FloatingPoint-32 to FloatingPoint-16

Questions
#1

When running the GraphSage code in the DGL library, it runs with FP-32 by default. What is to be done to run it with FP-16.
TIA

#2

This is a duplicate of thread 1372.