Error in cut.default(y, breaks, include.lowest = TRUE) : invalid number of intervals

Question

I use Iris dataset to build a neuralnet model.
My goal is to use the caret package from R to predict the species class of various Iris flowers.
I try to train model with leave one out cross validation by caret.

formula.bpn <- setosa+versicolor+virginica ~ Sepal.Length+Sepal.Width+Petal.Length+Petal.Width
smp.size <- floor(0.8*nrow(data))
set.seed(542)
train.ind <- sample(seq_len(nrow(data)), smp.size)
train <- data[train.ind,]
test <- data[-train.ind,]
train.control <- trainControl(method="LOOCV",
                              search="grid",
                              verboseIter=FALSE,
                              returnData=TRUE,
                              returnResamp="final",
                              savePredictions="final",
                              # classProbs=TRUE,
                              selectionFunction="best",
                              indexFinal=NULL,
                              allowParallel=TRUE
                              )
model <- train(form=formula.bpn, 
               data=train, 
               method="neuralnet",
               metric="RMSE",
               maximize=FALSE,
               trControl=train.control,
               tuneGrid=expand.grid(.layer1=c(1:4), .layer2=c(0:4), .layer3=c(0)), 
               na.action=na.omit,
               startweights=NULL,
               algorithm="rprop+",  # resilient backpropagation with weight backtracking
               err.fct="sse",
               act.fct="logistic",
               threshold=0.01, 
               stepmax=5e10, 
               linear.output=FALSE 
               )

But I got the error : Error in cut.default(y, breaks, include.lowest = TRUE) : invalid number of intervals
This is the traceback :

11: stop("invalid number of intervals")
10: cut.default(y, breaks, include.lowest = TRUE)
9: cut(y, breaks, include.lowest = TRUE)
8: createFolds(outcome, n, returnTrain = TRUE)
7: make_resamples(trControl, outcome = y)
6: with_preserve_seed({
       set_seed(list(seed = seed, rng_kind = rng_kind))
       code
   })
5: withr::with_seed(rs_seed, make_resamples(trControl, outcome = y))
4: train.default(x, y, weights = w, ...)
3: train(x, y, weights = w, ...)
2: train.formula(form = formula.bpn, data = train, method = "neuralnet", 
       metric = "RMSE", maximize = FALSE, trControl = train.control, 
       tuneGrid = expand.grid(.layer1 = c(1:4), .layer2 = c(0:4), 
           .layer3 = c(0)), na.action = na.omit, startweights = NULL, 
       algorithm = "rprop+", err.fct = "sse", act.fct = "logistic", 
       threshold = 0.01, stepmax = 5e+10, linear.output = FALSE)
1: train(form = formula.bpn, data = train, method = "neuralnet", 
       metric = "RMSE", maximize = FALSE, trControl = train.control, 
       tuneGrid = expand.grid(.layer1 = c(1:4), .layer2 = c(0:4), 
           .layer3 = c(0)), na.action = na.omit, startweights = NULL, 
       algorithm = "rprop+", err.fct = "sse", act.fct = "logistic", 
       threshold = 0.01, stepmax = 5e+10, linear.output = FALSE)

If I set form=setosa ~ Sepal.Length+Sepal.Width+Petal.Length+Petal.Width or versicolor ~ Sepal.Length+Sepal.Width+Petal.Length+Petal.Width or virginica ~ Sepal.Length+Sepal.Width+Petal.Length+Petal.Width, then train() can be performed normally. How can I fix this error ?