Print Ref compiled program in verify tests

[umangyadav]

Solution to #2525 and #2520 would mean changing compilation pipeline for ref program.
It is easier for debug purpose to print compiled ref program for comparison.

This PR allows to print it. Previously verify test printed uncompiled program for "ref".

[CharlieL7]

LGTM, doesn't solve the issue from #2525 however because the problem is that when ref runs with fp16 the accuracy is inherently lower than when running with higher floating point precision. Comparing other targets to ref in the verify tests then becomes as wild goose chase to figure out where the error is coming from.

[umangyadav]

LGTM, doesn't solve the issue from #2525 however because the problem is that when ref runs with fp16 the accuracy is inherently lower than when running with higher floating point precision. Comparing other targets to ref in the verify tests then becomes as wild goose chase to figure out where the error is coming from.

Yes this doesn't solve anything. This is just useful for debugging purposes.

[migraphx-bot]

Test	Batch	Rate new d11208	Rate old a09dc5	Diff	Compare
torchvision-resnet50	64	2,834.63	2,833.31	0.05%	✅
torchvision-resnet50_fp16	64	6,499.61	6,501.92	-0.04%	✅
torchvision-densenet121	32	2,093.02	2,094.20	-0.06%	✅
torchvision-densenet121_fp16	32	3,660.75	3,665.08	-0.12%	✅
torchvision-inceptionv3	32	1,598.67	1,594.40	0.27%	✅
torchvision-inceptionv3_fp16	32	2,561.17	2,557.61	0.14%	✅
cadene-inceptionv4	16	722.63	721.79	0.12%	✅
cadene-resnext64x4	16	692.81	692.75	0.01%	✅
slim-mobilenet	64	8,324.54	8,332.19	-0.09%	✅
slim-nasnetalarge	64	230.67	230.63	0.02%	✅
slim-resnet50v2	64	2,665.19	2,664.91	0.01%	✅
bert-mrpc-onnx	8	824.34	823.07	0.15%	✅
bert-mrpc-tf	1	390.56	389.27	0.33%	✅
pytorch-examples-wlang-gru	1	301.15	299.82	0.44%	✅
pytorch-examples-wlang-lstm	1	316.51	308.52	2.59%	✅
torchvision-resnet50_1	1	607.67	602.66	0.83%	✅
torchvision-inceptionv3_1	1	341.01	343.14	-0.62%	✅
cadene-dpn92_1	1	402.83	404.59	-0.44%	✅
cadene-resnext101_1	1	328.93	328.32	0.19%	✅
slim-vgg16_1	1	459.43	460.67	-0.27%	✅
slim-mobilenet_1	1	2,131.89	2,099.38	1.55%	✅
slim-inceptionv4_1	1	213.71	213.62	0.05%	✅
onnx-taau-downsample	1	305.89	304.86	0.34%	✅
dlrm-criteoterabyte	1	21.59	21.62	-0.15%	✅
dlrm-criteoterabyte_fp16	1	40.62	40.65	-0.08%	✅
agentmodel	1	6,021.05	5,935.95	1.43%	✅
unet_fp16	2	54.76	54.68	0.14%	✅
resnet50v1_fp16	1	948.01	932.98	1.61%	✅
bert_base_cased_fp16	64	903.62	903.11	0.06%	✅
bert_large_uncased_fp16	32	285.79	285.63	0.06%	✅
bert_large_fp16	1	166.57	166.70	-0.08%	✅
distilgpt2_fp16	16	1,281.94	1,281.25	0.05%	✅

This build is OK for 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

     ✅ torchvision-inceptionv3_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-dpn92_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-resnext101_1: PASSED: MIGraphX meets tolerance

     ✅ slim-vgg16_1: PASSED: MIGraphX meets tolerance

     ✅ slim-mobilenet_1: PASSED: MIGraphX meets tolerance

     ✅ slim-inceptionv4_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: FAILED: MIGraphX is not within tolerance - check verbose output

     ✅ bert_large_uncased_fp16: PASSED: MIGraphX meets tolerance

     ✅ bert_large: PASSED: MIGraphX meets tolerance

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