Dear Renaud,
I’m using your package in a foreach dopar loop to run a simulation. I’d like to manually split up the simulation to five (5) different machines to reduce computation time.
Entire simulation has 500k loops, but separated to different machine each will have 100k loops. I’d like to have a unique attr(df, "rng") for each of 500k loops using one registerDoRNG seed. For instance, for me to specify
- machine1 to use registerDoRNG(3.14), generate rngs from 1:100k
- machine2 to use registerDoRNG(3.14), generate rngs from 100,001:200k
- repeat for all other machines.
Is there any way I could specify this using doRNG?
Currently I have
registerDoRNG(3.14)
res <- foreach::foreach(i=1:500000,
.combine=rbind,
.packages = c("permute", "lattice", "matrixStats", "vegan", "HMP")
) %dopar%
{
df <- simulation code here
}
Dear Renaud,
I’m using your package in a foreach dopar loop to run a simulation. I’d like to manually split up the simulation to five (5) different machines to reduce computation time.
Entire simulation has 500k loops, but separated to different machine each will have 100k loops. I’d like to have a unique attr(df, "rng") for each of 500k loops using one registerDoRNG seed. For instance, for me to specify
Is there any way I could specify this using doRNG?
Currently I have