Skip "find_split_transpose" if split has more than one outputs

[umangyadav]

The current matcher is doing an assert on the number of outputs but instead, it should skip it if the number of outputs is greater than 1.

This matcher is doing

input --> slice --> transpose 
    | 
     ---->slice --> transpose 
    | 
    ---->slice --> transpose 

to

input --> transpose --> slice
                | 
                 ----> slice 
                | 
                ----> slice 

This is only useful if slice in original input graph has one output node (transpose in this case). If it has multiple outputs then doing transpose before the slice would change the layout for the slice's outputs (other than transpose) in original graph.

Partially solves #2116

[codecov]

Codecov Report

Merging #2126 (6a8d74f) into develop (ac96370) will not change coverage.
Report is 1 commits behind head on develop.
The diff coverage is 100.00%.

❗ Current head 6a8d74f differs from pull request most recent head 09f35e7. Consider uploading reports for the commit 09f35e7 to get more accurate results

@@           Coverage Diff            @@
##           develop    #2126   +/-   ##
========================================
  Coverage    91.43%   91.43%           
========================================
  Files          422      422           
  Lines        15771    15771           
========================================
  Hits         14420    14420           
  Misses        1351     1351           

Files Changed	Coverage Δ
src/simplify_algebra.cpp	96.79% <100.00%> (ø)

[migraphx-bot]

Test	Batch	Rate new 09f35e	Rate old ac9637	Diff	Compare
torchvision-resnet50	64	2,282.83	2,281.62	0.05%	✅
torchvision-resnet50_fp16	64	5,362.69	5,363.23	-0.01%	✅
torchvision-densenet121	32	1,837.46	1,826.57	0.60%	✅
torchvision-densenet121_fp16	32	3,390.99	3,387.24	0.11%	✅
torchvision-inceptionv3	32	1,342.27	1,340.34	0.14%	✅
torchvision-inceptionv3_fp16	32	2,592.18	2,582.06	0.39%	✅
cadene-inceptionv4	16	680.23	682.15	-0.28%	✅
cadene-resnext64x4	16	590.35	589.61	0.13%	✅
slim-mobilenet	64	7,213.16	7,225.62	-0.17%	✅
slim-nasnetalarge	64	236.67	236.89	-0.09%	✅
slim-resnet50v2	64	2,527.47	2,527.05	0.02%	✅
bert-mrpc-onnx	8	720.61	720.80	-0.03%	✅
bert-mrpc-tf	1	390.69	389.58	0.28%	✅
pytorch-examples-wlang-gru	1	302.21	301.37	0.28%	✅
pytorch-examples-wlang-lstm	1	305.98	313.89	-2.52%	✅
torchvision-resnet50_1	1	557.35	557.53	-0.03%	✅
torchvision-inceptionv3_1	1	306.45	306.50	-0.02%	✅
cadene-dpn92_1	1	351.99	345.97	1.74%	✅
cadene-resnext101_1	1	220.90	220.27	0.28%	✅
slim-vgg16_1	1	224.23	224.56	-0.15%	✅
slim-mobilenet_1	1	1,476.56	1,478.76	-0.15%	✅
slim-inceptionv4_1	1	223.35	222.77	0.26%	✅
onnx-taau-downsample	1	247.37	247.83	-0.18%	✅
dlrm-criteoterabyte	1	21.67	21.67	0.01%	✅
dlrm-criteoterabyte_fp16	1	40.63	40.63	-0.02%	✅
agentmodel	1	5,820.04	5,848.06	-0.48%	✅
unet_fp16	2	55.06	55.06	0.01%	✅

This build is OK for merge ✅

[migraphx-bot]

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

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

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

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

    :white_check_mark:torchvision-resnet50_1: PASSED: MIGraphX meets tolerance

🔴torchvision-inceptionv3_1: FAILED: MIGraphX is not within tolerance - check verbose output

🔴cadene-dpn92_1: FAILED: MIGraphX is not within tolerance - check verbose output

    :white_check_mark:cadene-resnext101_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:slim-vgg16_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:slim-mobilenet_1: PASSED: MIGraphX meets tolerance

🔴slim-inceptionv4_1: FAILED: MIGraphX is not within tolerance - check verbose output

    :white_check_mark:dlrm-criteoterabyte: PASSED: MIGraphX meets tolerance

    :white_check_mark:agentmodel: PASSED: MIGraphX meets tolerance

    :white_check_mark:unet: PASSED: MIGraphX meets tolerance

[kahmed10]

I think we need to add a unit test. I've seen this happen in distilgpt2 as well.

[pfultz2]

I think we need to add a unit test. I've seen this happen in distilgpt2 as well.

Yea a unit test would be good.

[umangyadav]

Added Unit-test