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

llama : minor style changes

Browse source 
blindly applied in online editor - hopefully I didn't break something
This commit is contained in:
Georgi Gerganov 2023-07-21 15:11:23 +03:00 • committed by GitHub
parent c047e8aec2
commit 11315b1d61
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 10 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

19
llama.cpp
View file

@ -2442,8 +2442,7 @@ void llama_sample_frequency_and_presence_penalties(struct llama_context * ctx, l
void llama_sample_grammar(struct llama_context * ctx, llama_token_data_array * candidates, const struct llama_grammar * grammar) { void llama_sample_grammar(struct llama_context * ctx, llama_token_data_array * candidates, const struct llama_grammar * grammar) {
assert(ctx); assert(ctx);
const int64_t t_start_sample_us = ggml_time_us(); const int64_t t_start_sample_us = ggml_time_us();
const llama_token eos = llama_token_eos();
bool allow_eos = false; bool allow_eos = false;
for (const auto & stack : grammar->stacks) { for (const auto & stack : grammar->stacks) {
@ -2453,8 +2452,10 @@ void llama_sample_grammar(struct llama_context * ctx, llama_token_data_array * c
} }
} }
std::vector<std::vector<uint32_t>> decoded_candidates; const llama_token eos = llama_token_eos();
std::vector<llama_grammar_candidate> grammar_candidates;
std::vector<std::vector<uint32_t>> candidates_decoded;
std::vector<llama_grammar_candidate> candidates_grammar;
for (size_t i = 0; i < candidates->size; ++i) { for (size_t i = 0; i < candidates->size; ++i) {
const llama_token id = candidates->data[i].id; const llama_token id = candidates->data[i].id;
@ -2466,14 +2467,14 @@ void llama_sample_grammar(struct llama_context * ctx, llama_token_data_array * c
} else if (*str == 0) { } else if (*str == 0) {
candidates->data[i].logit = -INFINITY; candidates->data[i].logit = -INFINITY;
} else { } else {
decoded_candidates.push_back(decode_utf8(str)); candidates_decoded.push_back(decode_utf8(str));
grammar_candidates.push_back({ i, decoded_candidates.back().data() }); candidates_grammar.push_back({ i, candidates_decoded.back().data() });
} }
} }
auto rejects = const auto rejects =
llama_grammar_reject_candidates(grammar->rules, grammar->stacks, grammar_candidates); llama_grammar_reject_candidates(grammar->rules, grammar->stacks, candidates_grammar);
for (auto reject : rejects) { for (auto & reject : rejects) {
candidates->data[reject.index].logit = -INFINITY; candidates->data[reject.index].logit = -INFINITY;
} }