The validateDataFrame function pulls entire dataset into Driver

[Josh Fleming]

The following code attempts to pull the entire dataset into the Driver before running validations:

photon-ml/photon-client/src/main/scala/com/linkedin/photon/ml/data/DataValidators.scala

Line 265 in 237c315

 val rows = dataSet.collect() 

We should instead map over the DataFrame rows so that the validations run in a distributed fasion, and then collect the results.

[Alex Shelkovnykov]

Resolve by #307