Jeremie Desgagne-Bouchard

@DhairyaLGandhi Regarding `JuliaBench.jl` and https://speed.fluxml.ai/, I'd be happy to take on the the duty of bringing the benchmarks on a Stipple.jl dashboard. I think this would help to bring a...

I'm afraid the proposed change results in hiding an error that arise from inconsistent eltype of input (X) and cell's initial state's (`state0`) when getting the gradients on GPU. For...

Maybe there's a necessary compromise here since if the Recur's state isn't parametrized (in the struct definition), then the promotion will happen successfully on the state, even on GPU. However,...

Regarding embeddings, although I haven't dealt with the potential caveats from weight norm and such, are there challenges I'm overlooking compared to doing a fairly trivial matrix indexing? Example: ```julia...

> the trivial indexing triggers scalar indexing on GPU arrays I recently used this approach for embedding and can confirm good performance on GPU, maybe there's been recent improvement in...

GIven the lack of both empirical and theoretical support for such feature, I'd close for now as I don't see a proper motivation to implement in a foreseeable future. The...

Given the improvements to memory footprint throughout various releases since the issue was open and the added support for categorical data through the new Tables API since v0.16, I'd close...

Implemented in v0.15

Following work for this tutorial on ranking: https://evovest.github.io/EvoTrees.jl/dev/tutorials/ranking-LTRC/, as well as considering the reference material from CatBoost ranking benchmarks: https://github.com/catboost/benchmarks/blob/master/ranking/Readme.md#4-results, my current take would be that adding a support for...

I think the `watch` config option was the wrong diagnosis. It actually looks like it's `format_julia_builds = true` that results in MD parsing error, sorry about that.