Fix MLIR input fusion non-std shapes from squeeze, flatten and unsqueeze

[manupak]

Currently, we see MLIR partition candidates recieving non-standard shape due to not fusing in squeeze, flatten and unsqueeze ops. These ops could be canonicalized to reshape without introducing additional ops as long as MLIR backend is concerned.

[codecov]

Codecov Report

Merging #2313 (833d28b) into develop (50c5984) will not change coverage.
Report is 2 commits behind head on develop.
The diff coverage is n/a.

❗ Current head 833d28b differs from pull request most recent head dc81fe4. Consider uploading reports for the commit dc81fe4 to get more accurate results

@@           Coverage Diff            @@
##           develop    #2313   +/-   ##
========================================
  Coverage    91.45%   91.45%           
========================================
  Files          433      433           
  Lines        16174    16174           
========================================
  Hits         14792    14792           
  Misses        1382     1382           

[migraphx-bot]

Test	Batch	Rate new be4658	Rate old c58e7d	Diff	Compare
torchvision-resnet50	64	2,323.67	2,324.91	-0.05%	✅
torchvision-resnet50_fp16	64	5,351.30	5,358.20	-0.13%	✅
torchvision-densenet121	32	1,834.15	1,847.11	-0.70%	✅
torchvision-densenet121_fp16	32	3,402.93	3,418.55	-0.46%	✅
torchvision-inceptionv3	32	1,297.28	1,294.11	0.24%	✅
torchvision-inceptionv3_fp16	32	2,526.91	2,526.65	0.01%	✅
cadene-inceptionv4	16	620.16	620.65	-0.08%	✅
cadene-resnext64x4	16	589.13	588.39	0.13%	✅
slim-mobilenet	64	7,208.27	7,222.03	-0.19%	✅
slim-nasnetalarge	64	236.28	236.42	-0.06%	✅
slim-resnet50v2	64	2,556.46	2,557.46	-0.04%	✅
bert-mrpc-onnx	8	825.45	825.89	-0.05%	✅
bert-mrpc-tf	1	389.89	388.50	0.36%	✅
pytorch-examples-wlang-gru	1	301.46	295.28	2.09%	✅
pytorch-examples-wlang-lstm	1	314.85	313.61	0.40%	✅
torchvision-resnet50_1	1	551.31	549.68	0.30%	✅
torchvision-inceptionv3_1	1	302.64	302.67	-0.01%	✅
cadene-dpn92_1	1	351.18	353.79	-0.74%	✅
cadene-resnext101_1	1	219.93	220.50	-0.26%	✅
slim-vgg16_1	1	223.78	223.86	-0.04%	✅
slim-mobilenet_1	1	1,517.03	1,473.02	2.99%	✅
slim-inceptionv4_1	1	217.76	216.23	0.71%	✅
onnx-taau-downsample	1	306.25	307.08	-0.27%	✅
dlrm-criteoterabyte	1	21.68	21.67	0.02%	✅
dlrm-criteoterabyte_fp16	1	40.71	40.73	-0.07%	✅
agentmodel	1	5,833.18	5,882.66	-0.84%	✅
unet_fp16	2	55.20	55.24	-0.08%	✅
resnet50v1_fp16	1	769.76	756.89	1.70%	✅
bert_base_cased_fp16	64	971.00	971.21	-0.02%	✅
bert_large_uncased_fp16	32	305.02	305.13	-0.04%	✅
bert_large_fp16	1	167.10	166.84	0.16%	✅
distilgpt2_fp16	16	1,279.63	1,279.91	-0.02%	✅

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

    :white_check_mark:torchvision-inceptionv3_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:cadene-dpn92_1: PASSED: MIGraphX meets tolerance

    :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

    :white_check_mark:slim-inceptionv4_1: PASSED: MIGraphX meets tolerance

    :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

    :white_check_mark:resnet50v1: PASSED: MIGraphX meets tolerance

🔴bert_base_cased_fp16: FAILED: MIGraphX is not within tolerance - check verbose output

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

    :white_check_mark:bert_large: PASSED: MIGraphX meets tolerance

🔴distilgpt2_fp16: FAILED: MIGraphX is not within tolerance - check verbose output

[krzysz00]

This seems fine to me, but I want to confirm that squeeze, unsqueeze, and flatten can all be replaced by reshape correctly like this

Also, can you add a test?

[giuseros]

I can confirm that the tests are passing on mi200: http://rocmhead.amd.com:8080/blue/organizations/jenkins/MLIR%2Ftest/detail/test/30/pipeline/54

[manupak]

This seems fine to me, but I want to confirm that squeeze, unsqueeze, and flatten can all be replaced by reshape correctly like this

Also, can you add a test?

They are special cases of 'reshape' where as in reshape just work all that out by just knowing the output shape.

I ve added tests now.

[pfultz2]

In the future, it would be to make these transformation when we are generating MLIR.

[krzysz00]

Yeah, but that'll be more possible once we have MIGraphX shapes in the MIGraphX dialect