Lize Coenen

Hi guys, While training a LambdaMART model for a dataset I'm working on, I experienced a NullPointerException (see stack trace at the end of this message). The dataset contains 10000+ rows and 100+ features and uses the same query-id (0) for all rows. I was able to reproduce this error by adapting the writeRandomData method in your EvaluatorTest (see attachment). This problem can be circumvented by splitting the dataset in a training and testing part before running RankLib.jar (and not asking RankLib...