[WIP]Qualcomm AI Engine Direct - Mimi Encoder/Decoder Enablement

[winskuo-quic]

Please Do NOT review this PR

This PR can execute Mimi Encoder/Decoder.
This PR is still WIP and still working on removing passes and resolving some edge cases.

Scripts to Execute

Setup scripts that I copied from #8753

• examples/qualcomm/oss_scripts/moshi/install_requirments.sh

To export and run the model

• python examples/qualcomm/oss_scripts/moshi/mimi.py -b build-android -s DEVICE -m MODEL

Summary

• Support Mimi Encoder
• Support Mimi Decoder
• Support OP:
  • exp
  • expm1
  • elu
  • transpose conv1d
  • bitwise_and
  • scalar_tensor
  • stack

Statistics

Inference Speed

SM8650 1 inference
QNN HTP Encoder: 224ms
QNN HTP Decoder: 75ms

Model Size

Encoder PTE: ~60mb
Decoder PTE: ~60mb

Test Plan

• UT

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/9349

• 📄 Preview Python docs built from this PR

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

❌ 1 New Failure

As of commit b681909 with merge base 366ad75 ():

NEW FAILURE - The following job has failed:

• Check Labels / Check labels (gh)
  RuntimeError: Error checking labels: PR does not have required labels

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

[github-actions]

This PR needs a release notes: label

If your changes are user facing and intended to be a part of release notes, please use a label starting with release notes:.

If not, please add the topic: not user facing label.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "topic: not user facing"

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 @billmguo, @iseeyuan, @cccclai,

The following PR is based on top of #8753.
This PR is currently able to export and execute Mimi Encoder/Decoder.
However, please do NOT review this PR since it is still WIP. I am still working on some edge cases and reduce some passes.
I have used the single sample to verify if it is working by listening the output .wav file. There are some background noises, however, I can still hear the person speaking and understanding what the person is saying.
I have included the execution instructions and some of the stats in the Summary Section above.
Please let me know if you run into any issues.
Thanks

[facebook-github-bot]

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