llama.cpp Fixes OOB Reads in UGM Tokenizer

Core summary

One sentence to understand this update

llama.cpp release b9917 addresses critical out-of-bounds read vulnerabilities in the UGM tokenizer, specifically related to precompiled_charsmap handling, preventing heap-buffer-overflows.

Impact & opportunity

What this could mean

Developers using llama.cpp benefit from enhanced security and stability, as these fixes prevent potential vulnerabilities in tokenization processes.