mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-05-07 09:34:07 +00:00
LLaMA doesn't support more than 2048 token context sizes, and going above that produces terrible results.
38 KiB
38 KiB