tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-27 12:05:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

tool-call: refactor common chat / tool-call api (+ tests / fixes) (#11900)

Browse Source 
* tool-call refactoring: moved common_chat_* to chat.h, common_chat_templates_init return a unique_ptr to opaque type

* addressed clang-tidy lints in [test-]chat.*

* rm minja deps from util & common & move it to common/minja/

* add name & tool_call_id to common_chat_msg

* add common_chat_tool

* added json <-> tools, msgs conversions to chat.h

* fix double bos/eos jinja avoidance hack (was preventing inner bos/eos tokens)

* fix deepseek r1 slow test (no longer <think> opening w/ new template)

* allow empty tools w/ auto + grammar

* fix & test server grammar & json_schema params w/ & w/o --jinja
This commit is contained in:
Olivier Chafik
2025-02-18 18:03:23 +00:00
committed by GitHub
parent 63ac128563
commit 63e489c025
18 changed files with 1385 additions and 993 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
examples/main/main.cpp
View File

@ -4,7 +4,7 @@
#include "log.h"
#include "sampling.h"
#include "llama.h"
#include "chat-template.hpp"
#include "chat.h"
#include <cstdio>
#include <cstring>
@ -158,7 +158,7 @@ int main(int argc, char ** argv) {
}
const llama_vocab * vocab = llama_model_get_vocab(model);
auto chat_templates = common_chat_templates_from_model(model, params.chat_template);
auto chat_templates = common_chat_templates_init(model, params.chat_template);
LOG_INF("%s: llama threadpool init, n_threads = %d\n", __func__, (int) params.cpuparams.n_threads);
@ -201,7 +201,7 @@ int main(int argc, char ** argv) {
}
// auto enable conversation mode if chat template is available
const bool has_chat_template = chat_templates.has_explicit_template && chat_templates.template_default;
const bool has_chat_template = common_chat_templates_was_explicit(chat_templates.get());
if (params.conversation_mode == COMMON_CONVERSATION_MODE_AUTO) {
if (has_chat_template) {
LOG_INF("%s: chat template is available, enabling conversation mode (disable it with -no-cnv)\n", __func__);
@ -219,7 +219,7 @@ int main(int argc, char ** argv) {
// print chat template example in conversation mode
if (params.conversation_mode) {
if (params.enable_chat_template) {
LOG_INF("%s: chat template example:\n%s\n", __func__, common_chat_format_example(*chat_templates.template_default, params.use_jinja).c_str());
LOG_INF("%s: chat template example:\n%s\n", __func__, common_chat_format_example(chat_templates.get(), params.use_jinja).c_str());
} else {
LOG_INF("%s: in-suffix/prefix is specified, chat template will be disabled\n", __func__);
}
@ -264,9 +264,11 @@ int main(int argc, char ** argv) {
std::vector<llama_token> embd_inp;
auto chat_add_and_format = [&chat_msgs, &chat_templates](const std::string & role, const std::string & content) {
common_chat_msg new_msg{role, content, {}};
auto formatted = common_chat_format_single(*chat_templates.template_default, chat_msgs, new_msg, role == "user", g_params->use_jinja);
chat_msgs.push_back({role, content, {}});
common_chat_msg new_msg;
new_msg.role = role;
new_msg.content = content;
auto formatted = common_chat_format_single(chat_templates.get(), chat_msgs, new_msg, role == "user", g_params->use_jinja);
chat_msgs.push_back(new_msg);
LOG_DBG("formatted: '%s'\n", formatted.c_str());
return formatted;
};
@ -755,11 +757,14 @@ int main(int argc, char ** argv) {
// check for reverse prompt using special tokens
llama_token last_token = common_sampler_last(smpl);
if (std::find(antiprompt_token.begin(), antiprompt_token.end(), last_token) != antiprompt_token.end()) {
if (params.interactive) {
is_interacting = true;
for (auto token : antiprompt_token) {
if (token == last_token) {
if (params.interactive) {
is_interacting = true;
}
is_antiprompt = true;
break;
}
is_antiprompt = true;
}
if (is_antiprompt) {