Add dynamic support to Shape and Expand

[turneram]

Motivation

To enable dynamic shape support for operators upstream from dynamic kv-cache.

Technical Details

Updates how Expand and Shape are parsed and dispatched so that dynamic past-sequence-length can propagate from input to attention.

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.

Follow the LLVM AI Tool Use Policy for contributions using AI.

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

Test	Batch	New Rate (d13df3)	Old Rate (1a5bb6)	Diff	Status
torchvision-resnet50	64	3,264.96	3,265.81	-0.03%	✅
torchvision-resnet50_fp16	64	7,550.67	7,548.56	0.03%	✅
torchvision-densenet121	32	2,451.54	2,481.85	-1.22%	✅
torchvision-densenet121_fp16	32	5,006.37	4,975.40	0.62%	✅
torchvision-inceptionv3	32	2,050.01	2,060.05	-0.49%	✅
torchvision-inceptionv3_fp16	32	4,412.99	4,414.40	-0.03%	✅
cadene-inceptionv4	16	821.09	584.00	40.60%	🔆
cadene-resnext64x4	16	783.25	782.84	0.05%	✅
slim-mobilenet	64	8,388.05	7,190.62	16.65%	🔆
slim-nasnetalarge	64	198.77	222.84	-10.80%	🔴
slim-resnet50v2	64	3,108.00	2,017.65	54.04%	🔆
bert-mrpc-onnx	8	1,107.56	1,168.85	-5.24%	🔴
bert-mrpc-tf	1	501.75	487.56	2.91%	✅
pytorch-examples-wlang-gru	1	480.38	477.33	0.64%	✅
pytorch-examples-wlang-lstm	1	447.10	387.83	15.28%	🔆
torchvision-resnet50_1	1	1,041.21	1,038.23	0.29%	✅
cadene-dpn92_1	1	471.70	449.51	4.94%	✅
cadene-resnext101_1	1	364.92	365.36	-0.12%	✅
onnx-taau-downsample	1	846.34	844.73	0.19%	✅
dlrm-criteoterabyte	1	13.94	17.44	-20.07%	🔴
dlrm-criteoterabyte_fp16	1	26.81	31.21	-14.08%	🔴
agentmodel	1	8,913.13	5,880.42	51.57%	🔆
unet_fp16	2	23.25	52.15	-55.42%	🔴
resnet50v1_fp16	1	1,394.91	1,385.99	0.64%	✅
resnet50v1_int8	1	1,825.84	1,657.40	10.16%	🔆
bert_base_cased_fp16	64	439.17	1,022.83	-57.06%	🔴
bert_large_uncased_fp16	32	124.42	339.30	-63.33%	🔴
bert_large_fp16	1	204.47	205.64	-0.57%	✅
distilgpt2_fp16	16	2,083.34	2,086.12	-0.13%	✅
yolov5s	1	557.44	560.85	-0.61%	✅
tinyllama	1	22.65	45.80	-50.56%	🔴
vicuna-fastchat	1	16.92	44.03	-61.56%	🔴
whisper-tiny-encoder	1	412.50	328.75	25.48%	🔆
whisper-tiny-decoder	1	414.62	214.20	93.57%	🔆
llama2_7b	1	6.10	14.70	-58.50%	🔴
qwen1.5-7b	1	10.14	23.49	-56.85%	🔴
phi3-3.8b	1	18.33	26.78	-31.54%	🔴
llama3-8b	1	19.05	21.70	-12.20%	🔴
whisper-large-encoder	1	10.20	8.77	16.32%	🔆
whisper-large-decoder	1	105.50	109.89	-3.99%	✅
mistral-7b	1	23.53	23.65	-0.50%	✅
FLUX.1-schnell	1	770.95	172.94	345.80%	🔆

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
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