Cross attention KV cache support in optimum-executorch
Open
@larryliu0820 is already working on this.
Since Dec 9, 2025.
- Dominant language
- Python
- Stars
- 5k
- Forks
- 1.2k
- Avg merge
- 2d 10h
- Merged PRs (30d)
- 581
Description
🚀 The feature, motivation and pitch
Module swap the cross attention in common seq-2-seq models (e.g., Whisper) so that it's exportable in ExecuTorch. This is important to allow CUDA and Metal backend to leverage cross attention KV cache and run efficiently.
Alternatives
No response
Additional context
No response
RFC (Optional)
No response
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.
Assessment
This issue has not been assessed yet.