forked from extern/whisper.cpp
whisper : add context param to disable gpu (#1293)
* whisper : check state->ctx_metal not null
* whisper : add whisper_context_params { use_gpu }
* whisper : new API with params & deprecate old API
* examples : use no-gpu param && whisper_init_from_file_with_params
* whisper.objc : enable metal & disable on simulator
* whisper.swiftui, metal : enable metal & support load default.metallib
* whisper.android : use new API
* bindings : use new API
* addon.node : fix build & test
* bindings : updata java binding
* bindings : add missing whisper_context_default_params_by_ref WHISPER_API for java
* metal : use SWIFTPM_MODULE_BUNDLE for GGML_SWIFT and reuse library load
* metal : move bundle var into block
* metal : use SWIFT_PACKAGE instead of GGML_SWIFT
* style : minor updates
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Jhen-Jie Hong
GitHub
@ -63,6 +63,7 @@ struct whisper_params {
|
||||
bool print_energy = false;
|
||||
bool no_timestamps = true;
|
||||
bool verbose_prompt = false;
|
||||
bool use_gpu = true;
|
||||
|
||||
std::string person = "Georgi";
|
||||
std::string language = "en";
|
||||
@ -84,25 +85,26 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
|
||||
whisper_print_usage(argc, argv, params);
|
||||
exit(0);
|
||||
}
|
||||
else if (arg == "-t" || arg == "--threads") { params.n_threads = std::stoi(argv[++i]); }
|
||||
else if (arg == "-vms" || arg == "--voice-ms") { params.voice_ms = std::stoi(argv[++i]); }
|
||||
else if (arg == "-c" || arg == "--capture") { params.capture_id = std::stoi(argv[++i]); }
|
||||
else if (arg == "-mt" || arg == "--max-tokens") { params.max_tokens = std::stoi(argv[++i]); }
|
||||
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 == "-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; }
|
||||
else if (arg == "-pe" || arg == "--print-energy") { params.print_energy = true; }
|
||||
else if (arg == "--verbose-prompt") { params.verbose_prompt = true; }
|
||||
else if (arg == "-p" || arg == "--person") { params.person = argv[++i]; }
|
||||
else if (arg == "--session") { params.path_session = argv[++i];}
|
||||
else if (arg == "-l" || arg == "--language") { params.language = argv[++i]; }
|
||||
else if (arg == "-mw" || arg == "--model-whisper") { params.model_wsp = argv[++i]; }
|
||||
else if (arg == "-ml" || arg == "--model-llama") { params.model_llama = argv[++i]; }
|
||||
else if (arg == "-s" || arg == "--speak") { params.speak = argv[++i]; }
|
||||
else if (arg == "--prompt-file") {
|
||||
else if (arg == "-t" || arg == "--threads") { params.n_threads = std::stoi(argv[++i]); }
|
||||
else if (arg == "-vms" || arg == "--voice-ms") { params.voice_ms = std::stoi(argv[++i]); }
|
||||
else if (arg == "-c" || arg == "--capture") { params.capture_id = std::stoi(argv[++i]); }
|
||||
else if (arg == "-mt" || arg == "--max-tokens") { params.max_tokens = std::stoi(argv[++i]); }
|
||||
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 == "-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; }
|
||||
else if (arg == "-pe" || arg == "--print-energy") { params.print_energy = true; }
|
||||
else if (arg == "-vp" || arg == "--verbose-prompt") { params.verbose_prompt = true; }
|
||||
else if (arg == "-ng" || arg == "--no-gpu") { params.use_gpu = false; }
|
||||
else if (arg == "-p" || arg == "--person") { params.person = argv[++i]; }
|
||||
else if (arg == "--session") { params.path_session = argv[++i];}
|
||||
else if (arg == "-l" || arg == "--language") { params.language = argv[++i]; }
|
||||
else if (arg == "-mw" || arg == "--model-whisper") { params.model_wsp = argv[++i]; }
|
||||
else if (arg == "-ml" || arg == "--model-llama") { params.model_llama = argv[++i]; }
|
||||
else if (arg == "-s" || arg == "--speak") { params.speak = argv[++i]; }
|
||||
else if (arg == "--prompt-file") {
|
||||
std::ifstream file(argv[++i]);
|
||||
std::copy(std::istreambuf_iterator<char>(file), std::istreambuf_iterator<char>(), back_inserter(params.prompt));
|
||||
if (params.prompt.back() == '\n') {
|
||||
@ -110,6 +112,7 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
|
||||
}
|
||||
}
|
||||
else if (arg == "-f" || arg == "--file") { params.fname_out = argv[++i]; }
|
||||
