_fp16 = mul(x = zero_mean_1_cast_fp16, y = zero_mean_1_cast_fp16)[name = tensor<string, []>("zero_mean_sq_1_cast_fp16")];
tensor<int32, [1]> var_88 = const()[name = tensor<string, []>("op_88"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_89_cast_fp16 = reduce_mean(axes = var_88, keep_dims = var_72, x = zero_mean_sq_1_cast_fp16)[name = tensor<string, []>("op_89_cast_fp16")];
tensor<fp16, []> var_90_to_fp16 = const()[name = tensor<string, []>("op_90_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_91_cast_fp16 = add(x = var_89_cast_fp16, y = var_90_to_fp16)[name = tensor<string, []>("op_91_cast_fp16")];
tensor<fp16, []> denom_1_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_1_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_1_cast_fp16 = rsqrt(epsilon = denom_1_epsilon_0_to_fp16, x = var_91_cast_fp16)[name = tensor<string, []>("denom_1_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_1_cast_fp16 = mul(x = zero_mean_1_cast_fp16, y = denom_1_cast_fp16)[name = tensor<string, []>("out_1_cast_fp16")];
tensor<fp16, [384]> obj_1_mean_0_to_fp16 = const()[name = tensor<string, []>("obj_1_mean_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40175808)))];
tensor<fp16, [384]> obj_1_variance_0_to_fp16 = const()[name = tensor<string, []>("obj_1_variance_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40176640)))];
tensor<fp16, [384]> obj_1_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_1_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40177472)))];
tensor<fp16, [384]> obj_1_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_1_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40178304)))];
tensor<fp16, []> obj_1_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_1_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_1_cast_fp16 = batch_norm(beta = obj_1_beta_0_to_fp16, epsilon = obj_1_epsilon_0_to_fp16, gamma = obj_1_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_1_cast_fp16)[name = tensor<string, []>("obj_1_cast_fp16")];
tensor<int32, [2]> var_106 = const()[name = tensor<string, []>("op_106"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_108 = const()[name = tensor<string, []>("op_108"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_1_pad_type_0 = const()[name = tensor<string, []>("query_1_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_1_pad_0 = const()[name = tensor<string, []>("query_1_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_self_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_self_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40179136)))];
tensor<fp16, [384]> layers_0_self_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40474112)))];
tensor<fp16, [1, 384, 1, 1]> query_1_cast_fp16 = conv(bias = layers_0_self_attn_q_proj_bias_to_fp16, dilations = var_108, groups = var_71, pad = query_1_pad_0, pad_type = query_1_pad_type_0, strides = var_106, weight = layers_0_self_attn_q_proj_weight_to_fp16, x = obj_1_cast_fp16)[name = tensor<string, []>("query_1_cast_fp16")];
tensor<int32, [2]> var_112 = const()[name = tensor<string, []>("op_112"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_114 = const()[name = tensor<string, []>("op_114"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_key_1_pad_type_0 = const()[name = tensor<string, []>("current_key_1_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_key_1_pad_0 = const()[name = tensor<string, []>("current_key_1_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_self_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_self_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40474944)))];
tensor<fp16, [1, 384, 1, 1]> current_key_1_cast_fp16 = conv(dilations = var_114, groups = var_71, pad = current_key_1_pad_0, pad_type = current_key_1_pad_type_0, strides = var_112, weight = layers_0_self_attn_k_proj_weight_to_fp16, x = obj_1_cast_fp16)[name = tensor<string, []>("current_key_1_cast_fp16")];
tensor<int32, [2]> var_119 = const()[name = tensor<string, []>("op_119"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_121 = const()[name = tensor<string, []>("op_121"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_value_1_pad_type_0 = const()[name = tensor<string, []>("current_value_1_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_value_1_pad_0 = const()[name = tensor<string, []>("current_value_1_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_self_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_self_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(40769920)))];
tensor<fp16, [384]> layers_0_self_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41064896)))];
tensor<fp16, [1, 384, 1, 1]> current_value_1_cast_fp16 = conv(bias = layers_0_self_attn_v_proj_bias_to_fp16, dilations = var_121, groups = var_71, pad = current_value_1_pad_0, pad_type = current_value_1_pad_type_0, strides = var_119, weight = layers_0_self_attn_v_proj_weight_to_fp16, x = obj_1_cast_fp16)[name = tensor<string, []>("current_value_1_cast_fp16")];
tensor<int32, [1]> var_125_axes_0 = const()[name = tensor<string, []>("op_125_axes_0"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 224]> var_125_cast_fp16 = expand_dims(axes = var_125_axes_0, x = kv_cache_update_mask)[name = tensor<string, []>("op_125_cast_fp16")];
tensor<int32, [1]> var_126_axes_0 = const()[name = tensor<string, []>("op_126_axes_0"), val = tensor<int32, [1]>([2])];
tensor<fp16, [1, 1, 1, 224]> var_126_cast_fp16 = expand_dims(axes = var_126_axes_0, x = var_125_cast_fp16)[name = tensor<string, []>("op_126_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_128_cast_fp16 = mul(x = current_key_1_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_128_cast_fp16")];
tensor<fp16, []> var_65_to_fp16 = const()[name = tensor<string, []>("op_65_to_fp16"), val = tensor<fp16, []>(0x1p+0)];
tensor<fp16, [1, 1, 1, 224]> var_129_cast_fp16 = sub(x = var_65_to_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_129_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_130_cast_fp16 = mul(x = var_47_cast_fp16_0, y = var_129_cast_fp16)[name = tensor<string, []>("op_130_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> key_1_cast_fp16 = add(x = var_128_cast_fp16, y = var_130_cast_fp16)[name = tensor<string, []>("key_1_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_132_cast_fp16 = mul(x = current_value_1_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_132_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_134_cast_fp16 = mul(x = var_54_cast_fp16_0, y = var_129_cast_fp16)[name = tensor<string, []>("op_134_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> value_1_cast_fp16 = add(x = var_132_cast_fp16, y = var_134_cast_fp16)[name = tensor<string, []>("value_1_cast_fp16")];
tensor<int32, [4]> var_137 = const()[name = tensor<string, []>("op_137"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_138_cast_fp16 = reshape(shape = var_137, x = query_1_cast_fp16)[name = tensor<string, []>("op_138_cast_fp16")];
tensor<fp16, []> var_139_to_fp16 = const()[name = tensor<string, []>("op_139_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_140_cast_fp16 = mul(x = var_138_cast_fp16, y = var_139_to_fp16)[name = tensor<string, []>("op_140_cast_fp16")];
tensor<int32, [4]> var_141 = const()[name = tensor<string, []>("op_141"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_142_cast_fp16 = reshape(shape = var_141, x = key_1_cast_fp16)[name = tensor<string, []>("op_142_cast_fp16")];
tensor<bool, []> mh_w_1_transpose_x_0 = const()[name = tensor<string, []>("mh_w_1_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_1_transpose_y_0 = const()[name = tensor<string, []>("mh_w_1_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 224]> mh_w_1_cast_fp16 = matmul(transpose_x = mh_w_1_transpose_x_0, transpose_y = mh_w_1_transpose_y_0, x = var_140_cast_fp16, y = var_142_cast_fp16)[name = tensor<string, []>("mh_w_1_cast_fp16")];
tensor<int32, [1]> var_146_axes_0 = const()[name = tensor<string, []>("op_146_axes_0"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 224]> var_146_cast_fp16 = expand_dims(axes = var_146_axes_0, x = decoder_key_padding_mask)[name = tensor<string, []>("op_146_cast_fp16")];
tensor<int32, [1]> var_147_axes_0 = const()[name = tensor<string, []>("op_147_axes_0"), val = tensor<int32, [1]>([2])];
tensor<fp16, [1, 1, 1, 224]> var_147_cast_fp16 = expand_dims(axes = var_147_axes_0, x = var_146_cast_fp16)[name = tensor<string, []>("op_147_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> mh_w_3_cast_fp16 = add(x = mh_w_1_cast_fp16, y = var_147_cast_fp16)[name = tensor<string, []>("mh_w_3_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> var_150_cast_fp16 = softmax(axis = var_64, x = mh_w_3_cast_fp16)[name = tensor<string, []>("op_150_cast_fp16")];
tensor<int32, [4]> var_151 = const()[name = tensor<string, []>("op_151"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_152_cast_fp16 = reshape(shape = var_151, x = value_1_cast_fp16)[name = tensor<string, []>("op_152_cast_fp16")];
tensor<bool, []> attn_1_transpose_x_0 = const()[name = tensor<string, []>("attn_1_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_1_transpose_y_0 = const()[name = tensor<string, []>("attn_1_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_1_cast_fp16 = matmul(transpose_x = attn_1_transpose_x_0, transpose_y = attn_1_transpose_y_0, x = var_152_cast_fp16, y = var_150_cast_fp16)[name = tensor<string, []>("attn_1_cast_fp16")];
tensor<int32, [4]> var_155 = const()[name = tensor<string, []>("op_155"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_1_cast_fp16 = reshape(shape = var_155, x = attn_1_cast_fp16)[name = tensor<string, []>("input_1_cast_fp16")];
tensor<int32, [2]> var_159 = const()[name = tensor<string, []>("op_159"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_161 = const()[name = tensor<string, []>("op_161"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_7_pad_type_0 = const()[name = tensor<string, []>("obj_7_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_7_pad_0 = const()[name = tensor<string, []>("obj_7_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_self_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_self_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41065728)))];
tensor<fp16, [384]> layers_0_self_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_self_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41360704)))];
tensor<fp16, [1, 384, 1, 1]> obj_7_cast_fp16 = conv(bias = layers_0_self_attn_o_proj_bias_to_fp16, dilations = var_161, groups = var_71, pad = obj_7_pad_0, pad_type = obj_7_pad_type_0, strides = var_159, weight = layers_0_self_attn_o_proj_weight_to_fp16, x = input_1_cast_fp16)[name = tensor<string, []>("obj_7_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_3_cast_fp16 = add(x = inputs_1_cast_fp16, y = obj_7_cast_fp16)[name = tensor<string, []>("inputs_3_cast_fp16")];
tensor<int32, [1]> var_171 = const()[name = tensor<string, []>("op_171"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_3_cast_fp16 = reduce_mean(axes = var_171, keep_dims = var_72, x = inputs_3_cast_fp16)[name = tensor<string, []>("channels_mean_3_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_3_cast_fp16 = sub(x = inputs_3_cast_fp16, y = channels_mean_3_cast_fp16)[name = tensor<string, []>("zero_mean_3_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_3_cast_fp16 = mul(x = zero_mean_3_cast_fp16, y = zero_mean_3_cast_fp16)[name = tensor<string, []>("zero_mean_sq_3_cast_fp16")];
tensor<int32, [1]> var_175 = const()[name = tensor<string, []>("op_175"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_176_cast_fp16 = reduce_mean(axes = var_175, keep_dims = var_72, x = zero_mean_sq_3_cast_fp16)[name = tensor<string, []>("op_176_cast_fp16")];
tensor<fp16, []> var_177_to_fp16 = const()[name = tensor<string, []>("op_177_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_178_cast_fp16 = add(x = var_176_cast_fp16, y = var_177_to_fp16)[name = tensor<string, []>("op_178_cast_fp16")];
tensor<fp16, []> denom_3_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_3_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_3_cast_fp16 = rsqrt(epsilon = denom_3_epsilon_0_to_fp16, x = var_178_cast_fp16)[name = tensor<string, []>("denom_3_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_3_cast_fp16 = mul(x = zero_mean_3_cast_fp16, y = denom_3_cast_fp16)[name = tensor<string, []>("out_3_cast_fp16")];
tensor<fp16, [384]> obj_9_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_9_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41361536)))];
tensor<fp16, [384]> obj_9_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_9_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41362368)))];
tensor<fp16, []> obj_9_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_9_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_9_cast_fp16 = batch_norm(beta = obj_9_beta_0_to_fp16, epsilon = obj_9_epsilon_0_to_fp16, gamma = obj_9_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_3_cast_fp16)[name = tensor<string, []>("obj_9_cast_fp16")];
tensor<int32, [2]> var_193 = const()[name = tensor<string, []>("op_193"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_195 = const()[name = tensor<string, []>("op_195"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_3_pad_type_0 = const()[name = tensor<string, []>("query_3_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_3_pad_0 = const()[name = tensor<string, []>("query_3_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_encoder_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_encoder_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41363200)))];
tensor<fp16, [384]> layers_0_encoder_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_encoder_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41658176)))];
tensor<fp16, [1, 384, 1, 1]> query_3_cast_fp16 = conv(bias = layers_0_encoder_attn_q_proj_bias_to_fp16, dilations = var_195, groups = var_71, pad = query_3_pad_0, pad_type = query_3_pad_type_0, strides = var_193, weight = layers_0_encoder_attn_q_proj_weight_to_fp16, x = obj_9_cast_fp16)[name = tensor<string, []>("query_3_cast_fp16")];
