File size: 156,698 Bytes
4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef 28d0147 4bce7ef | 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 | program(1.3)
[buildInfo = dict<string, string>({{"coremlc-component-MIL", "3520.4.1"}, {"coremlc-version", "3520.5.1"}})]
{
func main<ios18>(tensor<int32, [1, 77]> input_ids) {
int32 inputs_embeds_batch_dims_0 = const()[name = string("inputs_embeds_batch_dims_0"), val = int32(0)];
bool inputs_embeds_validate_indices_0 = const()[name = string("inputs_embeds_validate_indices_0"), val = bool(false)];
tensor<fp16, [49408, 768]> encoder_embeddings_token_embedding_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [49408, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64))), lut = tensor<fp16, [3088, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(28459136))))[name = string("encoder_embeddings_token_embedding_weight_to_fp16_palettized")];
int32 greater_equal_0_y_0 = const()[name = string("greater_equal_0_y_0"), val = int32(0)];
tensor<bool, [1, 77]> greater_equal_0 = greater_equal(x = input_ids, y = greater_equal_0_y_0)[name = string("greater_equal_0")];
int32 slice_by_index_0 = const()[name = string("slice_by_index_0"), val = int32(49408)];
tensor<int32, [1, 77]> add_12 = add(x = input_ids, y = slice_by_index_0)[name = string("add_12")];
tensor<int32, [1, 77]> select_0 = select(a = input_ids, b = add_12, cond = greater_equal_0)[name = string("select_0")];
int32 greater_equal_0_y_0_1 = const()[name = string("greater_equal_0_y_0_1"), val = int32(0)];
tensor<bool, [1, 77]> greater_equal_0_1 = greater_equal(x = select_0, y = greater_equal_0_y_0_1)[name = string("greater_equal_0_1")];
int32 slice_by_index_0_1 = const()[name = string("slice_by_index_0_1"), val = int32(49408)];
tensor<int32, [1, 77]> add_0 = add(x = select_0, y = slice_by_index_0_1)[name = string("add_0")];
tensor<int32, [1, 77]> select_0_1 = select(a = select_0, b = add_0, cond = greater_equal_0_1)[name = string("select_0_1")];
int32 inputs_embeds_cast_fp16_axis_0 = const()[name = string("inputs_embeds_cast_fp16_axis_0"), val = int32(0)];
tensor<fp16, [1, 77, 768]> inputs_embeds_cast_fp16 = gather(axis = inputs_embeds_cast_fp16_axis_0, batch_dims = inputs_embeds_batch_dims_0, indices = select_0_1, validate_indices = inputs_embeds_validate_indices_0, x = encoder_embeddings_token_embedding_weight_to_fp16_palettized)[name = string("inputs_embeds_cast_fp16")];
tensor<fp16, [1, 77, 768]> position_embeddings_to_fp16 = const()[name = string("position_embeddings_to_fp16"), val = tensor<fp16, [1, 77, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(28854464)))];
tensor<fp16, [1, 77, 768]> input_3_cast_fp16 = add(x = inputs_embeds_cast_fp16, y = position_embeddings_to_fp16)[name = string("input_3_cast_fp16")];
tensor<int32, [1]> hidden_states_1_axes_0 = const()[name = string("hidden_states_1_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_0_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_0_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(28972800)))];
tensor<fp16, [768]> encoder_encoder_layers_0_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(28974400)))];
fp16 var_11_to_fp16 = const()[name = string("op_11_to_fp16"), val = fp16(0x1.5p-17)];
tensor<fp16, [1, 77, 768]> hidden_states_1_cast_fp16 = layer_norm(axes = hidden_states_1_axes_0, beta = encoder_encoder_layers_0_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_0_layer_norm1_weight_to_fp16, x = input_3_cast_fp16)[name = string("hidden_states_1_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_0_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(28976000))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(29418432))))[name = string("encoder_encoder_layers_0_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_0_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(29424640)))];
tensor<fp16, [1, 77, 768]> linear_0_cast_fp16 = linear(bias = encoder_encoder_layers_0_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_0_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_1_cast_fp16)[name = string("linear_0_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_0_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(29426240))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(29868672))))[name = string("encoder_encoder_layers_0_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_0_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(29874880)))];
tensor<fp16, [1, 77, 768]> linear_1_cast_fp16 = linear(bias = encoder_encoder_layers_0_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_0_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_1_cast_fp16)[name = string("linear_1_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_0_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(29876480))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30318912))))[name = string("encoder_encoder_layers_0_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_0_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30325120)))];
tensor<fp16, [1, 77, 768]> linear_2_cast_fp16 = linear(bias = encoder_encoder_layers_0_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_0_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_1_cast_fp16)[name = string("linear_2_cast_fp16")];
tensor<int32, [4]> var_82 = const()[name = string("op_82"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_83_cast_fp16 = reshape(shape = var_82, x = linear_0_cast_fp16)[name = string("op_83_cast_fp16")];
tensor<int32, [4]> var_85 = const()[name = string("op_85"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_86_cast_fp16 = reshape(shape = var_85, x = linear_1_cast_fp16)[name = string("op_86_cast_fp16")];
tensor<int32, [4]> var_88 = const()[name = string("op_88"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_89_cast_fp16 = reshape(shape = var_88, x = linear_2_cast_fp16)[name = string("op_89_cast_fp16")];
tensor<int32, [4]> value_1_perm_0 = const()[name = string("value_1_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
fp16 var_9_to_fp16 = const()[name = string("op_9_to_fp16"), val = fp16(0x1p-3)];
tensor<fp16, [1, 77, 12, 64]> mul_1_cast_fp16 = mul(x = var_83_cast_fp16, y = var_9_to_fp16)[name = string("mul_1_cast_fp16")];
bool matmul_0_transpose_y_0 = const()[name = string("matmul_0_transpose_y_0"), val = bool(true)];
bool matmul_0_transpose_x_0 = const()[name = string("matmul_0_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_48_perm_0 = const()[name = string("transpose_48_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_49_perm_0 = const()[name = string("transpose_49_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_49 = transpose(perm = transpose_49_perm_0, x = var_86_cast_fp16)[name = string("transpose_118")];
tensor<fp16, [1, 12, 77, 64]> transpose_48 = transpose(perm = transpose_48_perm_0, x = mul_1_cast_fp16)[name = string("transpose_119")];
tensor<fp16, [1, 12, 77, 77]> matmul_0_cast_fp16 = matmul(transpose_x = matmul_0_transpose_x_0, transpose_y = matmul_0_transpose_y_0, x = transpose_48, y = transpose_49)[name = string("matmul_0_cast_fp16")];
tensor<fp16, [77, 77]> mul_0_to_fp16 = const()[name = string("mul_0_to_fp16"), val = tensor<fp16, [77, 77]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30326720)))];
tensor<fp16, [1, 12, 77, 77]> add_0_cast_fp16 = add(x = matmul_0_cast_fp16, y = mul_0_to_fp16)[name = string("add_0_cast_fp16")];
int32 softmax_0_axis_0 = const()[name = string("softmax_0_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_0_cast_fp16 = softmax(axis = softmax_0_axis_0, x = add_0_cast_fp16)[name = string("softmax_0_cast_fp16")];
bool attn_output_1_transpose_x_0 = const()[name = string("attn_output_1_transpose_x_0"), val = bool(false)];
bool attn_output_1_transpose_y_0 = const()[name = string("attn_output_1_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_1_cast_fp16 = transpose(perm = value_1_perm_0, x = var_89_cast_fp16)[name = string("transpose_117")];
tensor<fp16, [1, 12, 77, 64]> attn_output_1_cast_fp16 = matmul(transpose_x = attn_output_1_transpose_x_0, transpose_y = attn_output_1_transpose_y_0, x = softmax_0_cast_fp16, y = value_1_cast_fp16)[name = string("attn_output_1_cast_fp16")];
tensor<int32, [4]> var_92_perm_0 = const()[name = string("op_92_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_94 = const()[name = string("op_94"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_92_cast_fp16 = transpose(perm = var_92_perm_0, x = attn_output_1_cast_fp16)[name = string("transpose_116")];
tensor<fp16, [1, 77, 768]> var_95_cast_fp16 = reshape(shape = var_94, x = var_92_cast_fp16)[name = string("op_95_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_0_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30338688))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30781120))))[name = string("encoder_encoder_layers_0_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_0_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30787328)))];
tensor<fp16, [1, 77, 768]> linear_3_cast_fp16 = linear(bias = encoder_encoder_layers_0_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_0_self_attn_out_proj_weight_to_fp16_palettized, x = var_95_cast_fp16)[name = string("linear_3_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_7_cast_fp16 = add(x = input_3_cast_fp16, y = linear_3_cast_fp16)[name = string("input_7_cast_fp16")];
tensor<int32, [1]> input_9_axes_0 = const()[name = string("input_9_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_0_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_0_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30788928)))];
tensor<fp16, [768]> encoder_encoder_layers_0_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30790528)))];
tensor<fp16, [1, 77, 768]> input_9_cast_fp16 = layer_norm(axes = input_9_axes_0, beta = encoder_encoder_layers_0_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_0_layer_norm2_weight_to_fp16, x = input_7_cast_fp16)[name = string("input_9_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_0_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(30792128))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(32561664))))[name = string("encoder_encoder_layers_0_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_0_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(32586304)))];
tensor<fp16, [1, 77, 3072]> linear_4_cast_fp16 = linear(bias = encoder_encoder_layers_0_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_0_mlp_fc1_weight_to_fp16_palettized, x = input_9_cast_fp16)[name = string("linear_4_cast_fp16")];
fp16 var_110_to_fp16 = const()[name = string("op_110_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_111_cast_fp16 = mul(x = linear_4_cast_fp16, y = var_110_to_fp16)[name = string("op_111_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_112_cast_fp16 = sigmoid(x = var_111_cast_fp16)[name = string("op_112_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_13_cast_fp16 = mul(x = linear_4_cast_fp16, y = var_112_cast_fp16)[name = string("input_13_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_0_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(32592512))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34362048))))[name = string("encoder_encoder_layers_0_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_0_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_0_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34368256)))];
tensor<fp16, [1, 77, 768]> linear_5_cast_fp16 = linear(bias = encoder_encoder_layers_0_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_0_mlp_fc2_weight_to_fp16_palettized, x = input_13_cast_fp16)[name = string("linear_5_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_15_cast_fp16 = add(x = input_7_cast_fp16, y = linear_5_cast_fp16)[name = string("input_15_cast_fp16")];
tensor<int32, [1]> hidden_states_7_axes_0 = const()[name = string("hidden_states_7_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_1_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_1_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34369856)))];
tensor<fp16, [768]> encoder_encoder_layers_1_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34371456)))];
tensor<fp16, [1, 77, 768]> hidden_states_7_cast_fp16 = layer_norm(axes = hidden_states_7_axes_0, beta = encoder_encoder_layers_1_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_1_layer_norm1_weight_to_fp16, x = input_15_cast_fp16)[name = string("hidden_states_7_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_1_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34373056))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34815488))))[name = string("encoder_encoder_layers_1_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_1_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34821696)))];
tensor<fp16, [1, 77, 768]> linear_6_cast_fp16 = linear(bias = encoder_encoder_layers_1_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_1_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_7_cast_fp16)[name = string("linear_6_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_1_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(34823296))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(35265728))))[name = string("encoder_encoder_layers_1_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_1_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(35271936)))];
tensor<fp16, [1, 77, 768]> linear_7_cast_fp16 = linear(bias = encoder_encoder_layers_1_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_1_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_7_cast_fp16)[name = string("linear_7_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_1_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(35273536))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(35715968))))[name = string("encoder_encoder_layers_1_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_1_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(35722176)))];
