Error in importPosteriorFromHPC for GPP/Hmsc-hpc models with alignPost = TRUE

[elgabbas]

Hello,

I am receiving the following error when using importPosteriorFromHPC to merge posteriors for multiple MCMC chains.

Error in cpL[[j]]$Alpha[[r]] <- abind(cpL[[j]]$Alpha[[r]], rep(1, nfMax -  : 
  more elements supplied than there are to replace

I noticed that this error occurred when setting alignPost = TRUE for some models. The error comes from the alignPosterior(m) part of the function.

Could you please explain what alignPost does and is it necessary to set it to TRUE in the importPosteriorFromHPC function?

Thanks

[elgabbas]

UPDATE

I have uploaded the unfitted models along with the outputs of the Hmsc-HPC models in this link.

load("Mod_Init_Example.RData")
ListPost3 <- list.files("", pattern = "GPP30_NoTree_.+_post.rds$", full.names = TRUE) %>%
  sort() %>%
  snow::parLapply(cl = c1, x = ., fun = GetPosts)
# save(ListPost3, file = "ListPost3.RData")

# This failed
Model_Fit3 <- Hmsc::importPosteriorFromHPC(
  m = Mod_Init_DE, postList = ListPost3, nSamples = 1000, thin = 5, transient = 2000, alignPost = TRUE)

# This worked
Model_Fit3 <- Hmsc::importPosteriorFromHPC(
  m = Mod_Init_DE, postList = ListPost3, nSamples = 1000, thin = 5, transient = 2000, alignPost = FALSE)

This did not happen for me for other model variants created using almost identical scripts. Is there a specific reason for these model objects?