search

FluxML / Optimisers.jl

Optimisers.jl defines many standard optimisers and utilities for learning loops.
https://fluxml.ai/Optimisers.jl
MIT License
75 stars 22 forks source link

Don't load Yota at all #166

Closed mcabbott closed 8 months ago

mcabbott commented 8 months ago

Paging @dfdx -- what's the status of Yota on latest Julia? Should we hold out for a fix here?

dfdx commented 8 months ago

It can be fixed (see dfdx/Umlaut.jl#55 for the main required change), but I'm not sure Optimisers should depend on Yota at the moment. Yota is on hold until better times. Specifically, until more wide ecosystem for deep learning is available, so quite indefinite time. Till then, Optimisers may benefit more from a shorter list of more actively developed packages.

Yet, if you believe Yota will bring some unique benefits, please ping me and I will gladly update the package.

mcabbott commented 8 months ago

Certainly no pressure to work harder for the sake of this package! Just thought I shouldn't cut this without checking in.

We all had hopes that several AD packages might be viable at the same time, but remain stuck on Zygote...