cross-encoder/ms-marco-MiniLM-L6-v2
34788.6m
1<?xml version="1.0"?>
2<net name="Model110" version="11">
3 <layers>
4 <layer id="2" name="input_ids" type="Parameter" version="opset1">
5 <data shape="?,?" element_type="i64" />
6 <output>
7 <port id="0" precision="I64" names="input_ids">
8 <dim>-1</dim>
9 <dim>-1</dim>
10 </port>
11 </output>
12 </layer>
13 <layer id="1" name="attention_mask" type="Parameter" version="opset1">
14 <data shape="?,?" element_type="i64" />
15 <output>
16 <port id="0" precision="I64" names="attention_mask">
17 <dim>-1</dim>
18 <dim>-1</dim>
19 </port>
20 </output>
21 </layer>
22 <layer id="0" name="token_type_ids" type="Parameter" version="opset1">
23 <data shape="?,?" element_type="i64" />
24 <output>
25 <port id="0" precision="I64" names="token_type_ids">
26 <dim>-1</dim>
27 <dim>-1</dim>
28 </port>
29 </output>
30 </layer>
31 <layer id="3" name="Constant_299926" type="Const" version="opset1">
32 <data element_type="i8" shape="30522, 384" offset="0" size="11720448" />
33 <output>
34 <port id="0" precision="I8">
35 <dim>30522</dim>
36 <dim>384</dim>
37 </port>
38 </output>
39 </layer>
40 <layer id="4" name="Convert_299927" type="Convert" version="opset1">
41 <data destination_type="f32" />
42 <input>
43 <port id="0" precision="I8">
44 <dim>30522</dim>
45 <dim>384</dim>
46 </port>
47 </input>
48 <output>
49 <port id="1" precision="FP32">
50 <dim>30522</dim>
51 <dim>384</dim>
52 </port>
53 </output>
54 </layer>
55 <layer id="5" name="Constant_299928" type="Const" version="opset1">
56 <data element_type="f32" shape="30522, 1" offset="11720448" size="122088" />
57 <output>
58 <port id="0" precision="FP32">
59 <dim>30522</dim>
60 <dim>1</dim>
61 </port>
62 </output>
63 </layer>
64 <layer id="6" name="__module.bert.embeddings.word_embeddings/aten::embedding/Gather/fq_weights_0" type="Multiply" version="opset1">
65 <data auto_broadcast="numpy" />
66 <input>
67 <port id="0" precision="FP32">
68 <dim>30522</dim>
69 <dim>384</dim>
70 </port>
71 <port id="1" precision="FP32">
72 <dim>30522</dim>
73 <dim>1</dim>
74 </port>
75 </input>
76 <output>
77 <port id="2" precision="FP32">
78 <dim>30522</dim>
79 <dim>384</dim>
80 </port>
81 </output>
82 </layer>
83 <layer id="7" name="__module.bert.embeddings.word_embeddings/aten::embedding/Convert" type="Convert" version="opset1">
84 <data destination_type="i32" />
85 <input>
86 <port id="0" precision="I64">
87 <dim>-1</dim>
88 <dim>-1</dim>
89 </port>
90 </input>
91 <output>
92 <port id="1" precision="I32">
93 <dim>-1</dim>
94 <dim>-1</dim>
95 </port>
96 </output>
97 </layer>
98 <layer id="8" name="__module.bert.embeddings.word_embeddings/aten::embedding/Constant" type="Const" version="opset1">
99 <data element_type="i32" shape="" offset="11842536" size="4" />
100 <output>
101 <port id="0" precision="I32" />
102 </output>
103 </layer>
104 <layer id="9" name="__module.bert.embeddings.word_embeddings/aten::embedding/Gather" type="Gather" version="opset8">
105 <data batch_dims="0" />
106 <input>
107 <port id="0" precision="FP32">
108 <dim>30522</dim>
109 <dim>384</dim>
110 </port>
111 <port id="1" precision="I32">
112 <dim>-1</dim>
113 <dim>-1</dim>
114 </port>
115 <port id="2" precision="I32" />
116 </input>
117 <output>
118 <port id="3" precision="FP32" names="48,inputs_embeds">
119 <dim>-1</dim>
120 <dim>-1</dim>
121 <dim>384</dim>
122 </port>
123 </output>
124 </layer>
125 <layer id="10" name="Constant_299930" type="Const" version="opset1">
126 <data element_type="i8" shape="2, 384" offset="11842540" size="768" />
127 <output>
128 <port id="0" precision="I8">
129 <dim>2</dim>
130 <dim>384</dim>
131 </port>
132 </output>
133 </layer>
134 <layer id="11" name="Convert_299931" type="Convert" version="opset1">
135 <data destination_type="f32" />
136 <input>
137 <port id="0" precision="I8">
138 <dim>2</dim>
139 <dim>384</dim>
140 </port>
141 </input>
142 <output>
143 <port id="1" precision="FP32">
144 <dim>2</dim>
145 <dim>384</dim>
146 </port>
147 </output>
148 </layer>
149 <layer id="12" name="Constant_299932" type="Const" version="opset1">
