Brms/rstan on R 4.0

torkar · June 10, 2020, 6:27pm

Kill RStudio.
Make sure the file .Rprofile in your home directory contains exactly this,

if (Sys.getenv("RSTUDIO") == "1" && !nzchar(Sys.getenv("RSTUDIO_TERM")) && 
    Sys.info()["sysname"] == "Darwin") {
  parallel:::setDefaultClusterOptions(setup_strategy = "sequential")
}

Start RStudio

bh7 · June 10, 2020, 6:35pm

torkar:

if (Sys.getenv("RSTUDIO") == "1" && !nzchar(Sys.getenv("RSTUDIO_TERM")) && 
    Sys.info()["sysname"] == "Darwin") {
  parallel:::setDefaultClusterOptions(setup_strategy = "sequential")
}

Done, done, and done. Same error. Just shoot me now.

torkar · June 10, 2020, 6:39pm

Well, then we better wait for Ben before I get a heart attack ;)

bh7 · June 10, 2020, 6:41pm

lol. Many thanks again for your kindness and help. Will Ben be aware of this issue?

Stephen_Martin · June 10, 2020, 8:15pm

Does it run fine if you run it without Rstudio? Say, just in the terminal?

MauritsM · June 10, 2020, 8:53pm

One more thing you can try (for troubleshooting): restart RStudio and run this line:

parallel:::setDefaultClusterOptions(setup_strategy = "sequential")

Then start your rstan or brms code.

bh7 · June 10, 2020, 9:46pm

I just ran it both in terminal and RStudio and it ran both times. I have no idea why. That said, I used a different, simpler model: BRM1 <- brm(weight ~ Diet, data = ChickWeight). I don’t want to get too excited, so I’ll try running the more complex model now.

bh7 · June 10, 2020, 9:56pm

The more complex model did not run on either RStudio or R.
Error message from R: Compiling the C++ model
Start sampling
Error in makePSOCKcluster(names = spec, …) :
Cluster setup failed. 2 of 2 workers failed to connect.

bh7 · June 10, 2020, 9:57pm

This is the data for the complex model:

HtWtDataGenerator <- function(nSubj, rndsd = NULL, maleProb = 0.50) {
  # Random height, weight generator for males and females. Uses parameters from
  # Brainard, J. & Burmaster, D. E. (1992). Bivariate distributions for height and
  # weight of men and women in the United States. Risk Analysis, 12(2), 267-275.
  # Kruschke, J. K. (2011). Doing Bayesian data analysis:
  # A Tutorial with R and BUGS. Academic Press / Elsevier.
  # Kruschke, J. K. (2014). Doing Bayesian data analysis, 2nd Edition:
  # A Tutorial with R, JAGS and Stan. Academic Press / Elsevier.
  
  # require(MASS)
  
  # Specify parameters of multivariate normal (MVN) distributions.
  # Men:
  HtMmu   <- 69.18
  HtMsd   <- 2.87
  lnWtMmu <- 5.14
  lnWtMsd <- 0.17
  Mrho    <- 0.42
  Mmean   <- c(HtMmu, lnWtMmu)
  Msigma  <- matrix(c(HtMsd^2, Mrho * HtMsd * lnWtMsd,
                      Mrho * HtMsd * lnWtMsd, lnWtMsd^2), nrow = 2)
  # Women cluster 1:
  HtFmu1   <- 63.11
  HtFsd1   <- 2.76
  lnWtFmu1 <- 5.06
  lnWtFsd1 <- 0.24
  Frho1    <- 0.41
  prop1    <- 0.46
  Fmean1   <- c(HtFmu1, lnWtFmu1)
  Fsigma1  <- matrix(c(HtFsd1^2, Frho1 * HtFsd1 * lnWtFsd1,
                       Frho1 * HtFsd1 * lnWtFsd1, lnWtFsd1^2), nrow = 2)
  # Women cluster 2:
  HtFmu2   <- 64.36
  HtFsd2   <- 2.49
  lnWtFmu2 <- 4.86
  lnWtFsd2 <- 0.14
  Frho2    <- 0.44
  prop2    <- 1 - prop1
  Fmean2   <- c(HtFmu2, lnWtFmu2)
  Fsigma2  <- matrix(c(HtFsd2^2, Frho2 * HtFsd2 * lnWtFsd2,
                       Frho2 * HtFsd2 * lnWtFsd2, lnWtFsd2^2), nrow = 2)
  
