Use `numba` to accelerate numerical CPU code

[aseembits93]

Search before asking

• I have searched the Inference issues and found no similar feature requests.

Description

Hi all, I was wondering if there's a possibility to use numba to accelerate CPU code in this repository? With minor refactors, we can see massive performance boosts by adding numba's @jit decorator. Hope to hear your thoughts. Best,

Use case

No response

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[Yadav108]

Hey team,

I'd like to tackle this. I've profiled the codebase and can implement Numba acceleration for NMS/IoU operations in postprocess.py with expected 3-5x CPU speedups.

Approach:

• JIT-compiled NMS and IoU functions
• Optional dependency with NumPy fallback
• Full test coverage for numerical accuracy

I have ML optimization experience (quantization, on-device inference) and can deliver measurable performance gains. Can I be assigned?

@aseembits93 - Let me know if you're already working on this and want to collaborate.