I get Segmentation fault when generating samples from normal distribution

[matus.s...@privatdemail.net]

I want to generate normal samples with Stan. First I used this code:

import pystan
model = """
data{
    real dmx;
    real dvx;
}
model{
}
generated quantities {
    real rt[50];
    for (i in 1:50){
        rt[i]<-normal_rng(dmx,dvx);
    }
}
"""
sm = pystan.StanModel(model_code=model)
fit=sm.sampling(data={'dmx':0,'dvx':1},
        iter=1,chains=1,warmup=0,seed=1)

And I get:

RuntimeError: Must use algorithm="Fixed_param" for model that has no parameters.

Ok. So now I add the option

fit=sm.sampling(data={'dmx':0,'dvx':1},
        iter=1,chains=1,warmup=0,seed=1,algorithm='Fixed_param')

and I get

Segmentation fault

Looks like a bug to me.

I'm on Pystan 2.6.3, 32-bit Linux.

[Seth Flaxman]

FWIW, I just tried it with RStan (2.6.0 / 64-bit Windows) and all is fine...

> --
> You received this message because you are subscribed to the Google Groups
> "Stan users mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to stan-users+...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Bob Carpenter]

Matus --- which version of PyStan are you using?

We want to take seg fault issues very seriously due to their
tendency to crash their host environment.

Thanks for testing, Seth. Maybe Allen will have an idea
what's going on. This should probably be turned into an
issue.

Might this be a problem with 32-bit linux or PyStan?

- Bob

[matus.s...@privatdemail.net]

Today (before the segfault) I  upgraded to pystan 2.6.3 with

pip install pystan --upgrade

Maybe,  some dependency didn't get upgraded properly?

I investigated further with other models and the segfault occurs whenever I set algorithm='Fixed_param'

I will check the code on my other machines tomorrow.

m.

[Allen B. Riddell]

This isn't your fault. I'm pretty sure there's a bug in PyStan for Fixed
Param mode. I'll create the issue shortly.