]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
CUDA: only allocate FA tmp buffer if needed (llama/18564)
authorJohannes Gäßler <redacted>
Sat, 3 Jan 2026 12:55:53 +0000 (13:55 +0100)
committerGeorgi Gerganov <redacted>
Sun, 11 Jan 2026 09:02:08 +0000 (11:02 +0200)
commit5433aea73216d1724d7c21256b93cf251b31216f
tree85e196c1f817b11e4db51eb98e966f5cbcd5f558
parenta90aefcb055624e3e6786cce84c223acdad4cf3a
CUDA: only allocate FA tmp buffer if needed (llama/18564)
src/ggml-cuda/fattn-common.cuh