Seth Axen

With Enzyme, I think one could use `inactive`: https://enzyme.mit.edu/index.fcgi/julia/stable/generated/custom_rule/#Marking-functions-inactive. For source-to-source I think one would want it to also work for Tapir if possible. It should ideally impact all of...

> Is there something similar in `ArviZ.jl` to [`from_cmdstan()`](https://python.arviz.org/en/stable/api/generated/arviz.from_cmdstan.html) in the python library? If not, is there a specific way to plug into `InferenceData` that you would recommend? Integration with...

> Perfect, that looks like what I need. I quite like working with cmdstan directly on the command line so I'll probably try to extract the csv reading and conversion...

> Thanks, I'll take a closer look at `StanIO.jl`. At first sight, it seems to have quite a few dependencies that `ArviZ.jl` doesn't have. Would you be willing to take...

First the fingerprints are generated using a large number of bits, where a bit is 'on' if the 3D environment of an atom hashes to that bit (for details see...

Related to #2306 and proposes an alternative to the solution proposed there.

> Could we just make it a proper dependency on older Julia versions? It's still possible to make it a weak dependency on Julia 1.9 at the same time. Ah,...

> > For Chains the corresponding methods live in Turing proper. That might be cleaner. > > There have been long discussions (and even issues IIRC, maybe even in DynamicPPL?)...

> Is there a particular reason why we don't just add it to Turing for now? If the code lived in Turing it would entirely be type piracy. Other than...

> I guess short-term basically anything goes - it can just take a long time to move away and improve a suboptimal but somewhat working setup, in my experience. Okay,...