tensor<fp16, [1, 77, 768]> linear_8_cast_fp16 = linear(bias = encoder_encoder_layers_1_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_1_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_7_cast_fp16)[name = string("linear_8_cast_fp16")];
tensor<int32, [4]> var_141 = const()[name = string("op_141"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_142_cast_fp16 = reshape(shape = var_141, x = linear_6_cast_fp16)[name = string("op_142_cast_fp16")];
tensor<int32, [4]> var_144 = const()[name = string("op_144"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_145_cast_fp16 = reshape(shape = var_144, x = linear_7_cast_fp16)[name = string("op_145_cast_fp16")];
tensor<int32, [4]> var_147 = const()[name = string("op_147"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_148_cast_fp16 = reshape(shape = var_147, x = linear_8_cast_fp16)[name = string("op_148_cast_fp16")];
tensor<int32, [4]> value_3_perm_0 = const()[name = string("value_3_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_3_cast_fp16 = mul(x = var_142_cast_fp16, y = var_9_to_fp16)[name = string("mul_3_cast_fp16")];
bool matmul_1_transpose_y_0 = const()[name = string("matmul_1_transpose_y_0"), val = bool(true)];
bool matmul_1_transpose_x_0 = const()[name = string("matmul_1_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_50_perm_0 = const()[name = string("transpose_50_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_51_perm_0 = const()[name = string("transpose_51_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_51 = transpose(perm = transpose_51_perm_0, x = var_145_cast_fp16)[name = string("transpose_114")];
tensor<fp16, [1, 12, 77, 64]> transpose_50 = transpose(perm = transpose_50_perm_0, x = mul_3_cast_fp16)[name = string("transpose_115")];
tensor<fp16, [1, 12, 77, 77]> matmul_1_cast_fp16 = matmul(transpose_x = matmul_1_transpose_x_0, transpose_y = matmul_1_transpose_y_0, x = transpose_50, y = transpose_51)[name = string("matmul_1_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_1_cast_fp16 = add(x = matmul_1_cast_fp16, y = mul_0_to_fp16)[name = string("add_1_cast_fp16")];
int32 softmax_1_axis_0 = const()[name = string("softmax_1_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_1_cast_fp16 = softmax(axis = softmax_1_axis_0, x = add_1_cast_fp16)[name = string("softmax_1_cast_fp16")];
bool attn_output_5_transpose_x_0 = const()[name = string("attn_output_5_transpose_x_0"), val = bool(false)];
bool attn_output_5_transpose_y_0 = const()[name = string("attn_output_5_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_3_cast_fp16 = transpose(perm = value_3_perm_0, x = var_148_cast_fp16)[name = string("transpose_113")];
tensor<fp16, [1, 12, 77, 64]> attn_output_5_cast_fp16 = matmul(transpose_x = attn_output_5_transpose_x_0, transpose_y = attn_output_5_transpose_y_0, x = softmax_1_cast_fp16, y = value_3_cast_fp16)[name = string("attn_output_5_cast_fp16")];
tensor<int32, [4]> var_151_perm_0 = const()[name = string("op_151_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_153 = const()[name = string("op_153"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_151_cast_fp16 = transpose(perm = var_151_perm_0, x = attn_output_5_cast_fp16)[name = string("transpose_112")];
tensor<fp16, [1, 77, 768]> var_154_cast_fp16 = reshape(shape = var_153, x = var_151_cast_fp16)[name = string("op_154_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_1_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(35723776))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(36166208))))[name = string("encoder_encoder_layers_1_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_1_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(36172416)))];
tensor<fp16, [1, 77, 768]> linear_9_cast_fp16 = linear(bias = encoder_encoder_layers_1_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_1_self_attn_out_proj_weight_to_fp16_palettized, x = var_154_cast_fp16)[name = string("linear_9_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_19_cast_fp16 = add(x = input_15_cast_fp16, y = linear_9_cast_fp16)[name = string("input_19_cast_fp16")];
tensor<int32, [1]> input_21_axes_0 = const()[name = string("input_21_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_1_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_1_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(36174016)))];
tensor<fp16, [768]> encoder_encoder_layers_1_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(36175616)))];
tensor<fp16, [1, 77, 768]> input_21_cast_fp16 = layer_norm(axes = input_21_axes_0, beta = encoder_encoder_layers_1_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_1_layer_norm2_weight_to_fp16, x = input_19_cast_fp16)[name = string("input_21_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_1_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(36177216))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(37946752))))[name = string("encoder_encoder_layers_1_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_1_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(37971392)))];
tensor<fp16, [1, 77, 3072]> linear_10_cast_fp16 = linear(bias = encoder_encoder_layers_1_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_1_mlp_fc1_weight_to_fp16_palettized, x = input_21_cast_fp16)[name = string("linear_10_cast_fp16")];
fp16 var_169_to_fp16 = const()[name = string("op_169_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_170_cast_fp16 = mul(x = linear_10_cast_fp16, y = var_169_to_fp16)[name = string("op_170_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_171_cast_fp16 = sigmoid(x = var_170_cast_fp16)[name = string("op_171_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_25_cast_fp16 = mul(x = linear_10_cast_fp16, y = var_171_cast_fp16)[name = string("input_25_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_1_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(37977600))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(39747136))))[name = string("encoder_encoder_layers_1_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_1_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_1_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(39753344)))];
tensor<fp16, [1, 77, 768]> linear_11_cast_fp16 = linear(bias = encoder_encoder_layers_1_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_1_mlp_fc2_weight_to_fp16_palettized, x = input_25_cast_fp16)[name = string("linear_11_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_27_cast_fp16 = add(x = input_19_cast_fp16, y = linear_11_cast_fp16)[name = string("input_27_cast_fp16")];
tensor<int32, [1]> hidden_states_13_axes_0 = const()[name = string("hidden_states_13_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_2_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_2_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(39754944)))];
tensor<fp16, [768]> encoder_encoder_layers_2_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(39756544)))];
tensor<fp16, [1, 77, 768]> hidden_states_13_cast_fp16 = layer_norm(axes = hidden_states_13_axes_0, beta = encoder_encoder_layers_2_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_2_layer_norm1_weight_to_fp16, x = input_27_cast_fp16)[name = string("hidden_states_13_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_2_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(39758144))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(40200576))))[name = string("encoder_encoder_layers_2_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_2_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(40206784)))];
tensor<fp16, [1, 77, 768]> linear_12_cast_fp16 = linear(bias = encoder_encoder_layers_2_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_2_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_13_cast_fp16)[name = string("linear_12_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_2_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(40208384))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(40650816))))[name = string("encoder_encoder_layers_2_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_2_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(40657024)))];
tensor<fp16, [1, 77, 768]> linear_13_cast_fp16 = linear(bias = encoder_encoder_layers_2_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_2_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_13_cast_fp16)[name = string("linear_13_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_2_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(40658624))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41101056))))[name = string("encoder_encoder_layers_2_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_2_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41107264)))];
tensor<fp16, [1, 77, 768]> linear_14_cast_fp16 = linear(bias = encoder_encoder_layers_2_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_2_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_13_cast_fp16)[name = string("linear_14_cast_fp16")];
tensor<int32, [4]> var_200 = const()[name = string("op_200"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_201_cast_fp16 = reshape(shape = var_200, x = linear_12_cast_fp16)[name = string("op_201_cast_fp16")];
tensor<int32, [4]> var_203 = const()[name = string("op_203"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_204_cast_fp16 = reshape(shape = var_203, x = linear_13_cast_fp16)[name = string("op_204_cast_fp16")];
tensor<int32, [4]> var_206 = const()[name = string("op_206"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_207_cast_fp16 = reshape(shape = var_206, x = linear_14_cast_fp16)[name = string("op_207_cast_fp16")];
tensor<int32, [4]> value_5_perm_0 = const()[name = string("value_5_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_5_cast_fp16 = mul(x = var_201_cast_fp16, y = var_9_to_fp16)[name = string("mul_5_cast_fp16")];
bool matmul_2_transpose_y_0 = const()[name = string("matmul_2_transpose_y_0"), val = bool(true)];
bool matmul_2_transpose_x_0 = const()[name = string("matmul_2_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_52_perm_0 = const()[name = string("transpose_52_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_53_perm_0 = const()[name = string("transpose_53_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_53 = transpose(perm = transpose_53_perm_0, x = var_204_cast_fp16)[name = string("transpose_110")];
tensor<fp16, [1, 12, 77, 64]> transpose_52 = transpose(perm = transpose_52_perm_0, x = mul_5_cast_fp16)[name = string("transpose_111")];
tensor<fp16, [1, 12, 77, 77]> matmul_2_cast_fp16 = matmul(transpose_x = matmul_2_transpose_x_0, transpose_y = matmul_2_transpose_y_0, x = transpose_52, y = transpose_53)[name = string("matmul_2_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_2_cast_fp16 = add(x = matmul_2_cast_fp16, y = mul_0_to_fp16)[name = string("add_2_cast_fp16")];
int32 softmax_2_axis_0 = const()[name = string("softmax_2_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_2_cast_fp16 = softmax(axis = softmax_2_axis_0, x = add_2_cast_fp16)[name = string("softmax_2_cast_fp16")];
bool attn_output_9_transpose_x_0 = const()[name = string("attn_output_9_transpose_x_0"), val = bool(false)];
bool attn_output_9_transpose_y_0 = const()[name = string("attn_output_9_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_5_cast_fp16 = transpose(perm = value_5_perm_0, x = var_207_cast_fp16)[name = string("transpose_109")];
tensor<fp16, [1, 12, 77, 64]> attn_output_9_cast_fp16 = matmul(transpose_x = attn_output_9_transpose_x_0, transpose_y = attn_output_9_transpose_y_0, x = softmax_2_cast_fp16, y = value_5_cast_fp16)[name = string("attn_output_9_cast_fp16")];
tensor<int32, [4]> var_210_perm_0 = const()[name = string("op_210_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_212 = const()[name = string("op_212"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_210_cast_fp16 = transpose(perm = var_210_perm_0, x = attn_output_9_cast_fp16)[name = string("transpose_108")];
tensor<fp16, [1, 77, 768]> var_213_cast_fp16 = reshape(shape = var_212, x = var_210_cast_fp16)[name = string("op_213_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_2_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41108864))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41551296))))[name = string("encoder_encoder_layers_2_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_2_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41557504)))];
tensor<fp16, [1, 77, 768]> linear_15_cast_fp16 = linear(bias = encoder_encoder_layers_2_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_2_self_attn_out_proj_weight_to_fp16_palettized, x = var_213_cast_fp16)[name = string("linear_15_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_31_cast_fp16 = add(x = input_27_cast_fp16, y = linear_15_cast_fp16)[name = string("input_31_cast_fp16")];
tensor<int32, [1]> input_33_axes_0 = const()[name = string("input_33_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_2_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_2_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41559104)))];
tensor<fp16, [768]> encoder_encoder_layers_2_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41560704)))];
tensor<fp16, [1, 77, 768]> input_33_cast_fp16 = layer_norm(axes = input_33_axes_0, beta = encoder_encoder_layers_2_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_2_layer_norm2_weight_to_fp16, x = input_31_cast_fp16)[name = string("input_33_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_2_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(41562304))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(43331840))))[name = string("encoder_encoder_layers_2_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_2_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(43356480)))];
tensor<fp16, [1, 77, 3072]> linear_16_cast_fp16 = linear(bias = encoder_encoder_layers_2_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_2_mlp_fc1_weight_to_fp16_palettized, x = input_33_cast_fp16)[name = string("linear_16_cast_fp16")];