150 <data element_type="f32" shape="2, 1" offset="11843308" size="8" />
151 <output>
152 <port id="0" precision="FP32">
153 <dim>2</dim>
154 <dim>1</dim>
155 </port>
156 </output>
157 </layer>
158 <layer id="13" name="__module.bert.embeddings.token_type_embeddings/aten::embedding/Gather/fq_weights_0" type="Multiply" version="opset1">
159 <data auto_broadcast="numpy" />
160 <input>
161 <port id="0" precision="FP32">
162 <dim>2</dim>
163 <dim>384</dim>
164 </port>
165 <port id="1" precision="FP32">
166 <dim>2</dim>
167 <dim>1</dim>
168 </port>
169 </input>
170 <output>
171 <port id="2" precision="FP32">
172 <dim>2</dim>
173 <dim>384</dim>
174 </port>
175 </output>
176 </layer>
177 <layer id="14" name="__module.bert.embeddings.token_type_embeddings/aten::embedding/Convert" type="Convert" version="opset1">
178 <data destination_type="i32" />
179 <input>
180 <port id="0" precision="I64">
181 <dim>-1</dim>
182 <dim>-1</dim>
183 </port>
184 </input>
185 <output>
186 <port id="1" precision="I32">
187 <dim>-1</dim>
188 <dim>-1</dim>
189 </port>
190 </output>
191 </layer>
192 <layer id="15" name="__module.bert.embeddings.token_type_embeddings/aten::embedding/Constant" type="Const" version="opset1">
193 <data element_type="i32" shape="" offset="11842536" size="4" />
194 <output>
195 <port id="0" precision="I32" />
196 </output>
197 </layer>
198 <layer id="16" name="__module.bert.embeddings.token_type_embeddings/aten::embedding/Gather" type="Gather" version="opset8">
199 <data batch_dims="0" />
200 <input>
201 <port id="0" precision="FP32">
202 <dim>2</dim>
203 <dim>384</dim>
204 </port>
205 <port id="1" precision="I32">
206 <dim>-1</dim>
207 <dim>-1</dim>
208 </port>
209 <port id="2" precision="I32" />
210 </input>
211 <output>
212 <port id="3" precision="FP32" names="50,token_type_embeddings.1">
213 <dim>-1</dim>
214 <dim>-1</dim>
215 <dim>384</dim>
216 </port>
217 </output>
218 </layer>
219 <layer id="17" name="__module.bert.embeddings/aten::add/Add" type="Add" version="opset1">
220 <data auto_broadcast="numpy" />
221 <input>
222 <port id="0" precision="FP32">
223 <dim>-1</dim>
224 <dim>-1</dim>
225 <dim>384</dim>
226 </port>
227 <port id="1" precision="FP32">
228 <dim>-1</dim>
229 <dim>-1</dim>
230 <dim>384</dim>
231 </port>
232 </input>
233 <output>
234 <port id="2" precision="FP32" names="51_1">
235 <dim>-1</dim>
236 <dim>-1</dim>
237 <dim>384</dim>
238 </port>
239 </output>
240 </layer>
241 <layer id="18" name="Constant_299934" type="Const" version="opset1">
242 <data element_type="i8" shape="512, 384" offset="11843316" size="196608" />
243 <output>
244 <port id="0" precision="I8">
245 <dim>512</dim>
246 <dim>384</dim>
247 </port>
248 </output>
249 </layer>
250 <layer id="19" name="Convert_299935" type="Convert" version="opset1">
251 <data destination_type="f32" />
252 <input>
253 <port id="0" precision="I8">
254 <dim>512</dim>
255 <dim>384</dim>
256 </port>
257 </input>
258 <output>
259 <port id="1" precision="FP32">
260 <dim>512</dim>
261 <dim>384</dim>
262 </port>
263 </output>
264 </layer>
265 <layer id="20" name="Constant_299936" type="Const" version="opset1">
266 <data element_type="f32" shape="512, 1" offset="12039924" size="2048" />
267 <output>
268 <port id="0" precision="FP32">
269 <dim>512</dim>
270 <dim>1</dim>
271 </port>
272 </output>
273 </layer>
274 <layer id="21" name="__module.bert.embeddings.position_embeddings/aten::embedding/Gather/fq_weights_0" type="Multiply" version="opset1">
275 <data auto_broadcast="numpy" />
276 <input>
277 <port id="0" precision="FP32">
278 <dim>512</dim>
279 <dim>384</dim>
280 </port>
281 <port id="1" precision="FP32">
282 <dim>512</dim>
283 <dim>1</dim>
284 </port>
285 </input>
286 <output>
287 <port id="2" precision="FP32">
288 <dim>512</dim>
289 <dim>384</dim>
290 </port>
291 </output>
292 </layer>
293 <layer id="22" name="__module.bert.embeddings/aten::slice/Slice" type="Const" version="opset1">
294 <data element_type="i64" shape="1, 512" offset="12041972" size="4096" />
295 <output>
296 <port id="0" precision="I64" names="45">
297 <dim>1</dim>
