Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 13:58:46 +01:00
Code Issues Packages Projects Releases Wiki Activity

server : (proposal) allow user to customize chat template

Browse Source
This commit is contained in:
Xuan Son Nguyen 2024-11-20 20:36:58 +01:00
parent 42ae10bbcd
commit 1bc896fede
1 changed files with 21 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
examples/server/utils.hpp
View File

@ -300,8 +300,19 @@ static llama_tokens format_infill(
} }
// Format given chat. If tmpl is empty, we take the template from model metadata // Format given chat. If tmpl is empty, we take the template from model metadata
// If messages[i]["prefix"] or messages[i]["prefix"] is present, we format the chat with custom prefix/suffix
inline std::string format_chat(const struct llama_model * model, const std::string & tmpl, const std::vector<json> & messages) { inline std::string format_chat(const struct llama_model * model, const std::string & tmpl, const std::vector<json> & messages) {
std::vector<common_chat_msg> chat; std::vector<common_chat_msg> chat;
bool is_custom = false;
std::ostringstream oss;
// if at least one message has custom prefix/suffix, we switch to custom formatting
for (const auto & msg : messages) {
if (msg.contains("prefix") || msg.contains("suffix")) {
is_custom = true;
break;
}
}
for (size_t i = 0; i < messages.size(); ++i) { for (size_t i = 0; i < messages.size(); ++i) {
const auto & curr_msg = messages[i]; const auto & curr_msg = messages[i];
@ -325,10 +336,18 @@ inline std::string format_chat(const struct llama_model * model, const std::stri
throw std::runtime_error("Missing 'content' (ref: https://github.com/ggerganov/llama.cpp/issues/8367)"); throw std::runtime_error("Missing 'content' (ref: https://github.com/ggerganov/llama.cpp/issues/8367)");
} }
if (is_custom) {
std::string prefix = json_value(curr_msg, "prefix", std::string(""));
std::string suffix = json_value(curr_msg, "suffix", std::string(""));
oss << prefix << content << suffix;
} else {
chat.push_back({role, content}); chat.push_back({role, content});
} }
}
const auto formatted_chat = common_chat_apply_template(model, tmpl, chat, true); const auto formatted_chat = is_custom
? oss.str()
: common_chat_apply_template(model, tmpl, chat, true);
LOG_DBG("formatted_chat: '%s'\n", formatted_chat.c_str()); LOG_DBG("formatted_chat: '%s'\n", formatted_chat.c_str());
return formatted_chat; return formatted_chat;