fp16 var_228_to_fp16 = const()[name = string("op_228_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_229_cast_fp16 = mul(x = linear_16_cast_fp16, y = var_228_to_fp16)[name = string("op_229_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_230_cast_fp16 = sigmoid(x = var_229_cast_fp16)[name = string("op_230_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_37_cast_fp16 = mul(x = linear_16_cast_fp16, y = var_230_cast_fp16)[name = string("input_37_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_2_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(43362688))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45132224))))[name = string("encoder_encoder_layers_2_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_2_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_2_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45138432)))];
tensor<fp16, [1, 77, 768]> linear_17_cast_fp16 = linear(bias = encoder_encoder_layers_2_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_2_mlp_fc2_weight_to_fp16_palettized, x = input_37_cast_fp16)[name = string("linear_17_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_39_cast_fp16 = add(x = input_31_cast_fp16, y = linear_17_cast_fp16)[name = string("input_39_cast_fp16")];
tensor<int32, [1]> hidden_states_19_axes_0 = const()[name = string("hidden_states_19_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_3_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_3_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45140032)))];
tensor<fp16, [768]> encoder_encoder_layers_3_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45141632)))];
tensor<fp16, [1, 77, 768]> hidden_states_19_cast_fp16 = layer_norm(axes = hidden_states_19_axes_0, beta = encoder_encoder_layers_3_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_3_layer_norm1_weight_to_fp16, x = input_39_cast_fp16)[name = string("hidden_states_19_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_3_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45143232))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45585664))))[name = string("encoder_encoder_layers_3_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_3_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45591872)))];
tensor<fp16, [1, 77, 768]> linear_18_cast_fp16 = linear(bias = encoder_encoder_layers_3_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_3_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_19_cast_fp16)[name = string("linear_18_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_3_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(45593472))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46035904))))[name = string("encoder_encoder_layers_3_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_3_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46042112)))];
tensor<fp16, [1, 77, 768]> linear_19_cast_fp16 = linear(bias = encoder_encoder_layers_3_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_3_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_19_cast_fp16)[name = string("linear_19_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_3_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46043712))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46486144))))[name = string("encoder_encoder_layers_3_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_3_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46492352)))];
tensor<fp16, [1, 77, 768]> linear_20_cast_fp16 = linear(bias = encoder_encoder_layers_3_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_3_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_19_cast_fp16)[name = string("linear_20_cast_fp16")];
tensor<int32, [4]> var_259 = const()[name = string("op_259"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_260_cast_fp16 = reshape(shape = var_259, x = linear_18_cast_fp16)[name = string("op_260_cast_fp16")];
tensor<int32, [4]> var_262 = const()[name = string("op_262"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_263_cast_fp16 = reshape(shape = var_262, x = linear_19_cast_fp16)[name = string("op_263_cast_fp16")];
tensor<int32, [4]> var_265 = const()[name = string("op_265"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_266_cast_fp16 = reshape(shape = var_265, x = linear_20_cast_fp16)[name = string("op_266_cast_fp16")];
tensor<int32, [4]> value_7_perm_0 = const()[name = string("value_7_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_7_cast_fp16 = mul(x = var_260_cast_fp16, y = var_9_to_fp16)[name = string("mul_7_cast_fp16")];
bool matmul_3_transpose_y_0 = const()[name = string("matmul_3_transpose_y_0"), val = bool(true)];
bool matmul_3_transpose_x_0 = const()[name = string("matmul_3_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_54_perm_0 = const()[name = string("transpose_54_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_55_perm_0 = const()[name = string("transpose_55_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_55 = transpose(perm = transpose_55_perm_0, x = var_263_cast_fp16)[name = string("transpose_106")];
tensor<fp16, [1, 12, 77, 64]> transpose_54 = transpose(perm = transpose_54_perm_0, x = mul_7_cast_fp16)[name = string("transpose_107")];
tensor<fp16, [1, 12, 77, 77]> matmul_3_cast_fp16 = matmul(transpose_x = matmul_3_transpose_x_0, transpose_y = matmul_3_transpose_y_0, x = transpose_54, y = transpose_55)[name = string("matmul_3_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_3_cast_fp16 = add(x = matmul_3_cast_fp16, y = mul_0_to_fp16)[name = string("add_3_cast_fp16")];
int32 softmax_3_axis_0 = const()[name = string("softmax_3_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_3_cast_fp16 = softmax(axis = softmax_3_axis_0, x = add_3_cast_fp16)[name = string("softmax_3_cast_fp16")];
bool attn_output_13_transpose_x_0 = const()[name = string("attn_output_13_transpose_x_0"), val = bool(false)];
bool attn_output_13_transpose_y_0 = const()[name = string("attn_output_13_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_7_cast_fp16 = transpose(perm = value_7_perm_0, x = var_266_cast_fp16)[name = string("transpose_105")];
tensor<fp16, [1, 12, 77, 64]> attn_output_13_cast_fp16 = matmul(transpose_x = attn_output_13_transpose_x_0, transpose_y = attn_output_13_transpose_y_0, x = softmax_3_cast_fp16, y = value_7_cast_fp16)[name = string("attn_output_13_cast_fp16")];
tensor<int32, [4]> var_269_perm_0 = const()[name = string("op_269_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_271 = const()[name = string("op_271"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_269_cast_fp16 = transpose(perm = var_269_perm_0, x = attn_output_13_cast_fp16)[name = string("transpose_104")];
tensor<fp16, [1, 77, 768]> var_272_cast_fp16 = reshape(shape = var_271, x = var_269_cast_fp16)[name = string("op_272_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_3_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46493952))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46936384))))[name = string("encoder_encoder_layers_3_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_3_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46942592)))];
tensor<fp16, [1, 77, 768]> linear_21_cast_fp16 = linear(bias = encoder_encoder_layers_3_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_3_self_attn_out_proj_weight_to_fp16_palettized, x = var_272_cast_fp16)[name = string("linear_21_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_43_cast_fp16 = add(x = input_39_cast_fp16, y = linear_21_cast_fp16)[name = string("input_43_cast_fp16")];
tensor<int32, [1]> input_45_axes_0 = const()[name = string("input_45_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_3_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_3_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46944192)))];
tensor<fp16, [768]> encoder_encoder_layers_3_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46945792)))];
tensor<fp16, [1, 77, 768]> input_45_cast_fp16 = layer_norm(axes = input_45_axes_0, beta = encoder_encoder_layers_3_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_3_layer_norm2_weight_to_fp16, x = input_43_cast_fp16)[name = string("input_45_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_3_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(46947392))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(48716928))))[name = string("encoder_encoder_layers_3_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_3_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(48741568)))];
tensor<fp16, [1, 77, 3072]> linear_22_cast_fp16 = linear(bias = encoder_encoder_layers_3_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_3_mlp_fc1_weight_to_fp16_palettized, x = input_45_cast_fp16)[name = string("linear_22_cast_fp16")];
fp16 var_287_to_fp16 = const()[name = string("op_287_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_288_cast_fp16 = mul(x = linear_22_cast_fp16, y = var_287_to_fp16)[name = string("op_288_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_289_cast_fp16 = sigmoid(x = var_288_cast_fp16)[name = string("op_289_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_49_cast_fp16 = mul(x = linear_22_cast_fp16, y = var_289_cast_fp16)[name = string("input_49_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_3_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(48747776))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50517312))))[name = string("encoder_encoder_layers_3_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_3_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_3_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50523520)))];
tensor<fp16, [1, 77, 768]> linear_23_cast_fp16 = linear(bias = encoder_encoder_layers_3_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_3_mlp_fc2_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = string("linear_23_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_51_cast_fp16 = add(x = input_43_cast_fp16, y = linear_23_cast_fp16)[name = string("input_51_cast_fp16")];
tensor<int32, [1]> hidden_states_25_axes_0 = const()[name = string("hidden_states_25_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_4_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_4_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50525120)))];
tensor<fp16, [768]> encoder_encoder_layers_4_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50526720)))];
tensor<fp16, [1, 77, 768]> hidden_states_25_cast_fp16 = layer_norm(axes = hidden_states_25_axes_0, beta = encoder_encoder_layers_4_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_4_layer_norm1_weight_to_fp16, x = input_51_cast_fp16)[name = string("hidden_states_25_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_4_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50528320))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50970752))))[name = string("encoder_encoder_layers_4_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_4_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50976960)))];
tensor<fp16, [1, 77, 768]> linear_24_cast_fp16 = linear(bias = encoder_encoder_layers_4_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_4_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_25_cast_fp16)[name = string("linear_24_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_4_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(50978560))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(51420992))))[name = string("encoder_encoder_layers_4_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_4_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(51427200)))];
tensor<fp16, [1, 77, 768]> linear_25_cast_fp16 = linear(bias = encoder_encoder_layers_4_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_4_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_25_cast_fp16)[name = string("linear_25_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_4_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(51428800))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(51871232))))[name = string("encoder_encoder_layers_4_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_4_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(51877440)))];
tensor<fp16, [1, 77, 768]> linear_26_cast_fp16 = linear(bias = encoder_encoder_layers_4_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_4_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_25_cast_fp16)[name = string("linear_26_cast_fp16")];
tensor<int32, [4]> var_318 = const()[name = string("op_318"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_319_cast_fp16 = reshape(shape = var_318, x = linear_24_cast_fp16)[name = string("op_319_cast_fp16")];
tensor<int32, [4]> var_321 = const()[name = string("op_321"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_322_cast_fp16 = reshape(shape = var_321, x = linear_25_cast_fp16)[name = string("op_322_cast_fp16")];
tensor<int32, [4]> var_324 = const()[name = string("op_324"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_325_cast_fp16 = reshape(shape = var_324, x = linear_26_cast_fp16)[name = string("op_325_cast_fp16")];
tensor<int32, [4]> value_9_perm_0 = const()[name = string("value_9_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_9_cast_fp16 = mul(x = var_319_cast_fp16, y = var_9_to_fp16)[name = string("mul_9_cast_fp16")];
bool matmul_4_transpose_y_0 = const()[name = string("matmul_4_transpose_y_0"), val = bool(true)];
bool matmul_4_transpose_x_0 = const()[name = string("matmul_4_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_56_perm_0 = const()[name = string("transpose_56_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_57_perm_0 = const()[name = string("transpose_57_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_57 = transpose(perm = transpose_57_perm_0, x = var_322_cast_fp16)[name = string("transpose_102")];
tensor<fp16, [1, 12, 77, 64]> transpose_56 = transpose(perm = transpose_56_perm_0, x = mul_9_cast_fp16)[name = string("transpose_103")];
