QlinearGlobalAveragePool operator

[lakhinderwalia]

https://github.com/microsoft/onnxruntime/blob/main/docs/ContribOperators.md#com.microsoft.QLinearGlobalAveragePool

[codecov]

Codecov Report

Merging #2297 (bbcc73b) into develop (271eedd) will decrease coverage by 0.03%.
The diff coverage is 78.57%.

❗ Current head bbcc73b differs from pull request most recent head 5c3607b. Consider uploading reports for the commit 5c3607b to get more accurate results

@@             Coverage Diff             @@
##           develop    #2297      +/-   ##
===========================================
- Coverage    91.33%   91.31%   -0.03%     
===========================================
  Files          434      435       +1     
  Lines        16262    16290      +28     
===========================================
+ Hits         14853    14875      +22     
- Misses        1409     1415       +6     

Files	Coverage Δ
src/onnx/parse_pooling.cpp	97.75% <ø> (ø)
src/onnx/parse_qlinearglavgpool.cpp	78.57% <78.57%> (ø)

[migraphx-bot]

Test	Batch	Rate new 7e06b6	Rate old 1a1c1b	Diff	Compare
torchvision-resnet50	64	2,326.20	2,325.39	0.03%	✅
torchvision-resnet50_fp16	64	5,352.56	5,347.82	0.09%	✅
torchvision-densenet121	32	1,849.40	1,849.21	0.01%	✅
torchvision-densenet121_fp16	32	3,412.17	3,406.11	0.18%	✅
torchvision-inceptionv3	32	1,297.26	1,293.29	0.31%	✅
torchvision-inceptionv3_fp16	32	2,534.25	2,537.00	-0.11%	✅
cadene-inceptionv4	16	620.49	620.90	-0.06%	✅
cadene-resnext64x4	16	589.20	588.26	0.16%	✅
slim-mobilenet	64	7,214.58	7,213.68	0.01%	✅
slim-nasnetalarge	64	236.53	236.58	-0.02%	✅
slim-resnet50v2	64	2,557.86	2,557.33	0.02%	✅
bert-mrpc-onnx	8	825.05	824.42	0.08%	✅
bert-mrpc-tf	1	388.12	388.84	-0.19%	✅
pytorch-examples-wlang-gru	1	294.51	298.85	-1.45%	✅
pytorch-examples-wlang-lstm	1	306.40	314.28	-2.51%	✅
torchvision-resnet50_1	1	552.51	549.39	0.57%	✅
torchvision-inceptionv3_1	1	305.00	302.45	0.84%	✅
cadene-dpn92_1	1	354.83	353.66	0.33%	✅
cadene-resnext101_1	1	219.93	218.95	0.45%	✅
slim-vgg16_1	1	224.14	224.19	-0.02%	✅
slim-mobilenet_1	1	1,526.46	1,510.80	1.04%	✅
slim-inceptionv4_1	1	218.14	216.72	0.65%	✅
onnx-taau-downsample	1	307.03	306.03	0.33%	✅
dlrm-criteoterabyte	1	21.72	21.69	0.17%	✅
dlrm-criteoterabyte_fp16	1	40.75	40.72	0.06%	✅
agentmodel	1	5,849.32	5,795.96	0.92%	✅
unet_fp16	2	55.17	55.22	-0.08%	✅
resnet50v1_fp16	1	748.99	761.81	-1.68%	✅
bert_base_cased_fp16	64	971.44	971.55	-0.01%	✅
bert_large_uncased_fp16	32	305.27	305.24	0.01%	✅
bert_large_fp16	1	166.74	166.66	0.05%	✅
distilgpt2_fp16	16	1,278.86	1,279.47	-0.05%	✅

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

[pfultz2]

Maybe we should just name this parser as parse_qlinearpool so it can be used in the future for other types of pooling.

[lakhinderwalia]

Paul, (agreed for the future steps..) but, for expediency, I suggest we do it for the next quantized pooling operator. Else, it would mean adding code for operators() etc., I believe, and I will have to verify the code again.. with no additional value add for now. Thanks.