Mirror https://github.com/ggml-org/llama.cpp/pull/22121 - #103

Open

ngxson wants to merge 10 commits into

ngxson:masterfrom

Commits on May 26, 2026

use /// instead of // for llama_token_to_piece docstring
julmb
committed
74fbcb2
clarify llama_token_to_piece docstring in analogy to llama_detokenize
julmb
committed
72d9180
perform bounds checking on token in token_to_piece
julmb
committed
b6387d2
remove always-true conditional
julmb
committed
492a14f
handle INT32_MIN result of llama_token_to_piece in downstream uses
julmb
committed
3828da4
add limits includes in files that did not use it before
julmb
committed
af7af8a
add invalid token guard to token_to_piece_for_cache
julmb
committed
6b190ab
propagate token_to_piece failure
julmb
committed
0e9b56d
consistent punctuation
julmb
committed
36fe0b0
exit on llama_token_to_piece error
julmb
committed
2b00549