How to download a vetiver model object as an RDS object and use it on a different PC?

Machine Learning and Modeling
1

Hi there, I am working inside a research environment remote desktop and have built a cross-validated random forest model. I saved the workflow model as the vetiver model in a model board at the remote desktop and downloaded it from the remote desktop onto my own personal computer as a RDS file. However, I am unable to apply it in my PC to an external dataset. I would really appreciate any guidance on this.

2

Can you share a little more about the kind of code that is causing problems for you? If I were wanting to do this, I would start out by storing the model on the first machine with something like this:

library(tidymodels)
library(vetiver)
#> 
#> Attaching package: 'vetiver'
#> The following object is masked from 'package:tune':
#> 
#>     load_pkgs
library(pins)
data(Sacramento)

rf_spec <- rand_forest(mode = "regression")
rf_form <- price ~ type + sqft + beds + baths

rf_fit <- 
    workflow(rf_form, rf_spec) %>%
    fit(Sacramento)
v <- vetiver_model(rf_fit, "sacramento_rf")

model_board <- board_folder(path = "/tmp/test")
vetiver_pin_write(model_board, v)
#> Creating new version '20240925T201217Z-73f49'
#> Writing to pin 'sacramento_rf'
#> 
#> Create a Model Card for your published model
#> • Model Cards provide a framework for transparent, responsible reporting
#> • Use the vetiver `.Rmd` template as a place to start

Created on 2024-09-25 with reprex v2.1.1

And then on the second machine, I would read the model and predict() like this:

library(tidymodels)
library(vetiver)
#> 
#> Attaching package: 'vetiver'
#> The following object is masked from 'package:tune':
#> 
#>     load_pkgs
library(pins)
data(Sacramento)

model_board <- board_folder(path = "/tmp/test")
v <- vetiver_pin_read(model_board, "sacramento_rf")
predict(v, Sacramento[5:10,])
#> # A tibble: 6 × 1
#>     .pred
#>     <dbl>
#> 1 106683.
#> 2 133271.
#> 3 147360.
#> 4 158743.
#> 5 128204.
#> 6 155486.

Created on 2024-09-25 with reprex v2.1.1

Are you trying to do something different?

3

Thanks so much for your prompt reply. So I can't use the same board folder to read the model. I downloaded it in my PC. I have downloaded the "v" from the remote desktop as a .RDS file. When I tried to use readRDS and then use it for prediction, the error message I am getting is attached.

Screenshot 2024-09-25 191849
Screenshot 2024-09-25 191849833×130 10.4 KB

Would really appreciate help with this. Thanks again.

Data splitting

set.seed(23102013)
dta_split <- initial_split(dta1, prop = 3/5)
dta_train = training(dta_split)
dta_test = testing(dta_split)
dta_fold = mc_cv(dta_train, times = 50)

nrow(dta_train)
nrow(dta_test)

Model building

tree_rec = recipe(No_Nadir~., data = dta_train)%>%
  update_role(OS, All.Cause.Mortality, new_role = "id variable")%>%
  step_center(AGE,ECOGBL,PSABL,HGBBL,BASEBMI)%>%
  step_scale(AGE,ECOGBL,PSABL,HGBBL,BASEBMI)%>%
  step_dummy(all_nominal_predictors(), one_hot = F)


tune_spec = rand_forest(
  mtry = tune(),
  trees = 1000,
  min_n = tune())%>%
  set_mode("classification")%>%
  set_engine("ranger", importance = "impurity")
  

tune_wf = workflow()%>%
  add_model(tune_spec)%>%
  add_recipe(tree_rec)

Crossvalidation and tuning hyperparameters with grid search

library(parallel)
library(doParallel)

cl <- makePSOCKcluster(2)
registerDoParallel(cl)


set.seed(23102013)

tune_res = tune_grid(
  tune_wf,
  resamples = dta_fold,
  grid = 1000,
  control = control_resamples(save_pred = T)
)


tune_res%>%
  collect_metrics()%>%
  filter(.metric == "roc_auc")%>%
  dplyr::select(mean, min_n, mtry)%>%
  pivot_longer(min_n:mtry, values_to = "value", names_to = "parameters")%>%
  ggplot(aes(value,mean,color=parameters)) + 
  geom_point(show.legend = F) + 
  facet_wrap(~parameters, scales = "free_x") + 
  labs(x = NULL, y = "AUC")

Selecting the best model

### Selecting best parameters
best_auc <- select_best(tune_res)


final_rf = finalize_model(tune_spec,
                          best_auc)

model_board = board_folder("/document/psa_response"). 

v <- veiver_pin_read(model_board, "final_rf").

###Downloaded v from the remote environment as a RDS file. 

##Now in my Personal desktop
final_model = readRDS("v.RDS")
predict(final_model, dta_test3)
4

I would really appreciate it if someone could please help me understand where I was wrong.

5

Could you update your example to use the reprex package and perhaps a simple dataset that we both have access to? Using reprex makes it easier to see both the input and output, and for us to re-run the code in a local session. Thanks! :raised_hands: