llama.cpp
db0a8ce8 - ggml-metal: FWHT kernel for metal backend (#25924)

Commit

10 days ago

ggml-metal: FWHT kernel for metal backend (#25924)

References

#26561 - metal: per-op source split + parallel compile (replay)

#26570 - metal : per-device tuned (Q, NE) for flash-attn vec (replay)

Author

ggerganov

ggerganov

Committer

forforever73

forforever73

Parents

Loading