Tor Erlend Fjelde
Tor Erlend Fjelde
> Personally, I think it's clearer this way because it forces the user to recognise that the inference and prediction models should have the same structure. That's strictly not necessary...
All sensible arguments 👍 In the end, it's just a question of convenience vs. "don't allow incorrectness". Personally I've found the convenience very, well, convenient 🙃 so I'm obviously biased....
> I might still be open to enabling no-prefix as an option, but making no-prefix the default is not something I would like to be responsible for implementing. Definitively agree...
> Oh, one more thing, should it be 0.9.0 rather than 0.8.4? Since we're introducing new functions. It's a bit weird, but basically, at least to my understanding, when we...
> This relies on DynamicPPL gaining a method for link!! and invlink!! that takes a VarName rather than a sampler or a space. Is this meant to link only that...
This was the debug script I was using: ```julia using DynamicPPL, ReverseDiff, LogDensityProblems, LogDensityProblemsAD, Distributions s_global = nothing s_global_i = nothing @model function demo_assume_index_observe( x=[1.5, 2.0], ::Type{TV}=Vector{Float64} ) where {TV}...
Adding my voice to this issue. I mainly use a text-editor to take notes, but sometimes I need to draw something, in which case I use Xournal. Right now I...
@devmotion @yebai you might have some to add here
It was attempted in #589 but got downvoted for now, as it was deemed undesirable to add more macros (see discussion in that PR)
Indeed. I've had to use similar approaches in the past due to the computational complexity of the model. @yebai worth revisiting #589 ?