Ryan Curtin

@mlpack-jenkins test this please

This one's not ready for merge yet---I still have a Julia memory issue to debug. I'm fighting with that now...

I converted this back to a draft just so it doesn't get accidentally merged. I think I almost have the fix...

Okay, there was some significant overhaul of the Julia bindings memory handling necessary. Let me detail the changes... * First, incorporate the `GC.@preserve` changes from https://github.com/mlpack/mlpack.jl/pull/3; thanks @kpamnany! This prevents...

Haha! It works! Anyway, I'll merge this tomorrow since it's already approved. If anyone has any comments, I'm happy to address them first. :+1:

On the mlpack side we don't have any ratelimiting support set up. The system is just some 1U thing I threw in a rack somewhere and manually administrate; no nice...

Same issue and output here on a Lenovo Yoga 7i (same processor), running Debian sid. Happy to try any suggestions or provide any more output if it could be helpful....

I also commented out a test in 69ea3599 that suffers from the same issue.

> I don't think I get it completely. Though there are multiple calls to Evaluate() and Gradient() in a single step but we always update iterate just once during the...

Ok, great, now let's talk about how to solve it. :) >> Only recompute impostors when Shuffle() is called. > > I agree this could have been a good option,...