Qualcomm AI Engine Direct - wav2letter e2e example

[haowhsu-quic]

Summary

• enable wav2letter e2e example

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/5924

• 📄 Preview Python docs built from this PR

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

✅ No Failures

As of commit 9aceb17 with merge base 8fc3e20 ():
💚 Looks good so far! There are no failures yet. 💚

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

[haowhsu-quic]

Hi @cccclai, please have a look on this, thank you.
I wonder if we could land #5543? It will be helpful for us to continue enabling more oss models.

[cccclai]

Hi @cccclai, please have a look on this, thank you. I wonder if we could land #5543? It will be helpful for us to continue enabling more oss models.

Yes for sure. Sorry for getting late on this, and thank you for enabling more models!

[guangy10]

When this model is enabled can you please update the CI to cover this model?

1. Update the test script: https://github.com/pytorch/executorch/blob/main/.ci/scripts/test_model.sh#L147
2. Update the nightly benchmark workflow to see the perf: https://github.com/pytorch/executorch/blob/main/.github/workflows/android-perf.yml#L87

[haowhsu-quic]

Hi @guangy10, this model requires pretrained_weight as an input argument which requires getting checkpoint in advance.
I think the downloading process and extra argument may change the script logic now. Do you have suggestions?

[cccclai]

If there are extra efforts to add them to CI, maybe we can have a follow up on that. Thoughts? @guangy10

[facebook-github-bot]

@cccclai has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.