llama.cpp

fix: GLM-DSA crash in llama-tokenize when using vocab_only

#22102

Merged