Describe the bug
Problem with label standardization in regression processor. Label mean and std deviation are being calculated for train set, then those values are being overwritten with the same calculation for dev and test set link to code
After that, standardization is not actually performed on any data because of key mismatch over here. Instead of checking for the "label" key, you must look for self.tasks["regression"]["label_name"]

Error message
Silent error.

Expected behavior
RegressionProcessor should store regression label mean and std dev values only for train set. And standardization must be performed on the actual labels.

Additional context

To Reproduce
add print statements to view the output of this function and you will see that standardization of regression label is not taking place.

System:

• OS: Ubuntu 16.04.5 LTS
• GPU/CPU: Both
• FARM version: 0.4.5