ConceptAI & ML
Residual layer weight initialization
Scaling down the weights of residual projections by a factor proportional to the square root of twice the number of layers in order to maintain a stable standard deviation across the residual stream.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Lecturers
1Connections
4The lectures that define it or assume it, and the concepts next to it.