`top_p` should be on (0, 3).

[Steviey]

According to the example from Max, at the last conference in D.C. I try to tune top_p of step_select_vip(), (which by the way works)...I get warnings...

model_spec    <- linear_reg(penalty = NULL, mixture = NULL) %>% set_engine("lm")

      set.seed(1234)

      recipe_spec <- recipe(myFormula, data = df_train) %>%
          #step_normalize(all_predictors()) %>%
          colino::step_select_vip(all_predictors(),model=model_spec,outcome = "y",top_p=tune())

      folds <- vfold_cv(df_train,repeats=1,v=5,strata=y) # 

      model_wfl <- workflow() %>%
          add_model(model_spec) %>%
          add_recipe(recipe_spec) %>%
          tune_grid(resamples=folds,grid=25)

I get the warning... 

! Fold1: preprocessor 3/4: `top_p` should be on (0, 3).
! Fold1: preprocessor 4/4: `top_p` should be on (0, 3).
! Fold2: preprocessor 3/4: `top_p` should be on (0, 3).
! Fold2: preprocessor 4/4: `top_p` should be on (0, 3).
! Fold3: preprocessor 3/4: `top_p` should be on (0, 3).
! Fold3: preprocessor 4/4: `top_p` should be on (0, 3).
! Fold4: preprocessor 3/4: `top_p` should be on (0, 3).
! Fold4: preprocessor 4/4: `top_p` should be on (0, 3).
! Fold5: preprocessor 3/4: `top_p` should be on (0, 3).
! Fold5: preprocessor 4/4: `top_p` should be on (0, 3).
# Tuning results
# 5-fold cross-validation using stratification 
# A tibble: 5 × 4
  splits         id    .metrics         .notes          
  <list>         <chr> <list>           <list>          
1 <split [9/3]>  Fold1 <tibble [8 × 5]> <tibble [2 × 3]>
2 <split [9/3]>  Fold2 <tibble [8 × 5]> <tibble [2 × 3]>
3 <split [10/2]> Fold3 <tibble [8 × 5]> <tibble [2 × 3]>
4 <split [10/2]> Fold4 <tibble [8 × 5]> <tibble [2 × 3]>
5 <split [10/2]> Fold5 <tibble [8 × 5]> <tibble [2 × 3]>

There were issues with some computations:
  - Warning(s) x10: `top_p` should be on (0, 3).

What does this mean?

[stevenpawley]

Hello just a quick note that I'll try to look into this, once I get some time later this week

[stevenpawley]

Sorry for the hiatus but this appears to be related to the tuning grid which is attempting to select more features than are available in your dataset. In this case, if top_n > n_features, the filter steps will set top_n = n_features and will issue a warning.