assertion failure from diag_pre_multiply in rstan 1.3.0 about stan HOT 5 CLOSED

I also had this problem using Rstan 1.3 stable, but it worked fine using the latest revision of Github master branch.

from stan.

I just verified that this crashes in the most recent
RStan either in debug or fast mode. I don't even get
an error message, the R process crashes and the GUI disappears.
(I'm on: R 2.15.3 GUI 1.53 Leopard build 64-bit (6451))

This model runs without crashing or providing warnings
from the command line on the current master branch.
Printing out the product it looks OK.

Oh, and it's not the data. I changed that to transformed
data and made the num==5 so I could inspect the product
and it still crashed R.

• Bob

On 5/1/13 5:20 AM, Paul Metcalfe wrote:

This is a minimal version of a problem that has caught me in a real model. This model:

|data {
int<lower=0> num;
}

parameters {
vector<lower=0, upper=1>[num] scale;
matrix[num, num] fred;
}

transformed parameters {
matrix[num, num] out;

 out <- diag_pre_multiply(scale, fred);

}

model {
for (indx in 1:num) {
for (indy in 1:num) {
fred[indy, indx] ~ normal(0, 1);
}
}
}
|

dies with the error

|/home/pdm/.R/x86_64-unknown-linux-gnu/2.15/rstan/include//stanlib/eigen_3.1.2/Eigen/src/Core/DenseCoeffsBase.h:337: typename Eigen::internal::traits::Scalar& Eigen::DenseCoeffsBase<Derived, 1>::operator()(typename Eigen::internal::traits::Index, typename Eigen::internal::traits::Index) [with Derived = Eigen::Matrix<stan::agrad::var, -0x00000000000000001, -0x00000000000000001, 0, -0x00000000000000001, -0x00000000000000001>]: Assertion `row >= 0 && row < rows() && col >= 0 && col < cols()' failed.
Aborted (core dumped)
|

when run with

|library(rstan)

set_cppo("debug")

results <-
stan(file="dummy.txt",
data=list(num=100),
iter=1000,
chains=4)
|

The R version is 2.15.1, Rcpp is 0.10.3, gcc is

|Using built-in specs.
Target: x86_64-redhat-linux
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man
--infodir=/usr/share/info --with-bugurl=http://bugzilla.redhat.com/bugzilla
--enable-bootstrap --enable-shared --enable-threads=posix --enable-checking=release
--with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions
--enable-gnu-unique-object --enable-languages=c,c++,objc,obj-c++,java,fortran,ada
--enable-java-awt=gtk --disable-dssi
--with-java-home=/usr/lib/jvm/java-1.5.0-gcj-1.5.0.0/jre --enable-libgcj-multifile
--enable-java-maintainer-mode --with-ecj-jar=/usr/share/java/eclipse-ecj.jar
--disable-libjava-multilib --with-ppl --with-cloog --with-tune=generic --with-arch_32=i686
--build=x86_64-redhat-linux
Thread model: posix
gcc version 4.4.7 20120313 (Red Hat 4.4.7-3) (GCC)
|

and rstan is

|rstan (Version 1.3.0, packaged: 2013-04-12 21:12:02 UTC, GitRev: f574555)
|

The OS is linux amd64.

—
Reply to this email directly or view it on GitHub #34.

from stan.

I'm using a modified makefile to compile Rstan, so, this might be the reason it worked for me (I tried the model again and it worked):

I changed:
$(R) CMD rstan --md5 # --no-vignettes --no-manual
with
$(R) CMD build /home/marco/Documents/stan/rstan/rstan --md5 --no-vignettes --no-manual

And:
$(R) CMD INSTALL --merge-multiarch $(STANPKG)
with
$(R) CMD INSTALL --no-multiarch rstan_1.3.0.tar.gz

Maybe this is a clue. I'll trying using the default makefile and see if it runs. My OS is also Linux amd64.

EDIT: I wasn't able to reproduce this problem using the latest revision of Github master branch even after using the default makefile.

from stan.

If we use the current master branch, I think both stan from the command
line or rstan work. But using Stan 1.3.0 that we released, we have the same problem
as with RStan 1.3.0 that we released. So I do not think this is a problem with
RStan. This is the result of using stan-1.3.0 downloaded from mc-stan.org:

jq2@jq-tpw510:~/Desktop/stan-src-1.3.0/tmp$ more m1.stan

data {
int<lower=0> num;
}

parameters {
vector<lower=0, upper=1>[num] scale;
matrix[num, num] fred;
}

transformed parameters {
matrix[num, num] out;

out <- diag_pre_multiply(scale, fred);

}

model {
for (indx in 1:num) {
for (indy in 1:num) {
fred[indy, indx] ~ normal(0, 1);
}
}
}
jq2@jq-tpw510:~/Desktop/stan-src-1.3.0/tmp$ ./m1 --data=m1.rdump
m1: lib/eigen_3.1.2/Eigen/src/Core/DenseCoeffsBase.h:337: Eigen::DenseCoeffsBase<Derived, 1>::Scalar& Eigen::DenseCoeffsBase<Derived, 1>::operator()(Eigen::DenseCoeffsBase<Derived, 1>::Index, Eigen::DenseCoeffsBase<Derived, 1>::Index) [with Derived = Eigen::Matrix<stan::agrad::var, -1, -1>; Eigen::DenseCoeffsBase<Derived, 1>::Scalar = stan::agrad::var; Eigen::DenseCoeffsBase<Derived, 1>::Index = long int]: Assertion `row >= 0 && row < rows() && col >= 0 && col < cols()' failed.
Aborted

Jiqiang

from stan.

I guess the fix was 405bc3d ?

Thanks for the very wonderful stan!

from stan.