Add dynamic shape tests for xnnpack model tests

[Lidang-Jiang]

Summary

Add dynamic shape tests for xnnpack model tests that previously only tested with static inputs, as requested in #11585.

Models covered:

• DeepLab V3 — dynamic height/width with DynamicWrapper + interpolate
• EDSR — dynamic height/width with DynamicWrapper + interpolate
• Inception V3/V4 — dynamic height/width with DynamicWrapper + interpolate
• Emformer RNNT — dynamic batch dimension for joiner and transcriber
• MobileBERT — dynamic sequence length

All dynamic tests follow the established pattern from resnet.py (DynamicResNet) and torchvision_vit.py (DynamicViT): a wrapper module that resizes dynamic spatial inputs to fixed dimensions via F.interpolate before feeding into the model.

Models already covered by existing dynamic tests (no changes):

• MobileNet V2, MobileNet V3, ResNet, ViT

Models skipped:

• llama2_et_example.py — LLM, requires separate dynamic shape strategy
• very_big_model.py — synthetic test model

Before (only static tests)

$ python -m pytest backends/xnnpack/test/models/deeplab_v3.py -v
backends/xnnpack/test/models/deeplab_v3.py::TestDeepLabV3::test_fp32_dl3 PASSED

(no dynamic shape test exists)

After (static + dynamic tests pass)

$ python -m pytest backends/xnnpack/test/models/deeplab_v3.py::TestDeepLabV3::test_fp32_dl3_dynamic -v
backends/xnnpack/test/models/deeplab_v3.py::TestDeepLabV3::test_fp32_dl3_dynamic PASSED [100%]
======================== 1 passed in 28.84s ==============================

$ python -m pytest backends/xnnpack/test/models/mobilebert.py::TestMobilebert::test_fp32_mobilebert_dynamic -v
backends/xnnpack/test/models/mobilebert.py::TestMobilebert::test_fp32_mobilebert_dynamic PASSED [100%]
======================== 1 passed in 160.79s (0:02:40) ===================

Test plan

• test_fp32_dl3_dynamic passes locally
• test_fp32_mobilebert_dynamic passes locally
• CI xnnpack test suite passes

cc @GregoryComer @digantdesai @cbilgin

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18701

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• Rolling out OSDC (ARC) runners on pull workflow for PyTorch trunk commits

❌ 1 New Failure, 2 Unrelated Failures

As of commit 3e2193a with merge base ac68932 ():

NEW FAILURE - The following job has failed:

• pull / unittest / macos / macos-job (gh)
  export/tests/test_target_recipes.py::TestTargetRecipes::test_mv3_model

BROKEN TRUNK - The following jobs failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• pull / unittest / windows / windows-job (gh) (trunk failure)
  ##[error]The operation was canceled.
• pull / unittest-editable / windows / windows-job (gh) (trunk failure)
  ##[error]The operation was canceled.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[Lidang-Jiang]

@pytorchbot label "release notes: xnnpack"

[GregoryComer]

@Lidang-Jiang Thanks for the contribution. Looks good. Can you resolve the lint failure (run link)? I can go ahead and merge once the lint job is green.

[Lidang-Jiang]

Good catch, fixed in 3e2193a. I rebased onto latest main, applied the UFMT formatting fix in mobilebert.py, and force-pushed to rerun lint/CI.

[GregoryComer]

I'll go ahead and merge. Thanks for the contribution!