Qualcomm AI Engine Direct - GA model enablement (Swin Transformer)

[DannyYuyang-quic]

Summary

• e2e script / test case for GA Swin Transformer model (https://huggingface.co/microsoft/swin-tiny-patch4-window7-224)
  • perf: 8a8w 3.9ms/inf (SM8750)
  • acc: top1/5 ~= 70%/85% in ImageNet (https://www.kaggle.com/datasets/ifigotin/imagenetmini-1000)
• add pass to handle torch.roll
• properly annotate aten.masked_fill
• fix embedding UT in quantization flow

Test plan

python examples/qualcomm/oss_scripts/swin_transformer.py -b build-android -H $HOST -s $DEVICE -m $MODEL --dataset ${PATH_TO_DATASET}

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11099

• 📄 Preview Python docs built from this PR

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

✅ You can merge normally! (2 Unrelated Failures)

As of commit 4e6e746 with merge base a6e2961 ():

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 / linux / linux-job (gh) (trunk failure)
  examples/models/llama/tests/test_export_llama_lib.py::ExportLlamaLibTest::test_has_expected_ops_and_op_counts
• pull / unittest-editable / linux / linux-job (gh) (trunk failure)
  examples/models/llama/tests/test_export_llama_lib.py::ExportLlamaLibTest::test_has_expected_ops_and_op_counts

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

[DannyYuyang-quic]

@pytorchbot label "release notes: qualcomm"

[DannyYuyang-quic]

Hi @cccclai,
This PR is for GA model - Swin Transformer
It also includes some repro for newly added passes:

• test_qnn_backend_adaptive_avg_pool1d
• test_qnn_backend_masked_fill
• test_qnn_backend_roll

These tests are located under:
backends/qualcomm/tests/test_qnn_delegate.py

Please have a look, thanks!!