298 <dim>512</dim>
299 </port>
300 </output>
301 </layer>
302 <layer id="23" name="__module.bert.embeddings/aten::slice/Reshape" type="Const" version="opset1">
303 <data element_type="i64" shape="1" offset="12046068" size="8" />
304 <output>
305 <port id="0" precision="I64">
306 <dim>1</dim>
307 </port>
308 </output>
309 </layer>
310 <layer id="24" name="ShapeOf_90427" type="ShapeOf" version="opset3">
311 <data output_type="i64" />
312 <input>
313 <port id="0" precision="I64">
314 <dim>-1</dim>
315 <dim>-1</dim>
316 </port>
317 </input>
318 <output>
319 <port id="1" precision="I64">
320 <dim>2</dim>
321 </port>
322 </output>
323 </layer>
324 <layer id="25" name="Constant_90546" type="Const" version="opset1">
325 <data element_type="i64" shape="1" offset="12046076" size="8" />
326 <output>
327 <port id="0" precision="I64">
328 <dim>1</dim>
329 </port>
330 </output>
331 </layer>
332 <layer id="26" name="Constant_90429" type="Const" version="opset1">
333 <data element_type="i64" shape="" offset="12046068" size="8" />
334 <output>
335 <port id="0" precision="I64" />
336 </output>
337 </layer>
338 <layer id="27" name="Gather_90430" type="Gather" version="opset8">
339 <data batch_dims="0" />
340 <input>
341 <port id="0" precision="I64">
342 <dim>2</dim>
343 </port>
344 <port id="1" precision="I64">
345 <dim>1</dim>
346 </port>
347 <port id="2" precision="I64" />
348 </input>
349 <output>
350 <port id="3" precision="I64" names="35,41,43,44,61">
351 <dim>1</dim>
352 </port>
353 </output>
354 </layer>
355 <layer id="28" name="__module.bert.embeddings/aten::slice/Reshape_2" type="Const" version="opset1">
356 <data element_type="i64" shape="1" offset="12046076" size="8" />
357 <output>
358 <port id="0" precision="I64">
359 <dim>1</dim>
360 </port>
361 </output>
362 </layer>
363 <layer id="29" name="__module.bert.embeddings/aten::slice/Reshape_3" type="Const" version="opset1">
364 <data element_type="i64" shape="1" offset="12046076" size="8" />
365 <output>
366 <port id="0" precision="I64">
367 <dim>1</dim>
368 </port>
369 </output>
370 </layer>
371 <layer id="30" name="__module.bert.embeddings/aten::slice/Slice_1" type="Slice" version="opset8">
372 <input>
373 <port id="0" precision="I64">
374 <dim>1</dim>
375 <dim>512</dim>
376 </port>
377 <port id="1" precision="I64">
378 <dim>1</dim>
379 </port>
380 <port id="2" precision="I64">
381 <dim>1</dim>
382 </port>
383 <port id="3" precision="I64">
384 <dim>1</dim>
385 </port>
386 <port id="4" precision="I64">
387 <dim>1</dim>
388 </port>
389 </input>
390 <output>
391 <port id="5" precision="I64" names="46">
392 <dim>1</dim>
393 <dim>-1</dim>
394 </port>
395 </output>
396 </layer>
397 <layer id="31" name="__module.bert.embeddings.position_embeddings/aten::embedding/Convert" type="Convert" version="opset1">
398 <data destination_type="i32" />
399 <input>
400 <port id="0" precision="I64">
401 <dim>1</dim>
402 <dim>-1</dim>
403 </port>
404 </input>
405 <output>
406 <port id="1" precision="I32">
407 <dim>1</dim>
408 <dim>-1</dim>
409 </port>
410 </output>
411 </layer>
412 <layer id="32" name="__module.bert.embeddings.position_embeddings/aten::embedding/Constant" type="Const" version="opset1">
413 <data element_type="i32" shape="" offset="11842536" size="4" />
414 <output>
415 <port id="0" precision="I32" />
416 </output>
417 </layer>
418 <layer id="33" name="__module.bert.embeddings.position_embeddings/aten::embedding/Gather" type="Gather" version="opset8">
419 <data batch_dims="0" />
420 <input>
421 <port id="0" precision="FP32">
422 <dim>512</dim>
423 <dim>384</dim>
424 </port>
425 <port id="1" precision="I32">
426 <dim>1</dim>
427 <dim>-1</dim>
428 </port>
429 <port id="2" precision="I32" />
430 </input>
431 <output>
432 <port id="3" precision="FP32" names="53,position_embeddings.1">
433 <dim>1</dim>
434 <dim>-1</dim>
435 <dim>384</dim>
436 </port>
437 </output>
438 </layer>
439 <layer id="34" name="__module.bert.embeddings/aten::add_/Add" type="Add" version="opset1">
440 <data auto_broadcast="numpy" />
441 <input>
442 <port id="0" precision="FP32">
443 <dim>-1</dim>
