AaronPo/the-biergarten-app
1
0
Fork 0
mirror of https://github.com/aaronpo97/the-biergarten-app.git synced 2026-06-01 01:54:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix llama grammar

Browse Source
This commit is contained in:
Aaron Po
2026-04-15 23:28:27 -04:00
parent 62dfb5e14a
commit 6682b5de01
7 changed files with 23 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
pipeline/src/data_generation/llama/helpers.cc
View File

@@ -84,9 +84,8 @@ std::string PrepareRegionContext(std::string_view region_context,
std::string ToChatPrompt(const llama_model* model,
const std::string& system_prompt,
const std::string& user_prompt) {
std::string combined_prompt = system_prompt;
combined_prompt.append("\n\n");
combined_prompt.append(user_prompt);
std::string combined_prompt =
std::format("{}\n\n{}", system_prompt, user_prompt);
const char* tmpl = llama_model_chat_template(model, nullptr);
if (tmpl == nullptr) {
@@ -103,9 +102,9 @@ std::string ToChatPrompt(const llama_model* model,
constexpr std::size_t min_template_buffer_size = 1024;
std::vector<char> buffer(std::max<std::size_t>(
min_template_buffer_size,
(system_prompt.size() + user_prompt.size()) * 4));
std::vector<char> buffer(
std::max<std::size_t>(min_template_buffer_size,
(system_prompt.size() + user_prompt.size()) * 4));
auto apply_template_with_resize = [&](const llama_chat_message* chat_messages,
int32_t message_count) -> int32_t {