Padding and Masking

[npiroozan]

Hello,

I hope you are doing well. Please forgive the basic question here, but how I am dealing with an issue relating to a very large number of graph compiles due to dynamic tensor shapes.

The problem specifically concerns the Silicon example for allegro as defined in the tutorial section.

I think the solution to this problem would be padding and masking.

How and where in the code could this be done? Please note this issue concerns training of the model.

Thank you very much.