LLM inference in C/C++
C++ ggml
llama-grammar: fix numeric truncation for token_id parsing (#29382)
Daniel Kuts5cf3a3528716
diff --git a/src/llama-grammar.cpp b/src/llama-grammar.cppindex deffec8c07..0a8a61e7e8 100644--- a/src/llama-grammar.cpp+++ b/src/llama-grammar.cpp@@ -194,7 +194,11 @@ static std::pair<uint32_t, const char *> parse_token(const llama_vocab * vocab,if (*pos == '[') {pos++;const char * int_end = parse_int(pos);- uint32_t token_id = std::stoul(std::string(pos, int_end - pos));+ unsigned long id = std::stoul(std::string(pos, int_end - pos));+ if (id > std::numeric_limits<uint32_t>::max()) {+ throw std::runtime_error(std::string("parsed token id is too big at ") + pos);+ }+ uint32_t token_id = static_cast<uint32_t>(id);pos = int_end;if (*pos != ']') {throw std::runtime_error(std::string("expecting ']' at ") + pos);