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

Tool call support (generic + native for Llama, Functionary, Hermes, Mistral, Firefunction, DeepSeek) w/ lazy grammars (#9639)

Browse source 
---------

Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
This commit is contained in:
Olivier Chafik 2025-01-30 19:13:58 +00:00 • committed by GitHub
parent 27d135c970
commit 8b576b6c55
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
48 changed files with 3861 additions and 156 deletions
Download patch file Download diff file Expand all files Collapse all files

11
common/sampling.cpp
View file

@ -151,9 +151,18 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, co
lparams.no_perf = params.no_perf;
std::vector<const char *> trigger_words;
trigger_words.reserve(params.grammar_trigger_words.size());
for (const auto & str : params.grammar_trigger_words) {
trigger_words.push_back(str.word.c_str());
}
auto * result = new common_sampler {
/* .params = */ params,
/* .grmr = */ llama_sampler_init_grammar(vocab, params.grammar.c_str(), "root"),
/* .grmr = */ params.grammar_lazy
? llama_sampler_init_grammar_lazy(vocab, params.grammar.c_str(), "root",
trigger_words.data(), trigger_words.size(),
params.grammar_trigger_tokens.data(), params.grammar_trigger_tokens.size())
: llama_sampler_init_grammar(vocab, params.grammar.c_str(), "root"),
/* .chain = */ llama_sampler_chain_init(lparams),
/* .prev = */ ring_buffer<llama_token>(std::max(32, params.n_prev)),
/* .cur = */ {},