tensor<int32, [2]> var_199 = const()[name = tensor<string, []>("op_199"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_201 = const()[name = tensor<string, []>("op_201"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> key_3_pad_type_0 = const()[name = tensor<string, []>("key_3_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> key_3_pad_0 = const()[name = tensor<string, []>("key_3_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_encoder_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_encoder_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41659008)))];
tensor<fp16, [1, 384, 1, 1500]> key_3_cast_fp16 = conv(dilations = var_201, groups = var_71, pad = key_3_pad_0, pad_type = key_3_pad_type_0, strides = var_199, weight = layers_0_encoder_attn_k_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("key_3_cast_fp16")];
tensor<int32, [2]> var_206 = const()[name = tensor<string, []>("op_206"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_208 = const()[name = tensor<string, []>("op_208"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> value_3_pad_type_0 = const()[name = tensor<string, []>("value_3_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> value_3_pad_0 = const()[name = tensor<string, []>("value_3_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_encoder_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_encoder_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(41953984)))];
tensor<fp16, [384]> layers_0_encoder_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_encoder_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42248960)))];
tensor<fp16, [1, 384, 1, 1500]> value_3_cast_fp16 = conv(bias = layers_0_encoder_attn_v_proj_bias_to_fp16, dilations = var_208, groups = var_71, pad = value_3_pad_0, pad_type = value_3_pad_type_0, strides = var_206, weight = layers_0_encoder_attn_v_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("value_3_cast_fp16")];
tensor<int32, [4]> var_212 = const()[name = tensor<string, []>("op_212"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_213_cast_fp16 = reshape(shape = var_212, x = query_3_cast_fp16)[name = tensor<string, []>("op_213_cast_fp16")];
tensor<fp16, []> var_214_to_fp16 = const()[name = tensor<string, []>("op_214_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_215_cast_fp16 = mul(x = var_213_cast_fp16, y = var_214_to_fp16)[name = tensor<string, []>("op_215_cast_fp16")];
tensor<int32, [4]> var_216 = const()[name = tensor<string, []>("op_216"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_217_cast_fp16 = reshape(shape = var_216, x = key_3_cast_fp16)[name = tensor<string, []>("op_217_cast_fp16")];
tensor<bool, []> mh_w_5_transpose_x_0 = const()[name = tensor<string, []>("mh_w_5_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_5_transpose_y_0 = const()[name = tensor<string, []>("mh_w_5_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 1500]> mh_w_5_cast_fp16 = matmul(transpose_x = mh_w_5_transpose_x_0, transpose_y = mh_w_5_transpose_y_0, x = var_215_cast_fp16, y = var_217_cast_fp16)[name = tensor<string, []>("mh_w_5_cast_fp16")];
tensor<fp16, [1, 6, 1, 1500]> obj_13_cast_fp16 = softmax(axis = var_64, x = mh_w_5_cast_fp16)[name = tensor<string, []>("obj_13_cast_fp16")];
tensor<int32, [4]> var_221 = const()[name = tensor<string, []>("op_221"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_222_cast_fp16 = reshape(shape = var_221, x = value_3_cast_fp16)[name = tensor<string, []>("op_222_cast_fp16")];
tensor<bool, []> attn_3_transpose_x_0 = const()[name = tensor<string, []>("attn_3_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_3_transpose_y_0 = const()[name = tensor<string, []>("attn_3_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_3_cast_fp16 = matmul(transpose_x = attn_3_transpose_x_0, transpose_y = attn_3_transpose_y_0, x = var_222_cast_fp16, y = obj_13_cast_fp16)[name = tensor<string, []>("attn_3_cast_fp16")];
tensor<int32, [4]> var_225 = const()[name = tensor<string, []>("op_225"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_3_cast_fp16 = reshape(shape = var_225, x = attn_3_cast_fp16)[name = tensor<string, []>("input_3_cast_fp16")];
tensor<int32, [2]> var_229 = const()[name = tensor<string, []>("op_229"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_231 = const()[name = tensor<string, []>("op_231"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_11_pad_type_0 = const()[name = tensor<string, []>("obj_11_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_11_pad_0 = const()[name = tensor<string, []>("obj_11_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_0_encoder_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_encoder_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42249792)))];
tensor<fp16, [384]> layers_0_encoder_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_encoder_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42544768)))];
tensor<fp16, [1, 384, 1, 1]> obj_11_cast_fp16 = conv(bias = layers_0_encoder_attn_o_proj_bias_to_fp16, dilations = var_231, groups = var_71, pad = obj_11_pad_0, pad_type = obj_11_pad_type_0, strides = var_229, weight = layers_0_encoder_attn_o_proj_weight_to_fp16, x = input_3_cast_fp16)[name = tensor<string, []>("obj_11_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_5_cast_fp16 = add(x = inputs_3_cast_fp16, y = obj_11_cast_fp16)[name = tensor<string, []>("inputs_5_cast_fp16")];
tensor<int32, [1]> var_237 = const()[name = tensor<string, []>("op_237"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_5_cast_fp16 = reduce_mean(axes = var_237, keep_dims = var_72, x = inputs_5_cast_fp16)[name = tensor<string, []>("channels_mean_5_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_5_cast_fp16 = sub(x = inputs_5_cast_fp16, y = channels_mean_5_cast_fp16)[name = tensor<string, []>("zero_mean_5_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_5_cast_fp16 = mul(x = zero_mean_5_cast_fp16, y = zero_mean_5_cast_fp16)[name = tensor<string, []>("zero_mean_sq_5_cast_fp16")];
tensor<int32, [1]> var_241 = const()[name = tensor<string, []>("op_241"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_242_cast_fp16 = reduce_mean(axes = var_241, keep_dims = var_72, x = zero_mean_sq_5_cast_fp16)[name = tensor<string, []>("op_242_cast_fp16")];
tensor<fp16, []> var_243_to_fp16 = const()[name = tensor<string, []>("op_243_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_244_cast_fp16 = add(x = var_242_cast_fp16, y = var_243_to_fp16)[name = tensor<string, []>("op_244_cast_fp16")];
tensor<fp16, []> denom_5_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_5_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_5_cast_fp16 = rsqrt(epsilon = denom_5_epsilon_0_to_fp16, x = var_244_cast_fp16)[name = tensor<string, []>("denom_5_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_5_cast_fp16 = mul(x = zero_mean_5_cast_fp16, y = denom_5_cast_fp16)[name = tensor<string, []>("out_5_cast_fp16")];
tensor<fp16, [384]> input_5_gamma_0_to_fp16 = const()[name = tensor<string, []>("input_5_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42545600)))];
tensor<fp16, [384]> input_5_beta_0_to_fp16 = const()[name = tensor<string, []>("input_5_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42546432)))];
tensor<fp16, []> input_5_epsilon_0_to_fp16 = const()[name = tensor<string, []>("input_5_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> input_5_cast_fp16 = batch_norm(beta = input_5_beta_0_to_fp16, epsilon = input_5_epsilon_0_to_fp16, gamma = input_5_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_5_cast_fp16)[name = tensor<string, []>("input_5_cast_fp16")];
tensor<int32, [2]> var_255 = const()[name = tensor<string, []>("op_255"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_257 = const()[name = tensor<string, []>("op_257"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> input_7_pad_type_0 = const()[name = tensor<string, []>("input_7_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> input_7_pad_0 = const()[name = tensor<string, []>("input_7_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [1536, 384, 1, 1]> layers_0_fc1_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_fc1_weight_to_fp16"), val = tensor<fp16, [1536, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(42547264)))];
tensor<fp16, [1536]> layers_0_fc1_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_fc1_bias_to_fp16"), val = tensor<fp16, [1536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(43726976)))];
tensor<fp16, [1, 1536, 1, 1]> input_7_cast_fp16 = conv(bias = layers_0_fc1_bias_to_fp16, dilations = var_257, groups = var_71, pad = input_7_pad_0, pad_type = input_7_pad_type_0, strides = var_255, weight = layers_0_fc1_weight_to_fp16, x = input_5_cast_fp16)[name = tensor<string, []>("input_7_cast_fp16")];
tensor<string, []> input_9_mode_0 = const()[name = tensor<string, []>("input_9_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 1536, 1, 1]> input_9_cast_fp16 = gelu(mode = input_9_mode_0, x = input_7_cast_fp16)[name = tensor<string, []>("input_9_cast_fp16")];
tensor<int32, [2]> var_263 = const()[name = tensor<string, []>("op_263"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_265 = const()[name = tensor<string, []>("op_265"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> hidden_states_3_pad_type_0 = const()[name = tensor<string, []>("hidden_states_3_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> hidden_states_3_pad_0 = const()[name = tensor<string, []>("hidden_states_3_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 1536, 1, 1]> layers_0_fc2_weight_to_fp16 = const()[name = tensor<string, []>("layers_0_fc2_weight_to_fp16"), val = tensor<fp16, [384, 1536, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(43730112)))];
tensor<fp16, [384]> layers_0_fc2_bias_to_fp16 = const()[name = tensor<string, []>("layers_0_fc2_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(44909824)))];
tensor<fp16, [1, 384, 1, 1]> hidden_states_3_cast_fp16 = conv(bias = layers_0_fc2_bias_to_fp16, dilations = var_265, groups = var_71, pad = hidden_states_3_pad_0, pad_type = hidden_states_3_pad_type_0, strides = var_263, weight = layers_0_fc2_weight_to_fp16, x = input_9_cast_fp16)[name = tensor<string, []>("hidden_states_3_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_7_cast_fp16 = add(x = inputs_5_cast_fp16, y = hidden_states_3_cast_fp16)[name = tensor<string, []>("inputs_7_cast_fp16")];
tensor<int32, []> var_278 = const()[name = tensor<string, []>("op_278"), val = tensor<int32, []>(3)];
tensor<int32, []> var_285 = const()[name = tensor<string, []>("op_285"), val = tensor<int32, []>(1)];
tensor<bool, []> var_286 = const()[name = tensor<string, []>("op_286"), val = tensor<bool, []>(true)];
tensor<int32, [1]> var_298 = const()[name = tensor<string, []>("op_298"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_7_cast_fp16 = reduce_mean(axes = var_298, keep_dims = var_286, x = inputs_7_cast_fp16)[name = tensor<string, []>("channels_mean_7_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_7_cast_fp16 = sub(x = inputs_7_cast_fp16, y = channels_mean_7_cast_fp16)[name = tensor<string, []>("zero_mean_7_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_7_cast_fp16 = mul(x = zero_mean_7_cast_fp16, y = zero_mean_7_cast_fp16)[name = tensor<string, []>("zero_mean_sq_7_cast_fp16")];
tensor<int32, [1]> var_302 = const()[name = tensor<string, []>("op_302"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_303_cast_fp16 = reduce_mean(axes = var_302, keep_dims = var_286, x = zero_mean_sq_7_cast_fp16)[name = tensor<string, []>("op_303_cast_fp16")];
tensor<fp16, []> var_304_to_fp16 = const()[name = tensor<string, []>("op_304_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_305_cast_fp16 = add(x = var_303_cast_fp16, y = var_304_to_fp16)[name = tensor<string, []>("op_305_cast_fp16")];
tensor<fp16, []> denom_7_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_7_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_7_cast_fp16 = rsqrt(epsilon = denom_7_epsilon_0_to_fp16, x = var_305_cast_fp16)[name = tensor<string, []>("denom_7_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_7_cast_fp16 = mul(x = zero_mean_7_cast_fp16, y = denom_7_cast_fp16)[name = tensor<string, []>("out_7_cast_fp16")];
tensor<fp16, [384]> obj_15_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_15_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(44910656)))];
tensor<fp16, [384]> obj_15_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_15_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(44911488)))];
tensor<fp16, []> obj_15_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_15_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_15_cast_fp16 = batch_norm(beta = obj_15_beta_0_to_fp16, epsilon = obj_15_epsilon_0_to_fp16, gamma = obj_15_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_7_cast_fp16)[name = tensor<string, []>("obj_15_cast_fp16")];
tensor<int32, [2]> var_320 = const()[name = tensor<string, []>("op_320"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_322 = const()[name = tensor<string, []>("op_322"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_5_pad_type_0 = const()[name = tensor<string, []>("query_5_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_5_pad_0 = const()[name = tensor<string, []>("query_5_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_self_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_self_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(44912320)))];
tensor<fp16, [384]> layers_1_self_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(45207296)))];
tensor<fp16, [1, 384, 1, 1]> query_5_cast_fp16 = conv(bias = layers_1_self_attn_q_proj_bias_to_fp16, dilations = var_322, groups = var_285, pad = query_5_pad_0, pad_type = query_5_pad_type_0, strides = var_320, weight = layers_1_self_attn_q_proj_weight_to_fp16, x = obj_15_cast_fp16)[name = tensor<string, []>("query_5_cast_fp16")];
