scripts_r/precompile_model.R

In alexanderbuchholz/distbayesianmc: Distributed Bayesian Model Choice

# rscript compile logistic regression model

imodel <- "sparse_linear"
print(imodel)

if(imodel == "logit"){
  library(distbayesianmc)
  setwd("~/R_programming/distbayesianmc")
  source("~/R_programming/distbayesianmc/params_simulation/params_logit_higgs.R")
  
  if (sys.nframe() == 0){
    stan_code <- readChar(fileName, file.info(fileName)$size)
    mod <- rstan::stan_model(model_code = stan_code, auto_write = T, save_dso = FALSE)
    saveRDS(mod, file = "~/R_programming/distbayesianmc/stan_models/fit_logistic.rds")
    rm(mod)
  } else {
    mod <- readRDS(file = "~/R_programming/distbayesianmc/stan_models/fit_logistic.rds")
    
    rstan::rstan_options(auto_write = TRUE)
    
    dataset_loaded <- f_dataset_loader("higgs1_small")
    splitted_data <- f_pack_split_data(dataset_loaded$X, dataset_loaded$y, ssplits=1, iseed=1, typesplit="random")
    splitted_data <- f_prep_prior_logistic(splitted_data, scale = 1)
    res <- f_stan_sampling_single_split(mod, splitted_data[[1]], dataset = "pima", i_seed = 1, iter = 1, typesplit = "random", nchain = 2000, typeprior = "normal")
  }
  
} else if(imodel == "sparse_linear"){
  library(distbayesianmc)
  setwd("~/R_programming/distbayesianmc")
  source("~/R_programming/distbayesianmc/params_simulation/params_sparse.R")
  
  if (sys.nframe() == 0){
    stan_code <- readChar(fileName, file.info(fileName)$size)
    mod <- rstan::stan_model(model_code = stan_code, auto_write = T, save_dso = FALSE)
    saveRDS(mod, file = "~/R_programming/distbayesianmc/stan_models/fit_sparse_linear.rds")
    rm(mod)
  } else {
    mod <- readRDS(file = "~/R_programming/distbayesianmc/stan_models/fit_sparse_linear")
    
    rstan::rstan_options(auto_write = TRUE)
    
    dataset_loaded <- f_dataset_loader("higgs1_small")
    splitted_data <- f_pack_split_data(dataset_loaded$X, dataset_loaded$y, ssplits=1, iseed=1, typesplit="random")
    splitted_data <- f_prep_prior_logistic(splitted_data, scale = 1)
    res <- f_stan_sampling_single_split(mod, splitted_data[[1]], dataset = "pima", i_seed = 1, iter = 1, typesplit = "random", nchain = 2000, typeprior = "normal")
  }
  
}