tensor<fp16, [1, 12, 77, 77]> matmul_4_cast_fp16 = matmul(transpose_x = matmul_4_transpose_x_0, transpose_y = matmul_4_transpose_y_0, x = transpose_56, y = transpose_57)[name = string("matmul_4_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_4_cast_fp16 = add(x = matmul_4_cast_fp16, y = mul_0_to_fp16)[name = string("add_4_cast_fp16")];
int32 softmax_4_axis_0 = const()[name = string("softmax_4_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_4_cast_fp16 = softmax(axis = softmax_4_axis_0, x = add_4_cast_fp16)[name = string("softmax_4_cast_fp16")];
bool attn_output_17_transpose_x_0 = const()[name = string("attn_output_17_transpose_x_0"), val = bool(false)];
bool attn_output_17_transpose_y_0 = const()[name = string("attn_output_17_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_9_cast_fp16 = transpose(perm = value_9_perm_0, x = var_325_cast_fp16)[name = string("transpose_101")];
tensor<fp16, [1, 12, 77, 64]> attn_output_17_cast_fp16 = matmul(transpose_x = attn_output_17_transpose_x_0, transpose_y = attn_output_17_transpose_y_0, x = softmax_4_cast_fp16, y = value_9_cast_fp16)[name = string("attn_output_17_cast_fp16")];
tensor<int32, [4]> var_328_perm_0 = const()[name = string("op_328_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_330 = const()[name = string("op_330"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_328_cast_fp16 = transpose(perm = var_328_perm_0, x = attn_output_17_cast_fp16)[name = string("transpose_100")];
tensor<fp16, [1, 77, 768]> var_331_cast_fp16 = reshape(shape = var_330, x = var_328_cast_fp16)[name = string("op_331_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_4_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(51879040))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(52321472))))[name = string("encoder_encoder_layers_4_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_4_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(52327680)))];
tensor<fp16, [1, 77, 768]> linear_27_cast_fp16 = linear(bias = encoder_encoder_layers_4_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_4_self_attn_out_proj_weight_to_fp16_palettized, x = var_331_cast_fp16)[name = string("linear_27_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_55_cast_fp16 = add(x = input_51_cast_fp16, y = linear_27_cast_fp16)[name = string("input_55_cast_fp16")];
tensor<int32, [1]> input_57_axes_0 = const()[name = string("input_57_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_4_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_4_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(52329280)))];
tensor<fp16, [768]> encoder_encoder_layers_4_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(52330880)))];
tensor<fp16, [1, 77, 768]> input_57_cast_fp16 = layer_norm(axes = input_57_axes_0, beta = encoder_encoder_layers_4_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_4_layer_norm2_weight_to_fp16, x = input_55_cast_fp16)[name = string("input_57_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_4_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(52332480))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(54102016))))[name = string("encoder_encoder_layers_4_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_4_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(54126656)))];
tensor<fp16, [1, 77, 3072]> linear_28_cast_fp16 = linear(bias = encoder_encoder_layers_4_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_4_mlp_fc1_weight_to_fp16_palettized, x = input_57_cast_fp16)[name = string("linear_28_cast_fp16")];
fp16 var_346_to_fp16 = const()[name = string("op_346_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_347_cast_fp16 = mul(x = linear_28_cast_fp16, y = var_346_to_fp16)[name = string("op_347_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_348_cast_fp16 = sigmoid(x = var_347_cast_fp16)[name = string("op_348_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_61_cast_fp16 = mul(x = linear_28_cast_fp16, y = var_348_cast_fp16)[name = string("input_61_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_4_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(54132864))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(55902400))))[name = string("encoder_encoder_layers_4_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_4_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_4_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(55908608)))];
tensor<fp16, [1, 77, 768]> linear_29_cast_fp16 = linear(bias = encoder_encoder_layers_4_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_4_mlp_fc2_weight_to_fp16_palettized, x = input_61_cast_fp16)[name = string("linear_29_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_63_cast_fp16 = add(x = input_55_cast_fp16, y = linear_29_cast_fp16)[name = string("input_63_cast_fp16")];
tensor<int32, [1]> hidden_states_31_axes_0 = const()[name = string("hidden_states_31_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_5_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_5_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(55910208)))];
tensor<fp16, [768]> encoder_encoder_layers_5_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(55911808)))];
tensor<fp16, [1, 77, 768]> hidden_states_31_cast_fp16 = layer_norm(axes = hidden_states_31_axes_0, beta = encoder_encoder_layers_5_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_5_layer_norm1_weight_to_fp16, x = input_63_cast_fp16)[name = string("hidden_states_31_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_5_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(55913408))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(56355840))))[name = string("encoder_encoder_layers_5_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_5_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(56362048)))];
tensor<fp16, [1, 77, 768]> linear_30_cast_fp16 = linear(bias = encoder_encoder_layers_5_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_5_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_31_cast_fp16)[name = string("linear_30_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_5_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(56363648))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(56806080))))[name = string("encoder_encoder_layers_5_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_5_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(56812288)))];
tensor<fp16, [1, 77, 768]> linear_31_cast_fp16 = linear(bias = encoder_encoder_layers_5_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_5_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_31_cast_fp16)[name = string("linear_31_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_5_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(56813888))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57256320))))[name = string("encoder_encoder_layers_5_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_5_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57262528)))];
tensor<fp16, [1, 77, 768]> linear_32_cast_fp16 = linear(bias = encoder_encoder_layers_5_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_5_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_31_cast_fp16)[name = string("linear_32_cast_fp16")];
tensor<int32, [4]> var_377 = const()[name = string("op_377"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_378_cast_fp16 = reshape(shape = var_377, x = linear_30_cast_fp16)[name = string("op_378_cast_fp16")];
tensor<int32, [4]> var_380 = const()[name = string("op_380"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_381_cast_fp16 = reshape(shape = var_380, x = linear_31_cast_fp16)[name = string("op_381_cast_fp16")];
tensor<int32, [4]> var_383 = const()[name = string("op_383"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_384_cast_fp16 = reshape(shape = var_383, x = linear_32_cast_fp16)[name = string("op_384_cast_fp16")];
tensor<int32, [4]> value_11_perm_0 = const()[name = string("value_11_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_11_cast_fp16 = mul(x = var_378_cast_fp16, y = var_9_to_fp16)[name = string("mul_11_cast_fp16")];
bool matmul_5_transpose_y_0 = const()[name = string("matmul_5_transpose_y_0"), val = bool(true)];
bool matmul_5_transpose_x_0 = const()[name = string("matmul_5_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_58_perm_0 = const()[name = string("transpose_58_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_59_perm_0 = const()[name = string("transpose_59_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_59 = transpose(perm = transpose_59_perm_0, x = var_381_cast_fp16)[name = string("transpose_98")];
tensor<fp16, [1, 12, 77, 64]> transpose_58 = transpose(perm = transpose_58_perm_0, x = mul_11_cast_fp16)[name = string("transpose_99")];
tensor<fp16, [1, 12, 77, 77]> matmul_5_cast_fp16 = matmul(transpose_x = matmul_5_transpose_x_0, transpose_y = matmul_5_transpose_y_0, x = transpose_58, y = transpose_59)[name = string("matmul_5_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_5_cast_fp16 = add(x = matmul_5_cast_fp16, y = mul_0_to_fp16)[name = string("add_5_cast_fp16")];
int32 softmax_5_axis_0 = const()[name = string("softmax_5_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_5_cast_fp16 = softmax(axis = softmax_5_axis_0, x = add_5_cast_fp16)[name = string("softmax_5_cast_fp16")];
bool attn_output_21_transpose_x_0 = const()[name = string("attn_output_21_transpose_x_0"), val = bool(false)];
bool attn_output_21_transpose_y_0 = const()[name = string("attn_output_21_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_11_cast_fp16 = transpose(perm = value_11_perm_0, x = var_384_cast_fp16)[name = string("transpose_97")];
tensor<fp16, [1, 12, 77, 64]> attn_output_21_cast_fp16 = matmul(transpose_x = attn_output_21_transpose_x_0, transpose_y = attn_output_21_transpose_y_0, x = softmax_5_cast_fp16, y = value_11_cast_fp16)[name = string("attn_output_21_cast_fp16")];
tensor<int32, [4]> var_387_perm_0 = const()[name = string("op_387_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_389 = const()[name = string("op_389"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_387_cast_fp16 = transpose(perm = var_387_perm_0, x = attn_output_21_cast_fp16)[name = string("transpose_96")];
tensor<fp16, [1, 77, 768]> var_390_cast_fp16 = reshape(shape = var_389, x = var_387_cast_fp16)[name = string("op_390_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_5_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57264128))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57706560))))[name = string("encoder_encoder_layers_5_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_5_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57712768)))];
tensor<fp16, [1, 77, 768]> linear_33_cast_fp16 = linear(bias = encoder_encoder_layers_5_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_5_self_attn_out_proj_weight_to_fp16_palettized, x = var_390_cast_fp16)[name = string("linear_33_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_67_cast_fp16 = add(x = input_63_cast_fp16, y = linear_33_cast_fp16)[name = string("input_67_cast_fp16")];
tensor<int32, [1]> input_69_axes_0 = const()[name = string("input_69_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_5_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_5_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57714368)))];
tensor<fp16, [768]> encoder_encoder_layers_5_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57715968)))];
tensor<fp16, [1, 77, 768]> input_69_cast_fp16 = layer_norm(axes = input_69_axes_0, beta = encoder_encoder_layers_5_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_5_layer_norm2_weight_to_fp16, x = input_67_cast_fp16)[name = string("input_69_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_5_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(57717568))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(59487104))))[name = string("encoder_encoder_layers_5_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_5_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(59511744)))];
tensor<fp16, [1, 77, 3072]> linear_34_cast_fp16 = linear(bias = encoder_encoder_layers_5_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_5_mlp_fc1_weight_to_fp16_palettized, x = input_69_cast_fp16)[name = string("linear_34_cast_fp16")];
fp16 var_405_to_fp16 = const()[name = string("op_405_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_406_cast_fp16 = mul(x = linear_34_cast_fp16, y = var_405_to_fp16)[name = string("op_406_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_407_cast_fp16 = sigmoid(x = var_406_cast_fp16)[name = string("op_407_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_73_cast_fp16 = mul(x = linear_34_cast_fp16, y = var_407_cast_fp16)[name = string("input_73_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_5_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(59517952))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61287488))))[name = string("encoder_encoder_layers_5_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_5_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_5_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61293696)))];
tensor<fp16, [1, 77, 768]> linear_35_cast_fp16 = linear(bias = encoder_encoder_layers_5_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_5_mlp_fc2_weight_to_fp16_palettized, x = input_73_cast_fp16)[name = string("linear_35_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_75_cast_fp16 = add(x = input_67_cast_fp16, y = linear_35_cast_fp16)[name = string("input_75_cast_fp16")];
tensor<int32, [1]> hidden_states_37_axes_0 = const()[name = string("hidden_states_37_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_6_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_6_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61295296)))];
tensor<fp16, [768]> encoder_encoder_layers_6_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61296896)))];
