Qualcomm AI Engine Direct - VIT Optimization

[winskuo-quic]

Summary

QNN doesn't have much 5D permute optimization, which causes Vit running slower than CPU.

Switched pattern from unsqueeze->permute 5d->squeeze to permute 4d.

Improvements: 150ms/inference -> 4.2ms/inference.

Test plan

Pass Vit UT

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15696

• 📄 Preview Python docs built from this PR

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

❌ 4 New Failures, 1 Cancelled Job, 4 Unrelated Failures

As of commit da35397 with merge base d07a49a ():

NEW FAILURES - The following jobs have failed:

• pull / unittest / linux / linux-job (gh)
  backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_int8_static_quant_recipe
• pull / unittest / macos / macos-job (gh)
  backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_int8_static_quant_recipe
• pull / unittest-editable / linux / linux-job (gh)
  backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_int8_static_quant_recipe
• pull / unittest-editable / macos / macos-job (gh)
  backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_int8_static_quant_recipe

CANCELLED JOB - The following job was cancelled. Please retry:

• pull / test-openvino-linux / linux-job (gh)

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

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

• pull / test-binary-size-linux-gcc / linux-job (gh) (trunk failure)
  ##[error]The operation was canceled.
• pull / test-setup-linux-gcc / linux-job (gh) (trunk failure)
  ##[error]The operation was canceled.
• pull / unittest / windows / windows-job (gh) (trunk failure)
  backends/xnnpack/test/ops/test_conv1d.py::TestConv1d::test_qs8_conv1d_batchnorm_seq
• pull / unittest-editable / windows / windows-job (gh) (trunk failure)
  backends/xnnpack/test/ops/test_conv1d.py::TestConv1d::test_qs8_conv1d_batchnorm_seq

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

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

[winskuo-quic]

Hi @cccclai,
This PR is to address the issue you have reported about QNN running slower than CPU for Vit.
I have tested SM8750 with this PR, and inference speed improves from 140ms/inference -> 4.1ms/inference, which should be around 35 times faster.
I have made a source level transformation on the Vit model.
The bottleneck is caused by the 5D permute pattern here: https://github.com/pytorch/pytorch/blob/3cfbf98ea9d937d23f3700168b22706c957308ce/torch/nn/functional.py#L5825.
I did source level transformation because this pattern is a bit specific and it is hard to generalize.
Please have a look and let me know if you have any questions.
Thanks

[cccclai]

It can be a pass ideally, is it correct?