Reduce rank and set values according to 3rd dimension

agibsonccc September 27, 2022, 1:33pm 10

@SidneyLann is this for weighted loss? If not why not make that a placeholder?

"Attention Is All You Need" model implementation using dl4j

Topic		Replies	Views
Using gradient as an intermediate SDVariable SameDiff	11	392	June 14, 2022
Variable Length Output Layer for FeedForward Network DL4J	2	467	March 5, 2020
Modifying SD variables SameDiff	9	224	July 20, 2023
Using SameDiff for aCosineSimVertex SameDiff	4	221	October 7, 2022
Assign value to a placeholder SameDiff	55	574	September 24, 2023