Linux-Compatible Pre-built(.whl) PyPI Distribution for Library

NVIDIA / MinkowskiEngine

Minkowski Engine is an auto-diff neural network library for high-dimensional sparse tensors

https://nvidia.github.io/MinkowskiEngine

Other

2.43k stars 360 forks source link

Linux-Compatible Pre-built(.whl) PyPI Distribution for Library #561

Open Mayavan opened 1 year ago

Mayavan commented 1 year ago

I'm an avid user of your GitHub repository's library and appreciate the value it brings to my projects. I kindly request the possibility of uploading a Linux-specific Wheel (.whl) distribution package to PyPI.

Benefits:

Simplified Integration: A Linux-compatible package would streamline installation for Linux users like me.
Broader User Base: Accessible via common package management tools, attracting more Linux developers.
Improved Collaboration: Compatibility ensures seamless integration into various projects.

Thank you for considering this request. Looking forward to your response.

yangyangyang127 commented 1 year ago

good suggestion

lfxx commented 1 year ago

The build is really time consuming.

francotheengineer commented 1 year ago

@lfxx Very difficult to get a correct environment even with docker! I wonder If NVIDIA plans to continue to update this library with newer CUDA versions.

lfxx commented 1 year ago

@lfxx Very difficult to get a correct environment even with docker! I wonder If NVIDIA plans to continue to update this library with newer CUDA versions.

What's your environment?i can build this with cuda11.3 though it is time consuming.

francotheengineer commented 1 year ago

@lfxx Very difficult to get a correct environment even with docker! I wonder If NVIDIA plans to continue to update this library with newer CUDA versions.

What's your environment?i can build this with cuda11.3 though it is time consuming.

I can get a Docker working with: CUDA 11.8 Ubuntu 20.0x Python 3.8 PyTorch 2.0.1 Based off https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-22-12.html#rel-22-12

I have tried to get ME working with Cuda 12.x, It does not work. 11.8 is the highest you can go.