CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-21 12:16:57 +00:00
Code Issues Packages Projects Releases Wiki Activity

chat: fix int overflow, prevent size calculation in float/double (#17357)

Browse Source 
* chat: fix int overflow, prevent size calculation in float/double

* Update common/chat.cpp

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Xuan-Son Nguyen
2025-11-18 19:11:53 +01:00
committed by GitHub
parent a045492088
commit 10e9780154
1 changed files with 8 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
common/chat.cpp
View File

@@ -3359,7 +3359,7 @@ static common_chat_params common_chat_templates_apply_legacy(
const struct common_chat_templates * tmpls, const struct common_chat_templates * tmpls,
const struct common_chat_templates_inputs & inputs) const struct common_chat_templates_inputs & inputs)
{ {
int alloc_size = 0; size_t alloc_size = 0;
std::vector<llama_chat_message> chat; std::vector<llama_chat_message> chat;
std::vector<std::string> contents; std::vector<std::string> contents;
@@ -3381,7 +3381,8 @@ static common_chat_params common_chat_templates_apply_legacy(
const auto & msg = inputs.messages[i]; const auto & msg = inputs.messages[i];
const auto & content = contents[i]; const auto & content = contents[i];
chat.push_back({msg.role.c_str(), content.c_str()}); chat.push_back({msg.role.c_str(), content.c_str()});
alloc_size += (msg.role.size() + content.size()) * 1.25; size_t msg_size = msg.role.size() + content.size();
alloc_size += msg_size + (msg_size / 4); // == msg_size * 1.25 but avoiding float ops
} }
std::vector<char> buf(alloc_size); std::vector<char> buf(alloc_size);
@@ -3403,6 +3404,11 @@ static common_chat_params common_chat_templates_apply_legacy(
res = llama_chat_apply_template(src.c_str(), chat.data(), chat.size(), inputs.add_generation_prompt, buf.data(), buf.size()); res = llama_chat_apply_template(src.c_str(), chat.data(), chat.size(), inputs.add_generation_prompt, buf.data(), buf.size());
} }
// for safety, we check the result again
if (res < 0 || (size_t) res > buf.size()) {
throw std::runtime_error("failed to apply chat template, try using --jinja");
}
common_chat_params params; common_chat_params params;
params.prompt = std::string(buf.data(), res); params.prompt = std::string(buf.data(), res);
if (!inputs.json_schema.empty()) { if (!inputs.json_schema.empty()) {