tensor<fp16, [1, 77, 768]> hidden_states_37_cast_fp16 = layer_norm(axes = hidden_states_37_axes_0, beta = encoder_encoder_layers_6_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_6_layer_norm1_weight_to_fp16, x = input_75_cast_fp16)[name = string("hidden_states_37_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_6_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61298496))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61740928))))[name = string("encoder_encoder_layers_6_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_6_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61747136)))];
tensor<fp16, [1, 77, 768]> linear_36_cast_fp16 = linear(bias = encoder_encoder_layers_6_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_6_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_37_cast_fp16)[name = string("linear_36_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_6_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(61748736))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(62191168))))[name = string("encoder_encoder_layers_6_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_6_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(62197376)))];
tensor<fp16, [1, 77, 768]> linear_37_cast_fp16 = linear(bias = encoder_encoder_layers_6_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_6_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_37_cast_fp16)[name = string("linear_37_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_6_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(62198976))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(62641408))))[name = string("encoder_encoder_layers_6_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_6_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(62647616)))];
tensor<fp16, [1, 77, 768]> linear_38_cast_fp16 = linear(bias = encoder_encoder_layers_6_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_6_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_37_cast_fp16)[name = string("linear_38_cast_fp16")];
tensor<int32, [4]> var_436 = const()[name = string("op_436"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_437_cast_fp16 = reshape(shape = var_436, x = linear_36_cast_fp16)[name = string("op_437_cast_fp16")];
tensor<int32, [4]> var_439 = const()[name = string("op_439"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_440_cast_fp16 = reshape(shape = var_439, x = linear_37_cast_fp16)[name = string("op_440_cast_fp16")];
tensor<int32, [4]> var_442 = const()[name = string("op_442"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_443_cast_fp16 = reshape(shape = var_442, x = linear_38_cast_fp16)[name = string("op_443_cast_fp16")];
tensor<int32, [4]> value_13_perm_0 = const()[name = string("value_13_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_13_cast_fp16 = mul(x = var_437_cast_fp16, y = var_9_to_fp16)[name = string("mul_13_cast_fp16")];
bool matmul_6_transpose_y_0 = const()[name = string("matmul_6_transpose_y_0"), val = bool(true)];
bool matmul_6_transpose_x_0 = const()[name = string("matmul_6_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_60_perm_0 = const()[name = string("transpose_60_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_61_perm_0 = const()[name = string("transpose_61_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_61 = transpose(perm = transpose_61_perm_0, x = var_440_cast_fp16)[name = string("transpose_94")];
tensor<fp16, [1, 12, 77, 64]> transpose_60 = transpose(perm = transpose_60_perm_0, x = mul_13_cast_fp16)[name = string("transpose_95")];
tensor<fp16, [1, 12, 77, 77]> matmul_6_cast_fp16 = matmul(transpose_x = matmul_6_transpose_x_0, transpose_y = matmul_6_transpose_y_0, x = transpose_60, y = transpose_61)[name = string("matmul_6_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_6_cast_fp16 = add(x = matmul_6_cast_fp16, y = mul_0_to_fp16)[name = string("add_6_cast_fp16")];
int32 softmax_6_axis_0 = const()[name = string("softmax_6_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_6_cast_fp16 = softmax(axis = softmax_6_axis_0, x = add_6_cast_fp16)[name = string("softmax_6_cast_fp16")];
bool attn_output_25_transpose_x_0 = const()[name = string("attn_output_25_transpose_x_0"), val = bool(false)];
bool attn_output_25_transpose_y_0 = const()[name = string("attn_output_25_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_13_cast_fp16 = transpose(perm = value_13_perm_0, x = var_443_cast_fp16)[name = string("transpose_93")];
tensor<fp16, [1, 12, 77, 64]> attn_output_25_cast_fp16 = matmul(transpose_x = attn_output_25_transpose_x_0, transpose_y = attn_output_25_transpose_y_0, x = softmax_6_cast_fp16, y = value_13_cast_fp16)[name = string("attn_output_25_cast_fp16")];
tensor<int32, [4]> var_446_perm_0 = const()[name = string("op_446_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_448 = const()[name = string("op_448"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_446_cast_fp16 = transpose(perm = var_446_perm_0, x = attn_output_25_cast_fp16)[name = string("transpose_92")];
tensor<fp16, [1, 77, 768]> var_449_cast_fp16 = reshape(shape = var_448, x = var_446_cast_fp16)[name = string("op_449_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_6_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(62649216))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(63091648))))[name = string("encoder_encoder_layers_6_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_6_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(63097856)))];
tensor<fp16, [1, 77, 768]> linear_39_cast_fp16 = linear(bias = encoder_encoder_layers_6_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_6_self_attn_out_proj_weight_to_fp16_palettized, x = var_449_cast_fp16)[name = string("linear_39_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_79_cast_fp16 = add(x = input_75_cast_fp16, y = linear_39_cast_fp16)[name = string("input_79_cast_fp16")];
tensor<int32, [1]> input_81_axes_0 = const()[name = string("input_81_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_6_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_6_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(63099456)))];
tensor<fp16, [768]> encoder_encoder_layers_6_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(63101056)))];
tensor<fp16, [1, 77, 768]> input_81_cast_fp16 = layer_norm(axes = input_81_axes_0, beta = encoder_encoder_layers_6_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_6_layer_norm2_weight_to_fp16, x = input_79_cast_fp16)[name = string("input_81_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_6_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(63102656))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64872192))))[name = string("encoder_encoder_layers_6_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_6_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64896832)))];
tensor<fp16, [1, 77, 3072]> linear_40_cast_fp16 = linear(bias = encoder_encoder_layers_6_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_6_mlp_fc1_weight_to_fp16_palettized, x = input_81_cast_fp16)[name = string("linear_40_cast_fp16")];
fp16 var_464_to_fp16 = const()[name = string("op_464_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_465_cast_fp16 = mul(x = linear_40_cast_fp16, y = var_464_to_fp16)[name = string("op_465_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_466_cast_fp16 = sigmoid(x = var_465_cast_fp16)[name = string("op_466_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_85_cast_fp16 = mul(x = linear_40_cast_fp16, y = var_466_cast_fp16)[name = string("input_85_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_6_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(64903040))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(66672576))))[name = string("encoder_encoder_layers_6_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_6_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_6_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(66678784)))];
tensor<fp16, [1, 77, 768]> linear_41_cast_fp16 = linear(bias = encoder_encoder_layers_6_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_6_mlp_fc2_weight_to_fp16_palettized, x = input_85_cast_fp16)[name = string("linear_41_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_87_cast_fp16 = add(x = input_79_cast_fp16, y = linear_41_cast_fp16)[name = string("input_87_cast_fp16")];
tensor<int32, [1]> hidden_states_43_axes_0 = const()[name = string("hidden_states_43_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_7_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_7_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(66680384)))];
tensor<fp16, [768]> encoder_encoder_layers_7_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(66681984)))];
tensor<fp16, [1, 77, 768]> hidden_states_43_cast_fp16 = layer_norm(axes = hidden_states_43_axes_0, beta = encoder_encoder_layers_7_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_7_layer_norm1_weight_to_fp16, x = input_87_cast_fp16)[name = string("hidden_states_43_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_7_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(66683584))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67126016))))[name = string("encoder_encoder_layers_7_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_7_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67132224)))];
tensor<fp16, [1, 77, 768]> linear_42_cast_fp16 = linear(bias = encoder_encoder_layers_7_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_7_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_43_cast_fp16)[name = string("linear_42_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_7_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67133824))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67576256))))[name = string("encoder_encoder_layers_7_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_7_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67582464)))];
tensor<fp16, [1, 77, 768]> linear_43_cast_fp16 = linear(bias = encoder_encoder_layers_7_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_7_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_43_cast_fp16)[name = string("linear_43_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_7_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(67584064))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68026496))))[name = string("encoder_encoder_layers_7_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_7_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68032704)))];
tensor<fp16, [1, 77, 768]> linear_44_cast_fp16 = linear(bias = encoder_encoder_layers_7_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_7_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_43_cast_fp16)[name = string("linear_44_cast_fp16")];
tensor<int32, [4]> var_495 = const()[name = string("op_495"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_496_cast_fp16 = reshape(shape = var_495, x = linear_42_cast_fp16)[name = string("op_496_cast_fp16")];
tensor<int32, [4]> var_498 = const()[name = string("op_498"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_499_cast_fp16 = reshape(shape = var_498, x = linear_43_cast_fp16)[name = string("op_499_cast_fp16")];
tensor<int32, [4]> var_501 = const()[name = string("op_501"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_502_cast_fp16 = reshape(shape = var_501, x = linear_44_cast_fp16)[name = string("op_502_cast_fp16")];
tensor<int32, [4]> value_15_perm_0 = const()[name = string("value_15_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_15_cast_fp16 = mul(x = var_496_cast_fp16, y = var_9_to_fp16)[name = string("mul_15_cast_fp16")];
bool matmul_7_transpose_y_0 = const()[name = string("matmul_7_transpose_y_0"), val = bool(true)];
bool matmul_7_transpose_x_0 = const()[name = string("matmul_7_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_62_perm_0 = const()[name = string("transpose_62_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_63_perm_0 = const()[name = string("transpose_63_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_63 = transpose(perm = transpose_63_perm_0, x = var_499_cast_fp16)[name = string("transpose_90")];
tensor<fp16, [1, 12, 77, 64]> transpose_62 = transpose(perm = transpose_62_perm_0, x = mul_15_cast_fp16)[name = string("transpose_91")];
tensor<fp16, [1, 12, 77, 77]> matmul_7_cast_fp16 = matmul(transpose_x = matmul_7_transpose_x_0, transpose_y = matmul_7_transpose_y_0, x = transpose_62, y = transpose_63)[name = string("matmul_7_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_7_cast_fp16 = add(x = matmul_7_cast_fp16, y = mul_0_to_fp16)[name = string("add_7_cast_fp16")];
int32 softmax_7_axis_0 = const()[name = string("softmax_7_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_7_cast_fp16 = softmax(axis = softmax_7_axis_0, x = add_7_cast_fp16)[name = string("softmax_7_cast_fp16")];
bool attn_output_29_transpose_x_0 = const()[name = string("attn_output_29_transpose_x_0"), val = bool(false)];
bool attn_output_29_transpose_y_0 = const()[name = string("attn_output_29_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_15_cast_fp16 = transpose(perm = value_15_perm_0, x = var_502_cast_fp16)[name = string("transpose_89")];
tensor<fp16, [1, 12, 77, 64]> attn_output_29_cast_fp16 = matmul(transpose_x = attn_output_29_transpose_x_0, transpose_y = attn_output_29_transpose_y_0, x = softmax_7_cast_fp16, y = value_15_cast_fp16)[name = string("attn_output_29_cast_fp16")];
tensor<int32, [4]> var_505_perm_0 = const()[name = string("op_505_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_507 = const()[name = string("op_507"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_505_cast_fp16 = transpose(perm = var_505_perm_0, x = attn_output_29_cast_fp16)[name = string("transpose_88")];
tensor<fp16, [1, 77, 768]> var_508_cast_fp16 = reshape(shape = var_507, x = var_505_cast_fp16)[name = string("op_508_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_7_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68034304))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68476736))))[name = string("encoder_encoder_layers_7_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_7_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68482944)))];
