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

llama : minor grammar refactor (#10897)

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-12-19 17:42:13 +02:00 • committed by GitHub
parent 36319dec5d
commit 5cab3e4aaa
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 26 additions and 37 deletions
Download patch file Download diff file Expand all files Collapse all files

30
src/llama-grammar.cpp
View file

@ -822,15 +822,11 @@ llama_grammar_stacks & llama_grammar_get_stacks(struct llama_grammar * grammar)
return grammar->stacks;
}
void llama_grammar_accept(
const llama_grammar_rules & rules,
const llama_grammar_stacks & stacks,
const uint32_t chr,
llama_grammar_stacks & stacks_new) {
stacks_new.clear();
stacks_new.reserve(stacks.size());
void llama_grammar_accept(struct llama_grammar * grammar, uint32_t chr) {
llama_grammar_stacks stacks_new;
stacks_new.reserve(grammar->stacks.size());
for (const auto & stack : stacks) {
for (const auto & stack : grammar->stacks) {
if (stack.empty()) {
continue;
}
@ -844,9 +840,11 @@ void llama_grammar_accept(
if (!llama_grammar_is_end_of_sequence(pos)) {
new_stack.push_back(pos);
}
llama_grammar_advance_stack(rules, new_stack, stacks_new);
llama_grammar_advance_stack(grammar->rules, new_stack, stacks_new);
}
}
grammar->stacks = std::move(stacks_new);
}
llama_grammar_candidates llama_grammar_reject_candidates_for_stack(
@ -1051,7 +1049,12 @@ void llama_grammar_free_impl(struct llama_grammar * grammar) {
}
struct llama_grammar * llama_grammar_clone_impl(const struct llama_grammar & grammar) {
llama_grammar * result = new llama_grammar { grammar.vocab, grammar.rules, grammar.stacks, grammar.partial_utf8, };
llama_grammar * result = new llama_grammar {
grammar.vocab,
grammar.rules,
grammar.stacks,
grammar.partial_utf8,
};
// redirect elements in stacks to point to new rules
for (size_t is = 0; is < result->stacks.size(); is++) {
@ -1059,7 +1062,7 @@ struct llama_grammar * llama_grammar_clone_impl(const struct llama_grammar & gra
for (size_t ir0 = 0; ir0 < grammar.rules.size(); ir0++) {
for (size_t ir1 = 0; ir1 < grammar.rules[ir0].size(); ir1++) {
if (grammar.stacks[is][ie] == &grammar.rules[ir0][ir1]) {
result->stacks[is][ie] = &result->rules[ir0][ir1];
result->stacks[is][ie] = &result->rules[ir0][ir1];
}
}
}
@ -1126,11 +1129,8 @@ void llama_grammar_accept_impl(struct llama_grammar & grammar, llama_token token
const auto decoded = decode_utf8(piece, grammar.partial_utf8);
const auto & code_points = decoded.first;
llama_grammar_stacks stacks_new;
for (auto it = code_points.begin(), end = code_points.end() - 1; it != end; ++it) {
llama_grammar_accept(grammar.rules, grammar.stacks, *it, stacks_new);
grammar.stacks = std::move(stacks_new);
llama_grammar_accept(&grammar, *it);
}
grammar.partial_utf8 = decoded.second;