How to pass null/Nan values into the dataframe passed into batch transform

Question

I am trying to make inference using my Xgboost model on a dataset which has NaN values, now inherently Xgboost handles the NaN values, it does not throw any error while training with NaN values whereas the batch transform job gives the error **''could not convert string to float''** when it encounters NaN values into the dataset that is to be transformed.  
Can anyone help me as to how can I pass NaN values into my input dataset for the batch transform job?  
Thanks a ton!

Answer

Hi bharat-patidar,  
  
Thanks for your interest in Amazon SageMaker XGBoost. We recently made a change to address empty values in the dataset. Now you should be able to pass empty strings or "NaN" strings with CSV data. Can you please verify and let us know if the change addresses the issue at your end?  
  
Thanks!  
Regards,  
Amazon SageMaker Team

How to pass null/Nan values into the dataframe passed into batch transform

Relevant content