thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

common/parser: handle reasoning budget (#20297)

Browse source 
* v1

* Finished!

* Handlie cli

* Reasoning sampler

* Apply suggestions from code review

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

* Less explosive terminology :)

* Add utf-8 case and tests

* common : migrate reasoning budget sampler to common

* cont : clean up

* cont : expose state and allow passing as initial state

* cont : remove unused imports

* cont : update state machine doc string

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
Co-authored-by: Alde Rojas <hello@alde.dev>
This commit is contained in:
Piotr Wilkin (ilintar) 2026-03-11 10:26:12 +01:00 • committed by GitHub
parent 5f91b1d5d5
commit acb7c79069
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
18 changed files with 670 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

22
common/chat.cpp
View file

@ -857,7 +857,9 @@ static common_chat_params common_chat_params_init_ministral_3(const common_chat_
auto extract_reasoning = inputs.reasoning_format != COMMON_REASONING_FORMAT_NONE;
auto include_grammar = true;
data.supports_thinking = true;
data.supports_thinking = true;
data.thinking_start_tag = "[THINK]";
data.thinking_end_tag = "[/THINK]";
data.prompt = common_chat_template_direct_apply(tmpl, inputs, /* messages_override = */ adjusted_messages);
data.format = COMMON_CHAT_FORMAT_PEG_NATIVE;
data.preserved_tokens = {
@ -1165,9 +1167,11 @@ static common_chat_params common_chat_params_init_kimi_k2(const common_chat_temp
const autoparser::templates_params & inputs) {
common_chat_params data;
data.prompt = common_chat_template_direct_apply(tmpl, inputs);
data.format = COMMON_CHAT_FORMAT_PEG_NATIVE;
data.supports_thinking = true;
data.prompt = common_chat_template_direct_apply(tmpl, inputs);
data.format = COMMON_CHAT_FORMAT_PEG_NATIVE;
data.supports_thinking = true;
data.thinking_start_tag = "<think>";
data.thinking_end_tag = "</think>";
data.preserved_tokens = {
"<|tool_calls_section_begin|>",
"<|tool_calls_section_end|>",
@ -1527,6 +1531,16 @@ static common_chat_params common_chat_templates_apply_jinja(const struct common_
autoparser.analyze_template(tmpl);
auto auto_params = autoparser::peg_generator::generate_parser(tmpl, params, autoparser);
auto_params.supports_thinking = autoparser.reasoning.mode != autoparser::reasoning_mode::NONE;
if (auto_params.supports_thinking) {
auto_params.thinking_start_tag = autoparser.reasoning.start;
auto_params.thinking_end_tag = autoparser.reasoning.end;
// FORCED_OPEN and FORCED_CLOSED both put <think> in the generation prompt
// (FORCED_CLOSED forces empty <think></think> when thinking is disabled,
// but forces <think> open when thinking is enabled)
auto_params.thinking_forced_open =
autoparser.reasoning.mode == autoparser::reasoning_mode::FORCED_OPEN ||
autoparser.reasoning.mode == autoparser::reasoning_mode::FORCED_CLOSED;
}
return auto_params;
} catch (const std::exception & e) {
throw std::invalid_argument(std::string("Unable to generate parser for this template. Automatic parser generation failed: ") + e.what());