mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 29 Packages Projects Releases Wiki Activity

common : determine generation prompt using longest common prefix (#22657)

Browse Source
This commit is contained in:
Aldehir Rojas
2026-05-03 17:18:23 -05:00
committed by GitHub
parent 048a490f76
commit e48034dfc9
1 changed files with 27 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/chat.cpp
+17 -8
View File
@@ -2116,6 +2116,21 @@ std::optional<common_chat_params> common_chat_try_specialized_template(
return std::nullopt; return std::nullopt;
} }
static std::string common_chat_templates_generation_prompt(const common_chat_template & tmpl, const autoparser::generation_params & inputs) {
autoparser::generation_params params = inputs;
params.add_generation_prompt = false;
std::string no_gen_prompt = common_chat_template_direct_apply_impl(tmpl, params);
params.add_generation_prompt = true;
std::string gen_prompt = common_chat_template_direct_apply_impl(tmpl, params);
size_t prefix_len = 0;
size_t min_size = std::min(no_gen_prompt.size(), gen_prompt.size());
while (prefix_len < min_size && no_gen_prompt[prefix_len] == gen_prompt[prefix_len]) {
prefix_len++;
}
return gen_prompt.substr(prefix_len);
}
static common_chat_params common_chat_templates_apply_jinja(const struct common_chat_templates * tmpls, static common_chat_params common_chat_templates_apply_jinja(const struct common_chat_templates * tmpls,
const struct common_chat_templates_inputs & inputs) { const struct common_chat_templates_inputs & inputs) {
autoparser::generation_params params; autoparser::generation_params params;
@@ -2130,6 +2145,7 @@ static common_chat_params common_chat_templates_apply_jinja(const struct common_
params.enable_thinking = inputs.enable_thinking; params.enable_thinking = inputs.enable_thinking;
params.grammar = inputs.grammar; params.grammar = inputs.grammar;
params.now = inputs.now; params.now = inputs.now;
params.add_generation_prompt = inputs.add_generation_prompt;
params.add_bos = tmpls->add_bos; params.add_bos = tmpls->add_bos;
params.add_eos = tmpls->add_eos; params.add_eos = tmpls->add_eos;
@@ -2153,14 +2169,7 @@ static common_chat_params common_chat_templates_apply_jinja(const struct common_
workaround::func_args_not_string(params.messages); workaround::func_args_not_string(params.messages);
} }
params.add_generation_prompt = false; params.generation_prompt = common_chat_templates_generation_prompt(tmpl, params);
std::string no_gen_prompt = common_chat_template_direct_apply_impl(tmpl, params);
params.add_generation_prompt = true;
std::string gen_prompt = common_chat_template_direct_apply_impl(tmpl, params);
auto diff = calculate_diff_split(no_gen_prompt, gen_prompt);
params.generation_prompt = diff.right + diff.suffix;
params.add_generation_prompt = inputs.add_generation_prompt;
params.extra_context = common_chat_extra_context(); params.extra_context = common_chat_extra_context();
for (auto el : inputs.chat_template_kwargs) { for (auto el : inputs.chat_template_kwargs) {