ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-06-14 12:58:10 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

talk : improve prompting

Browse Source
This commit is contained in:
Georgi Gerganov
2022-12-12 23:44:36 +02:00
parent 930c693989
commit a613f16aec
4 changed files with 39 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
examples/talk/gpt-2.cpp
View File

@ -139,7 +139,7 @@ gpt_vocab::id gpt_sample_top_k_top_p(
}
//printf("\n");
//for (int i = 0; i < (int)logits_id.size(); i++) {
//for (int i = 0; i < (int) logits_id.size(); i++) {
// printf("%d: '%s' %f\n", i, vocab.id_to_token.at(logits_id[i].second).c_str(), logits_id[i].first);
//}
//exit(0);
@ -825,8 +825,8 @@ Me too.
int32_t n_threads = std::min(N_THREAD, (int) std::thread::hardware_concurrency());
// sampling parameters
int32_t top_k = 20;
float top_p = 0.98f;
int32_t top_k = 5;
float top_p = 0.9f;
float temp = 1.0f;
};
@ -840,7 +840,7 @@ struct gpt2_context * gpt2_init(const char * path_model) {
const int64_t t_start_us = ggml_time_us();
if (!gpt2_model_load(path_model, ctx->model, ctx->vocab)) {
fprintf(stderr, "%s: failed to load model from '%s'\n", __func__, "gpt-2.bin");
fprintf(stderr, "%s: failed to load model from '%s'\n", __func__, path_model);
return nullptr;
}
@ -913,10 +913,7 @@ std::string gpt2_gen_text(gpt2_context * ctx, const char * text, int max_tokens)
result += ctx->vocab.id_to_token[embd[0]];
// end of text token
if (embd.back() == 50256 ||
ctx->vocab.id_to_token[embd.back()] == "." ||
ctx->vocab.id_to_token[embd.back()] == "!" ||
ctx->vocab.id_to_token[embd.back()] == "?") {
if (embd.back() == 50256) {
break;
}
}