How to ensemble list of nn object from result of parallelly and neuralnet

I'm creating a backpropagation neural network in R using neural net The data I use are montly u and v wind, sea surface, relative humidity data from era5 in netcdf (.nc) format to predict monthly rainfall. Initially I built a model with neuralnet with code like this:

# Build Model NN
model = neuralnet(
  rainfall ~ wind_u + wind_v,
  data = data_train,
  hidden = c(3,2),
  threshold = 0.1,
  learningrate = 0.01,
  linear.output = TRUE,
  stepmax = 1e6
)

but the program takes a long time to run. so I reduced my train data to only contain u and v wind data to predict rainfall. then I use parallelly to speed up the process with code like this

ncores = parallelly::availableCores() - 6
cl <- makeCluster(ncores)

# Export libraries and data to each cluster node
clusterExport(cl, list("neuralnet", "data_train"))

# start time
start_time <- Sys.time()
cat("Waktu mulai:", format(start_time, "%H:%M:%S"), "\n")


# Train model with parallelly
models <- parLapply(cl, 1:ncores, function(i) {
  neuralnet(rainfall ~ wind_u + wind_v,
            data = data_train,
            hidden = c(3, 2),
            threshold = 0.1,
            learningrate = 0.01,
            linear.output = TRUE,
            stepmax = 1e6)
})

# stop cluster 
stopCluster(cl)

# end time
end_time <- Sys.time()
cat("Waktu selesai:", format(end_time, "%H:%M:%S"), "\n")

# calculate the time of processing model
execution_time <- end_time - start_time
cat("Durasi eksekusi:", execution_time, "\n")

The result is a model containing a list of nn objects from models[[1]] to models[[10]] How can I use a combination of these models so that I can predict rainfall just by

predict(models, data_test)

The following is the program code for preparing the data that I used

# set working directory
setwd('D:\\temp\\stdm\\')

# Load data
nc_ch = nc_open('input\\ch_train.nc')
nc_wind_rh = nc_open('input\\rh_wind_train.nc')
nc_ch_test = nc_open('input\\ch_test.nc')
nc_wind_rh_test = nc_open('input\\rh_wind_test.nc')

# extract values from variables
# train data
# rh = ncvar_get(nc_wind_rh, "r")  
wind_u = ncvar_get(nc_wind_rh, "u")               
wind_v = ncvar_get(nc_wind_rh, "v")                
rainfall = ncvar_get(nc_ch, "tp") *1000 #convert from m to mm            

# test data
# rh_test = ncvar_get(nc_wind_rh_test, "r")  
wind_u_test = ncvar_get(nc_wind_rh_test, "u")                
wind_v_test = ncvar_get(nc_wind_rh_test, "v")               
rainfall_test = ncvar_get(nc_ch_test, "tp") *1000          

# close .nc
nc_close(nc_ch)
nc_close(nc_wind_rh)
nc_close(nc_ch_test)
nc_close(nc_wind_rh_test)

# convert to 2D data
# rh = as.vector(rh)
wind_u = as.vector(wind_u)
wind_v = as.vector(wind_v)
rainfall = as.vector(rainfall)

# rh_test = as.vector(rh_test)
wind_u_test = as.vector(wind_u_test)
wind_v_test = as.vector(wind_v_test)
rainfall_test = as.vector(rainfall_test)

# convert to dataframe
data_train = data.frame(wind_u, wind_v)
data_test = data.frame(wind_u_test, wind_v_test)

# Normalize data (optional)
data_train = scale(data_train)
data_train = cbind(data_train,rainfall)
data_test = scale(data_test)