tensor<int32, [2]> var_326 = const()[name = tensor<string, []>("op_326"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_328 = const()[name = tensor<string, []>("op_328"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_key_3_pad_type_0 = const()[name = tensor<string, []>("current_key_3_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_key_3_pad_0 = const()[name = tensor<string, []>("current_key_3_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_self_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_self_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(45208128)))];
tensor<fp16, [1, 384, 1, 1]> current_key_3_cast_fp16 = conv(dilations = var_328, groups = var_285, pad = current_key_3_pad_0, pad_type = current_key_3_pad_type_0, strides = var_326, weight = layers_1_self_attn_k_proj_weight_to_fp16, x = obj_15_cast_fp16)[name = tensor<string, []>("current_key_3_cast_fp16")];
tensor<int32, [2]> var_333 = const()[name = tensor<string, []>("op_333"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_335 = const()[name = tensor<string, []>("op_335"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_value_3_pad_type_0 = const()[name = tensor<string, []>("current_value_3_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_value_3_pad_0 = const()[name = tensor<string, []>("current_value_3_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_self_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_self_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(45503104)))];
tensor<fp16, [384]> layers_1_self_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(45798080)))];
tensor<fp16, [1, 384, 1, 1]> current_value_3_cast_fp16 = conv(bias = layers_1_self_attn_v_proj_bias_to_fp16, dilations = var_335, groups = var_285, pad = current_value_3_pad_0, pad_type = current_value_3_pad_type_0, strides = var_333, weight = layers_1_self_attn_v_proj_weight_to_fp16, x = obj_15_cast_fp16)[name = tensor<string, []>("current_value_3_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_342_cast_fp16 = mul(x = current_key_3_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_342_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_344_cast_fp16 = mul(x = var_47_cast_fp16_1, y = var_129_cast_fp16)[name = tensor<string, []>("op_344_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> key_5_cast_fp16 = add(x = var_342_cast_fp16, y = var_344_cast_fp16)[name = tensor<string, []>("key_5_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_346_cast_fp16 = mul(x = current_value_3_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_346_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_348_cast_fp16 = mul(x = var_54_cast_fp16_1, y = var_129_cast_fp16)[name = tensor<string, []>("op_348_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> value_5_cast_fp16 = add(x = var_346_cast_fp16, y = var_348_cast_fp16)[name = tensor<string, []>("value_5_cast_fp16")];
tensor<int32, [4]> var_351 = const()[name = tensor<string, []>("op_351"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_352_cast_fp16 = reshape(shape = var_351, x = query_5_cast_fp16)[name = tensor<string, []>("op_352_cast_fp16")];
tensor<fp16, []> var_353_to_fp16 = const()[name = tensor<string, []>("op_353_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_354_cast_fp16 = mul(x = var_352_cast_fp16, y = var_353_to_fp16)[name = tensor<string, []>("op_354_cast_fp16")];
tensor<int32, [4]> var_355 = const()[name = tensor<string, []>("op_355"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_356_cast_fp16 = reshape(shape = var_355, x = key_5_cast_fp16)[name = tensor<string, []>("op_356_cast_fp16")];
tensor<bool, []> mh_w_7_transpose_x_0 = const()[name = tensor<string, []>("mh_w_7_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_7_transpose_y_0 = const()[name = tensor<string, []>("mh_w_7_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 224]> mh_w_7_cast_fp16 = matmul(transpose_x = mh_w_7_transpose_x_0, transpose_y = mh_w_7_transpose_y_0, x = var_354_cast_fp16, y = var_356_cast_fp16)[name = tensor<string, []>("mh_w_7_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> mh_w_9_cast_fp16 = add(x = mh_w_7_cast_fp16, y = var_147_cast_fp16)[name = tensor<string, []>("mh_w_9_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> var_364_cast_fp16 = softmax(axis = var_278, x = mh_w_9_cast_fp16)[name = tensor<string, []>("op_364_cast_fp16")];
tensor<int32, [4]> var_365 = const()[name = tensor<string, []>("op_365"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_366_cast_fp16 = reshape(shape = var_365, x = value_5_cast_fp16)[name = tensor<string, []>("op_366_cast_fp16")];
tensor<bool, []> attn_5_transpose_x_0 = const()[name = tensor<string, []>("attn_5_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_5_transpose_y_0 = const()[name = tensor<string, []>("attn_5_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_5_cast_fp16 = matmul(transpose_x = attn_5_transpose_x_0, transpose_y = attn_5_transpose_y_0, x = var_366_cast_fp16, y = var_364_cast_fp16)[name = tensor<string, []>("attn_5_cast_fp16")];
tensor<int32, [4]> var_369 = const()[name = tensor<string, []>("op_369"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_11_cast_fp16 = reshape(shape = var_369, x = attn_5_cast_fp16)[name = tensor<string, []>("input_11_cast_fp16")];
tensor<int32, [2]> var_373 = const()[name = tensor<string, []>("op_373"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_375 = const()[name = tensor<string, []>("op_375"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_21_pad_type_0 = const()[name = tensor<string, []>("obj_21_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_21_pad_0 = const()[name = tensor<string, []>("obj_21_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_self_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_self_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(45798912)))];
tensor<fp16, [384]> layers_1_self_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_self_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46093888)))];
tensor<fp16, [1, 384, 1, 1]> obj_21_cast_fp16 = conv(bias = layers_1_self_attn_o_proj_bias_to_fp16, dilations = var_375, groups = var_285, pad = obj_21_pad_0, pad_type = obj_21_pad_type_0, strides = var_373, weight = layers_1_self_attn_o_proj_weight_to_fp16, x = input_11_cast_fp16)[name = tensor<string, []>("obj_21_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_9_cast_fp16 = add(x = inputs_7_cast_fp16, y = obj_21_cast_fp16)[name = tensor<string, []>("inputs_9_cast_fp16")];
tensor<int32, [1]> var_385 = const()[name = tensor<string, []>("op_385"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_9_cast_fp16 = reduce_mean(axes = var_385, keep_dims = var_286, x = inputs_9_cast_fp16)[name = tensor<string, []>("channels_mean_9_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_9_cast_fp16 = sub(x = inputs_9_cast_fp16, y = channels_mean_9_cast_fp16)[name = tensor<string, []>("zero_mean_9_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_9_cast_fp16 = mul(x = zero_mean_9_cast_fp16, y = zero_mean_9_cast_fp16)[name = tensor<string, []>("zero_mean_sq_9_cast_fp16")];
tensor<int32, [1]> var_389 = const()[name = tensor<string, []>("op_389"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_390_cast_fp16 = reduce_mean(axes = var_389, keep_dims = var_286, x = zero_mean_sq_9_cast_fp16)[name = tensor<string, []>("op_390_cast_fp16")];
tensor<fp16, []> var_391_to_fp16 = const()[name = tensor<string, []>("op_391_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_392_cast_fp16 = add(x = var_390_cast_fp16, y = var_391_to_fp16)[name = tensor<string, []>("op_392_cast_fp16")];
tensor<fp16, []> denom_9_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_9_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_9_cast_fp16 = rsqrt(epsilon = denom_9_epsilon_0_to_fp16, x = var_392_cast_fp16)[name = tensor<string, []>("denom_9_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_9_cast_fp16 = mul(x = zero_mean_9_cast_fp16, y = denom_9_cast_fp16)[name = tensor<string, []>("out_9_cast_fp16")];
tensor<fp16, [384]> obj_23_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_23_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46094720)))];
tensor<fp16, [384]> obj_23_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_23_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46095552)))];
tensor<fp16, []> obj_23_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_23_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_23_cast_fp16 = batch_norm(beta = obj_23_beta_0_to_fp16, epsilon = obj_23_epsilon_0_to_fp16, gamma = obj_23_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_9_cast_fp16)[name = tensor<string, []>("obj_23_cast_fp16")];
tensor<int32, [2]> var_407 = const()[name = tensor<string, []>("op_407"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_409 = const()[name = tensor<string, []>("op_409"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_7_pad_type_0 = const()[name = tensor<string, []>("query_7_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_7_pad_0 = const()[name = tensor<string, []>("query_7_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_encoder_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_encoder_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46096384)))];
tensor<fp16, [384]> layers_1_encoder_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_encoder_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46391360)))];
tensor<fp16, [1, 384, 1, 1]> query_7_cast_fp16 = conv(bias = layers_1_encoder_attn_q_proj_bias_to_fp16, dilations = var_409, groups = var_285, pad = query_7_pad_0, pad_type = query_7_pad_type_0, strides = var_407, weight = layers_1_encoder_attn_q_proj_weight_to_fp16, x = obj_23_cast_fp16)[name = tensor<string, []>("query_7_cast_fp16")];
tensor<int32, [2]> var_413 = const()[name = tensor<string, []>("op_413"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_415 = const()[name = tensor<string, []>("op_415"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> key_7_pad_type_0 = const()[name = tensor<string, []>("key_7_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> key_7_pad_0 = const()[name = tensor<string, []>("key_7_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_encoder_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_encoder_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46392192)))];
tensor<fp16, [1, 384, 1, 1500]> key_7_cast_fp16 = conv(dilations = var_415, groups = var_285, pad = key_7_pad_0, pad_type = key_7_pad_type_0, strides = var_413, weight = layers_1_encoder_attn_k_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("key_7_cast_fp16")];
tensor<int32, [2]> var_420 = const()[name = tensor<string, []>("op_420"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_422 = const()[name = tensor<string, []>("op_422"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> value_7_pad_type_0 = const()[name = tensor<string, []>("value_7_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> value_7_pad_0 = const()[name = tensor<string, []>("value_7_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_encoder_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_encoder_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46687168)))];
tensor<fp16, [384]> layers_1_encoder_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_encoder_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46982144)))];
tensor<fp16, [1, 384, 1, 1500]> value_7_cast_fp16 = conv(bias = layers_1_encoder_attn_v_proj_bias_to_fp16, dilations = var_422, groups = var_285, pad = value_7_pad_0, pad_type = value_7_pad_type_0, strides = var_420, weight = layers_1_encoder_attn_v_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("value_7_cast_fp16")];
tensor<int32, [4]> var_426 = const()[name = tensor<string, []>("op_426"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_427_cast_fp16 = reshape(shape = var_426, x = query_7_cast_fp16)[name = tensor<string, []>("op_427_cast_fp16")];
tensor<fp16, []> var_428_to_fp16 = const()[name = tensor<string, []>("op_428_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_429_cast_fp16 = mul(x = var_427_cast_fp16, y = var_428_to_fp16)[name = tensor<string, []>("op_429_cast_fp16")];
tensor<int32, [4]> var_430 = const()[name = tensor<string, []>("op_430"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_431_cast_fp16 = reshape(shape = var_430, x = key_7_cast_fp16)[name = tensor<string, []>("op_431_cast_fp16")];
tensor<bool, []> mh_w_11_transpose_x_0 = const()[name = tensor<string, []>("mh_w_11_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_11_transpose_y_0 = const()[name = tensor<string, []>("mh_w_11_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 1500]> mh_w_11_cast_fp16 = matmul(transpose_x = mh_w_11_transpose_x_0, transpose_y = mh_w_11_transpose_y_0, x = var_429_cast_fp16, y = var_431_cast_fp16)[name = tensor<string, []>("mh_w_11_cast_fp16")];
tensor<fp16, [1, 6, 1, 1500]> obj_27_cast_fp16 = softmax(axis = var_278, x = mh_w_11_cast_fp16)[name = tensor<string, []>("obj_27_cast_fp16")];
tensor<int32, [4]> var_435 = const()[name = tensor<string, []>("op_435"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_436_cast_fp16 = reshape(shape = var_435, x = value_7_cast_fp16)[name = tensor<string, []>("op_436_cast_fp16")];
tensor<bool, []> attn_7_transpose_x_0 = const()[name = tensor<string, []>("attn_7_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_7_transpose_y_0 = const()[name = tensor<string, []>("attn_7_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_7_cast_fp16 = matmul(transpose_x = attn_7_transpose_x_0, transpose_y = attn_7_transpose_y_0, x = var_436_cast_fp16, y = obj_27_cast_fp16)[name = tensor<string, []>("attn_7_cast_fp16")];
tensor<int32, [4]> var_439 = const()[name = tensor<string, []>("op_439"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_13_cast_fp16 = reshape(shape = var_439, x = attn_7_cast_fp16)[name = tensor<string, []>("input_13_cast_fp16")];