444 <dim>-1</dim>
445 <dim>384</dim>
446 </port>
447 <port id="1" precision="FP32">
448 <dim>1</dim>
449 <dim>-1</dim>
450 <dim>384</dim>
451 </port>
452 </input>
453 <output>
454 <port id="2" precision="FP32" names="51,embeddings.1">
455 <dim>-1</dim>
456 <dim>-1</dim>
457 <dim>384</dim>
458 </port>
459 </output>
460 </layer>
461 <layer id="35" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1">
462 <data element_type="i32" shape="1" offset="12046084" size="4" />
463 <output>
464 <port id="0" precision="I32">
465 <dim>1</dim>
466 </port>
467 </output>
468 </layer>
469 <layer id="36" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6">
470 <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" />
471 <input>
472 <port id="0" precision="FP32">
473 <dim>-1</dim>
474 <dim>-1</dim>
475 <dim>384</dim>
476 </port>
477 <port id="1" precision="I32">
478 <dim>1</dim>
479 </port>
480 </input>
481 <output>
482 <port id="2" precision="FP32">
483 <dim>-1</dim>
484 <dim>-1</dim>
485 <dim>384</dim>
486 </port>
487 </output>
488 </layer>
489 <layer id="37" name="Constant_90301" type="Const" version="opset1">
490 <data element_type="f32" shape="1, 1, 384" offset="12046088" size="1536" />
491 <output>
492 <port id="0" precision="FP32">
493 <dim>1</dim>
494 <dim>1</dim>
495 <dim>384</dim>
496 </port>
497 </output>
498 </layer>
499 <layer id="38" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
500 <data auto_broadcast="numpy" />
501 <input>
502 <port id="0" precision="FP32">
503 <dim>-1</dim>
504 <dim>-1</dim>
505 <dim>384</dim>
506 </port>
507 <port id="1" precision="FP32">
508 <dim>1</dim>
509 <dim>1</dim>
510 <dim>384</dim>
511 </port>
512 </input>
513 <output>
514 <port id="2" precision="FP32">
515 <dim>-1</dim>
516 <dim>-1</dim>
517 <dim>384</dim>
518 </port>
519 </output>
520 </layer>
521 <layer id="39" name="Constant_90302" type="Const" version="opset1">
522 <data element_type="f32" shape="1, 1, 384" offset="12047624" size="1536" />
523 <output>
524 <port id="0" precision="FP32">
525 <dim>1</dim>
526 <dim>1</dim>
527 <dim>384</dim>
528 </port>
529 </output>
530 </layer>
531 <layer id="40" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1">
532 <data auto_broadcast="numpy" />
533 <input>
534 <port id="0" precision="FP32">
535 <dim>-1</dim>
536 <dim>-1</dim>
537 <dim>384</dim>
538 </port>
539 <port id="1" precision="FP32">
540 <dim>1</dim>
541 <dim>1</dim>
542 <dim>384</dim>
543 </port>
544 </input>
545 <output>
546 <port id="2" precision="FP32" names="58,input.1">
547 <dim>-1</dim>
548 <dim>-1</dim>
549 <dim>384</dim>
550 </port>
551 </output>
552 </layer>
553 <layer id="41" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1">
554 <data element_type="f32" shape="1, 1, 384" offset="12049160" size="1536" />
555 <output>
556 <port id="0" precision="FP32">
557 <dim>1</dim>
558 <dim>1</dim>
559 <dim>384</dim>
560 </port>
561 </output>
562 </layer>
563 <layer id="42" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1">
564 <data auto_broadcast="numpy" />
565 <input>
566 <port id="0" precision="FP32">
567 <dim>-1</dim>
568 <dim>-1</dim>
569 <dim>384</dim>
570 </port>
571 <port id="1" precision="FP32">
572 <dim>1</dim>
573 <dim>1</dim>
574 <dim>384</dim>
575 </port>
576 </input>
577 <output>
578 <port id="2" precision="FP32">
579 <dim>-1</dim>
580 <dim>-1</dim>
581 <dim>384</dim>
582 </port>
583 </output>
584 </layer>
585 <layer id="43" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1">
586 <data element_type="f32" shape="" offset="12050696" size="4" />
587 <output>
588 <port id="0" precision="FP32" />
589 </output>
590 </layer>
591 <layer id="44" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1">
592 <data element_type="f32" shape="" offset="12050700" size="4" />
593 <output>
594 <port id="0" precision="FP32" />
595 </output>
596 </layer>
597 <layer id="45" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1">
598 <data element_type="f32" shape="" offset="12050696" size="4" />
599 <output>
600 <port id="0" precision="FP32" />
601 </output>
602 </layer>
