]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
CUDA: fuse adds, fuse add with rms norm (llama/15631)
authorAman Gupta <redacted>
Fri, 29 Aug 2025 03:35:58 +0000 (11:35 +0800)
committerGeorgi Gerganov <redacted>
Fri, 5 Sep 2025 09:54:07 +0000 (12:54 +0300)
commit81415da35a8659aafea925aae1c718b5bbd08192
tree8110588cf268374f1dfdf85bdf67064f89a018e6
parent70d536e2130c7e9cae27c3442d5a56430fa49a42
CUDA: fuse adds, fuse add with rms norm (llama/15631)

* CUDA: fused add with rms_norm_mul

* Non-broadcast fuse works

* Add fused adds

* format

* Remove n_fuse from template params

* Address review comments

* Move template inside binbcast
src/ggml-cuda/binbcast.cu
src/ggml-cuda/binbcast.cuh
src/ggml-cuda/ggml-cuda.cu
src/ggml-cuda/norm.cu
src/ggml-cuda/norm.cuh