tensor<int32, [2]> var_443 = const()[name = tensor<string, []>("op_443"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_445 = const()[name = tensor<string, []>("op_445"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_25_pad_type_0 = const()[name = tensor<string, []>("obj_25_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_25_pad_0 = const()[name = tensor<string, []>("obj_25_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_1_encoder_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_encoder_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(46982976)))];
tensor<fp16, [384]> layers_1_encoder_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_encoder_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(47277952)))];
tensor<fp16, [1, 384, 1, 1]> obj_25_cast_fp16 = conv(bias = layers_1_encoder_attn_o_proj_bias_to_fp16, dilations = var_445, groups = var_285, pad = obj_25_pad_0, pad_type = obj_25_pad_type_0, strides = var_443, weight = layers_1_encoder_attn_o_proj_weight_to_fp16, x = input_13_cast_fp16)[name = tensor<string, []>("obj_25_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_11_cast_fp16 = add(x = inputs_9_cast_fp16, y = obj_25_cast_fp16)[name = tensor<string, []>("inputs_11_cast_fp16")];
tensor<int32, [1]> var_454 = const()[name = tensor<string, []>("op_454"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_11_cast_fp16 = reduce_mean(axes = var_454, keep_dims = var_286, x = inputs_11_cast_fp16)[name = tensor<string, []>("channels_mean_11_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_11_cast_fp16 = sub(x = inputs_11_cast_fp16, y = channels_mean_11_cast_fp16)[name = tensor<string, []>("zero_mean_11_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_11_cast_fp16 = mul(x = zero_mean_11_cast_fp16, y = zero_mean_11_cast_fp16)[name = tensor<string, []>("zero_mean_sq_11_cast_fp16")];
tensor<int32, [1]> var_458 = const()[name = tensor<string, []>("op_458"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_459_cast_fp16 = reduce_mean(axes = var_458, keep_dims = var_286, x = zero_mean_sq_11_cast_fp16)[name = tensor<string, []>("op_459_cast_fp16")];
tensor<fp16, []> var_460_to_fp16 = const()[name = tensor<string, []>("op_460_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_461_cast_fp16 = add(x = var_459_cast_fp16, y = var_460_to_fp16)[name = tensor<string, []>("op_461_cast_fp16")];
tensor<fp16, []> denom_11_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_11_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_11_cast_fp16 = rsqrt(epsilon = denom_11_epsilon_0_to_fp16, x = var_461_cast_fp16)[name = tensor<string, []>("denom_11_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_11_cast_fp16 = mul(x = zero_mean_11_cast_fp16, y = denom_11_cast_fp16)[name = tensor<string, []>("out_11_cast_fp16")];
tensor<fp16, [384]> input_15_gamma_0_to_fp16 = const()[name = tensor<string, []>("input_15_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(47278784)))];
tensor<fp16, [384]> input_15_beta_0_to_fp16 = const()[name = tensor<string, []>("input_15_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(47279616)))];
tensor<fp16, []> input_15_epsilon_0_to_fp16 = const()[name = tensor<string, []>("input_15_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> input_15_cast_fp16 = batch_norm(beta = input_15_beta_0_to_fp16, epsilon = input_15_epsilon_0_to_fp16, gamma = input_15_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_11_cast_fp16)[name = tensor<string, []>("input_15_cast_fp16")];
tensor<int32, [2]> var_472 = const()[name = tensor<string, []>("op_472"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_474 = const()[name = tensor<string, []>("op_474"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> input_17_pad_type_0 = const()[name = tensor<string, []>("input_17_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> input_17_pad_0 = const()[name = tensor<string, []>("input_17_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [1536, 384, 1, 1]> layers_1_fc1_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_fc1_weight_to_fp16"), val = tensor<fp16, [1536, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(47280448)))];
tensor<fp16, [1536]> layers_1_fc1_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_fc1_bias_to_fp16"), val = tensor<fp16, [1536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(48460160)))];
tensor<fp16, [1, 1536, 1, 1]> input_17_cast_fp16 = conv(bias = layers_1_fc1_bias_to_fp16, dilations = var_474, groups = var_285, pad = input_17_pad_0, pad_type = input_17_pad_type_0, strides = var_472, weight = layers_1_fc1_weight_to_fp16, x = input_15_cast_fp16)[name = tensor<string, []>("input_17_cast_fp16")];
tensor<string, []> input_19_mode_0 = const()[name = tensor<string, []>("input_19_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 1536, 1, 1]> input_19_cast_fp16 = gelu(mode = input_19_mode_0, x = input_17_cast_fp16)[name = tensor<string, []>("input_19_cast_fp16")];
tensor<int32, [2]> var_480 = const()[name = tensor<string, []>("op_480"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_482 = const()[name = tensor<string, []>("op_482"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> hidden_states_5_pad_type_0 = const()[name = tensor<string, []>("hidden_states_5_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> hidden_states_5_pad_0 = const()[name = tensor<string, []>("hidden_states_5_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 1536, 1, 1]> layers_1_fc2_weight_to_fp16 = const()[name = tensor<string, []>("layers_1_fc2_weight_to_fp16"), val = tensor<fp16, [384, 1536, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(48463296)))];
tensor<fp16, [384]> layers_1_fc2_bias_to_fp16 = const()[name = tensor<string, []>("layers_1_fc2_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49643008)))];
tensor<fp16, [1, 384, 1, 1]> hidden_states_5_cast_fp16 = conv(bias = layers_1_fc2_bias_to_fp16, dilations = var_482, groups = var_285, pad = hidden_states_5_pad_0, pad_type = hidden_states_5_pad_type_0, strides = var_480, weight = layers_1_fc2_weight_to_fp16, x = input_19_cast_fp16)[name = tensor<string, []>("hidden_states_5_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_13_cast_fp16 = add(x = inputs_11_cast_fp16, y = hidden_states_5_cast_fp16)[name = tensor<string, []>("inputs_13_cast_fp16")];
tensor<int32, []> var_496 = const()[name = tensor<string, []>("op_496"), val = tensor<int32, []>(3)];
tensor<int32, []> var_503 = const()[name = tensor<string, []>("op_503"), val = tensor<int32, []>(1)];
tensor<bool, []> var_504 = const()[name = tensor<string, []>("op_504"), val = tensor<bool, []>(true)];
tensor<int32, [1]> var_516 = const()[name = tensor<string, []>("op_516"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_13_cast_fp16 = reduce_mean(axes = var_516, keep_dims = var_504, x = inputs_13_cast_fp16)[name = tensor<string, []>("channels_mean_13_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_13_cast_fp16 = sub(x = inputs_13_cast_fp16, y = channels_mean_13_cast_fp16)[name = tensor<string, []>("zero_mean_13_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_13_cast_fp16 = mul(x = zero_mean_13_cast_fp16, y = zero_mean_13_cast_fp16)[name = tensor<string, []>("zero_mean_sq_13_cast_fp16")];
tensor<int32, [1]> var_520 = const()[name = tensor<string, []>("op_520"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_521_cast_fp16 = reduce_mean(axes = var_520, keep_dims = var_504, x = zero_mean_sq_13_cast_fp16)[name = tensor<string, []>("op_521_cast_fp16")];
tensor<fp16, []> var_522_to_fp16 = const()[name = tensor<string, []>("op_522_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_523_cast_fp16 = add(x = var_521_cast_fp16, y = var_522_to_fp16)[name = tensor<string, []>("op_523_cast_fp16")];
tensor<fp16, []> denom_13_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_13_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_13_cast_fp16 = rsqrt(epsilon = denom_13_epsilon_0_to_fp16, x = var_523_cast_fp16)[name = tensor<string, []>("denom_13_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_13_cast_fp16 = mul(x = zero_mean_13_cast_fp16, y = denom_13_cast_fp16)[name = tensor<string, []>("out_13_cast_fp16")];
tensor<fp16, [384]> obj_29_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_29_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49643840)))];
tensor<fp16, [384]> obj_29_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_29_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49644672)))];
tensor<fp16, []> obj_29_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_29_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_29_cast_fp16 = batch_norm(beta = obj_29_beta_0_to_fp16, epsilon = obj_29_epsilon_0_to_fp16, gamma = obj_29_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_13_cast_fp16)[name = tensor<string, []>("obj_29_cast_fp16")];
tensor<int32, [2]> var_538 = const()[name = tensor<string, []>("op_538"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_540 = const()[name = tensor<string, []>("op_540"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_9_pad_type_0 = const()[name = tensor<string, []>("query_9_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_9_pad_0 = const()[name = tensor<string, []>("query_9_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_self_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_self_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49645504)))];
tensor<fp16, [384]> layers_2_self_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49940480)))];
tensor<fp16, [1, 384, 1, 1]> query_9_cast_fp16 = conv(bias = layers_2_self_attn_q_proj_bias_to_fp16, dilations = var_540, groups = var_503, pad = query_9_pad_0, pad_type = query_9_pad_type_0, strides = var_538, weight = layers_2_self_attn_q_proj_weight_to_fp16, x = obj_29_cast_fp16)[name = tensor<string, []>("query_9_cast_fp16")];
tensor<int32, [2]> var_544 = const()[name = tensor<string, []>("op_544"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_546 = const()[name = tensor<string, []>("op_546"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_key_5_pad_type_0 = const()[name = tensor<string, []>("current_key_5_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_key_5_pad_0 = const()[name = tensor<string, []>("current_key_5_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_self_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_self_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(49941312)))];
tensor<fp16, [1, 384, 1, 1]> current_key_5_cast_fp16 = conv(dilations = var_546, groups = var_503, pad = current_key_5_pad_0, pad_type = current_key_5_pad_type_0, strides = var_544, weight = layers_2_self_attn_k_proj_weight_to_fp16, x = obj_29_cast_fp16)[name = tensor<string, []>("current_key_5_cast_fp16")];
tensor<int32, [2]> var_551 = const()[name = tensor<string, []>("op_551"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_553 = const()[name = tensor<string, []>("op_553"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_value_5_pad_type_0 = const()[name = tensor<string, []>("current_value_5_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_value_5_pad_0 = const()[name = tensor<string, []>("current_value_5_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_self_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_self_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50236288)))];
tensor<fp16, [384]> layers_2_self_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50531264)))];
tensor<fp16, [1, 384, 1, 1]> current_value_5_cast_fp16 = conv(bias = layers_2_self_attn_v_proj_bias_to_fp16, dilations = var_553, groups = var_503, pad = current_value_5_pad_0, pad_type = current_value_5_pad_type_0, strides = var_551, weight = layers_2_self_attn_v_proj_weight_to_fp16, x = obj_29_cast_fp16)[name = tensor<string, []>("current_value_5_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_560_cast_fp16 = mul(x = current_key_5_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_560_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_562_cast_fp16 = mul(x = var_47_cast_fp16_2, y = var_129_cast_fp16)[name = tensor<string, []>("op_562_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> key_9_cast_fp16 = add(x = var_560_cast_fp16, y = var_562_cast_fp16)[name = tensor<string, []>("key_9_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_564_cast_fp16 = mul(x = current_value_5_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_564_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_566_cast_fp16 = mul(x = var_54_cast_fp16_2, y = var_129_cast_fp16)[name = tensor<string, []>("op_566_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> value_9_cast_fp16 = add(x = var_564_cast_fp16, y = var_566_cast_fp16)[name = tensor<string, []>("value_9_cast_fp16")];
tensor<int32, [4]> var_569 = const()[name = tensor<string, []>("op_569"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_570_cast_fp16 = reshape(shape = var_569, x = query_9_cast_fp16)[name = tensor<string, []>("op_570_cast_fp16")];
tensor<fp16, []> var_571_to_fp16 = const()[name = tensor<string, []>("op_571_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_572_cast_fp16 = mul(x = var_570_cast_fp16, y = var_571_to_fp16)[name = tensor<string, []>("op_572_cast_fp16")];
tensor<int32, [4]> var_573 = const()[name = tensor<string, []>("op_573"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_574_cast_fp16 = reshape(shape = var_573, x = key_9_cast_fp16)[name = tensor<string, []>("op_574_cast_fp16")];
tensor<bool, []> mh_w_13_transpose_x_0 = const()[name = tensor<string, []>("mh_w_13_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_13_transpose_y_0 = const()[name = tensor<string, []>("mh_w_13_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 224]> mh_w_13_cast_fp16 = matmul(transpose_x = mh_w_13_transpose_x_0, transpose_y = mh_w_13_transpose_y_0, x = var_572_cast_fp16, y = var_574_cast_fp16)[name = tensor<string, []>("mh_w_13_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> mh_w_15_cast_fp16 = add(x = mh_w_13_cast_fp16, y = var_147_cast_fp16)[name = tensor<string, []>("mh_w_15_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> var_582_cast_fp16 = softmax(axis = var_496, x = mh_w_15_cast_fp16)[name = tensor<string, []>("op_582_cast_fp16")];
