search

GICodeWarrior / fir

Foxhole Inventory Report
MIT License
21 stars 9 forks source link

Question Using Pixi for Dependency Management #14

Closed gregorywaynepower closed 3 months ago

gregorywaynepower commented 3 months ago

Hey @GICodeWarrior, thanks for creating this.

Would you be open to a pull request with rolling all of the required dependencies into a pixi.toml file? It'll help with managing your dependencies from npm, PyPI, and conda repos if you use those.

It can also resolve dependencies across build platforms as well.

GICodeWarrior commented 3 months ago

Thanks for reaching out.

I'm open to build improvements.

What are the advantages of pixi versus the existing dependency management?

gregorywaynepower commented 3 months ago

You get faster resolving of dependencies and you're able to add tasks (like running or installing packages via npm) that users can execute via the pixi shell. It's pretty similar to the conda package manager, it'll also let you install from PyPI.

The only issue that I see is so far is that sed isn't available for Windows users.

GICodeWarrior commented 3 months ago

I'm not understanding the advantages.

Is the goal to run the build natively on Windows?

My understanding is that tensorflow doesn't support GPU acceleration natively on Windows.

gregorywaynepower commented 3 months ago

My goal was to roll all of the required dependencies under a package manager that isn't tied to the OS or language ecosystem. I wasn't sure what environments you were developing and running this on which was why I mentioned Windows.

I was just trying to help folks to be able to run the build.sh script without folks having to navigate the complexity of installing package managers to install software from different ecosystems.