[executorch][gemma4] fuse MLP gate/up at GGUF load #20481

Gasoonjia wants to merge 7 commits into

gemma4_31b-cuda-decode-speeduppytorch/executorch:gemma4_31b-cuda-decode-speedupfrom

gemma4_31b-mlp-fusion-unifiedpytorch/executorch:gemma4_31b-mlp-fusion-unifiedCopy head branch name to clipboard