vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

--think to force any model to return reasoning_content (or just parse <think> for deepseek r1)

Browse source
This commit is contained in:
ochafik 2025-02-05 12:16:37 +00:00
parent 5d60cebbcc
commit 9d7c3cc51b
9 changed files with 306 additions and 145 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/server/server.cpp
View file

@ -4052,7 +4052,7 @@ int main(int argc, char ** argv) {
}
auto body = json::parse(req.body);
json data = oaicompat_completion_params_parse(body, params.use_jinja, ctx_server.chat_templates);
json data = oaicompat_completion_params_parse(body, params.use_jinja, params.think, ctx_server.chat_templates);
return handle_completions_impl(
SERVER_TASK_TYPE_COMPLETION,
@ -4065,7 +4065,7 @@ int main(int argc, char ** argv) {
// same with handle_chat_completions, but without inference part
const auto handle_apply_template = [&ctx_server, &params, &res_ok](const httplib::Request & req, httplib::Response & res) {
auto body = json::parse(req.body);
json data = oaicompat_completion_params_parse(body, params.use_jinja, ctx_server.chat_templates);
json data = oaicompat_completion_params_parse(body, params.use_jinja, params.think, ctx_server.chat_templates);
res_ok(res, {{ "prompt", std::move(data.at("prompt")) }});
};