Enable non-packed inputs for mlir

[pfultz2]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.17%. Comparing base (04ac9fc) to head (639d003).
Report is 1 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #3541   +/-   ##
========================================
  Coverage    92.17%   92.17%           
========================================
  Files          512      512           
  Lines        21393    21393           
========================================
  Hits         19720    19720           
  Misses        1673     1673           

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

[migraphx-bot]

Test	Batch	Rate new 639d00	Rate old 275f85	Diff	Compare
torchvision-resnet50	64	3,259.88	3,258.32	0.05%	✅
torchvision-resnet50_fp16	64	6,991.70	6,998.28	-0.09%	✅
torchvision-densenet121	32	2,435.83	2,439.53	-0.15%	✅
torchvision-densenet121_fp16	32	4,078.01	4,083.72	-0.14%	✅
torchvision-inceptionv3	32	1,638.80	1,639.50	-0.04%	✅
torchvision-inceptionv3_fp16	32	2,762.86	2,761.57	0.05%	✅
cadene-inceptionv4	16	776.36	776.25	0.01%	✅
cadene-resnext64x4	16	810.43	811.85	-0.18%	✅
slim-mobilenet	64	7,534.50	7,537.50	-0.04%	✅
slim-nasnetalarge	64	212.16	211.56	0.28%	✅
slim-resnet50v2	64	3,503.11	3,504.39	-0.04%	✅
bert-mrpc-onnx	8	1,151.61	1,148.92	0.23%	✅
bert-mrpc-tf	1	494.07	467.60	5.66%	🔆
pytorch-examples-wlang-gru	1	430.00	422.65	1.74%	✅
pytorch-examples-wlang-lstm	1	395.98	380.77	3.99%	🔆
torchvision-resnet50_1	1	786.19	815.97	-3.65%	🔴
cadene-dpn92_1	1	408.66	400.35	2.07%	✅
cadene-resnext101_1	1	384.86	384.70	0.04%	✅
onnx-taau-downsample	1	342.82	343.42	-0.17%	✅
dlrm-criteoterabyte	1	33.33	33.32	0.03%	✅
dlrm-criteoterabyte_fp16	1	52.71	52.72	-0.03%	✅
agentmodel	1	8,590.92	8,139.01	5.55%	🔆
unet_fp16	2	58.90	58.87	0.05%	✅
resnet50v1_fp16	1	923.74	930.44	-0.72%	✅
resnet50v1_int8	1	1,009.62	994.18	1.55%	✅
bert_base_cased_fp16	64	1,171.37	1,172.52	-0.10%	✅
bert_large_uncased_fp16	32	363.50	363.39	0.03%	✅
bert_large_fp16	1	198.51	198.49	0.01%	✅
distilgpt2_fp16	16	2,202.16	2,199.99	0.10%	✅
yolov5s	1	541.06	543.07	-0.37%	✅
tinyllama	1	43.50	43.73	-0.52%	✅
vicuna-fastchat	1	177.91	172.56	3.10%	🔆
whisper-tiny-encoder	1	418.74	417.97	0.19%	✅
whisper-tiny-decoder	1	429.69	428.62	0.25%	✅

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: PASSED: MIGraphX meets tolerance

     ✅ 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: PASSED: MIGraphX meets tolerance

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

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

     ✅ distilgpt2_fp16: PASSED: MIGraphX meets tolerance

[pfultz2]

@krzysz00 This is checking if the shapes can be used by mlir(ie can be constructed from slice/reshape/transpose/broadcast). I know there are similar checks in mlir. Are you able to check that this is checking the shapes correctly?

[krzysz00]

https://github.com/ROCm/rocMLIR/blob/develop/mlir/lib/Dialect/MIGraphX/IR/MIGraphX.cpp#L182 is our logic.

This code feels similar, though I'd need to run through some examples with broadcast dimensions

[krzysz00]

Assuming this doesn't read to non-packed outputs, seems fine to me