CLBlast: Add broadcast support for matrix multiplication (#3402)
Broadcast src0 into src1 across dimensions 2 and 3 when needed. This is required for models that use GQA.
S
shibe2 committed
665018c749101e81c816675198e731e47d6b1dbe
Parent: 29a404a
Committed by GitHub <noreply@github.com>
on 10/2/2023, 7:26:15 PM