603 <layer id="46" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1">
604 <data element_type="f32" shape="" offset="12050700" size="4" />
605 <output>
606 <port id="0" precision="FP32" />
607 </output>
608 </layer>
609 <layer id="47" name="__module.bert.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1">
610 <data levels="256" auto_broadcast="numpy" />
611 <input>
612 <port id="0" precision="FP32">
613 <dim>-1</dim>
614 <dim>-1</dim>
615 <dim>384</dim>
616 </port>
617 <port id="1" precision="FP32" />
618 <port id="2" precision="FP32" />
619 <port id="3" precision="FP32" />
620 <port id="4" precision="FP32" />
621 </input>
622 <output>
623 <port id="5" precision="FP32">
624 <dim>-1</dim>
625 <dim>-1</dim>
626 <dim>384</dim>
627 </port>
628 </output>
629 </layer>
630 <layer id="48" name="Constant_299938" type="Const" version="opset1">
631 <data element_type="i8" shape="384, 384" offset="12050704" size="147456" />
632 <output>
633 <port id="0" precision="I8">
634 <dim>384</dim>
635 <dim>384</dim>
636 </port>
637 </output>
638 </layer>
639 <layer id="49" name="Convert_299939" type="Convert" version="opset1">
640 <data destination_type="f32" />
641 <input>
642 <port id="0" precision="I8">
643 <dim>384</dim>
644 <dim>384</dim>
645 </port>
646 </input>
647 <output>
648 <port id="1" precision="FP32">
649 <dim>384</dim>
650 <dim>384</dim>
651 </port>
652 </output>
653 </layer>
654 <layer id="50" name="Constant_299940" type="Const" version="opset1">
655 <data element_type="f32" shape="384, 1" offset="12198160" size="1536" />
656 <output>
657 <port id="0" precision="FP32">
658 <dim>384</dim>
659 <dim>1</dim>
660 </port>
661 </output>
662 </layer>
663 <layer id="51" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1">
664 <data auto_broadcast="numpy" />
665 <input>
666 <port id="0" precision="FP32">
667 <dim>384</dim>
668 <dim>384</dim>
669 </port>
670 <port id="1" precision="FP32">
671 <dim>384</dim>
672 <dim>1</dim>
673 </port>
674 </input>
675 <output>
676 <port id="2" precision="FP32">
677 <dim>384</dim>
678 <dim>384</dim>
679 </port>
680 </output>
681 </layer>
682 <layer id="52" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1">
683 <data transpose_a="false" transpose_b="true" />
684 <input>
685 <port id="0" precision="FP32">
686 <dim>-1</dim>
687 <dim>-1</dim>
688 <dim>384</dim>
689 </port>
690 <port id="1" precision="FP32">
691 <dim>384</dim>
692 <dim>384</dim>
693 </port>
694 </input>
695 <output>
696 <port id="2" precision="FP32">
697 <dim>-1</dim>
698 <dim>-1</dim>
699 <dim>384</dim>
700 </port>
701 </output>
702 </layer>
703 <layer id="53" name="Constant_90303" type="Const" version="opset1">
704 <data element_type="f32" shape="1, 1, 384" offset="12199696" size="1536" />
705 <output>
706 <port id="0" precision="FP32">
707 <dim>1</dim>
708 <dim>1</dim>
709 <dim>384</dim>
710 </port>
711 </output>
712 </layer>
713 <layer id="54" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/Add" type="Add" version="opset1">
714 <data auto_broadcast="numpy" />
715 <input>
716 <port id="0" precision="FP32">
717 <dim>-1</dim>
718 <dim>-1</dim>
719 <dim>384</dim>
720 </port>
721 <port id="1" precision="FP32">
722 <dim>1</dim>
723 <dim>1</dim>
724 <dim>384</dim>
725 </port>
726 </input>
727 <output>
728 <port id="2" precision="FP32" names="96,x.1">
729 <dim>-1</dim>
730 <dim>-1</dim>
731 <dim>384</dim>
732 </port>
733 </output>
734 </layer>
735 <layer id="55" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1">
736 <data element_type="f32" shape="" offset="12201232" size="4" />
737 <output>
738 <port id="0" precision="FP32" />
739 </output>
740 </layer>
741 <layer id="56" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1">
742 <data element_type="f32" shape="" offset="12201236" size="4" />
743 <output>
744 <port id="0" precision="FP32" />
745 </output>
746 </layer>
747 <layer id="57" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1">
748 <data element_type="f32" shape="" offset="12201232" size="4" />
749 <output>
750 <port id="0" precision="FP32" />
751 </output>
752 </layer>
753 <layer id="58" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1">
