Missing output variable path__ with brms and pathfinder algorithm

Modeling
,
1

I have tried to experiment with the pathfinder algorithm with brms (2.23.0) and the cmdstan (2.38.0) backend, and run into
Error: Can’t find the following variable(s) in the output: path__

Here is a minimalist example :

library(brms)
library(cmdstanr)

set.seed(1)

N <- 100
dat <- data.frame(
x = rnorm(N)
)
dat$y <- 1 + 2 * dat$x + rnorm(N, sd = 0.5)
m_frm <- bf(y ~ 1 + x)

m_prior <- c(
prior(normal(0, 5), class = "Intercept"),
prior(normal(0, 5), class = "b"),
prior(exponential(1), class = "sigma")
)

fit_pf <- brm(
formula = m_frm,
data = dat,
family = gaussian(),
prior = m_prior,
backend = "cmdstanr",
algorithm = "pathfinder",
seed = 123
)

And here is the putput:

Model executable is up to date!
Start samplingPath [1] :Initial log joint density = -245.677287 
Path [1] : Iter      log prob        ||dx||      ||grad||     alpha      alpha0      # evals       ELBO    Best ELBO        Notes  
              8      -7.416e+01      2.190e-03   4.058e-03    1.000e+00  1.000e+00       277 -7.992e+01 -7.990e+01                   
Path [1] :Best Iter: [8] ELBO (-79.898117) evaluations: (277) 
Path [2] :Initial log joint density = -256.288260 
Path [2] : Iter      log prob        ||dx||      ||grad||     alpha      alpha0      # evals       ELBO    Best ELBO        Notes  
              9      -7.416e+01      1.305e-04   3.638e-03    1.000e+00  1.000e+00       372 -7.992e+01 -8.005e+01                   
Path [2] :Best Iter: [8] ELBO (-79.921340) evaluations: (372) 
Path [3] :Initial log joint density = -322.887346 
Path [3] : Iter      log prob        ||dx||      ||grad||     alpha      alpha0      # evals       ELBO    Best ELBO        Notes  
             17      -7.416e+01      9.268e-05   3.530e-04    1.000e+00  1.000e+00       643 -8.004e+01 -8.003e+01                   
Path [3] :Best Iter: [17] ELBO (-80.026751) evaluations: (643) 
Path [4] :Initial log joint density = -1581.358110 
Path [4] : Iter      log prob        ||dx||      ||grad||     alpha      alpha0      # evals       ELBO    Best ELBO        Notes  
             14      -7.416e+01      7.399e-05   3.752e-03    1.000e+00  1.000e+00       566 -7.997e+01 -8.003e+01                   
Path [4] :Best Iter: [12] ELBO (-79.972363) evaluations: (566) 
Finished in  0.3 seconds.


Error: Can't find the following variable(s) in the output: path__
2

I think this was fixed by

but there hasn’t been a new brms release yet. If you install brms from GitHub I think it should work. Let me know if you try that and still get the error.

3

Indeed, upgrading to brms 2.23.1 did the trick. Thanks!

install.packages("pak")
pak::pkg_install("paul-buerkner/brms")