llama.cpp
f4f9367f - less code duplication, offload k and v separately

Commit

2 years ago

less code duplication, offload k and v separately

References

#4309 - llama : per-layer KV cache

Author

slaren

slaren

Parents

Loading