754 <data element_type="f32" shape="" offset="12201236" size="4" />
755 <output>
756 <port id="0" precision="FP32" />
757 </output>
758 </layer>
759 <layer id="59" name="__module.bert.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1">
760 <data levels="256" auto_broadcast="numpy" />
761 <input>
762 <port id="0" precision="FP32">
763 <dim>-1</dim>
764 <dim>-1</dim>
765 <dim>384</dim>
766 </port>
767 <port id="1" precision="FP32" />
768 <port id="2" precision="FP32" />
769 <port id="3" precision="FP32" />
770 <port id="4" precision="FP32" />
771 </input>
772 <output>
773 <port id="5" precision="FP32">
774 <dim>-1</dim>
775 <dim>-1</dim>
776 <dim>384</dim>
777 </port>
778 </output>
779 </layer>
780 <layer id="60" name="__module.bert.encoder.layer.0.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1">
781 <data element_type="i64" shape="4" offset="12201240" size="32" />
782 <output>
783 <port id="0" precision="I64">
784 <dim>4</dim>
785 </port>
786 </output>
787 </layer>
788 <layer id="61" name="__module.bert.encoder.layer.0.attention.self/aten::view/Reshape" type="Reshape" version="opset1">
789 <data special_zero="true" />
790 <input>
791 <port id="0" precision="FP32">
792 <dim>-1</dim>
793 <dim>-1</dim>
794 <dim>384</dim>
795 </port>
796 <port id="1" precision="I64">
797 <dim>4</dim>
798 </port>
799 </input>
800 <output>
801 <port id="2" precision="FP32" names="100,x.3">
802 <dim>-1</dim>
803 <dim>-1</dim>
804 <dim>12</dim>
805 <dim>32</dim>
806 </port>
807 </output>
808 </layer>
809 <layer id="62" name="Constant_84452" type="Const" version="opset1">
810 <data element_type="i64" shape="4" offset="12201272" size="32" />
811 <output>
812 <port id="0" precision="I64" names="101">
813 <dim>4</dim>
814 </port>
815 </output>
816 </layer>
817 <layer id="63" name="__module.bert.encoder.layer.0.attention.self/aten::permute/Transpose" type="Transpose" version="opset1">
818 <input>
819 <port id="0" precision="FP32">
820 <dim>-1</dim>
821 <dim>-1</dim>
822 <dim>12</dim>
823 <dim>32</dim>
824 </port>
825 <port id="1" precision="I64">
826 <dim>4</dim>
827 </port>
828 </input>
829 <output>
830 <port id="2" precision="FP32" names="102">
831 <dim>-1</dim>
832 <dim>12</dim>
833 <dim>-1</dim>
834 <dim>32</dim>
835 </port>
836 </output>
837 </layer>
838 <layer id="64" name="Constant_299942" type="Const" version="opset1">
839 <data element_type="i8" shape="384, 384" offset="12201304" size="147456" />
840 <output>
841 <port id="0" precision="I8">
842 <dim>384</dim>
843 <dim>384</dim>
844 </port>
845 </output>
846 </layer>
847 <layer id="65" name="Convert_299943" type="Convert" version="opset1">
848 <data destination_type="f32" />
849 <input>
850 <port id="0" precision="I8">
851 <dim>384</dim>
852 <dim>384</dim>
853 </port>
854 </input>
855 <output>
856 <port id="1" precision="FP32">
857 <dim>384</dim>
858 <dim>384</dim>
859 </port>
860 </output>
861 </layer>
862 <layer id="66" name="Constant_299944" type="Const" version="opset1">
863 <data element_type="f32" shape="384, 1" offset="12348760" size="1536" />
864 <output>
865 <port id="0" precision="FP32">
866 <dim>384</dim>
867 <dim>1</dim>
868 </port>
869 </output>
870 </layer>
871 <layer id="67" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1">
872 <data auto_broadcast="numpy" />
873 <input>
874 <port id="0" precision="FP32">
875 <dim>384</dim>
876 <dim>384</dim>
877 </port>
878 <port id="1" precision="FP32">
879 <dim>384</dim>
880 <dim>1</dim>
881 </port>
882 </input>
883 <output>
884 <port id="2" precision="FP32">
885 <dim>384</dim>
886 <dim>384</dim>
887 </port>
888 </output>
889 </layer>
890 <layer id="68" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1">
891 <data transpose_a="false" transpose_b="true" />
892 <input>
893 <port id="0" precision="FP32">
894 <dim>-1</dim>
895 <dim>-1</dim>
896 <dim>384</dim>
897 </port>
898 <port id="1" precision="FP32">
899 <dim>384</dim>
900 <dim>384</dim>
901 </port>
902 </input>
903 <output>
904 <port id="2" precision="FP32">
905 <dim>-1</dim>
906 <dim>-1</dim>
907 <dim>384</dim>