tensor<int32, [4]> var_583 = const()[name = tensor<string, []>("op_583"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_584_cast_fp16 = reshape(shape = var_583, x = value_9_cast_fp16)[name = tensor<string, []>("op_584_cast_fp16")];
tensor<bool, []> attn_9_transpose_x_0 = const()[name = tensor<string, []>("attn_9_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_9_transpose_y_0 = const()[name = tensor<string, []>("attn_9_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_9_cast_fp16 = matmul(transpose_x = attn_9_transpose_x_0, transpose_y = attn_9_transpose_y_0, x = var_584_cast_fp16, y = var_582_cast_fp16)[name = tensor<string, []>("attn_9_cast_fp16")];
tensor<int32, [4]> var_587 = const()[name = tensor<string, []>("op_587"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_21_cast_fp16 = reshape(shape = var_587, x = attn_9_cast_fp16)[name = tensor<string, []>("input_21_cast_fp16")];
tensor<int32, [2]> var_591 = const()[name = tensor<string, []>("op_591"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_593 = const()[name = tensor<string, []>("op_593"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_35_pad_type_0 = const()[name = tensor<string, []>("obj_35_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_35_pad_0 = const()[name = tensor<string, []>("obj_35_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_self_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_self_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50532096)))];
tensor<fp16, [384]> layers_2_self_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_self_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50827072)))];
tensor<fp16, [1, 384, 1, 1]> obj_35_cast_fp16 = conv(bias = layers_2_self_attn_o_proj_bias_to_fp16, dilations = var_593, groups = var_503, pad = obj_35_pad_0, pad_type = obj_35_pad_type_0, strides = var_591, weight = layers_2_self_attn_o_proj_weight_to_fp16, x = input_21_cast_fp16)[name = tensor<string, []>("obj_35_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_15_cast_fp16 = add(x = inputs_13_cast_fp16, y = obj_35_cast_fp16)[name = tensor<string, []>("inputs_15_cast_fp16")];
tensor<int32, [1]> var_603 = const()[name = tensor<string, []>("op_603"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_15_cast_fp16 = reduce_mean(axes = var_603, keep_dims = var_504, x = inputs_15_cast_fp16)[name = tensor<string, []>("channels_mean_15_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_15_cast_fp16 = sub(x = inputs_15_cast_fp16, y = channels_mean_15_cast_fp16)[name = tensor<string, []>("zero_mean_15_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_15_cast_fp16 = mul(x = zero_mean_15_cast_fp16, y = zero_mean_15_cast_fp16)[name = tensor<string, []>("zero_mean_sq_15_cast_fp16")];
tensor<int32, [1]> var_607 = const()[name = tensor<string, []>("op_607"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_608_cast_fp16 = reduce_mean(axes = var_607, keep_dims = var_504, x = zero_mean_sq_15_cast_fp16)[name = tensor<string, []>("op_608_cast_fp16")];
tensor<fp16, []> var_609_to_fp16 = const()[name = tensor<string, []>("op_609_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_610_cast_fp16 = add(x = var_608_cast_fp16, y = var_609_to_fp16)[name = tensor<string, []>("op_610_cast_fp16")];
tensor<fp16, []> denom_15_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_15_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_15_cast_fp16 = rsqrt(epsilon = denom_15_epsilon_0_to_fp16, x = var_610_cast_fp16)[name = tensor<string, []>("denom_15_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_15_cast_fp16 = mul(x = zero_mean_15_cast_fp16, y = denom_15_cast_fp16)[name = tensor<string, []>("out_15_cast_fp16")];
tensor<fp16, [384]> obj_37_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_37_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50827904)))];
tensor<fp16, [384]> obj_37_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_37_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50828736)))];
tensor<fp16, []> obj_37_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_37_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_37_cast_fp16 = batch_norm(beta = obj_37_beta_0_to_fp16, epsilon = obj_37_epsilon_0_to_fp16, gamma = obj_37_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_15_cast_fp16)[name = tensor<string, []>("obj_37_cast_fp16")];
tensor<int32, [2]> var_625 = const()[name = tensor<string, []>("op_625"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_627 = const()[name = tensor<string, []>("op_627"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_11_pad_type_0 = const()[name = tensor<string, []>("query_11_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_11_pad_0 = const()[name = tensor<string, []>("query_11_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_encoder_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_encoder_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(50829568)))];
tensor<fp16, [384]> layers_2_encoder_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_encoder_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51124544)))];
tensor<fp16, [1, 384, 1, 1]> query_11_cast_fp16 = conv(bias = layers_2_encoder_attn_q_proj_bias_to_fp16, dilations = var_627, groups = var_503, pad = query_11_pad_0, pad_type = query_11_pad_type_0, strides = var_625, weight = layers_2_encoder_attn_q_proj_weight_to_fp16, x = obj_37_cast_fp16)[name = tensor<string, []>("query_11_cast_fp16")];
tensor<int32, [2]> var_631 = const()[name = tensor<string, []>("op_631"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_633 = const()[name = tensor<string, []>("op_633"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> key_11_pad_type_0 = const()[name = tensor<string, []>("key_11_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> key_11_pad_0 = const()[name = tensor<string, []>("key_11_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_encoder_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_encoder_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51125376)))];
tensor<fp16, [1, 384, 1, 1500]> key_11_cast_fp16 = conv(dilations = var_633, groups = var_503, pad = key_11_pad_0, pad_type = key_11_pad_type_0, strides = var_631, weight = layers_2_encoder_attn_k_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("key_11_cast_fp16")];
tensor<int32, [2]> var_638 = const()[name = tensor<string, []>("op_638"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_640 = const()[name = tensor<string, []>("op_640"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> value_11_pad_type_0 = const()[name = tensor<string, []>("value_11_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> value_11_pad_0 = const()[name = tensor<string, []>("value_11_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_encoder_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_encoder_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51420352)))];
tensor<fp16, [384]> layers_2_encoder_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_encoder_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51715328)))];
tensor<fp16, [1, 384, 1, 1500]> value_11_cast_fp16 = conv(bias = layers_2_encoder_attn_v_proj_bias_to_fp16, dilations = var_640, groups = var_503, pad = value_11_pad_0, pad_type = value_11_pad_type_0, strides = var_638, weight = layers_2_encoder_attn_v_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("value_11_cast_fp16")];
tensor<int32, [4]> var_644 = const()[name = tensor<string, []>("op_644"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_645_cast_fp16 = reshape(shape = var_644, x = query_11_cast_fp16)[name = tensor<string, []>("op_645_cast_fp16")];
tensor<fp16, []> var_646_to_fp16 = const()[name = tensor<string, []>("op_646_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_647_cast_fp16 = mul(x = var_645_cast_fp16, y = var_646_to_fp16)[name = tensor<string, []>("op_647_cast_fp16")];
tensor<int32, [4]> var_648 = const()[name = tensor<string, []>("op_648"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_649_cast_fp16 = reshape(shape = var_648, x = key_11_cast_fp16)[name = tensor<string, []>("op_649_cast_fp16")];
tensor<bool, []> mh_w_17_transpose_x_0 = const()[name = tensor<string, []>("mh_w_17_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_17_transpose_y_0 = const()[name = tensor<string, []>("mh_w_17_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 1500]> mh_w_17_cast_fp16 = matmul(transpose_x = mh_w_17_transpose_x_0, transpose_y = mh_w_17_transpose_y_0, x = var_647_cast_fp16, y = var_649_cast_fp16)[name = tensor<string, []>("mh_w_17_cast_fp16")];
tensor<fp16, [1, 6, 1, 1500]> obj_41_cast_fp16 = softmax(axis = var_496, x = mh_w_17_cast_fp16)[name = tensor<string, []>("obj_41_cast_fp16")];
tensor<int32, [4]> var_653 = const()[name = tensor<string, []>("op_653"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_654_cast_fp16 = reshape(shape = var_653, x = value_11_cast_fp16)[name = tensor<string, []>("op_654_cast_fp16")];
tensor<bool, []> attn_11_transpose_x_0 = const()[name = tensor<string, []>("attn_11_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_11_transpose_y_0 = const()[name = tensor<string, []>("attn_11_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_11_cast_fp16 = matmul(transpose_x = attn_11_transpose_x_0, transpose_y = attn_11_transpose_y_0, x = var_654_cast_fp16, y = obj_41_cast_fp16)[name = tensor<string, []>("attn_11_cast_fp16")];
tensor<int32, [4]> var_657 = const()[name = tensor<string, []>("op_657"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_23_cast_fp16 = reshape(shape = var_657, x = attn_11_cast_fp16)[name = tensor<string, []>("input_23_cast_fp16")];
tensor<int32, [2]> var_661 = const()[name = tensor<string, []>("op_661"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_663 = const()[name = tensor<string, []>("op_663"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_39_pad_type_0 = const()[name = tensor<string, []>("obj_39_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_39_pad_0 = const()[name = tensor<string, []>("obj_39_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_2_encoder_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_encoder_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(51716160)))];
tensor<fp16, [384]> layers_2_encoder_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_encoder_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(52011136)))];
tensor<fp16, [1, 384, 1, 1]> obj_39_cast_fp16 = conv(bias = layers_2_encoder_attn_o_proj_bias_to_fp16, dilations = var_663, groups = var_503, pad = obj_39_pad_0, pad_type = obj_39_pad_type_0, strides = var_661, weight = layers_2_encoder_attn_o_proj_weight_to_fp16, x = input_23_cast_fp16)[name = tensor<string, []>("obj_39_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_17_cast_fp16 = add(x = inputs_15_cast_fp16, y = obj_39_cast_fp16)[name = tensor<string, []>("inputs_17_cast_fp16")];
tensor<int32, [1]> var_672 = const()[name = tensor<string, []>("op_672"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_17_cast_fp16 = reduce_mean(axes = var_672, keep_dims = var_504, x = inputs_17_cast_fp16)[name = tensor<string, []>("channels_mean_17_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_17_cast_fp16 = sub(x = inputs_17_cast_fp16, y = channels_mean_17_cast_fp16)[name = tensor<string, []>("zero_mean_17_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_17_cast_fp16 = mul(x = zero_mean_17_cast_fp16, y = zero_mean_17_cast_fp16)[name = tensor<string, []>("zero_mean_sq_17_cast_fp16")];
tensor<int32, [1]> var_676 = const()[name = tensor<string, []>("op_676"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_677_cast_fp16 = reduce_mean(axes = var_676, keep_dims = var_504, x = zero_mean_sq_17_cast_fp16)[name = tensor<string, []>("op_677_cast_fp16")];
tensor<fp16, []> var_678_to_fp16 = const()[name = tensor<string, []>("op_678_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_679_cast_fp16 = add(x = var_677_cast_fp16, y = var_678_to_fp16)[name = tensor<string, []>("op_679_cast_fp16")];
tensor<fp16, []> denom_17_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_17_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_17_cast_fp16 = rsqrt(epsilon = denom_17_epsilon_0_to_fp16, x = var_679_cast_fp16)[name = tensor<string, []>("denom_17_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_17_cast_fp16 = mul(x = zero_mean_17_cast_fp16, y = denom_17_cast_fp16)[name = tensor<string, []>("out_17_cast_fp16")];
tensor<fp16, [384]> input_25_gamma_0_to_fp16 = const()[name = tensor<string, []>("input_25_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(52011968)))];
tensor<fp16, [384]> input_25_beta_0_to_fp16 = const()[name = tensor<string, []>("input_25_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(52012800)))];
tensor<fp16, []> input_25_epsilon_0_to_fp16 = const()[name = tensor<string, []>("input_25_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> input_25_cast_fp16 = batch_norm(beta = input_25_beta_0_to_fp16, epsilon = input_25_epsilon_0_to_fp16, gamma = input_25_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_17_cast_fp16)[name = tensor<string, []>("input_25_cast_fp16")];
tensor<int32, [2]> var_690 = const()[name = tensor<string, []>("op_690"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_692 = const()[name = tensor<string, []>("op_692"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> input_27_pad_type_0 = const()[name = tensor<string, []>("input_27_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> input_27_pad_0 = const()[name = tensor<string, []>("input_27_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [1536, 384, 1, 1]> layers_2_fc1_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_fc1_weight_to_fp16"), val = tensor<fp16, [1536, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(52013632)))];
tensor<fp16, [1536]> layers_2_fc1_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_fc1_bias_to_fp16"), val = tensor<fp16, [1536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(53193344)))];
