Carlos Parada
Carlos Parada
> IMO ideally we should just throw better error messages for common errors. I think users shouldn't receive long and undecipherable stack traces if possible in the first place. An...
> Often the information in the stacktrace is useful for developers, so I don't think it should be hidden. I'd like to offer this as a counterargument that long error...
@yebai @cpfiffer did I misunderstand what we were talking about yesterday?
Do we have code that checks the compile times of models? How long has Zygote compilation been taking this long?
> > How long has Zygote compilation been taking this long? > > Since Julia 1.6 afaik. As in, the 1.6 update caused it? Is the compilation faster on 1.5?
@ToucheSir @Keno Any progress on this?
It might be a good idea to check on MeasureTheory.jl about this. Since it's optimized to work specifically with PPLs, it might have some version of this already in place.
@devmotion any updates on this?
> ``` > Warning: 666 proposed samples were rejected due to numerical errors (80% of total). > Too many numerical errors may reflect problems in the model and/or sampler (increasing...
> IMO we should just do what's already done in [TuringLang/DynamicPPL.jl#275](https://github.com/TuringLang/DynamicPPL.jl/pull/275). > > The reason why I'm in favour of that instead of something like a wrapper-type will require changing...