908 </port>
909 </output>
910 </layer>
911 <layer id="69" name="Constant_90304" type="Const" version="opset1">
912 <data element_type="f32" shape="1, 1, 384" offset="12350296" size="1536" />
913 <output>
914 <port id="0" precision="FP32">
915 <dim>1</dim>
916 <dim>1</dim>
917 <dim>384</dim>
918 </port>
919 </output>
920 </layer>
921 <layer id="70" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/Add" type="Add" version="opset1">
922 <data auto_broadcast="numpy" />
923 <input>
924 <port id="0" precision="FP32">
925 <dim>-1</dim>
926 <dim>-1</dim>
927 <dim>384</dim>
928 </port>
929 <port id="1" precision="FP32">
930 <dim>1</dim>
931 <dim>1</dim>
932 <dim>384</dim>
933 </port>
934 </input>
935 <output>
936 <port id="2" precision="FP32" names="105,x.5">
937 <dim>-1</dim>
938 <dim>-1</dim>
939 <dim>384</dim>
940 </port>
941 </output>
942 </layer>
943 <layer id="71" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1">
944 <data element_type="f32" shape="" offset="12351832" size="4" />
945 <output>
946 <port id="0" precision="FP32" />
947 </output>
948 </layer>
949 <layer id="72" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1">
950 <data element_type="f32" shape="" offset="12351836" size="4" />
951 <output>
952 <port id="0" precision="FP32" />
953 </output>
954 </layer>
955 <layer id="73" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1">
956 <data element_type="f32" shape="" offset="12351832" size="4" />
957 <output>
958 <port id="0" precision="FP32" />
959 </output>
960 </layer>
961 <layer id="74" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1">
962 <data element_type="f32" shape="" offset="12351836" size="4" />
963 <output>
964 <port id="0" precision="FP32" />
965 </output>
966 </layer>
967 <layer id="75" name="__module.bert.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1">
968 <data levels="256" auto_broadcast="numpy" />
969 <input>
970 <port id="0" precision="FP32">
971 <dim>-1</dim>
972 <dim>-1</dim>
973 <dim>384</dim>
974 </port>
975 <port id="1" precision="FP32" />
976 <port id="2" precision="FP32" />
977 <port id="3" precision="FP32" />
978 <port id="4" precision="FP32" />
979 </input>
980 <output>
981 <port id="5" precision="FP32">
982 <dim>-1</dim>
983 <dim>-1</dim>
984 <dim>384</dim>
985 </port>
986 </output>
987 </layer>
988 <layer id="76" name="__module.bert.encoder.layer.0.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1">
989 <data element_type="i64" shape="4" offset="12201240" size="32" />
990 <output>
991 <port id="0" precision="I64">
992 <dim>4</dim>
993 </port>
994 </output>
995 </layer>
996 <layer id="77" name="__module.bert.encoder.layer.0.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1">
997 <data special_zero="true" />
998 <input>
999 <port id="0" precision="FP32">
1000 <dim>-1</dim>
1001 <dim>-1</dim>
1002 <dim>384</dim>
1003 </port>
1004 <port id="1" precision="I64">
1005 <dim>4</dim>
1006 </port>
1007 </input>
1008 <output>
1009 <port id="2" precision="FP32" names="109,x.7">
1010 <dim>-1</dim>
1011 <dim>-1</dim>
1012 <dim>12</dim>
1013 <dim>32</dim>
1014 </port>
1015 </output>
1016 </layer>
1017 <layer id="78" name="Constant_84477" type="Const" version="opset1">
1018 <data element_type="i64" shape="4" offset="12201272" size="32" />
1019 <output>
1020 <port id="0" precision="I64" names="110">
1021 <dim>4</dim>
1022 </port>
1023 </output>
1024 </layer>
1025 <layer id="79" name="__module.bert.encoder.layer.0.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1">
1026 <input>
1027 <port id="0" precision="FP32">
1028 <dim>-1</dim>
1029 <dim>-1</dim>
1030 <dim>12</dim>
1031 <dim>32</dim>
1032 </port>
1033 <port id="1" precision="I64">
1034 <dim>4</dim>
1035 </port>
1036 </input>
1037 <output>
1038 <port id="2" precision="FP32" names="111">
1039 <dim>-1</dim>
1040 <dim>12</dim>
1041 <dim>-1</dim>
1042 <dim>32</dim>
1043 </port>
1044 </output>
1045 </layer>
1046 <layer id="80" name="Constant_299946" type="Const" version="opset1">
1047 <data element_type="i8" shape="384, 384" offset="12351840" size="147456" />
1048 <output>
