ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

tool-call: fix Qwen 2.5 Coder support, add micro benchmarks, support trigger patterns for lazy grammars (#12034)

Browse source 
* sampler: turn lazy grammar trigger words to regexes

* add scripts/tool_bench.sh & .py

* constrain llama json output regardless of function name if matches at beginning

* update relaxed newline space rule in grammar tests

* support add_generation_prompt query parameter (useful for /apply_template)

* Update src/llama-grammar.cpp

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Olivier Chafik 2025-03-05 13:05:13 +00:00 • committed by GitHub
parent fa31c438e0
commit 669912d9a5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
26 changed files with 1314 additions and 408 deletions
Download patch file Download diff file Expand all files Collapse all files

15
src/llama-grammar.h
View file

@ -3,6 +3,7 @@
#include "llama.h"
#include <map>
#include <regex>
#include <string>
#include <vector>
@ -105,6 +106,11 @@ struct llama_grammar_parser {
void print(FILE * file);
};
struct llama_grammar_trigger_pattern {
std::string pattern;
std::regex regex;
};
struct llama_grammar {
// note: allow null vocab for testing (not great)
const llama_vocab * vocab;
@ -122,7 +128,10 @@ struct llama_grammar {
bool awaiting_trigger = false; // Initialized to true for lazy grammars only
std::string trigger_buffer; // Output buffered by lazy grammar. Will be cleared once trigger is found.
std::vector<llama_token> trigger_tokens; // Tokens that trigger a lazy grammar, or tokens to force printing of (even if special).
std::vector<std::string> trigger_words;
std::vector<llama_grammar_trigger_pattern>
trigger_patterns; // Regular expressions that trigger a lazy grammar. Must be a full match of the entire generated
// string, and the grammar will be given the string from the first match group onwards.
};
//
@ -141,8 +150,8 @@ struct llama_grammar * llama_grammar_init_impl(
const char * grammar_str,
const char * grammar_root,
bool lazy,
const char ** trigger_words,
size_t num_trigger_words,
const char ** trigger_patterns,
size_t num_trigger_patterns,
const llama_token * trigger_tokens,
size_t num_trigger_tokens);