Inconsistencies between F# and C# samples

We reviewed and found some inconsistencies between the C# and F# samples.
This issue is just a convenient list for having the issues in a single place so we can all take it into account while updating the F# samples in the future.

If you fix any inconsistency below, please answer this thread, so we remove it, ok?

Please see the below table comparing the execution of F# and C# samples.

Sample Name | Comments | F# sample output |  c# sample output
-- | -- | -- | --
**Credit card Fraud Detection**  | 1)Accuracy is 1 in f# sample which is not correct. 2) The way o/p is displayed is not formatted properly in f# sample. No headers, no line breaks etc 3)The probability is showing in different units for c# and f#. for ex probability =0.9993f in f# and 4.303323E-05 in C# | Trainer and Predictor: | Trainer:Predictor: 
**TensorFlow Estimator** | Comment: Can see an exception in f# o/p while finding path of filel |   |  
**Github Labeler** | Comment: Getting build error while executing github labeler sample |  | running fine for c# sample
**TesnorFlow Scorer** | Comment: Toaster 2 and Toaster 3 are shwoing accuracy as 1 |   |