I want to embed 10 columns of categorical features to different dim sizes, so I write a “for” loop to enumerate each column and feed to its Embedding layer (I initialized 10 Embedding layers).
Here is my question:
For each iteration, I have to go through the whole “for” loop. But it is a SMALL for loop because only 10 operations are involved. I would like to know if code like this could slow down the efficiency of GPU？
To be more clear, imagine I have a code that could use 100% of GPU, then I insert a small for loop in this code, is it efficient?
Thanks for answers!