File Limit Request: leip-pylre-onnx - 450MiB

[AchachayJurado]

Project URL

https://pypi.org/project/leip-pylre-onnx/

Does this project already exist?

• Yes

New Limit

450 MiB

Update issue title

• I have updated the title.

Which indexes

PyPI

About the project

leip-pylre-onnx is a runtime engine for onnx models optimized/encrypted/watermarked for a target with leip-forge.
Most of it is written in C++, with a Python API on top. pip install leip-pylre-onnx allows any Python user to easily run inference on specific target devices, this means we are packaging target dependent compiled libraries to be accessed at runtime which makes a larger binary.

• Originally wheels were over 1 GiB as some of them are requiring cuda libraries such as cudnn, but now down to ~450 MiB for x86_64and ~120MiB for aarch64.
• Future plans to reduce size by deprecating cuda11 support

Reasons for the request

LEIP Deploy was vastly used by C++ users (hosted in our apt registry) but now more and more python API requests are coming and we want to enable the usage of models optimized with leip-forge.
Two wheels (one for each x86_64 & aarch64) for each of the supported python versions (cp3.8, 3.9, 3.10, 3.11 and 3.12) will be published for each release every -roughly- month.

Code of Conduct

• I agree to follow the PSF Code of Conduct

[ewdurbin]

Limit of 450 granted