ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-06-12 20:18:08 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

talk-llama : sync llama.cpp (#2709)

Browse Source
This commit is contained in:
Georgi Gerganov
2025-01-13 08:55:48 +02:00
committed by GitHub
parent 45d3faf961
commit 35d0e02c72
36 changed files with 12960 additions and 11341 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
examples/talk-llama/llama-grammar.cpp
View File

@ -1,5 +1,6 @@
#include "llama-grammar.h"
#include "llama-impl.h"
#include "llama-vocab.h"
#include "llama-sampling.h"
@ -822,15 +823,11 @@ 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 +841,11 @@ 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 +1050,12 @@ 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 +1063,7 @@ 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 +1130,8 @@ 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;