  # Randomly generate data values from those MVN distributions.
  if (!is.null(rndsd)) {set.seed(rndsd)}
  datamatrix <- matrix(0, nrow = nSubj, ncol = 3)
  colnames(datamatrix) <- c("male", "height", "weight")
  maleval <- 1; femaleval <- 0 # arbitrary coding values
  for (i in 1:nSubj)  {
    # Flip coin to decide sex
    sex <- sample(c(maleval, femaleval), size = 1, replace = TRUE,
                  prob = c(maleProb, 1 - maleProb))
    if (sex == maleval) {datum = MASS::mvrnorm(n = 1, mu = Mmean, Sigma = Msigma)}
    if (sex == femaleval) {
      Fclust = sample(c(1, 2), size = 1, replace = TRUE, prob = c(prop1, prop2))
      if (Fclust == 1) {datum = MASS::mvrnorm(n = 1, mu = Fmean1, Sigma = Fsigma1)}
      if (Fclust == 2) {datum = MASS::mvrnorm(n = 1, mu = Fmean2, Sigma = Fsigma2)}
    }
    datamatrix[i, ] = c(sex, round(c(datum[1], exp(datum[2])), 1))
  }
  
  return(datamatrix)
} # end function

bh7 · June 10, 2020, 9:58pm

And this is the code:

d <-
HtWtDataGenerator(nSubj = 57, maleProb = .5) %>%
as_tibble()

fit1 <-
brm(data = d,
family = gaussian,
weight ~ 1 + height,
prior = c(prior(normal(0, 100), class = Intercept),
prior(normal(0, 100), class = b),
prior(cauchy(0, 10), class = sigma)),
chains = 4, cores = 2, iter = 2000, warmup = 1000,
seed = 2)

MauritsM · June 10, 2020, 10:09pm

Sorry that it didn’t work. I fear you’ll have to wait for @bgoodri to have a look - I don’t know what else you could try.

bgoodri · June 11, 2020, 12:14am

cores = 1?

JosephBulbulia · June 11, 2020, 12:35pm

Stab in the dark: could it be a problem with your Xcode installation, not R 4.0

bh7 · June 11, 2020, 1:55pm

Today is a Day of Days!! The complex model ran without problem.
I did two things differently:
I changed ‘cores = 2’ to ‘cores = 1’ (not sure what the difference is…) and in Xcode (my version, at least) I clicked on brms in the project panel (which was hidden until I looked for it).
Thank you all so much for your help, patience, and guidance - very much appreciated!!

gferraz · June 16, 2020, 10:55am

The parentheses seem to be wrong. The second ‘)’ after RSTUDIO_TERM") and the ‘)’ after “Darwin” have no matching ‘(’. No?

torkar · June 16, 2020, 11:03am

@gferraz you mean this?

if (Sys.getenv("RSTUDIO") == "1" && !nzchar(Sys.getenv("RSTUDIO_TERM")) && 
    Sys.info()["sysname"] == "Darwin" ) {
  parallel:::setDefaultClusterOptions(setup_strategy = "sequential")
}

gferraz · June 16, 2020, 11:34am

No, I meant this:

if (Sys.getenv(“RSTUDIO”) == “1” && RSTUDIO_TERM")) &&
Sys.info()[“sysname”] == “Darwin”) {
parallel:::setDefaultClusterOptions(setup_strategy = “sequential”)
}

I am having the same problem as bh7

torkar · June 16, 2020, 11:37am

Hmm, this is what I have in my ~/.Rprofile

if (Sys.getenv("RSTUDIO") == "1" && !nzchar(Sys.getenv("RSTUDIO_TERM")) && 
    Sys.info()["sysname"] == "Darwin" ) {
  parallel:::setDefaultClusterOptions(setup_strategy = "sequential")
}

gferraz · June 16, 2020, 11:39am

Yes, I see. That is the first bit of code you recommended and all the brackets are matching here. But it didn’t work out for me. And I am positive that I put it on .Rprofile in my home folder, and re-started RStudio. I still get the error message:

Error in unserialize(node$con) : error reading from connection
Calls: -> slaveLoop -> makeSOCKmaster
Execution halted
Error in unserialize(node$con) : error reading from connection
Calls: -> slaveLoop -> makeSOCKmaster
Execution halted
Error in makePSOCKcluster(names = spec, …) :
Cluster setup failed. 2 of 2 workers failed to connect.

torkar · June 16, 2020, 11:45am

If you open a terminal on your computer (search for Terminal in Spotlight on OS X) and cut and paste this (press return afterwards):

cat " ## WORKAROUND: https://github.com/rstudio/rstudio/issues/6692
## Revert to 'sequential' setup of PSOCK cluster in RStudio Console on macOS and R 4.0.0
if (Sys.getenv("RSTUDIO") == "1" && RSTUDIO_TERM")) &&
    Sys.info()["sysname"] == "Darwin" ) {
  parallel:::setDefaultClusterOptions(setup_strategy = "sequential")
} " >> ~/.Rprofile

then it should, for sure, put the above text in ~/.Rprofile. Then restart RStudio.

Topic		Replies	Views
Unable to run RStan (and brms) General rstan , brms	6	639	February 4, 2023
Error when running Rstan with update RStan	28	3246	March 28, 2020
Compilation error after installing rstan on new arm Mac with macOS Monterey Modeling rstan , brms	2	687	April 19, 2022
Error running BRMS in R after updates Modeling	4	544	January 19, 2020
Compilation error BRMS General	2	525	June 17, 2019

Brms/rstan on R 4.0

Related topics