Preserve shape ops when removing QDQ pairs

[ikalinic]

Motivation

Fix simplify_qdq so Q/DQ pair removal does not accidentally drop intervening shape operations or produce shape-changing rewrites.

Technical Details

Updated Q/DQ removal to replay supported shape operations when removing non-FP4 fake quantization chains, while continuing to drop FP4 pack/unpack adapter operations where appropriate. Added guards for concat QLinear splitting when sliced scale or zero-point shapes do not match the corresponding concat input. Added a regression test for preserving slice and reshape through Q/DQ removal.

Changelog Category

Add a CHANGELOG.md entry for any option other than Not Applicable

• • Added: New functionality.
• • Changed: Changes to existing functionality.
• • Removed: Functionality or support that has been removed. (Compared to a previous release)
• • Optimized: Component performance that has been optimized or improved.
• • Resolved Issues: Known issues from a previous version that have been resolved.
• • Not Applicable: This PR is not to be included in the changelog.

[codecov]

Codecov Report

❌ Patch coverage is 90.76923% with 6 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/simplify_qdq.cpp	91.80%	5 Missing ⚠️
src/include/migraphx/instruction_traversal.hpp	75.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #5048      +/-   ##
===========================================
- Coverage    93.27%   93.26%   -0.01%     
===========================================
  Files          623      623              
  Lines        32889    33020     +131     
===========================================
+ Hits         30674    30793     +119     
- Misses        2215     2227      +12     

Files with missing lines	Coverage Δ
src/include/migraphx/instruction_traversal.hpp	87.50% <75.00%> (-12.50%)	⬇️
src/simplify_qdq.cpp	94.23% <91.80%> (-0.39%)	⬇️

... and 5 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[gh-app-migraphx-bot-pr-write]

Test	Batch	New Rate (53fefd)	Old Rate (e1e03e)	Diff	Status
torchvision-resnet50	64	3,137.65	2,693.41	16.49%	🔆
torchvision-resnet50_fp16	64	6,623.21	6,415.71	3.23%	✅
torchvision-densenet121	32	2,694.15	2,702.20	-0.30%	✅
torchvision-densenet121_fp16	32	4,532.58	4,553.39	-0.46%	✅
torchvision-inceptionv3	32	1,795.49	950.94	88.81%	🔆
torchvision-inceptionv3_fp16	32	2,814.52	2,269.97	23.99%	🔆
cadene-inceptionv4	16	820.92	538.17	52.54%	🔆
cadene-resnext64x4	16	782.17	459.28	70.30%	🔆
slim-mobilenet	64	3,834.84	8,429.89	-54.51%	🔴
slim-nasnetalarge	64	226.95	89.65	153.15%	🔆
slim-resnet50v2	64	3,133.92	1,563.17	100.49%	🔆
bert-mrpc-onnx	8	1,171.68	1,165.41	0.54%	✅
bert-mrpc-tf	1	496.32	486.44	2.03%	✅
pytorch-examples-wlang-gru	1	492.04	487.62	0.91%	✅
pytorch-examples-wlang-lstm	1	257.91	392.17	-34.24%	🔴
torchvision-resnet50_1	1	765.43	771.94	-0.84%	✅
cadene-dpn92_1	1	110.83	442.37	-74.95%	🔴
cadene-resnext101_1	1	364.68	364.35	0.09%	✅
onnx-taau-downsample	1	400.04	400.76	-0.18%	✅
dlrm-criteoterabyte	1	32.41	32.68	-0.81%	✅
dlrm-criteoterabyte_fp16	1	51.82	52.60	-1.48%	✅
agentmodel	1	11,896.04	9,162.71	29.83%	🔆
unet_fp16	2	14.15	57.35	-75.33%	🔴
resnet50v1_fp16	1	864.45	1,017.75	-15.06%	🔴
resnet50v1_int8	1	251.69	937.30	-73.15%	🔴
bert_base_cased_fp16	64	932.78	1,102.37	-15.38%	🔴
bert_large_uncased_fp16	32	345.53	346.79	-0.36%	✅
bert_large_fp16	1	203.00	203.94	-0.46%	✅
distilgpt2_fp16	16	2,072.68	2,094.08	-1.02%	✅
yolov5s	1	557.77	560.76	-0.53%	✅
tinyllama	1	34.87	45.87	-23.98%	🔴
vicuna-fastchat	1	43.41	44.20	-1.78%	✅
whisper-tiny-encoder	1	412.83	415.10	-0.55%	✅
whisper-tiny-decoder	1	409.15	409.89	-0.18%	✅
llama2_7b	1	8.68	20.96	-58.61%	🔴
qwen1.5-7b	1	10.09	23.66	-57.36%	🔴
phi3-3.8b	1	26.68	26.82	-0.52%	✅
llama3-8b	1	21.75	21.88	-0.60%	✅
whisper-large-encoder	1	7.20	10.21	-29.42%	🔴
whisper-large-decoder	1	22.76	103.92	-78.10%	🔴
mistral-7b	1	23.79	23.89	-0.41%	✅
FLUX.1-schnell	1	330.56	757.77	-56.38%	🔴

Regressions detected 🔴

[gh-app-migraphx-bot-pr-write]

Test	Status	Result
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
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
llama2_7b	✅	PASSED: MIGraphX meets tolerance
qwen1.5-7b	✅	PASSED: MIGraphX meets tolerance
phi3-3.8b	✅	PASSED: MIGraphX meets tolerance
llama3-8b	✅	PASSED: MIGraphX meets tolerance
whisper-large-encoder	✅	PASSED: MIGraphX meets tolerance
whisper-large-decoder	✅	PASSED: MIGraphX meets tolerance
mistral-7b	✅	PASSED: MIGraphX meets tolerance
FLUX.1-schnell	✅	PASSED: MIGraphX meets tolerance

[TedThemistokleous]

Hey thanks for your contribution!

I have a few suggested changes and some initial questions

[TedThemistokleous]

Adding Shiv since he's looked a lot of the QDQ stuff

[ikalinic]

Hi @TedThemistokleous and @shivadbhavsar, I've addressed the suggested changes. Can I get a re-review on this?