Setting missing value code for TFDV


I’m trying to use TFDV with my own dataset that has specific codes for missing values, so but I’m not sure how to communicate to tensorflow that this and that code should be considered as a missing value, now it just somehow assumes that a random question mark in a column otherwise full of integers is not a missing data point but something else.

(this is the first time I’m using this Discourse so sorry if this question went to wrong place!)

To fill missing values, you can use tensorflow directly on a dataframe. Please read about tf.where.