Will Tebbutt
Will Tebbutt
Does this just require a patch bump?
> Speed of tracing, i.e. how fast trace(f, args...) itself works. This is what I meant in the first comment. The last time I profiled the tracer, most time was...
@dfdx could we please make a release? Not sure if I have permission to do it / if you have some reason for holding off making a release
> Now, whether that is indeed the correct approach, is unclear 🙃 For example, I think maybe just 10 arguments is a bit "too few", while I'd be hard pressed...
> My general stance is that NamedTuples, Tuples etc. are just not the right tool for larger collections I agree with this in general. However, I'm unaware of another approach...
> Naive question, what is if @generated ... pattern? See [here](https://docs.julialang.org/en/v1/manual/metaprogramming/#Optionally-generated-functions) :)
Hmmmm. Do you have a particular use case for `cov(f, x, y)` with a BLR? I don't have any super strong objections, I'm more just intriuged as to where it's...
Hmm yeah, why no. I can't see what the harm would be -- lets do it.
This would indeed be something that would be good to have. We don't have it in this package at this point, but I would definitely be open to a proposal
> Distances.jl doesn't work with CuArrays Quite different performance of pairwise on CPU vs GPU JuliaStats/Distances.jl#143 Since we define our own version of `pairwise` etc in KernelFunctions.jl, we could (for...