ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-06-13 04:28:07 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

talk-llama : sync latest llama.cpp (close #922, close #954)

Browse Source
This commit is contained in:
Georgi Gerganov
2023-05-23 14:04:39 +03:00
parent 041be06d58
commit 77eab3fbfe
4 changed files with 287 additions and 167 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
examples/talk-llama/talk-llama.cpp
View File

@ -33,8 +33,6 @@ struct whisper_params {
int32_t max_tokens = 32;
int32_t audio_ctx = 0;
int32_t n_parts_llama = -1;
float vad_thold = 0.6f;
float freq_thold = 100.0f;
@ -72,7 +70,6 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
else if (arg == "-ac" || arg == "--audio-ctx") { params.audio_ctx = std::stoi(argv[++i]); }
else if (arg == "-vth" || arg == "--vad-thold") { params.vad_thold = std::stof(argv[++i]); }
else if (arg == "-fth" || arg == "--freq-thold") { params.freq_thold = std::stof(argv[++i]); }
else if (arg == "--n-parts-llama") { params.n_parts_llama = std::stoi(argv[++i]); }
else if (arg == "-su" || arg == "--speed-up") { params.speed_up = true; }
else if (arg == "-tr" || arg == "--translate") { params.translate = true; }
else if (arg == "-ps" || arg == "--print-special") { params.print_special = true; }
@ -123,7 +120,6 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
fprintf(stderr, " -l LANG, --language LANG [%-7s] spoken language\n", params.language.c_str());
fprintf(stderr, " -mw FILE, --model-whisper [%-7s] whisper model file\n", params.model_wsp.c_str());
fprintf(stderr, " -ml FILE, --model-llama [%-7s] llama model file\n", params.model_llama.c_str());
fprintf(stderr, " --n-parts-llama N [%-7d] num parts in llama model file\n", params.n_parts_llama);
fprintf(stderr, " -s FILE, --speak TEXT [%-7s] command for TTS\n", params.speak.c_str());
fprintf(stderr, " --prompt-file FNAME [%-7s] file with custom prompt to start dialog\n", "");
fprintf(stderr, " --session FNAME file to cache model state in (may be large!) (default: none)\n");
@ -239,13 +235,14 @@ int main(int argc, char ** argv) {
// llama init
llama_init_backend();
auto lparams = llama_context_default_params();
// tune these to your liking
lparams.n_ctx = 2048;
lparams.seed = 1;
lparams.f16_kv = true;
lparams.n_parts = params.n_parts_llama;
struct llama_context * ctx_llama = llama_init_from_file(params.model_llama.c_str(), lparams);