tensor<fp16, [1, 1536, 1, 1]> input_27_cast_fp16 = conv(bias = layers_2_fc1_bias_to_fp16, dilations = var_692, groups = var_503, pad = input_27_pad_0, pad_type = input_27_pad_type_0, strides = var_690, weight = layers_2_fc1_weight_to_fp16, x = input_25_cast_fp16)[name = tensor<string, []>("input_27_cast_fp16")];
tensor<string, []> input_29_mode_0 = const()[name = tensor<string, []>("input_29_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 1536, 1, 1]> input_29_cast_fp16 = gelu(mode = input_29_mode_0, x = input_27_cast_fp16)[name = tensor<string, []>("input_29_cast_fp16")];
tensor<int32, [2]> var_698 = const()[name = tensor<string, []>("op_698"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_700 = const()[name = tensor<string, []>("op_700"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> hidden_states_7_pad_type_0 = const()[name = tensor<string, []>("hidden_states_7_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> hidden_states_7_pad_0 = const()[name = tensor<string, []>("hidden_states_7_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 1536, 1, 1]> layers_2_fc2_weight_to_fp16 = const()[name = tensor<string, []>("layers_2_fc2_weight_to_fp16"), val = tensor<fp16, [384, 1536, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(53196480)))];
tensor<fp16, [384]> layers_2_fc2_bias_to_fp16 = const()[name = tensor<string, []>("layers_2_fc2_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(54376192)))];
tensor<fp16, [1, 384, 1, 1]> hidden_states_7_cast_fp16 = conv(bias = layers_2_fc2_bias_to_fp16, dilations = var_700, groups = var_503, pad = hidden_states_7_pad_0, pad_type = hidden_states_7_pad_type_0, strides = var_698, weight = layers_2_fc2_weight_to_fp16, x = input_29_cast_fp16)[name = tensor<string, []>("hidden_states_7_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_19_cast_fp16 = add(x = inputs_17_cast_fp16, y = hidden_states_7_cast_fp16)[name = tensor<string, []>("inputs_19_cast_fp16")];
tensor<int32, []> var_714 = const()[name = tensor<string, []>("op_714"), val = tensor<int32, []>(3)];
tensor<int32, []> var_721 = const()[name = tensor<string, []>("op_721"), val = tensor<int32, []>(1)];
tensor<bool, []> var_722 = const()[name = tensor<string, []>("op_722"), val = tensor<bool, []>(true)];
tensor<int32, [1]> var_734 = const()[name = tensor<string, []>("op_734"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_19_cast_fp16 = reduce_mean(axes = var_734, keep_dims = var_722, x = inputs_19_cast_fp16)[name = tensor<string, []>("channels_mean_19_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_19_cast_fp16 = sub(x = inputs_19_cast_fp16, y = channels_mean_19_cast_fp16)[name = tensor<string, []>("zero_mean_19_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_19_cast_fp16 = mul(x = zero_mean_19_cast_fp16, y = zero_mean_19_cast_fp16)[name = tensor<string, []>("zero_mean_sq_19_cast_fp16")];
tensor<int32, [1]> var_738 = const()[name = tensor<string, []>("op_738"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_739_cast_fp16 = reduce_mean(axes = var_738, keep_dims = var_722, x = zero_mean_sq_19_cast_fp16)[name = tensor<string, []>("op_739_cast_fp16")];
tensor<fp16, []> var_740_to_fp16 = const()[name = tensor<string, []>("op_740_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_741_cast_fp16 = add(x = var_739_cast_fp16, y = var_740_to_fp16)[name = tensor<string, []>("op_741_cast_fp16")];
tensor<fp16, []> denom_19_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_19_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_19_cast_fp16 = rsqrt(epsilon = denom_19_epsilon_0_to_fp16, x = var_741_cast_fp16)[name = tensor<string, []>("denom_19_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_19_cast_fp16 = mul(x = zero_mean_19_cast_fp16, y = denom_19_cast_fp16)[name = tensor<string, []>("out_19_cast_fp16")];
tensor<fp16, [384]> obj_43_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_43_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(54377024)))];
tensor<fp16, [384]> obj_43_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_43_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(54377856)))];
tensor<fp16, []> obj_43_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_43_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_43_cast_fp16 = batch_norm(beta = obj_43_beta_0_to_fp16, epsilon = obj_43_epsilon_0_to_fp16, gamma = obj_43_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_19_cast_fp16)[name = tensor<string, []>("obj_43_cast_fp16")];
tensor<int32, [2]> var_756 = const()[name = tensor<string, []>("op_756"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_758 = const()[name = tensor<string, []>("op_758"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_13_pad_type_0 = const()[name = tensor<string, []>("query_13_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_13_pad_0 = const()[name = tensor<string, []>("query_13_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_self_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_self_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(54378688)))];
tensor<fp16, [384]> layers_3_self_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(54673664)))];
tensor<fp16, [1, 384, 1, 1]> query_13_cast_fp16 = conv(bias = layers_3_self_attn_q_proj_bias_to_fp16, dilations = var_758, groups = var_721, pad = query_13_pad_0, pad_type = query_13_pad_type_0, strides = var_756, weight = layers_3_self_attn_q_proj_weight_to_fp16, x = obj_43_cast_fp16)[name = tensor<string, []>("query_13_cast_fp16")];
tensor<int32, [2]> var_762 = const()[name = tensor<string, []>("op_762"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_764 = const()[name = tensor<string, []>("op_764"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_key_pad_type_0 = const()[name = tensor<string, []>("current_key_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_key_pad_0 = const()[name = tensor<string, []>("current_key_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_self_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_self_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(54674496)))];
tensor<fp16, [1, 384, 1, 1]> current_key_cast_fp16 = conv(dilations = var_764, groups = var_721, pad = current_key_pad_0, pad_type = current_key_pad_type_0, strides = var_762, weight = layers_3_self_attn_k_proj_weight_to_fp16, x = obj_43_cast_fp16)[name = tensor<string, []>("current_key_cast_fp16")];
tensor<int32, [2]> var_769 = const()[name = tensor<string, []>("op_769"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_771 = const()[name = tensor<string, []>("op_771"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> current_value_pad_type_0 = const()[name = tensor<string, []>("current_value_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> current_value_pad_0 = const()[name = tensor<string, []>("current_value_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_self_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_self_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(54969472)))];
tensor<fp16, [384]> layers_3_self_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55264448)))];
tensor<fp16, [1, 384, 1, 1]> current_value_cast_fp16 = conv(bias = layers_3_self_attn_v_proj_bias_to_fp16, dilations = var_771, groups = var_721, pad = current_value_pad_0, pad_type = current_value_pad_type_0, strides = var_769, weight = layers_3_self_attn_v_proj_weight_to_fp16, x = obj_43_cast_fp16)[name = tensor<string, []>("current_value_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_778_cast_fp16 = mul(x = current_key_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_778_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_780_cast_fp16 = mul(x = var_47_cast_fp16_3, y = var_129_cast_fp16)[name = tensor<string, []>("op_780_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> key_13_cast_fp16 = add(x = var_778_cast_fp16, y = var_780_cast_fp16)[name = tensor<string, []>("key_13_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_782_cast_fp16 = mul(x = current_value_cast_fp16, y = var_126_cast_fp16)[name = tensor<string, []>("op_782_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> var_784_cast_fp16 = mul(x = var_54_cast_fp16_3, y = var_129_cast_fp16)[name = tensor<string, []>("op_784_cast_fp16")];
tensor<fp16, [1, 384, 1, 224]> value_13_cast_fp16 = add(x = var_782_cast_fp16, y = var_784_cast_fp16)[name = tensor<string, []>("value_13_cast_fp16")];
tensor<int32, [4]> var_787 = const()[name = tensor<string, []>("op_787"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_788_cast_fp16 = reshape(shape = var_787, x = query_13_cast_fp16)[name = tensor<string, []>("op_788_cast_fp16")];
tensor<fp16, []> var_789_to_fp16 = const()[name = tensor<string, []>("op_789_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_790_cast_fp16 = mul(x = var_788_cast_fp16, y = var_789_to_fp16)[name = tensor<string, []>("op_790_cast_fp16")];
tensor<int32, [4]> var_791 = const()[name = tensor<string, []>("op_791"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_792_cast_fp16 = reshape(shape = var_791, x = key_13_cast_fp16)[name = tensor<string, []>("op_792_cast_fp16")];
tensor<bool, []> mh_w_19_transpose_x_0 = const()[name = tensor<string, []>("mh_w_19_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_19_transpose_y_0 = const()[name = tensor<string, []>("mh_w_19_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 224]> mh_w_19_cast_fp16 = matmul(transpose_x = mh_w_19_transpose_x_0, transpose_y = mh_w_19_transpose_y_0, x = var_790_cast_fp16, y = var_792_cast_fp16)[name = tensor<string, []>("mh_w_19_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> mh_w_21_cast_fp16 = add(x = mh_w_19_cast_fp16, y = var_147_cast_fp16)[name = tensor<string, []>("mh_w_21_cast_fp16")];
tensor<fp16, [1, 6, 1, 224]> var_800_cast_fp16 = softmax(axis = var_714, x = mh_w_21_cast_fp16)[name = tensor<string, []>("op_800_cast_fp16")];
tensor<int32, [4]> var_801 = const()[name = tensor<string, []>("op_801"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 224]> var_802_cast_fp16 = reshape(shape = var_801, x = value_13_cast_fp16)[name = tensor<string, []>("op_802_cast_fp16")];
tensor<bool, []> attn_13_transpose_x_0 = const()[name = tensor<string, []>("attn_13_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_13_transpose_y_0 = const()[name = tensor<string, []>("attn_13_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_13_cast_fp16 = matmul(transpose_x = attn_13_transpose_x_0, transpose_y = attn_13_transpose_y_0, x = var_802_cast_fp16, y = var_800_cast_fp16)[name = tensor<string, []>("attn_13_cast_fp16")];
tensor<int32, [4]> var_805 = const()[name = tensor<string, []>("op_805"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_31_cast_fp16 = reshape(shape = var_805, x = attn_13_cast_fp16)[name = tensor<string, []>("input_31_cast_fp16")];
tensor<int32, [2]> var_809 = const()[name = tensor<string, []>("op_809"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_811 = const()[name = tensor<string, []>("op_811"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_49_pad_type_0 = const()[name = tensor<string, []>("obj_49_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_49_pad_0 = const()[name = tensor<string, []>("obj_49_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_self_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_self_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55265280)))];
tensor<fp16, [384]> layers_3_self_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_self_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55560256)))];
tensor<fp16, [1, 384, 1, 1]> obj_49_cast_fp16 = conv(bias = layers_3_self_attn_o_proj_bias_to_fp16, dilations = var_811, groups = var_721, pad = obj_49_pad_0, pad_type = obj_49_pad_type_0, strides = var_809, weight = layers_3_self_attn_o_proj_weight_to_fp16, x = input_31_cast_fp16)[name = tensor<string, []>("obj_49_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_21_cast_fp16 = add(x = inputs_19_cast_fp16, y = obj_49_cast_fp16)[name = tensor<string, []>("inputs_21_cast_fp16")];
tensor<int32, [1]> var_821 = const()[name = tensor<string, []>("op_821"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_21_cast_fp16 = reduce_mean(axes = var_821, keep_dims = var_722, x = inputs_21_cast_fp16)[name = tensor<string, []>("channels_mean_21_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_21_cast_fp16 = sub(x = inputs_21_cast_fp16, y = channels_mean_21_cast_fp16)[name = tensor<string, []>("zero_mean_21_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_21_cast_fp16 = mul(x = zero_mean_21_cast_fp16, y = zero_mean_21_cast_fp16)[name = tensor<string, []>("zero_mean_sq_21_cast_fp16")];
tensor<int32, [1]> var_825 = const()[name = tensor<string, []>("op_825"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_826_cast_fp16 = reduce_mean(axes = var_825, keep_dims = var_722, x = zero_mean_sq_21_cast_fp16)[name = tensor<string, []>("op_826_cast_fp16")];
tensor<fp16, []> var_827_to_fp16 = const()[name = tensor<string, []>("op_827_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_828_cast_fp16 = add(x = var_826_cast_fp16, y = var_827_to_fp16)[name = tensor<string, []>("op_828_cast_fp16")];
tensor<fp16, []> denom_21_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_21_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_21_cast_fp16 = rsqrt(epsilon = denom_21_epsilon_0_to_fp16, x = var_828_cast_fp16)[name = tensor<string, []>("denom_21_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_21_cast_fp16 = mul(x = zero_mean_21_cast_fp16, y = denom_21_cast_fp16)[name = tensor<string, []>("out_21_cast_fp16")];
tensor<fp16, [384]> obj_51_gamma_0_to_fp16 = const()[name = tensor<string, []>("obj_51_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55561088)))];
tensor<fp16, [384]> obj_51_beta_0_to_fp16 = const()[name = tensor<string, []>("obj_51_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55561920)))];
