tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-27 03:55:20 +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

56
common/common.h
View File

@ -616,62 +616,6 @@ std::string common_detokenize(
const std::vector<llama_token> & tokens,
bool special = true);
//
// Chat template utils
//
struct common_tool_call {
std::string name;
std::string arguments;
std::string id;
};
// same with llama_chat_message, but uses std::string
struct common_chat_msg {
std::string role;
std::string content;
std::vector<common_tool_call> tool_calls;
std::string reasoning_content = "";
};
// Check if the template supplied via "--chat-template" is supported or not. Returns true if it's valid
bool common_chat_verify_template(const std::string & tmpl, bool use_jinja);
namespace minja {
class chat_template;
}
typedef minja::chat_template common_chat_template;
struct common_chat_templates {
bool has_explicit_template; // Model had builtin template or template overridde was specified.
std::unique_ptr<common_chat_template> template_default; // always set (defaults to chatml)
std::unique_ptr<common_chat_template> template_tool_use;
};
// CPP wrapper for llama_chat_apply_template
// If the built-in template is not supported, we default to chatml
// If the custom "tmpl" is not supported, we throw an error
std::string common_chat_apply_template(
const common_chat_template & tmpl,
const std::vector<common_chat_msg> & chat,
bool add_ass,
bool use_jinja);
// Format single message, while taking into account the position of that message in chat history
std::string common_chat_format_single(
const common_chat_template & tmpl,
const std::vector<common_chat_msg> & past_msg,
const common_chat_msg & new_msg,
bool add_ass,
bool use_jinja);
// Returns an example of formatted chat
std::string common_chat_format_example(
const common_chat_template & tmpl, bool use_jinja);
common_chat_templates common_chat_templates_from_model(const struct llama_model * model, const std::string & chat_template_override);
//
// KV cache utils
//