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

stream : "-kc" now enables context keeping from previous segment (#90)

Browse Source 
By default, the context keeping is disabled
This commit is contained in:
Georgi Gerganov
2022-11-22 18:20:05 +02:00
parent 63ae03b8e0
commit 385236d1d3
3 changed files with 17 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
examples/stream/stream.cpp
View File

@ -336,7 +336,7 @@ int main(int argc, char ** argv) {
wparams.print_realtime = false;
wparams.print_timestamps = !params.no_timestamps;
wparams.translate = params.translate;
wparams.no_context = params.no_context;
wparams.no_context = true;
wparams.single_segment = true;
wparams.max_tokens = params.max_tokens;
wparams.language = params.language.c_str();
@ -345,9 +345,9 @@ int main(int argc, char ** argv) {
wparams.audio_ctx = params.audio_ctx;
wparams.speed_up = params.speed_up;
wparams.prompt_tokens = prompt_tokens.data();
wparams.prompt_n_tokens = prompt_tokens.size();
wparams.prompt_tokens = params.no_context ? nullptr : prompt_tokens.data();
wparams.prompt_n_tokens = params.no_context ? 0 : prompt_tokens.size();
if (whisper_full(ctx, wparams, pcmf32.data(), pcmf32.size()) != 0) {
fprintf(stderr, "%s: failed to process audio\n", argv[0]);
return 6;
@ -399,12 +399,15 @@ int main(int argc, char ** argv) {
pcmf32_old = std::vector<float>(pcmf32.end() - n_samples_keep, pcmf32.end());
// Add tokens of the last full length segment as the prompt
prompt_tokens.clear();
const int n_segments = whisper_full_n_segments(ctx);
for (int i = 0; i < n_segments; ++i) {
const int token_count = whisper_full_n_tokens(ctx, i);
for (int j = 0; j < token_count; ++j) {
prompt_tokens.push_back(whisper_full_get_token_id(ctx, i, j));
if (!params.no_context) {
prompt_tokens.clear();
const int n_segments = whisper_full_n_segments(ctx);
for (int i = 0; i < n_segments; ++i) {
const int token_count = whisper_full_n_tokens(ctx, i);
for (int j = 0; j < token_count; ++j) {
prompt_tokens.push_back(whisper_full_get_token_id(ctx, i, j));
}
}
}
}