tensor<fp16, [1, 77, 768]> linear_45_cast_fp16 = linear(bias = encoder_encoder_layers_7_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_7_self_attn_out_proj_weight_to_fp16_palettized, x = var_508_cast_fp16)[name = string("linear_45_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_91_cast_fp16 = add(x = input_87_cast_fp16, y = linear_45_cast_fp16)[name = string("input_91_cast_fp16")];
tensor<int32, [1]> input_93_axes_0 = const()[name = string("input_93_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_7_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_7_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68484544)))];
tensor<fp16, [768]> encoder_encoder_layers_7_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68486144)))];
tensor<fp16, [1, 77, 768]> input_93_cast_fp16 = layer_norm(axes = input_93_axes_0, beta = encoder_encoder_layers_7_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_7_layer_norm2_weight_to_fp16, x = input_91_cast_fp16)[name = string("input_93_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_7_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(68487744))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(70257280))))[name = string("encoder_encoder_layers_7_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_7_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(70281920)))];
tensor<fp16, [1, 77, 3072]> linear_46_cast_fp16 = linear(bias = encoder_encoder_layers_7_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_7_mlp_fc1_weight_to_fp16_palettized, x = input_93_cast_fp16)[name = string("linear_46_cast_fp16")];
fp16 var_523_to_fp16 = const()[name = string("op_523_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_524_cast_fp16 = mul(x = linear_46_cast_fp16, y = var_523_to_fp16)[name = string("op_524_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_525_cast_fp16 = sigmoid(x = var_524_cast_fp16)[name = string("op_525_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_97_cast_fp16 = mul(x = linear_46_cast_fp16, y = var_525_cast_fp16)[name = string("input_97_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_7_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(70288128))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72057664))))[name = string("encoder_encoder_layers_7_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_7_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_7_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72063872)))];
tensor<fp16, [1, 77, 768]> linear_47_cast_fp16 = linear(bias = encoder_encoder_layers_7_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_7_mlp_fc2_weight_to_fp16_palettized, x = input_97_cast_fp16)[name = string("linear_47_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_99_cast_fp16 = add(x = input_91_cast_fp16, y = linear_47_cast_fp16)[name = string("input_99_cast_fp16")];
tensor<int32, [1]> hidden_states_49_axes_0 = const()[name = string("hidden_states_49_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_8_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_8_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72065472)))];
tensor<fp16, [768]> encoder_encoder_layers_8_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72067072)))];
tensor<fp16, [1, 77, 768]> hidden_states_49_cast_fp16 = layer_norm(axes = hidden_states_49_axes_0, beta = encoder_encoder_layers_8_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_8_layer_norm1_weight_to_fp16, x = input_99_cast_fp16)[name = string("hidden_states_49_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_8_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72068672))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72511104))))[name = string("encoder_encoder_layers_8_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_8_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72517312)))];
tensor<fp16, [1, 77, 768]> linear_48_cast_fp16 = linear(bias = encoder_encoder_layers_8_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_8_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_49_cast_fp16)[name = string("linear_48_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_8_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72518912))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72961344))))[name = string("encoder_encoder_layers_8_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_8_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72967552)))];
tensor<fp16, [1, 77, 768]> linear_49_cast_fp16 = linear(bias = encoder_encoder_layers_8_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_8_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_49_cast_fp16)[name = string("linear_49_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_8_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(72969152))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73411584))))[name = string("encoder_encoder_layers_8_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_8_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73417792)))];
tensor<fp16, [1, 77, 768]> linear_50_cast_fp16 = linear(bias = encoder_encoder_layers_8_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_8_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_49_cast_fp16)[name = string("linear_50_cast_fp16")];
tensor<int32, [4]> var_554 = const()[name = string("op_554"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_555_cast_fp16 = reshape(shape = var_554, x = linear_48_cast_fp16)[name = string("op_555_cast_fp16")];
tensor<int32, [4]> var_557 = const()[name = string("op_557"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_558_cast_fp16 = reshape(shape = var_557, x = linear_49_cast_fp16)[name = string("op_558_cast_fp16")];
tensor<int32, [4]> var_560 = const()[name = string("op_560"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_561_cast_fp16 = reshape(shape = var_560, x = linear_50_cast_fp16)[name = string("op_561_cast_fp16")];
tensor<int32, [4]> value_17_perm_0 = const()[name = string("value_17_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_17_cast_fp16 = mul(x = var_555_cast_fp16, y = var_9_to_fp16)[name = string("mul_17_cast_fp16")];
bool matmul_8_transpose_y_0 = const()[name = string("matmul_8_transpose_y_0"), val = bool(true)];
bool matmul_8_transpose_x_0 = const()[name = string("matmul_8_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_64_perm_0 = const()[name = string("transpose_64_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_65_perm_0 = const()[name = string("transpose_65_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_65 = transpose(perm = transpose_65_perm_0, x = var_558_cast_fp16)[name = string("transpose_86")];
tensor<fp16, [1, 12, 77, 64]> transpose_64 = transpose(perm = transpose_64_perm_0, x = mul_17_cast_fp16)[name = string("transpose_87")];
tensor<fp16, [1, 12, 77, 77]> matmul_8_cast_fp16 = matmul(transpose_x = matmul_8_transpose_x_0, transpose_y = matmul_8_transpose_y_0, x = transpose_64, y = transpose_65)[name = string("matmul_8_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_8_cast_fp16 = add(x = matmul_8_cast_fp16, y = mul_0_to_fp16)[name = string("add_8_cast_fp16")];
int32 softmax_8_axis_0 = const()[name = string("softmax_8_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_8_cast_fp16 = softmax(axis = softmax_8_axis_0, x = add_8_cast_fp16)[name = string("softmax_8_cast_fp16")];
bool attn_output_33_transpose_x_0 = const()[name = string("attn_output_33_transpose_x_0"), val = bool(false)];
bool attn_output_33_transpose_y_0 = const()[name = string("attn_output_33_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_17_cast_fp16 = transpose(perm = value_17_perm_0, x = var_561_cast_fp16)[name = string("transpose_85")];
tensor<fp16, [1, 12, 77, 64]> attn_output_33_cast_fp16 = matmul(transpose_x = attn_output_33_transpose_x_0, transpose_y = attn_output_33_transpose_y_0, x = softmax_8_cast_fp16, y = value_17_cast_fp16)[name = string("attn_output_33_cast_fp16")];
tensor<int32, [4]> var_564_perm_0 = const()[name = string("op_564_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_566 = const()[name = string("op_566"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_564_cast_fp16 = transpose(perm = var_564_perm_0, x = attn_output_33_cast_fp16)[name = string("transpose_84")];
tensor<fp16, [1, 77, 768]> var_567_cast_fp16 = reshape(shape = var_566, x = var_564_cast_fp16)[name = string("op_567_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_8_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73419392))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73861824))))[name = string("encoder_encoder_layers_8_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_8_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73868032)))];
tensor<fp16, [1, 77, 768]> linear_51_cast_fp16 = linear(bias = encoder_encoder_layers_8_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_8_self_attn_out_proj_weight_to_fp16_palettized, x = var_567_cast_fp16)[name = string("linear_51_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_103_cast_fp16 = add(x = input_99_cast_fp16, y = linear_51_cast_fp16)[name = string("input_103_cast_fp16")];
tensor<int32, [1]> input_105_axes_0 = const()[name = string("input_105_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_8_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_8_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73869632)))];
tensor<fp16, [768]> encoder_encoder_layers_8_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73871232)))];
tensor<fp16, [1, 77, 768]> input_105_cast_fp16 = layer_norm(axes = input_105_axes_0, beta = encoder_encoder_layers_8_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_8_layer_norm2_weight_to_fp16, x = input_103_cast_fp16)[name = string("input_105_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_8_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(73872832))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(75642368))))[name = string("encoder_encoder_layers_8_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_8_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(75667008)))];
tensor<fp16, [1, 77, 3072]> linear_52_cast_fp16 = linear(bias = encoder_encoder_layers_8_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_8_mlp_fc1_weight_to_fp16_palettized, x = input_105_cast_fp16)[name = string("linear_52_cast_fp16")];
fp16 var_582_to_fp16 = const()[name = string("op_582_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_583_cast_fp16 = mul(x = linear_52_cast_fp16, y = var_582_to_fp16)[name = string("op_583_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_584_cast_fp16 = sigmoid(x = var_583_cast_fp16)[name = string("op_584_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_109_cast_fp16 = mul(x = linear_52_cast_fp16, y = var_584_cast_fp16)[name = string("input_109_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_8_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(75673216))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77442752))))[name = string("encoder_encoder_layers_8_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_8_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_8_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77448960)))];
tensor<fp16, [1, 77, 768]> linear_53_cast_fp16 = linear(bias = encoder_encoder_layers_8_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_8_mlp_fc2_weight_to_fp16_palettized, x = input_109_cast_fp16)[name = string("linear_53_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_111_cast_fp16 = add(x = input_103_cast_fp16, y = linear_53_cast_fp16)[name = string("input_111_cast_fp16")];
tensor<int32, [1]> hidden_states_55_axes_0 = const()[name = string("hidden_states_55_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_9_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_9_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77450560)))];
tensor<fp16, [768]> encoder_encoder_layers_9_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77452160)))];
tensor<fp16, [1, 77, 768]> hidden_states_55_cast_fp16 = layer_norm(axes = hidden_states_55_axes_0, beta = encoder_encoder_layers_9_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_9_layer_norm1_weight_to_fp16, x = input_111_cast_fp16)[name = string("hidden_states_55_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_9_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77453760))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77896192))))[name = string("encoder_encoder_layers_9_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_9_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77902400)))];
tensor<fp16, [1, 77, 768]> linear_54_cast_fp16 = linear(bias = encoder_encoder_layers_9_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_9_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_55_cast_fp16)[name = string("linear_54_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_9_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(77904000))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78346432))))[name = string("encoder_encoder_layers_9_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_9_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78352640)))];
tensor<fp16, [1, 77, 768]> linear_55_cast_fp16 = linear(bias = encoder_encoder_layers_9_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_9_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_55_cast_fp16)[name = string("linear_55_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_9_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78354240))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78796672))))[name = string("encoder_encoder_layers_9_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_9_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78802880)))];
