Tamas K. Papp
Tamas K. Papp
But isn't that what ```julia @view a[begin+2:end-3] ``` is for? Please don't get me wrong, I found this package very useful in the past, but I am genuinely curious what...
@mcreel, here is the PR, currently a draft. I just moved everything to a single file, but lack the context about this; if you could contribute explanations that would be...
Thanks, I will incorporate the explanation, and do some optimization when time permits. > The comparison is not exactly fair The relevant comparison is *effective* samples (ESS) / time.
Will look into it today.
I think it is an issue with AD producing NaNs or similar. Ideally one would investigate, but I added a quick fix in the form of a wrapper, I made...
Also, you can investigate corner cases of log densities with `LogDensityProblems.stresstest`.
That's fine, keep it open as long as you like and feel free to ask if there is anything I can help with.
You should try `stresstest` with the `∇P` (plain vanilla `ADgradient`, no `LogDensityRejectErrors`). It should indeed return the vector of locations which are not OK (eg evaluate with `NaN`s). Plotting is...
If I understand correctly, there are two issues: 1. the script errors, 2. the results are different. Did these happen after the recent (minor) updates to LogDensityProblems and DynamicHMC, or...
Are you using the latest LogDensityProblems and DynamicHMC? I could not reproduce the errors above. Also, I am not sure how I would detect the discrepancy, should I look at...