[executorch][cuda] fuse gate/up MLP projections #20482

Gasoonjia wants to merge 1 commit into

gemma4_31b_export_under_32gbpytorch/executorch:gemma4_31b_export_under_32gbfrom

gemma4_31b-cuda-decode-speeduppytorch/executorch:gemma4_31b-cuda-decode-speedupCopy head branch name to clipboard