tensor<fp16, [1, 77, 768]> linear_56_cast_fp16 = linear(bias = encoder_encoder_layers_9_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_9_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_55_cast_fp16)[name = string("linear_56_cast_fp16")];
tensor<int32, [4]> var_613 = const()[name = string("op_613"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_614_cast_fp16 = reshape(shape = var_613, x = linear_54_cast_fp16)[name = string("op_614_cast_fp16")];
tensor<int32, [4]> var_616 = const()[name = string("op_616"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_617_cast_fp16 = reshape(shape = var_616, x = linear_55_cast_fp16)[name = string("op_617_cast_fp16")];
tensor<int32, [4]> var_619 = const()[name = string("op_619"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_620_cast_fp16 = reshape(shape = var_619, x = linear_56_cast_fp16)[name = string("op_620_cast_fp16")];
tensor<int32, [4]> value_19_perm_0 = const()[name = string("value_19_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_19_cast_fp16 = mul(x = var_614_cast_fp16, y = var_9_to_fp16)[name = string("mul_19_cast_fp16")];
bool matmul_9_transpose_y_0 = const()[name = string("matmul_9_transpose_y_0"), val = bool(true)];
bool matmul_9_transpose_x_0 = const()[name = string("matmul_9_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_66_perm_0 = const()[name = string("transpose_66_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_67_perm_0 = const()[name = string("transpose_67_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_67 = transpose(perm = transpose_67_perm_0, x = var_617_cast_fp16)[name = string("transpose_82")];
tensor<fp16, [1, 12, 77, 64]> transpose_66 = transpose(perm = transpose_66_perm_0, x = mul_19_cast_fp16)[name = string("transpose_83")];
tensor<fp16, [1, 12, 77, 77]> matmul_9_cast_fp16 = matmul(transpose_x = matmul_9_transpose_x_0, transpose_y = matmul_9_transpose_y_0, x = transpose_66, y = transpose_67)[name = string("matmul_9_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_9_cast_fp16 = add(x = matmul_9_cast_fp16, y = mul_0_to_fp16)[name = string("add_9_cast_fp16")];
int32 softmax_9_axis_0 = const()[name = string("softmax_9_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_9_cast_fp16 = softmax(axis = softmax_9_axis_0, x = add_9_cast_fp16)[name = string("softmax_9_cast_fp16")];
bool attn_output_37_transpose_x_0 = const()[name = string("attn_output_37_transpose_x_0"), val = bool(false)];
bool attn_output_37_transpose_y_0 = const()[name = string("attn_output_37_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_19_cast_fp16 = transpose(perm = value_19_perm_0, x = var_620_cast_fp16)[name = string("transpose_81")];
tensor<fp16, [1, 12, 77, 64]> attn_output_37_cast_fp16 = matmul(transpose_x = attn_output_37_transpose_x_0, transpose_y = attn_output_37_transpose_y_0, x = softmax_9_cast_fp16, y = value_19_cast_fp16)[name = string("attn_output_37_cast_fp16")];
tensor<int32, [4]> var_623_perm_0 = const()[name = string("op_623_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_625 = const()[name = string("op_625"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_623_cast_fp16 = transpose(perm = var_623_perm_0, x = attn_output_37_cast_fp16)[name = string("transpose_80")];
tensor<fp16, [1, 77, 768]> var_626_cast_fp16 = reshape(shape = var_625, x = var_623_cast_fp16)[name = string("op_626_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_9_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(78804480))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(79246912))))[name = string("encoder_encoder_layers_9_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_9_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(79253120)))];
tensor<fp16, [1, 77, 768]> linear_57_cast_fp16 = linear(bias = encoder_encoder_layers_9_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_9_self_attn_out_proj_weight_to_fp16_palettized, x = var_626_cast_fp16)[name = string("linear_57_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_115_cast_fp16 = add(x = input_111_cast_fp16, y = linear_57_cast_fp16)[name = string("input_115_cast_fp16")];
tensor<int32, [1]> input_117_axes_0 = const()[name = string("input_117_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_9_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_9_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(79254720)))];
tensor<fp16, [768]> encoder_encoder_layers_9_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(79256320)))];
tensor<fp16, [1, 77, 768]> input_117_cast_fp16 = layer_norm(axes = input_117_axes_0, beta = encoder_encoder_layers_9_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_9_layer_norm2_weight_to_fp16, x = input_115_cast_fp16)[name = string("input_117_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_9_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(79257920))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(81027456))))[name = string("encoder_encoder_layers_9_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_9_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(81052096)))];
tensor<fp16, [1, 77, 3072]> linear_58_cast_fp16 = linear(bias = encoder_encoder_layers_9_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_9_mlp_fc1_weight_to_fp16_palettized, x = input_117_cast_fp16)[name = string("linear_58_cast_fp16")];
fp16 var_641_to_fp16 = const()[name = string("op_641_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_642_cast_fp16 = mul(x = linear_58_cast_fp16, y = var_641_to_fp16)[name = string("op_642_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_643_cast_fp16 = sigmoid(x = var_642_cast_fp16)[name = string("op_643_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_121_cast_fp16 = mul(x = linear_58_cast_fp16, y = var_643_cast_fp16)[name = string("input_121_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_9_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(81058304))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(82827840))))[name = string("encoder_encoder_layers_9_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_9_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_9_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(82834048)))];
tensor<fp16, [1, 77, 768]> linear_59_cast_fp16 = linear(bias = encoder_encoder_layers_9_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_9_mlp_fc2_weight_to_fp16_palettized, x = input_121_cast_fp16)[name = string("linear_59_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_123_cast_fp16 = add(x = input_115_cast_fp16, y = linear_59_cast_fp16)[name = string("input_123_cast_fp16")];
tensor<int32, [1]> hidden_states_61_axes_0 = const()[name = string("hidden_states_61_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_10_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_10_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(82835648)))];
tensor<fp16, [768]> encoder_encoder_layers_10_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(82837248)))];
tensor<fp16, [1, 77, 768]> hidden_states_61_cast_fp16 = layer_norm(axes = hidden_states_61_axes_0, beta = encoder_encoder_layers_10_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_10_layer_norm1_weight_to_fp16, x = input_123_cast_fp16)[name = string("hidden_states_61_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_10_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(82838848))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(83281280))))[name = string("encoder_encoder_layers_10_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_10_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(83287488)))];
tensor<fp16, [1, 77, 768]> linear_60_cast_fp16 = linear(bias = encoder_encoder_layers_10_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_10_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_61_cast_fp16)[name = string("linear_60_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_10_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(83289088))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(83731520))))[name = string("encoder_encoder_layers_10_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_10_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(83737728)))];
tensor<fp16, [1, 77, 768]> linear_61_cast_fp16 = linear(bias = encoder_encoder_layers_10_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_10_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_61_cast_fp16)[name = string("linear_61_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_10_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(83739328))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84181760))))[name = string("encoder_encoder_layers_10_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_10_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84187968)))];
tensor<fp16, [1, 77, 768]> linear_62_cast_fp16 = linear(bias = encoder_encoder_layers_10_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_10_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_61_cast_fp16)[name = string("linear_62_cast_fp16")];
tensor<int32, [4]> var_672 = const()[name = string("op_672"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_673_cast_fp16 = reshape(shape = var_672, x = linear_60_cast_fp16)[name = string("op_673_cast_fp16")];
tensor<int32, [4]> var_675 = const()[name = string("op_675"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_676_cast_fp16 = reshape(shape = var_675, x = linear_61_cast_fp16)[name = string("op_676_cast_fp16")];
tensor<int32, [4]> var_678 = const()[name = string("op_678"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_679_cast_fp16 = reshape(shape = var_678, x = linear_62_cast_fp16)[name = string("op_679_cast_fp16")];
tensor<int32, [4]> value_21_perm_0 = const()[name = string("value_21_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_21_cast_fp16 = mul(x = var_673_cast_fp16, y = var_9_to_fp16)[name = string("mul_21_cast_fp16")];
bool matmul_10_transpose_y_0 = const()[name = string("matmul_10_transpose_y_0"), val = bool(true)];
bool matmul_10_transpose_x_0 = const()[name = string("matmul_10_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_68_perm_0 = const()[name = string("transpose_68_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_69_perm_0 = const()[name = string("transpose_69_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_69 = transpose(perm = transpose_69_perm_0, x = var_676_cast_fp16)[name = string("transpose_78")];
tensor<fp16, [1, 12, 77, 64]> transpose_68 = transpose(perm = transpose_68_perm_0, x = mul_21_cast_fp16)[name = string("transpose_79")];
tensor<fp16, [1, 12, 77, 77]> matmul_10_cast_fp16 = matmul(transpose_x = matmul_10_transpose_x_0, transpose_y = matmul_10_transpose_y_0, x = transpose_68, y = transpose_69)[name = string("matmul_10_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_10_cast_fp16 = add(x = matmul_10_cast_fp16, y = mul_0_to_fp16)[name = string("add_10_cast_fp16")];
int32 softmax_10_axis_0 = const()[name = string("softmax_10_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_10_cast_fp16 = softmax(axis = softmax_10_axis_0, x = add_10_cast_fp16)[name = string("softmax_10_cast_fp16")];
bool attn_output_41_transpose_x_0 = const()[name = string("attn_output_41_transpose_x_0"), val = bool(false)];
bool attn_output_41_transpose_y_0 = const()[name = string("attn_output_41_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_21_cast_fp16 = transpose(perm = value_21_perm_0, x = var_679_cast_fp16)[name = string("transpose_77")];
tensor<fp16, [1, 12, 77, 64]> attn_output_41_cast_fp16 = matmul(transpose_x = attn_output_41_transpose_x_0, transpose_y = attn_output_41_transpose_y_0, x = softmax_10_cast_fp16, y = value_21_cast_fp16)[name = string("attn_output_41_cast_fp16")];
tensor<int32, [4]> var_682_perm_0 = const()[name = string("op_682_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_684 = const()[name = string("op_684"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_682_cast_fp16 = transpose(perm = var_682_perm_0, x = attn_output_41_cast_fp16)[name = string("transpose_76")];
tensor<fp16, [1, 77, 768]> var_685_cast_fp16 = reshape(shape = var_684, x = var_682_cast_fp16)[name = string("op_685_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_10_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84189568))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84632000))))[name = string("encoder_encoder_layers_10_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_10_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84638208)))];
tensor<fp16, [1, 77, 768]> linear_63_cast_fp16 = linear(bias = encoder_encoder_layers_10_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_10_self_attn_out_proj_weight_to_fp16_palettized, x = var_685_cast_fp16)[name = string("linear_63_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_127_cast_fp16 = add(x = input_123_cast_fp16, y = linear_63_cast_fp16)[name = string("input_127_cast_fp16")];
tensor<int32, [1]> input_129_axes_0 = const()[name = string("input_129_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_10_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_10_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84639808)))];
tensor<fp16, [768]> encoder_encoder_layers_10_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84641408)))];
tensor<fp16, [1, 77, 768]> input_129_cast_fp16 = layer_norm(axes = input_129_axes_0, beta = encoder_encoder_layers_10_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_10_layer_norm2_weight_to_fp16, x = input_127_cast_fp16)[name = string("input_129_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_10_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(84643008))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(86412544))))[name = string("encoder_encoder_layers_10_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_10_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(86437184)))];