tensor<fp16, []> obj_51_epsilon_0_to_fp16 = const()[name = tensor<string, []>("obj_51_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> obj_51_cast_fp16 = batch_norm(beta = obj_51_beta_0_to_fp16, epsilon = obj_51_epsilon_0_to_fp16, gamma = obj_51_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_21_cast_fp16)[name = tensor<string, []>("obj_51_cast_fp16")];
tensor<int32, [2]> var_843 = const()[name = tensor<string, []>("op_843"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_845 = const()[name = tensor<string, []>("op_845"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> query_pad_type_0 = const()[name = tensor<string, []>("query_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> query_pad_0 = const()[name = tensor<string, []>("query_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_encoder_attn_q_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_encoder_attn_q_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55562752)))];
tensor<fp16, [384]> layers_3_encoder_attn_q_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_encoder_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55857728)))];
tensor<fp16, [1, 384, 1, 1]> query_cast_fp16 = conv(bias = layers_3_encoder_attn_q_proj_bias_to_fp16, dilations = var_845, groups = var_721, pad = query_pad_0, pad_type = query_pad_type_0, strides = var_843, weight = layers_3_encoder_attn_q_proj_weight_to_fp16, x = obj_51_cast_fp16)[name = tensor<string, []>("query_cast_fp16")];
tensor<int32, [2]> var_849 = const()[name = tensor<string, []>("op_849"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_851 = const()[name = tensor<string, []>("op_851"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> key_pad_type_0 = const()[name = tensor<string, []>("key_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> key_pad_0 = const()[name = tensor<string, []>("key_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_encoder_attn_k_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_encoder_attn_k_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(55858560)))];
tensor<fp16, [1, 384, 1, 1500]> key_cast_fp16 = conv(dilations = var_851, groups = var_721, pad = key_pad_0, pad_type = key_pad_type_0, strides = var_849, weight = layers_3_encoder_attn_k_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("key_cast_fp16")];
tensor<int32, [2]> var_856 = const()[name = tensor<string, []>("op_856"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_858 = const()[name = tensor<string, []>("op_858"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> value_pad_type_0 = const()[name = tensor<string, []>("value_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> value_pad_0 = const()[name = tensor<string, []>("value_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_encoder_attn_v_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_encoder_attn_v_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56153536)))];
tensor<fp16, [384]> layers_3_encoder_attn_v_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_encoder_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56448512)))];
tensor<fp16, [1, 384, 1, 1500]> value_cast_fp16 = conv(bias = layers_3_encoder_attn_v_proj_bias_to_fp16, dilations = var_858, groups = var_721, pad = value_pad_0, pad_type = value_pad_type_0, strides = var_856, weight = layers_3_encoder_attn_v_proj_weight_to_fp16, x = encoder_output_embeds)[name = tensor<string, []>("value_cast_fp16")];
tensor<int32, [4]> var_862 = const()[name = tensor<string, []>("op_862"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1]> var_863_cast_fp16 = reshape(shape = var_862, x = query_cast_fp16)[name = tensor<string, []>("op_863_cast_fp16")];
tensor<fp16, []> var_864_to_fp16 = const()[name = tensor<string, []>("op_864_to_fp16"), val = tensor<fp16, []>(0x1p-3)];
tensor<fp16, [1, 6, 64, 1]> var_865_cast_fp16 = mul(x = var_863_cast_fp16, y = var_864_to_fp16)[name = tensor<string, []>("op_865_cast_fp16")];
tensor<int32, [4]> var_866 = const()[name = tensor<string, []>("op_866"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_867_cast_fp16 = reshape(shape = var_866, x = key_cast_fp16)[name = tensor<string, []>("op_867_cast_fp16")];
tensor<bool, []> mh_w_transpose_x_0 = const()[name = tensor<string, []>("mh_w_transpose_x_0"), val = tensor<bool, []>(true)];
tensor<bool, []> mh_w_transpose_y_0 = const()[name = tensor<string, []>("mh_w_transpose_y_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 6, 1, 1500]> mh_w_cast_fp16 = matmul(transpose_x = mh_w_transpose_x_0, transpose_y = mh_w_transpose_y_0, x = var_865_cast_fp16, y = var_867_cast_fp16)[name = tensor<string, []>("mh_w_cast_fp16")];
tensor<fp16, [1, 6, 1, 1500]> obj_55_cast_fp16 = softmax(axis = var_714, x = mh_w_cast_fp16)[name = tensor<string, []>("obj_55_cast_fp16")];
tensor<int32, [4]> var_871 = const()[name = tensor<string, []>("op_871"), val = tensor<int32, [4]>([1, 6, 64, -1])];
tensor<fp16, [1, 6, 64, 1500]> var_872_cast_fp16 = reshape(shape = var_871, x = value_cast_fp16)[name = tensor<string, []>("op_872_cast_fp16")];
tensor<bool, []> attn_transpose_x_0 = const()[name = tensor<string, []>("attn_transpose_x_0"), val = tensor<bool, []>(false)];
tensor<bool, []> attn_transpose_y_0 = const()[name = tensor<string, []>("attn_transpose_y_0"), val = tensor<bool, []>(true)];
tensor<fp16, [1, 6, 64, 1]> attn_cast_fp16 = matmul(transpose_x = attn_transpose_x_0, transpose_y = attn_transpose_y_0, x = var_872_cast_fp16, y = obj_55_cast_fp16)[name = tensor<string, []>("attn_cast_fp16")];
tensor<int32, [4]> var_875 = const()[name = tensor<string, []>("op_875"), val = tensor<int32, [4]>([1, 384, 1, -1])];
tensor<fp16, [1, 384, 1, 1]> input_33_cast_fp16 = reshape(shape = var_875, x = attn_cast_fp16)[name = tensor<string, []>("input_33_cast_fp16")];
tensor<int32, [2]> var_879 = const()[name = tensor<string, []>("op_879"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_881 = const()[name = tensor<string, []>("op_881"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> obj_53_pad_type_0 = const()[name = tensor<string, []>("obj_53_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> obj_53_pad_0 = const()[name = tensor<string, []>("obj_53_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 384, 1, 1]> layers_3_encoder_attn_o_proj_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_encoder_attn_o_proj_weight_to_fp16"), val = tensor<fp16, [384, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56449344)))];
tensor<fp16, [384]> layers_3_encoder_attn_o_proj_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_encoder_attn_o_proj_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56744320)))];
tensor<fp16, [1, 384, 1, 1]> obj_53_cast_fp16 = conv(bias = layers_3_encoder_attn_o_proj_bias_to_fp16, dilations = var_881, groups = var_721, pad = obj_53_pad_0, pad_type = obj_53_pad_type_0, strides = var_879, weight = layers_3_encoder_attn_o_proj_weight_to_fp16, x = input_33_cast_fp16)[name = tensor<string, []>("obj_53_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_23_cast_fp16 = add(x = inputs_21_cast_fp16, y = obj_53_cast_fp16)[name = tensor<string, []>("inputs_23_cast_fp16")];
tensor<int32, [1]> var_890 = const()[name = tensor<string, []>("op_890"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_23_cast_fp16 = reduce_mean(axes = var_890, keep_dims = var_722, x = inputs_23_cast_fp16)[name = tensor<string, []>("channels_mean_23_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_23_cast_fp16 = sub(x = inputs_23_cast_fp16, y = channels_mean_23_cast_fp16)[name = tensor<string, []>("zero_mean_23_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_23_cast_fp16 = mul(x = zero_mean_23_cast_fp16, y = zero_mean_23_cast_fp16)[name = tensor<string, []>("zero_mean_sq_23_cast_fp16")];
tensor<int32, [1]> var_894 = const()[name = tensor<string, []>("op_894"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_895_cast_fp16 = reduce_mean(axes = var_894, keep_dims = var_722, x = zero_mean_sq_23_cast_fp16)[name = tensor<string, []>("op_895_cast_fp16")];
tensor<fp16, []> var_896_to_fp16 = const()[name = tensor<string, []>("op_896_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_897_cast_fp16 = add(x = var_895_cast_fp16, y = var_896_to_fp16)[name = tensor<string, []>("op_897_cast_fp16")];
tensor<fp16, []> denom_23_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_23_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_23_cast_fp16 = rsqrt(epsilon = denom_23_epsilon_0_to_fp16, x = var_897_cast_fp16)[name = tensor<string, []>("denom_23_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_23_cast_fp16 = mul(x = zero_mean_23_cast_fp16, y = denom_23_cast_fp16)[name = tensor<string, []>("out_23_cast_fp16")];
tensor<fp16, [384]> input_35_gamma_0_to_fp16 = const()[name = tensor<string, []>("input_35_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56745152)))];
tensor<fp16, [384]> input_35_beta_0_to_fp16 = const()[name = tensor<string, []>("input_35_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56745984)))];
tensor<fp16, []> input_35_epsilon_0_to_fp16 = const()[name = tensor<string, []>("input_35_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> input_35_cast_fp16 = batch_norm(beta = input_35_beta_0_to_fp16, epsilon = input_35_epsilon_0_to_fp16, gamma = input_35_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_23_cast_fp16)[name = tensor<string, []>("input_35_cast_fp16")];
tensor<int32, [2]> var_908 = const()[name = tensor<string, []>("op_908"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_910 = const()[name = tensor<string, []>("op_910"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> input_37_pad_type_0 = const()[name = tensor<string, []>("input_37_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> input_37_pad_0 = const()[name = tensor<string, []>("input_37_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [1536, 384, 1, 1]> layers_3_fc1_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_fc1_weight_to_fp16"), val = tensor<fp16, [1536, 384, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(56746816)))];
tensor<fp16, [1536]> layers_3_fc1_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_fc1_bias_to_fp16"), val = tensor<fp16, [1536]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(57926528)))];
tensor<fp16, [1, 1536, 1, 1]> input_37_cast_fp16 = conv(bias = layers_3_fc1_bias_to_fp16, dilations = var_910, groups = var_721, pad = input_37_pad_0, pad_type = input_37_pad_type_0, strides = var_908, weight = layers_3_fc1_weight_to_fp16, x = input_35_cast_fp16)[name = tensor<string, []>("input_37_cast_fp16")];
tensor<string, []> input_mode_0 = const()[name = tensor<string, []>("input_mode_0"), val = tensor<string, []>("EXACT")];
tensor<fp16, [1, 1536, 1, 1]> input_cast_fp16 = gelu(mode = input_mode_0, x = input_37_cast_fp16)[name = tensor<string, []>("input_cast_fp16")];
tensor<int32, [2]> var_916 = const()[name = tensor<string, []>("op_916"), val = tensor<int32, [2]>([1, 1])];
tensor<int32, [2]> var_918 = const()[name = tensor<string, []>("op_918"), val = tensor<int32, [2]>([1, 1])];
tensor<string, []> hidden_states_9_pad_type_0 = const()[name = tensor<string, []>("hidden_states_9_pad_type_0"), val = tensor<string, []>("custom")];
tensor<int32, [4]> hidden_states_9_pad_0 = const()[name = tensor<string, []>("hidden_states_9_pad_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<fp16, [384, 1536, 1, 1]> layers_3_fc2_weight_to_fp16 = const()[name = tensor<string, []>("layers_3_fc2_weight_to_fp16"), val = tensor<fp16, [384, 1536, 1, 1]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(57929664)))];
tensor<fp16, [384]> layers_3_fc2_bias_to_fp16 = const()[name = tensor<string, []>("layers_3_fc2_bias_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(59109376)))];
tensor<fp16, [1, 384, 1, 1]> hidden_states_9_cast_fp16 = conv(bias = layers_3_fc2_bias_to_fp16, dilations = var_918, groups = var_721, pad = hidden_states_9_pad_0, pad_type = hidden_states_9_pad_type_0, strides = var_916, weight = layers_3_fc2_weight_to_fp16, x = input_cast_fp16)[name = tensor<string, []>("hidden_states_9_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> inputs_cast_fp16 = add(x = inputs_23_cast_fp16, y = hidden_states_9_cast_fp16)[name = tensor<string, []>("inputs_cast_fp16")];
tensor<bool, []> var_929 = const()[name = tensor<string, []>("op_929"), val = tensor<bool, []>(true)];
tensor<int32, [1]> var_933 = const()[name = tensor<string, []>("op_933"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> channels_mean_cast_fp16 = reduce_mean(axes = var_933, keep_dims = var_929, x = inputs_cast_fp16)[name = tensor<string, []>("channels_mean_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_cast_fp16 = sub(x = inputs_cast_fp16, y = channels_mean_cast_fp16)[name = tensor<string, []>("zero_mean_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> zero_mean_sq_cast_fp16 = mul(x = zero_mean_cast_fp16, y = zero_mean_cast_fp16)[name = tensor<string, []>("zero_mean_sq_cast_fp16")];
tensor<int32, [1]> var_937 = const()[name = tensor<string, []>("op_937"), val = tensor<int32, [1]>([1])];
tensor<fp16, [1, 1, 1, 1]> var_938_cast_fp16 = reduce_mean(axes = var_937, keep_dims = var_929, x = zero_mean_sq_cast_fp16)[name = tensor<string, []>("op_938_cast_fp16")];
tensor<fp16, []> var_939_to_fp16 = const()[name = tensor<string, []>("op_939_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 1, 1, 1]> var_940_cast_fp16 = add(x = var_938_cast_fp16, y = var_939_to_fp16)[name = tensor<string, []>("op_940_cast_fp16")];
tensor<fp16, []> denom_epsilon_0_to_fp16 = const()[name = tensor<string, []>("denom_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1p-24)];
tensor<fp16, [1, 1, 1, 1]> denom_cast_fp16 = rsqrt(epsilon = denom_epsilon_0_to_fp16, x = var_940_cast_fp16)[name = tensor<string, []>("denom_cast_fp16")];
tensor<fp16, [1, 384, 1, 1]> out_cast_fp16 = mul(x = zero_mean_cast_fp16, y = denom_cast_fp16)[name = tensor<string, []>("out_cast_fp16")];
tensor<fp16, [384]> hidden_states_gamma_0_to_fp16 = const()[name = tensor<string, []>("hidden_states_gamma_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(59110208)))];
