Defining Training and Testing data instead of random split

Question

I would like the training and testing spilt to not be random instead it should be decided on the value of the column for eg training set = base_data[base_data[\'columnA\']==0] t