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

Save and load example adjust

Browse source
This commit is contained in:
Ivan Stepanov 2023-04-28 20:19:17 +03:00
parent 6c4c88d54f
commit 416f49182a
1 changed files with 18 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

34
examples/save-load-state/save-load-state.cpp
View file

@ -64,14 +64,15 @@ int main(int argc, char ** argv) {
// first run // first run
printf("\n%s", params.prompt.c_str()); printf("\n%s", params.prompt.c_str());
for (auto i = 0; i < params.n_predict; i++) { for (auto i = 0; i < params.n_predict; i++) {
auto next_token = llama_sample_top_p_top_k( auto logits = llama_get_logits(ctx);
ctx, auto n_vocab = llama_n_vocab(ctx);
&last_n_tokens_data.back() - params.repeat_last_n, std::vector<llama_token_data> candidates;
params.repeat_last_n, candidates.reserve(n_vocab);
40, for (llama_token token_id = 0; token_id < n_vocab; token_id++) {
1.0, candidates.emplace_back(llama_token_data{token_id, logits[token_id], 0.0f});
1.0, }
1.1); llama_token_data_array candidates_p = { candidates.data(), candidates.size(), false };
auto next_token = llama_sample_token(ctx, &candidates_p);
auto next_token_str = llama_token_to_str(ctx, next_token); auto next_token_str = llama_token_to_str(ctx, next_token);
last_n_tokens_data.push_back(next_token); last_n_tokens_data.push_back(next_token);
printf("%s", next_token_str); printf("%s", next_token_str);
@ -106,14 +107,15 @@ int main(int argc, char ** argv) {
// second run // second run
for (auto i = 0; i < params.n_predict; i++) { for (auto i = 0; i < params.n_predict; i++) {
auto next_token = llama_sample_top_p_top_k( auto logits = llama_get_logits(ctx2);
ctx2, auto n_vocab = llama_n_vocab(ctx2);
&last_n_tokens_data.back() - params.repeat_last_n, std::vector<llama_token_data> candidates;
params.repeat_last_n, candidates.reserve(n_vocab);
40, for (llama_token token_id = 0; token_id < n_vocab; token_id++) {
1.0, candidates.emplace_back(llama_token_data{token_id, logits[token_id], 0.0f});
1.0, }
1.1); llama_token_data_array candidates_p = { candidates.data(), candidates.size(), false };
auto next_token = llama_sample_token(ctx2, &candidates_p);
auto next_token_str = llama_token_to_str(ctx2, next_token); auto next_token_str = llama_token_to_str(ctx2, next_token);
last_n_tokens_data.push_back(next_token); last_n_tokens_data.push_back(next_token);
printf("%s", next_token_str); printf("%s", next_token_str);