Move qlinear before concat to allow output fusion

[shivadbhavsar]

OLD:
q -> conv -> dq -> add -> relu -> q .......... -> q -> conv -> dq -> add -> relu
			   |						     |
		           -> step ----------------------------------------------> concat -> q -> conv -> ...
							
NEW:

q -> conv -> dq -> add -> relu -> q .......... -> q -> conv -> dq -> add -> relu -> q
			   |	                                                    |
			   -> step -> q -----------------------------------------> concat -> conv -> ...

For now this brings down resnet50 int8 (nhwc enabled) on navi31 by ~0.05ms. Will do more work around the step -> q part in another PR.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.30%. Comparing base (79f95da) to head (d4233ed).
Report is 2 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #3782   +/-   ##
========================================
  Coverage    92.29%   92.30%           
========================================
  Files          519      519           
  Lines        22233    22261   +28     
========================================
+ Hits         20520    20548   +28     
  Misses        1713     1713           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[migraphx-bot]

Test	Batch	Rate new d4233e	Rate old 5dc019	Diff	Compare
torchvision-resnet50	64	3,232.80	3,233.22	-0.01%	✅
torchvision-resnet50_fp16	64	6,864.01	6,861.00	0.04%	✅
torchvision-densenet121	32	2,431.06	2,431.85	-0.03%	✅
torchvision-densenet121_fp16	32	4,179.85	4,166.12	0.33%	✅
torchvision-inceptionv3	32	1,612.44	1,612.38	0.00%	✅
torchvision-inceptionv3_fp16	32	2,673.89	2,684.23	-0.39%	✅
cadene-inceptionv4	16	748.62	749.29	-0.09%	✅
cadene-resnext64x4	16	809.11	809.05	0.01%	✅
slim-mobilenet	64	6,659.29	6,659.22	0.00%	✅
slim-nasnetalarge	64	198.94	198.92	0.01%	✅
slim-resnet50v2	64	3,424.86	3,425.06	-0.01%	✅
bert-mrpc-onnx	8	1,139.96	1,138.85	0.10%	✅
bert-mrpc-tf	1	471.55	473.11	-0.33%	✅
pytorch-examples-wlang-gru	1	428.91	468.31	-8.41%	🔴
pytorch-examples-wlang-lstm	1	399.26	394.75	1.14%	✅
torchvision-resnet50_1	1	771.48	781.23	-1.25%	✅
cadene-dpn92_1	1	414.83	411.07	0.91%	✅
cadene-resnext101_1	1	389.13	389.62	-0.13%	✅
onnx-taau-downsample	1	371.01	372.07	-0.28%	✅
dlrm-criteoterabyte	1	30.53	30.53	-0.00%	✅
dlrm-criteoterabyte_fp16	1	49.03	49.10	-0.15%	✅
agentmodel	1	7,691.72	7,684.45	0.09%	✅
unet_fp16	2	57.61	57.79	-0.32%	✅
resnet50v1_fp16	1	979.96	991.06	-1.12%	✅
resnet50v1_int8	1	789.43	767.32	2.88%	✅
bert_base_cased_fp16	64	1,172.27	1,172.46	-0.02%	✅
bert_large_uncased_fp16	32	362.25	362.48	-0.06%	✅
bert_large_fp16	1	199.40	197.81	0.81%	✅
distilgpt2_fp16	16	2,213.95	2,214.12	-0.01%	✅
yolov5s	1	515.80	516.64	-0.16%	✅
tinyllama	1	43.41	43.46	-0.10%	✅
vicuna-fastchat	1	43.74	43.82	-0.19%	✅
whisper-tiny-encoder	1	411.10	410.74	0.09%	✅
whisper-tiny-decoder	1	405.55	402.68	0.71%	✅

This build is not recommended to merge 🔴

[migraphx-bot]

     ✅ bert-mrpc-onnx: PASSED: MIGraphX meets tolerance

     ✅ bert-mrpc-tf: PASSED: MIGraphX meets tolerance

     ✅ pytorch-examples-wlang-gru: PASSED: MIGraphX meets tolerance

     ✅ pytorch-examples-wlang-lstm: PASSED: MIGraphX meets tolerance

     ✅ torchvision-resnet50_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-dpn92_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-resnext101_1: PASSED: MIGraphX meets tolerance

❌dlrm-criteoterabyte: ERROR - check error output

usage: accuracy_checker.py [-h] [--onnx ONNX] [--tf TF] [--provider PROVIDER]
[--batch BATCH] [--fill1] [--fill0] [--fp16]
[--argmax] [--verbose] [--tolerance TOLERANCE]
[--input-dim INPUT_DIM] [--target TARGET]
[--ort-run] [--ort-logging]
[--disable-offload-copy] [--disable-fast-math]
[--exhaustive_tune]
accuracy_checker.py: error: unrecognized arguments: input lS_i 307200 13 @lS_i 26 307200

     ✅ agentmodel: PASSED: MIGraphX meets tolerance

     ✅ unet: PASSED: MIGraphX meets tolerance

     ✅ resnet50v1: PASSED: MIGraphX meets tolerance

     ✅ bert_base_cased_fp16: PASSED: MIGraphX meets tolerance

🔴bert_large_uncased_fp16: FAILED: MIGraphX is not within tolerance - check verbose output

     ✅ bert_large: PASSED: MIGraphX meets tolerance

     ✅ yolov5s: PASSED: MIGraphX meets tolerance

     ✅ tinyllama: PASSED: MIGraphX meets tolerance

❌vicuna-fastchat: ERROR - check error output

usage: accuracy_checker.py [-h] [--onnx ONNX] [--tf TF] [--provider PROVIDER]
[--batch BATCH] [--fill1] [--fill0] [--fp16]
[--argmax] [--verbose] [--tolerance TOLERANCE]
[--input-dim INPUT_DIM] [--target TARGET]
[--ort-run] [--ort-logging]
[--disable-offload-copy] [--disable-fast-math]
[--exhaustive_tune]
accuracy_checker.py: error: unrecognized arguments: input_ids attention_mask 1 256 @attention_mask 1 256

     ✅ whisper-tiny-encoder: PASSED: MIGraphX meets tolerance

     ✅ whisper-tiny-decoder: PASSED: MIGraphX meets tolerance

     ✅ distilgpt2_fp16: PASSED: MIGraphX meets tolerance