1049 <port id="0" precision="I8">
1050 <dim>384</dim>
1051 <dim>384</dim>
1052 </port>
1053 </output>
1054 </layer>
1055 <layer id="81" name="Convert_299947" type="Convert" version="opset1">
1056 <data destination_type="f32" />
1057 <input>
1058 <port id="0" precision="I8">
1059 <dim>384</dim>
1060 <dim>384</dim>
1061 </port>
1062 </input>
1063 <output>
1064 <port id="1" precision="FP32">
1065 <dim>384</dim>
1066 <dim>384</dim>
1067 </port>
1068 </output>
1069 </layer>
1070 <layer id="82" name="Constant_299948" type="Const" version="opset1">
1071 <data element_type="f32" shape="384, 1" offset="12499296" size="1536" />
1072 <output>
1073 <port id="0" precision="FP32">
1074 <dim>384</dim>
1075 <dim>1</dim>
1076 </port>
1077 </output>
1078 </layer>
1079 <layer id="83" name="__module.bert.encoder.layer.0.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1">
1080 <data auto_broadcast="numpy" />
1081 <input>
1082 <port id="0" precision="FP32">
1083 <dim>384</dim>
1084 <dim>384</dim>
1085 </port>
1086 <port id="1" precision="FP32">
1087 <dim>384</dim>
1088 <dim>1</dim>
1089 </port>
1090 </input>
1091 <output>
1092 <port id="2" precision="FP32">
1093 <dim>384</dim>
1094 <dim>384</dim>
1095 </port>
1096 </output>
1097 </layer>
1098 <layer id="84" name="__module.bert.encoder.layer.0.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1">
1099 <data transpose_a="false" transpose_b="true" />
1100 <input>
1101 <port id="0" precision="FP32">
1102 <dim>-1</dim>
1103 <dim>-1</dim>
1104 <dim>384</dim>
1105 </port>
1106 <port id="1" precision="FP32">
1107 <dim>384</dim>
1108 <dim>384</dim>
1109 </port>
1110 </input>
1111 <output>
1112 <port id="2" precision="FP32">
1113 <dim>-1</dim>
1114 <dim>-1</dim>
1115 <dim>384</dim>
1116 </port>
1117 </output>
1118 </layer>
1119 <layer id="85" name="Constant_90305" type="Const" version="opset1">
1120 <data element_type="f32" shape="1, 1, 384" offset="12500832" size="1536" />
1121 <output>
1122 <port id="0" precision="FP32">
1123 <dim>1</dim>
1124 <dim>1</dim>
1125 <dim>384</dim>
1126 </port>
1127 </output>
1128 </layer>
1129 <layer id="86" name="__module.bert.encoder.layer.0.attention.self.value/aten::linear/Add" type="Add" version="opset1">
1130 <data auto_broadcast="numpy" />
1131 <input>
1132 <port id="0" precision="FP32">
1133 <dim>-1</dim>
1134 <dim>-1</dim>
1135 <dim>384</dim>
1136 </port>
1137 <port id="1" precision="FP32">
1138 <dim>1</dim>
1139 <dim>1</dim>
1140 <dim>384</dim>
1141 </port>
1142 </input>
1143 <output>
1144 <port id="2" precision="FP32" names="114,x.9">
1145 <dim>-1</dim>
1146 <dim>-1</dim>
1147 <dim>384</dim>
1148 </port>
1149 </output>
1150 </layer>
1151 <layer id="87" name="__module.bert.encoder.layer.0.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1">
1152 <data element_type="i64" shape="4" offset="12201240" size="32" />
1153 <output>
1154 <port id="0" precision="I64">
1155 <dim>4</dim>
1156 </port>
1157 </output>
1158 </layer>
1159 <layer id="88" name="__module.bert.encoder.layer.0.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1">
1160 <data special_zero="true" />
1161 <input>
1162 <port id="0" precision="FP32">
1163 <dim>-1</dim>
1164 <dim>-1</dim>
1165 <dim>384</dim>
1166 </port>
1167 <port id="1" precision="I64">
1168 <dim>4</dim>
1169 </port>
1170 </input>
1171 <output>
1172 <port id="2" precision="FP32" names="118,x.11">
1173 <dim>-1</dim>
1174 <dim>-1</dim>
1175 <dim>12</dim>
1176 <dim>32</dim>
1177 </port>
1178 </output>
1179 </layer>
1180 <layer id="89" name="Constant_84502" type="Const" version="opset1">
1181 <data element_type="i64" shape="4" offset="12201272" size="32" />
1182 <output>
1183 <port id="0" precision="I64" names="119">
1184 <dim>4</dim>
1185 </port>
1186 </output>
1187 </layer>
1188 <layer id="90" name="__module.bert.encoder.layer.0.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1">
1189 <input>
1190 <port id="0" precision="FP32">
1191 <dim>-1</dim>
1192 <dim>-1</dim>
1193 <dim>12</dim>
1194 <dim>32</dim>
1195 </port>
1196 <port id="1" precision="I64">
1197 <dim>4</dim>
1198 </port>
1199 </input>
1200 <output>