step-by-step creation of models with accumulation of predictors vs GridSearch

Can you please tell me if step linear models of the independent variable "ols_step_both_p()" (R) are possible with the accumulation of predictors in the amount of 58, 220 and 299, naturally with the Bonferoni corrections, which are also accumulated? can all intermediate results be analyzed in the article or not? or is it allowed to publish only the result of one stage (any)?

Can not understand GridSearch (and the like):

rf <- RFTrainer$new()
gst <-GridSearchCV$new(trainer = rf,
parameters = list(n_estimators = c(100),
max_depth = c(5,2,10)),
n_folds = 3,
scoring = c('accuracy','auc'))
data("DATA1")
gst$fit(DATA1, "pc2")

Output:

[1] "entering grid search"
[1] "In total, 3 models will be trained"
> gst[["trainer"]]
<RFTrainer>
Public:
always_split: NULL
class_weights: NULL
classification: 1
clone: function (deep = FALSE)
criterion: gini
fit: function (X, y)
get_importance: function ()
importance: impurity
initialize: function (n_estimators, max_depth, max_features, min_node_size,
max_depth: 10
max_features: 17
min_node_size: 1
n_estimators: 298
predict: function (df)
seed: 42
verbose: TRUE
Private:
trained_model: ranger

Please help me write code to find a linear relationship model by choosing from 297 predictors (and if possible + 2 more categorical predictors)