vargonis

How do you work around this problem? I've resorted to using itikz to generate an SVG and then just drop the cell and embed the result (as an image). I...

That's right, one would need a reversible jump MCMC. But how about SMC? In the tutorial on [infinite mixture models](https://turing.ml/dev/tutorials/6-infinitemixturemodel/) a stochastic dimension model is run with SMC. However, for...

On a related note, since by the nature of the format the output of `pdPageExtractText` is not fully determined, it would be useful to: 1. Have access to character level...

I've been pondering this and I'm not convinced. If the element type was intended to be the type of the parameters (of which there can be more than one!), I...

By the way, I just realized that there's the `ValueSupport` type which can be either `Int` or `Float64`. So, here I'm really colliding with the limitations of Distributions' design... :/

Looking forward to see the new design! I do hope it will allow for distributions with arbitrary support. Having just `Int` and `Float64` is annoyingly limiting.

Any update on this? Also, I'd really appreciate if the potential implementation above is turned into a complete example to build upon (for users like me who know nothing about...

Didn't try with an A100, but the `--cuda` option for the llamacpp server works, thanks!

Any updates here? Channel dimension is naturally `N-1` in space-like models, but for sequential or time-like models it is natural to have it first... Or more generally `N-2`, if one...

Hey, I'm wondering about that too. If I've understood things correctly, the internal representation of `gmsh` is not really a mesh, so it's perhaps not possible to directly "import" meshes...