Feras Saad
Feras Saad
> 2\. plus one extra fix for stability: when all scores passed to `log-categorical` are negative infinity, we won't error. In the case that all scores are negative infinity it...
> 1. Should the implementation of importance resampling use log-categorical? I agree with you that the answer is yes. However, the behavior of log-categorical should be reasoned about independently of...
Related https://github.com/probcomp/sppl/issues/92, since min/max can be of the form min(X1, X2, ..., Xn).
Thank you @MikeTam1021, we will consider relaxing the versions to major only.
@MikeTam1021 Which versions ultimately worked for you?
Note that sympy hangs only if `x
Generally it seems we need to migrate the solver in `transforms.py` to use python floats, since sympy has unreliable performance for edge cases which will be too complicated to try...
Try to use the reverse compiler from the compiled SPE `m` to an SPPL program source with equivalent semantics and then append it to the original program. Below is a...
> consider changing Gen.initialize to Gen.trace I think `Gen.trace` can be confusing if we are going to use it as both a noun ad a verb. In Gen we commonly...
> Any chance GenFluxOptizers.jl can be merged with Gen.jl? @SamWitty You could install `GenFluxOptimizers.jl` into your local installation by writing the following into the Julia REPL ``` $ julia julia>...