tensor<fp16, [384]> hidden_states_beta_0_to_fp16 = const()[name = tensor<string, []>("hidden_states_beta_0_to_fp16"), val = tensor<fp16, [384]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(59111040)))];
tensor<fp16, []> hidden_states_epsilon_0_to_fp16 = const()[name = tensor<string, []>("hidden_states_epsilon_0_to_fp16"), val = tensor<fp16, []>(0x1.5p-17)];
tensor<fp16, [1, 384, 1, 1]> hidden_states_cast_fp16 = batch_norm(beta = hidden_states_beta_0_to_fp16, epsilon = hidden_states_epsilon_0_to_fp16, gamma = hidden_states_gamma_0_to_fp16, mean = obj_1_mean_0_to_fp16, variance = obj_1_variance_0_to_fp16, x = out_cast_fp16)[name = tensor<string, []>("hidden_states_cast_fp16")];
tensor<int32, [1]> var_950_axes_0 = const()[name = tensor<string, []>("op_950_axes_0"), val = tensor<int32, [1]>([2])];
tensor<fp16, [1, 384, 1]> var_950_cast_fp16 = squeeze(axes = var_950_axes_0, x = hidden_states_cast_fp16)[name = tensor<string, []>("op_950_cast_fp16")];
tensor<int32, [3]> var_953_perm_0 = const()[name = tensor<string, []>("op_953_perm_0"), val = tensor<int32, [3]>([0, 2, 1])];
tensor<fp16, [51864]> linear_0_bias_0_to_fp16 = const()[name = tensor<string, []>("linear_0_bias_0_to_fp16"), val = tensor<fp16, [51864]>(BLOBFILE(path = tensor<string, []>("@model_path/weights/weight.bin"), offset = tensor<uint64, []>(59111872)))];
tensor<fp16, [1, 1, 384]> transpose_0 = transpose(perm = var_953_perm_0, x = var_950_cast_fp16)[name = tensor<string, []>("transpose_0")];
tensor<fp16, [1, 1, 51864]> logits = linear(bias = linear_0_bias_0_to_fp16, weight = embed_tokens_weight_to_fp16, x = transpose_0)[name = tensor<string, []>("linear_0_cast_fp16")];
tensor<int32, []> var_957 = const()[name = tensor<string, []>("op_957"), val = tensor<int32, []>(1)];
tensor<bool, []> obj_59_interleave_0 = const()[name = tensor<string, []>("obj_59_interleave_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 1536, 1, 1]> key_cache_updates = concat(axis = var_957, interleave = obj_59_interleave_0, values = (current_key_1_cast_fp16, current_key_3_cast_fp16, current_key_5_cast_fp16, current_key_cast_fp16))[name = tensor<string, []>("obj_59_cast_fp16")];
tensor<int32, []> var_960 = const()[name = tensor<string, []>("op_960"), val = tensor<int32, []>(1)];
tensor<bool, []> obj_61_interleave_0 = const()[name = tensor<string, []>("obj_61_interleave_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 1536, 1, 1]> value_cache_updates = concat(axis = var_960, interleave = obj_61_interleave_0, values = (current_value_1_cast_fp16, current_value_3_cast_fp16, current_value_5_cast_fp16, current_value_cast_fp16))[name = tensor<string, []>("obj_61_cast_fp16")];
tensor<int32, [4]> var_971_begin_0 = const()[name = tensor<string, []>("op_971_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_971_end_0 = const()[name = tensor<string, []>("op_971_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_971_end_mask_0 = const()[name = tensor<string, []>("op_971_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_971_cast_fp16 = slice_by_index(begin = var_971_begin_0, end = var_971_end_0, end_mask = var_971_end_mask_0, x = obj_27_cast_fp16)[name = tensor<string, []>("op_971_cast_fp16")];
tensor<int32, [4]> var_974_begin_0 = const()[name = tensor<string, []>("op_974_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_974_end_0 = const()[name = tensor<string, []>("op_974_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_974_end_mask_0 = const()[name = tensor<string, []>("op_974_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_974_squeeze_mask_0 = const()[name = tensor<string, []>("op_974_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_974_cast_fp16 = slice_by_index(begin = var_974_begin_0, end = var_974_end_0, end_mask = var_974_end_mask_0, squeeze_mask = var_974_squeeze_mask_0, x = var_971_cast_fp16)[name = tensor<string, []>("op_974_cast_fp16")];
tensor<int32, [4]> var_989_begin_0 = const()[name = tensor<string, []>("op_989_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_989_end_0 = const()[name = tensor<string, []>("op_989_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_989_end_mask_0 = const()[name = tensor<string, []>("op_989_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_989_cast_fp16 = slice_by_index(begin = var_989_begin_0, end = var_989_end_0, end_mask = var_989_end_mask_0, x = obj_41_cast_fp16)[name = tensor<string, []>("op_989_cast_fp16")];
tensor<int32, [4]> var_992_begin_0 = const()[name = tensor<string, []>("op_992_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_992_end_0 = const()[name = tensor<string, []>("op_992_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_992_end_mask_0 = const()[name = tensor<string, []>("op_992_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_992_squeeze_mask_0 = const()[name = tensor<string, []>("op_992_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_992_cast_fp16 = slice_by_index(begin = var_992_begin_0, end = var_992_end_0, end_mask = var_992_end_mask_0, squeeze_mask = var_992_squeeze_mask_0, x = var_989_cast_fp16)[name = tensor<string, []>("op_992_cast_fp16")];
tensor<int32, [4]> var_1007_begin_0 = const()[name = tensor<string, []>("op_1007_begin_0"), val = tensor<int32, [4]>([0, 5, 0, 0])];
tensor<int32, [4]> var_1007_end_0 = const()[name = tensor<string, []>("op_1007_end_0"), val = tensor<int32, [4]>([1, 6, 1, 1500])];
tensor<bool, [4]> var_1007_end_mask_0 = const()[name = tensor<string, []>("op_1007_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_1007_cast_fp16 = slice_by_index(begin = var_1007_begin_0, end = var_1007_end_0, end_mask = var_1007_end_mask_0, x = obj_41_cast_fp16)[name = tensor<string, []>("op_1007_cast_fp16")];
tensor<int32, [4]> var_1010_begin_0 = const()[name = tensor<string, []>("op_1010_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_1010_end_0 = const()[name = tensor<string, []>("op_1010_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_1010_end_mask_0 = const()[name = tensor<string, []>("op_1010_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_1010_squeeze_mask_0 = const()[name = tensor<string, []>("op_1010_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_1010_cast_fp16 = slice_by_index(begin = var_1010_begin_0, end = var_1010_end_0, end_mask = var_1010_end_mask_0, squeeze_mask = var_1010_squeeze_mask_0, x = var_1007_cast_fp16)[name = tensor<string, []>("op_1010_cast_fp16")];
tensor<int32, [4]> var_1025_begin_0 = const()[name = tensor<string, []>("op_1025_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_1025_end_0 = const()[name = tensor<string, []>("op_1025_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_1025_end_mask_0 = const()[name = tensor<string, []>("op_1025_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_1025_cast_fp16 = slice_by_index(begin = var_1025_begin_0, end = var_1025_end_0, end_mask = var_1025_end_mask_0, x = obj_55_cast_fp16)[name = tensor<string, []>("op_1025_cast_fp16")];
tensor<int32, [4]> var_1028_begin_0 = const()[name = tensor<string, []>("op_1028_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_1028_end_0 = const()[name = tensor<string, []>("op_1028_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_1028_end_mask_0 = const()[name = tensor<string, []>("op_1028_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_1028_squeeze_mask_0 = const()[name = tensor<string, []>("op_1028_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_1028_cast_fp16 = slice_by_index(begin = var_1028_begin_0, end = var_1028_end_0, end_mask = var_1028_end_mask_0, squeeze_mask = var_1028_squeeze_mask_0, x = var_1025_cast_fp16)[name = tensor<string, []>("op_1028_cast_fp16")];
tensor<int32, [4]> var_1043_begin_0 = const()[name = tensor<string, []>("op_1043_begin_0"), val = tensor<int32, [4]>([0, 1, 0, 0])];
tensor<int32, [4]> var_1043_end_0 = const()[name = tensor<string, []>("op_1043_end_0"), val = tensor<int32, [4]>([1, 2, 1, 1500])];
tensor<bool, [4]> var_1043_end_mask_0 = const()[name = tensor<string, []>("op_1043_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_1043_cast_fp16 = slice_by_index(begin = var_1043_begin_0, end = var_1043_end_0, end_mask = var_1043_end_mask_0, x = obj_55_cast_fp16)[name = tensor<string, []>("op_1043_cast_fp16")];
tensor<int32, [4]> var_1046_begin_0 = const()[name = tensor<string, []>("op_1046_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_1046_end_0 = const()[name = tensor<string, []>("op_1046_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_1046_end_mask_0 = const()[name = tensor<string, []>("op_1046_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_1046_squeeze_mask_0 = const()[name = tensor<string, []>("op_1046_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_1046_cast_fp16 = slice_by_index(begin = var_1046_begin_0, end = var_1046_end_0, end_mask = var_1046_end_mask_0, squeeze_mask = var_1046_squeeze_mask_0, x = var_1043_cast_fp16)[name = tensor<string, []>("op_1046_cast_fp16")];
tensor<int32, [4]> var_1061_begin_0 = const()[name = tensor<string, []>("op_1061_begin_0"), val = tensor<int32, [4]>([0, 2, 0, 0])];
tensor<int32, [4]> var_1061_end_0 = const()[name = tensor<string, []>("op_1061_end_0"), val = tensor<int32, [4]>([1, 3, 1, 1500])];
tensor<bool, [4]> var_1061_end_mask_0 = const()[name = tensor<string, []>("op_1061_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_1061_cast_fp16 = slice_by_index(begin = var_1061_begin_0, end = var_1061_end_0, end_mask = var_1061_end_mask_0, x = obj_55_cast_fp16)[name = tensor<string, []>("op_1061_cast_fp16")];
tensor<int32, [4]> var_1064_begin_0 = const()[name = tensor<string, []>("op_1064_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_1064_end_0 = const()[name = tensor<string, []>("op_1064_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_1064_end_mask_0 = const()[name = tensor<string, []>("op_1064_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_1064_squeeze_mask_0 = const()[name = tensor<string, []>("op_1064_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_1064_cast_fp16 = slice_by_index(begin = var_1064_begin_0, end = var_1064_end_0, end_mask = var_1064_end_mask_0, squeeze_mask = var_1064_squeeze_mask_0, x = var_1061_cast_fp16)[name = tensor<string, []>("op_1064_cast_fp16")];
tensor<int32, [4]> var_1079_begin_0 = const()[name = tensor<string, []>("op_1079_begin_0"), val = tensor<int32, [4]>([0, 3, 0, 0])];
tensor<int32, [4]> var_1079_end_0 = const()[name = tensor<string, []>("op_1079_end_0"), val = tensor<int32, [4]>([1, 4, 1, 1500])];
tensor<bool, [4]> var_1079_end_mask_0 = const()[name = tensor<string, []>("op_1079_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_1079_cast_fp16 = slice_by_index(begin = var_1079_begin_0, end = var_1079_end_0, end_mask = var_1079_end_mask_0, x = obj_55_cast_fp16)[name = tensor<string, []>("op_1079_cast_fp16")];
tensor<int32, [4]> var_1082_begin_0 = const()[name = tensor<string, []>("op_1082_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_1082_end_0 = const()[name = tensor<string, []>("op_1082_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_1082_end_mask_0 = const()[name = tensor<string, []>("op_1082_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_1082_squeeze_mask_0 = const()[name = tensor<string, []>("op_1082_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_1082_cast_fp16 = slice_by_index(begin = var_1082_begin_0, end = var_1082_end_0, end_mask = var_1082_end_mask_0, squeeze_mask = var_1082_squeeze_mask_0, x = var_1079_cast_fp16)[name = tensor<string, []>("op_1082_cast_fp16")];
tensor<int32, [4]> var_1097_begin_0 = const()[name = tensor<string, []>("op_1097_begin_0"), val = tensor<int32, [4]>([0, 4, 0, 0])];
tensor<int32, [4]> var_1097_end_0 = const()[name = tensor<string, []>("op_1097_end_0"), val = tensor<int32, [4]>([1, 5, 1, 1500])];
tensor<bool, [4]> var_1097_end_mask_0 = const()[name = tensor<string, []>("op_1097_end_mask_0"), val = tensor<bool, [4]>([true, false, true, true])];
tensor<fp16, [1, 1, 1, 1500]> var_1097_cast_fp16 = slice_by_index(begin = var_1097_begin_0, end = var_1097_end_0, end_mask = var_1097_end_mask_0, x = obj_55_cast_fp16)[name = tensor<string, []>("op_1097_cast_fp16")];
tensor<int32, [4]> var_1100_begin_0 = const()[name = tensor<string, []>("op_1100_begin_0"), val = tensor<int32, [4]>([0, 0, 0, 0])];
tensor<int32, [4]> var_1100_end_0 = const()[name = tensor<string, []>("op_1100_end_0"), val = tensor<int32, [4]>([1, 1, 1, 1500])];
tensor<bool, [4]> var_1100_end_mask_0 = const()[name = tensor<string, []>("op_1100_end_mask_0"), val = tensor<bool, [4]>([true, true, false, true])];
tensor<bool, [4]> var_1100_squeeze_mask_0 = const()[name = tensor<string, []>("op_1100_squeeze_mask_0"), val = tensor<bool, [4]>([false, false, true, false])];
tensor<fp16, [1, 1, 1500]> var_1100_cast_fp16 = slice_by_index(begin = var_1100_begin_0, end = var_1100_end_0, end_mask = var_1100_end_mask_0, squeeze_mask = var_1100_squeeze_mask_0, x = var_1097_cast_fp16)[name = tensor<string, []>("op_1100_cast_fp16")];
tensor<int32, []> var_1107 = const()[name = tensor<string, []>("op_1107"), val = tensor<int32, []>(1)];
tensor<bool, []> var_1108_interleave_0 = const()[name = tensor<string, []>("op_1108_interleave_0"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 8, 1500]> var_1108_cast_fp16 = concat(axis = var_1107, interleave = var_1108_interleave_0, values = (var_974_cast_fp16, var_992_cast_fp16, var_1010_cast_fp16, var_1028_cast_fp16, var_1046_cast_fp16, var_1064_cast_fp16, var_1082_cast_fp16, var_1100_cast_fp16))[name = tensor<string, []>("op_1108_cast_fp16")];
tensor<int32, [1]> var_1110 = const()[name = tensor<string, []>("op_1110"), val = tensor<int32, [1]>([1])];
tensor<bool, []> var_1111 = const()[name = tensor<string, []>("op_1111"), val = tensor<bool, []>(false)];
tensor<fp16, [1, 1500]> alignment_heads_weights = reduce_mean(axes = var_1110, keep_dims = var_1111, x = var_1108_cast_fp16)[name = tensor<string, []>("obj_cast_fp16")];
} -> (logits, key_cache_updates, value_cache_updates, alignment_heads_weights);
} |