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

fixes based on review @cebtenzzre

Browse source
This commit is contained in:
Minsoo Cheong 2024-03-23 00:24:01 +09:00
parent 27f2e85520
commit 0a243da7d4
2 changed files with 5 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

4
common/sampling.cpp
View file

@ -173,7 +173,7 @@ static llama_token llama_sampling_sample_impl(
const float mirostat_tau = params.mirostat_tau;
const float mirostat_eta = params.mirostat_eta;
std::vector<float> original_logits = {};
std::vector<float> original_logits;
auto cur_p = llama_sampling_configure_token_candidates(ctx_sampling, ctx_main, ctx_cfg, idx, !is_resampling, &original_logits);
if (!is_resampling) {
GGML_ASSERT(!original_logits.empty());
@ -271,7 +271,7 @@ static llama_token_data_array llama_sampling_configure_token_candidates_impl(
if (apply_grammar && original_logits != NULL) {
// Only make a copy of the original logits if we are not applying grammar checks, not sure if I actually have to do this.
(*original_logits).insert((*original_logits).end(), logits, logits + llama_n_vocab(llama_get_model(ctx_main)));
*original_logits = {logits, logits + llama_n_vocab(llama_get_model(ctx_main))};
}
// apply params.logit_bias map