tensor<fp16, [1, 77, 3072]> linear_64_cast_fp16 = linear(bias = encoder_encoder_layers_10_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_10_mlp_fc1_weight_to_fp16_palettized, x = input_129_cast_fp16)[name = string("linear_64_cast_fp16")];
fp16 var_700_to_fp16 = const()[name = string("op_700_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_701_cast_fp16 = mul(x = linear_64_cast_fp16, y = var_700_to_fp16)[name = string("op_701_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_702_cast_fp16 = sigmoid(x = var_701_cast_fp16)[name = string("op_702_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_133_cast_fp16 = mul(x = linear_64_cast_fp16, y = var_702_cast_fp16)[name = string("input_133_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_10_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(86443392))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88212928))))[name = string("encoder_encoder_layers_10_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_10_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_10_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88219136)))];
tensor<fp16, [1, 77, 768]> linear_65_cast_fp16 = linear(bias = encoder_encoder_layers_10_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_10_mlp_fc2_weight_to_fp16_palettized, x = input_133_cast_fp16)[name = string("linear_65_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_135_cast_fp16 = add(x = input_127_cast_fp16, y = linear_65_cast_fp16)[name = string("input_135_cast_fp16")];
tensor<int32, [1]> hidden_states_67_axes_0 = const()[name = string("hidden_states_67_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_11_layer_norm1_weight_to_fp16 = const()[name = string("encoder_encoder_layers_11_layer_norm1_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88220736)))];
tensor<fp16, [768]> encoder_encoder_layers_11_layer_norm1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_layer_norm1_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88222336)))];
tensor<fp16, [1, 77, 768]> hidden_states_67_cast_fp16 = layer_norm(axes = hidden_states_67_axes_0, beta = encoder_encoder_layers_11_layer_norm1_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_11_layer_norm1_weight_to_fp16, x = input_135_cast_fp16)[name = string("hidden_states_67_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_11_self_attn_q_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88223936))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88666368))))[name = string("encoder_encoder_layers_11_self_attn_q_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_11_self_attn_q_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_self_attn_q_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88672576)))];
tensor<fp16, [1, 77, 768]> linear_66_cast_fp16 = linear(bias = encoder_encoder_layers_11_self_attn_q_proj_bias_to_fp16, weight = encoder_encoder_layers_11_self_attn_q_proj_weight_to_fp16_palettized, x = hidden_states_67_cast_fp16)[name = string("linear_66_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_11_self_attn_k_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(88674176))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89116608))))[name = string("encoder_encoder_layers_11_self_attn_k_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_11_self_attn_k_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_self_attn_k_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89122816)))];
tensor<fp16, [1, 77, 768]> linear_67_cast_fp16 = linear(bias = encoder_encoder_layers_11_self_attn_k_proj_bias_to_fp16, weight = encoder_encoder_layers_11_self_attn_k_proj_weight_to_fp16_palettized, x = hidden_states_67_cast_fp16)[name = string("linear_67_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_11_self_attn_v_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89124416))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89566848))))[name = string("encoder_encoder_layers_11_self_attn_v_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_11_self_attn_v_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_self_attn_v_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89573056)))];
tensor<fp16, [1, 77, 768]> linear_68_cast_fp16 = linear(bias = encoder_encoder_layers_11_self_attn_v_proj_bias_to_fp16, weight = encoder_encoder_layers_11_self_attn_v_proj_weight_to_fp16_palettized, x = hidden_states_67_cast_fp16)[name = string("linear_68_cast_fp16")];
tensor<int32, [4]> var_731 = const()[name = string("op_731"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_732_cast_fp16 = reshape(shape = var_731, x = linear_66_cast_fp16)[name = string("op_732_cast_fp16")];
tensor<int32, [4]> var_734 = const()[name = string("op_734"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_735_cast_fp16 = reshape(shape = var_734, x = linear_67_cast_fp16)[name = string("op_735_cast_fp16")];
tensor<int32, [4]> var_737 = const()[name = string("op_737"), val = tensor<int32, [4]>([1, 77, -1, 64])];
tensor<fp16, [1, 77, 12, 64]> var_738_cast_fp16 = reshape(shape = var_737, x = linear_68_cast_fp16)[name = string("op_738_cast_fp16")];
tensor<int32, [4]> value_perm_0 = const()[name = string("value_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 77, 12, 64]> mul_23_cast_fp16 = mul(x = var_732_cast_fp16, y = var_9_to_fp16)[name = string("mul_23_cast_fp16")];
bool matmul_11_transpose_y_0 = const()[name = string("matmul_11_transpose_y_0"), val = bool(true)];
bool matmul_11_transpose_x_0 = const()[name = string("matmul_11_transpose_x_0"), val = bool(false)];
tensor<int32, [4]> transpose_70_perm_0 = const()[name = string("transpose_70_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<int32, [4]> transpose_71_perm_0 = const()[name = string("transpose_71_perm_0"), val = tensor<int32, [4]>([0, 2, -3, -1])];
tensor<fp16, [1, 12, 77, 64]> transpose_71 = transpose(perm = transpose_71_perm_0, x = var_735_cast_fp16)[name = string("transpose_74")];
tensor<fp16, [1, 12, 77, 64]> transpose_70 = transpose(perm = transpose_70_perm_0, x = mul_23_cast_fp16)[name = string("transpose_75")];
tensor<fp16, [1, 12, 77, 77]> matmul_11_cast_fp16 = matmul(transpose_x = matmul_11_transpose_x_0, transpose_y = matmul_11_transpose_y_0, x = transpose_70, y = transpose_71)[name = string("matmul_11_cast_fp16")];
tensor<fp16, [1, 12, 77, 77]> add_11_cast_fp16 = add(x = matmul_11_cast_fp16, y = mul_0_to_fp16)[name = string("add_11_cast_fp16")];
int32 softmax_11_axis_0 = const()[name = string("softmax_11_axis_0"), val = int32(-1)];
tensor<fp16, [1, 12, 77, 77]> softmax_11_cast_fp16 = softmax(axis = softmax_11_axis_0, x = add_11_cast_fp16)[name = string("softmax_11_cast_fp16")];
bool attn_output_45_transpose_x_0 = const()[name = string("attn_output_45_transpose_x_0"), val = bool(false)];
bool attn_output_45_transpose_y_0 = const()[name = string("attn_output_45_transpose_y_0"), val = bool(false)];
tensor<fp16, [1, 12, 77, 64]> value_cast_fp16 = transpose(perm = value_perm_0, x = var_738_cast_fp16)[name = string("transpose_73")];
tensor<fp16, [1, 12, 77, 64]> attn_output_45_cast_fp16 = matmul(transpose_x = attn_output_45_transpose_x_0, transpose_y = attn_output_45_transpose_y_0, x = softmax_11_cast_fp16, y = value_cast_fp16)[name = string("attn_output_45_cast_fp16")];
tensor<int32, [4]> var_741_perm_0 = const()[name = string("op_741_perm_0"), val = tensor<int32, [4]>([0, 2, 1, 3])];
tensor<int32, [3]> var_743 = const()[name = string("op_743"), val = tensor<int32, [3]>([1, 77, -1])];
tensor<fp16, [1, 77, 12, 64]> var_741_cast_fp16 = transpose(perm = var_741_perm_0, x = attn_output_45_cast_fp16)[name = string("transpose_72")];
tensor<fp16, [1, 77, 768]> var_744_cast_fp16 = reshape(shape = var_743, x = var_741_cast_fp16)[name = string("op_744_cast_fp16")];
tensor<fp16, [768, 768]> encoder_encoder_layers_11_self_attn_out_proj_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(89574656))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(90017088))))[name = string("encoder_encoder_layers_11_self_attn_out_proj_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_11_self_attn_out_proj_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_self_attn_out_proj_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(90023296)))];
tensor<fp16, [1, 77, 768]> linear_69_cast_fp16 = linear(bias = encoder_encoder_layers_11_self_attn_out_proj_bias_to_fp16, weight = encoder_encoder_layers_11_self_attn_out_proj_weight_to_fp16_palettized, x = var_744_cast_fp16)[name = string("linear_69_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_139_cast_fp16 = add(x = input_135_cast_fp16, y = linear_69_cast_fp16)[name = string("input_139_cast_fp16")];
tensor<int32, [1]> input_141_axes_0 = const()[name = string("input_141_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_encoder_layers_11_layer_norm2_weight_to_fp16 = const()[name = string("encoder_encoder_layers_11_layer_norm2_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(90024896)))];
tensor<fp16, [768]> encoder_encoder_layers_11_layer_norm2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_layer_norm2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(90026496)))];
tensor<fp16, [1, 77, 768]> input_141_cast_fp16 = layer_norm(axes = input_141_axes_0, beta = encoder_encoder_layers_11_layer_norm2_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_encoder_layers_11_layer_norm2_weight_to_fp16, x = input_139_cast_fp16)[name = string("input_141_cast_fp16")];
tensor<fp16, [3072, 768]> encoder_encoder_layers_11_mlp_fc1_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [3072, 768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(90028096))), lut = tensor<fp16, [192, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(91797632))))[name = string("encoder_encoder_layers_11_mlp_fc1_weight_to_fp16_palettized")];
tensor<fp16, [3072]> encoder_encoder_layers_11_mlp_fc1_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_mlp_fc1_bias_to_fp16"), val = tensor<fp16, [3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(91822272)))];
tensor<fp16, [1, 77, 3072]> linear_70_cast_fp16 = linear(bias = encoder_encoder_layers_11_mlp_fc1_bias_to_fp16, weight = encoder_encoder_layers_11_mlp_fc1_weight_to_fp16_palettized, x = input_141_cast_fp16)[name = string("linear_70_cast_fp16")];
fp16 var_759_to_fp16 = const()[name = string("op_759_to_fp16"), val = fp16(0x1.b3cp+0)];
tensor<fp16, [1, 77, 3072]> var_760_cast_fp16 = mul(x = linear_70_cast_fp16, y = var_759_to_fp16)[name = string("op_760_cast_fp16")];
tensor<fp16, [1, 77, 3072]> var_761_cast_fp16 = sigmoid(x = var_760_cast_fp16)[name = string("op_761_cast_fp16")];
tensor<fp16, [1, 77, 3072]> input_145_cast_fp16 = mul(x = linear_70_cast_fp16, y = var_761_cast_fp16)[name = string("input_145_cast_fp16")];
tensor<fp16, [768, 3072]> encoder_encoder_layers_11_mlp_fc2_weight_to_fp16_palettized = constexpr_lut_to_dense(indices = tensor<uint6, [768, 3072]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(91828480))), lut = tensor<fp16, [48, 1, 64, 1]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(93598016))))[name = string("encoder_encoder_layers_11_mlp_fc2_weight_to_fp16_palettized")];
tensor<fp16, [768]> encoder_encoder_layers_11_mlp_fc2_bias_to_fp16 = const()[name = string("encoder_encoder_layers_11_mlp_fc2_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(93604224)))];
tensor<fp16, [1, 77, 768]> linear_71_cast_fp16 = linear(bias = encoder_encoder_layers_11_mlp_fc2_bias_to_fp16, weight = encoder_encoder_layers_11_mlp_fc2_weight_to_fp16_palettized, x = input_145_cast_fp16)[name = string("linear_71_cast_fp16")];
tensor<fp16, [1, 77, 768]> input_cast_fp16 = add(x = input_139_cast_fp16, y = linear_71_cast_fp16)[name = string("input_cast_fp16")];
tensor<int32, [1]> last_hidden_state_axes_0 = const()[name = string("last_hidden_state_axes_0"), val = tensor<int32, [1]>([-1])];
tensor<fp16, [768]> encoder_final_layer_norm_weight_to_fp16 = const()[name = string("encoder_final_layer_norm_weight_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(93605824)))];
tensor<fp16, [768]> encoder_final_layer_norm_bias_to_fp16 = const()[name = string("encoder_final_layer_norm_bias_to_fp16"), val = tensor<fp16, [768]>(BLOBFILE(path = string("@model_path/weights/weight.bin"), offset = uint64(93607424)))];
tensor<fp16, [1, 77, 768]> hidden_embeds = layer_norm(axes = last_hidden_state_axes_0, beta = encoder_final_layer_norm_bias_to_fp16, epsilon = var_11_to_fp16, gamma = encoder_final_layer_norm_weight_to_fp16, x = input_cast_fp16)[name = string("last_hidden_state_cast_fp16")];
tensor<int32, [1]> var_772 = const()[name = string("op_772"), val = tensor<int32, [1]>([0])];
int32 var_774_axis_0 = const()[name = string("op_774_axis_0"), val = int32(-1)];
bool var_774_keep_dims_0 = const()[name = string("op_774_keep_dims_0"), val = bool(false)];
string var_774_output_dtype_0 = const()[name = string("op_774_output_dtype_0"), val = string("int32")];
tensor<int32, [1]> var_774 = reduce_argmax(axis = var_774_axis_0, keep_dims = var_774_keep_dims_0, output_dtype = var_774_output_dtype_0, x = input_ids)[name = string("op_774")];
int32 stack_0_axis_0 = const()[name = string("stack_0_axis_0"), val = int32(1)];
tensor<int32, [1, 2]> stack_0 = stack(axis = stack_0_axis_0, values = (var_772, var_774))[name = string("stack_0")];
int32 var_776_transpose_batch_dims_0 = const()[name = string("op_776_transpose_batch_dims_0"), val = int32(0)];
bool var_776_transpose_validate_indices_0 = const()[name = string("op_776_transpose_validate_indices_0"), val = bool(false)];
string stack_0_to_uint16_dtype_0 = const()[name = string("stack_0_to_uint16_dtype_0"), val = string("uint16")];
tensor<uint16, [1, 2]> stack_0_to_uint16 = cast(dtype = stack_0_to_uint16_dtype_0, x = stack_0)[name = string("cast_0")];
tensor<fp16, [1, 768]> pooled_outputs = gather_nd(batch_dims = var_776_transpose_batch_dims_0, indices = stack_0_to_uint16, validate_indices = var_776_transpose_validate_indices_0, x = hidden_embeds)[name = string("op_776_transpose_cast_fp16_cast_uint16")];
} -> (hidden_embeds, pooled_outputs);
} |