Fix llama.cpp chat token ID parity - #143

Merged

Qubitium merged 3 commits into

mainfrom

fix/llamacpp-chat-token-ids

Sep 4, 2026

Commits on Sep 4, 2026

fix(llama-cpp): preserve rendered chat token ids
Qubitium
committed
b1b5728
perf(llama-cpp): reuse shared multiple-choice prefixes
Qubitium
committed
570d6bc
perf(llama-cpp): batch distinct scoring prefixes
Qubitium
committed
1a7b4da