ultralytics / ultralytics

Ultralytics YOLO11 πŸš€
https://docs.ultralytics.com
GNU Affero General Public License v3.0
36.33k stars 7k forks source link

Apple M4 NeuralEngine support for MacMini, Mac, iPhone, iPad #19074

Open sakrig opened 5 days ago

sakrig commented 5 days ago

Search before asking

Description

Hello, what are the plans for supporting the Apple M4 NeuralEngine across Apple Products for the App and the CLI?

At least the cli is a good tool to run across Apple M4 products, even iPhone has a terminal app for the CLI to run in :) The app - not needed so much, except for the portable devices like the iPhone and iPad.

[app cli]. MacMini [app cli]. Mac
[app cli]. iPad [app cli] iPhone ... other

Use case

At least the cli is a good tool to run across Apple M4 products, even iPhone has a terminal app for the CLI to run in :) The app - not needed so much, except for the portable devices like the iPhone and iPad.

Additional

The NeuralEngine API seems well designed to support all Apple M4 devices, across iPhone, iPad, MacMini, Mac. Maybe this is easier than one might imagine - Probably the NeuralEngine code can be built for any Apple platform from XCODE - I have not tried.

Are you willing to submit a PR?

UltralyticsAssistant commented 5 days ago

πŸ‘‹ Hello @sakrig, thank you for sharing this feature request and for your interest in Ultralytics πŸš€! Supporting the Apple M4 NeuralEngine across all Apple products is an exciting proposal, and we appreciate your input.

If you're new to Ultralytics, we recommend visiting the Docs where you can explore Python and CLI usage examples and see if our current features meet your requirements.

To address your suggestion, Ultralytics engineers will review your request and review the feasibility of integrating the Apple M4 NeuralEngine capabilities. In the meantime, you can explore more about Ultralytics' current system safeguards, usage options, and environments:

Upgrade

To ensure compatibility with the newest features and fixes, upgrade to the latest ultralytics package. Install it with:


pip install -Umaxcdn
glenn-jocher commented 4 days ago

@sakrig thanks for your question! Ultralytics YOLO models already support Apple Neural Engine acceleration via CoreML exports, which are compatible with M-series chips including M4. You can export models to CoreML format using model.export(format='coreml') and they'll leverage available Neural Engine hardware automatically. For iOS deployment, our existing HUB app demonstrates ANE integration - check our CoreML and iOS app docs for implementation details. Community contributions to enhance Apple platform support are welcome via PRs!