Deep learning in Julia

[cpfiffer]

A very thorough analysis of nice-to-haves:

State of machine learning in Julia

I’ll offer a perspective from someone who (as a conscious choice) primarily uses Python over Julia. I work with, and maintain libraries for, all of PyTorch, JAX, and Julia. For context my answers will draw some parallels between: JAX, with Equinox for neural networks; Julia, with Flux for neural networks; as these actually feel remarkably similar. JAX and Julia are both based around jit-compilers; both ubiquitously perform program transforms via homoiconicity. Equinox and Flux both build mod…