All GA models are supported by CoreML and have similar performance to using CoreML via torch.export
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 5k
- Forks
- 1.2k
- Avg merge
- 2d 10h
- Merged PRs (30d)
- 581
Description
GA model list: https://docs.google.com/spreadsheets/d/18HbvYhnhPfkHih0oRPGvUloRVyZpTwKtZxaJrfkD_9s/edit?gid=0#gid=0
We should be able to follow this flow verbatim and get good perf: https://github.com/pytorch/executorch/blob/main/docs/source/backends-coreml.md
We want to see better performance in CoreML than XNNPACK for models that can take advantage of the GPU and ANE.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the GA model list in the linked spreadsheet and read docs/source/backends-coreml.md to follow the documented CoreML flow. Compare the resulting CoreML performance with XNNPACK for models that can use the GPU and ANE. Done means all listed GA models are supported and show similar or better performance through CoreML than via torch.export.
Written by the indexing model from the issue text.
Assessment
- Domain
- machine-learning, mobile-dev, performance
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100