RMLSE validation between rpart model and test set showing Na and zero - r

Question

I am seeking a little help as I have hit a wall.
I have trained a model (CART) with a train dataset and am looking to validate the accuracy of the model with RMLSE on a test set.
I have the following:
data.frame': 5463 obs. of 15 variables:
$ Start_date: chr "2011-01-20 02:00:00" "2011-01-20 05:00:00" "2011-01-20 06:00:00"
$ Season : Factor w/ 4 levels "spring","summer",..: 1 1 1 1 1 1 1 1 1 1
$ Holiday : Factor w/ 2 levels "Not","Holiday": 1 1 1 1 1 1 1 1 1 1 ...
$ Workingday: int 1 1 1 1 1 1 1 1 1 1 ...
$ Weather : Factor w/ 4 levels "Clear","Cloudy",..: 1 1 1 1 1 2 1 2 2 2..
$ Temp : num 10.66 9.84 9.02 9.02 9.02 ...
$ Humidity : int 56 60 60 55 55 52 48 45 42 45 ...
$ Windspeed : num 0 15 15 15 19 ...
$ Count : num 1 1 1 8 18 6 3 4 5 3 ...
$ Date : chr "2011-01-20" "2011-01-20" "2011-01-20" "2011-01-20" ...
$ Hour : Factor w/ 24 levels "00","01","02",..: 3 6 7 8 9 10 11 12 .
$ Year : chr "2011" "2011" "2011" "2011" ...
$ Month : chr "01" "01" "01" "01" ...
$ Weekday : Factor w/ 7 levels "Friday","Monday",..: 5 5 5 5 5 5 5 5 5 5
$ Hour_Bin : num 0 0 0 0 0 0 0 0 0 0 ...
$ temp_Bin : num 1 1 1 2 2 2 2 2 2 2 ...
$ year_Bin : num 1 1 1 1 1 1 1 1 1 1 ...
The predicted values is vector of:
Named num [1:5463] 9 9 9 9 9 9 9 9 9 9 ...
- attr(*, "names")= chr [1:5463] "9266" "9267" "9268" "9269" ...
I have used the function:
Evaluate_Model <- function (test, pred) {
return(sqrt(1/nrow(test)*sum((log(pred+1)-log(test$Count+1))^2)))
}
and also tried the matrix package
library('Metrics')
rmsle(test$Count, pred)
when I try to get the Root Mean Squared Logarithmic Error, I am returned either [0] or [Na].
I gone through the process of converting the count variable to different data types, and also tried putting the prediction into a dataframe and evaluate it from their.
I have also trained a model with one attribute and tried to evaluate these models, but am still hetting the same result.
My target variable (count) and the other attributes have zero values, but these are real values, not na's.
IS it the training of the algorithm, the data types???
Any help would be appreciated.
A sample of the model code:
model3 <- rpart(Count~Month+Temp, data = train)
# round prediction
pred <- round(predict(model3, newdata = test))
Evaluate_Model(test, pred)
Thanks in advance.