else if (arg == "-ng" || arg == "--no-gpu") { params.use_gpu = false; }
|
||||
else {
|
||||
fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
|
||||
whisper_print_usage(argc, argv, params);
|
||||
@ -125,27 +128,28 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
|
||||
fprintf(stderr, "usage: %s [options]\n", argv[0]);
|
||||
fprintf(stderr, "\n");
|
||||
fprintf(stderr, "options:\n");
|
||||
fprintf(stderr, " -h, --help [default] show this help message and exit\n");
|
||||
fprintf(stderr, " -t N, --threads N [%-7d] number of threads to use during computation\n", params.n_threads);
|
||||
fprintf(stderr, " -vms N, --voice-ms N [%-7d] voice duration in milliseconds\n", params.voice_ms);
|
||||
fprintf(stderr, " -c ID, --capture ID [%-7d] capture device ID\n", params.capture_id);
|
||||
fprintf(stderr, " -mt N, --max-tokens N [%-7d] maximum number of tokens per audio chunk\n", params.max_tokens);
|
||||
fprintf(stderr, " -ac N, --audio-ctx N [%-7d] audio context size (0 - all)\n", params.audio_ctx);
|
||||
fprintf(stderr, " -vth N, --vad-thold N [%-7.2f] voice activity detection threshold\n", params.vad_thold);
|
||||
fprintf(stderr, " -fth N, --freq-thold N [%-7.2f] high-pass frequency cutoff\n", params.freq_thold);
|
||||
fprintf(stderr, " -su, --speed-up [%-7s] speed up audio by x2 (reduced accuracy)\n", params.speed_up ? "true" : "false");
|
||||
fprintf(stderr, " -tr, --translate [%-7s] translate from source language to english\n", params.translate ? "true" : "false");
|
||||
fprintf(stderr, " -ps, --print-special [%-7s] print special tokens\n", params.print_special ? "true" : "false");
|
||||
fprintf(stderr, " -pe, --print-energy [%-7s] print sound energy (for debugging)\n", params.print_energy ? "true" : "false");
|
||||
fprintf(stderr, " -p NAME, --person NAME [%-7s] person name (for prompt selection)\n", params.person.c_str());
|
||||
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, " -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");
|
||||
fprintf(stderr, " --verbose-prompt [%-7s] print prompt at start\n", params.verbose_prompt ? "true" : "false");
|
||||
fprintf(stderr, " -f FNAME, --file FNAME [%-7s] text output file name\n", params.fname_out.c_str());
|
||||
fprintf(stderr, " -h, --help [default] show this help message and exit\n");
|
||||
fprintf(stderr, " -t N, --threads N [%-7d] number of threads to use during computation\n", params.n_threads);
|
||||
fprintf(stderr, " -vms N, --voice-ms N [%-7d] voice duration in milliseconds\n", params.voice_ms);
|
||||
fprintf(stderr, " -c ID, --capture ID [%-7d] capture device ID\n", params.capture_id);
|
||||
fprintf(stderr, " -mt N, --max-tokens N [%-7d] maximum number of tokens per audio chunk\n", params.max_tokens);
|
||||
fprintf(stderr, " -ac N, --audio-ctx N [%-7d] audio context size (0 - all)\n", params.audio_ctx);
|
||||
fprintf(stderr, " -vth N, --vad-thold N [%-7.2f] voice activity detection threshold\n", params.vad_thold);
|
||||
fprintf(stderr, " -fth N, --freq-thold N [%-7.2f] high-pass frequency cutoff\n", params.freq_thold);
|
||||
fprintf(stderr, " -su, --speed-up [%-7s] speed up audio by x2 (reduced accuracy)\n", params.speed_up ? "true" : "false");
|
||||
fprintf(stderr, " -tr, --translate [%-7s] translate from source language to english\n", params.translate ? "true" : "false");
|
||||
fprintf(stderr, " -ps, --print-special [%-7s] print special tokens\n", params.print_special ? "true" : "false");
|
||||
fprintf(stderr, " -pe, --print-energy [%-7s] print sound energy (for debugging)\n", params.print_energy ? "true" : "false");
|
||||
fprintf(stderr, " -vp, --verbose-prompt [%-7s] print prompt at start\n", params.verbose_prompt ? "true" : "false");
|
||||
fprintf(stderr, " -ng, --no-gpu [%-7s] disable GPU\n", params.use_gpu ? "false" : "true");
|
||||
fprintf(stderr, " -p NAME, --person NAME [%-7s] person name (for prompt selection)\n", params.person.c_str());
|
||||
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, " -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");
|
||||
fprintf(stderr, " -f FNAME, --file FNAME [%-7s] text output file name\n", params.fname_out.c_str());
|
||||
fprintf(stderr, "\n");
|
||||
}
|
||||
|
||||
@ -252,7 +256,10 @@ int main(int argc, char ** argv) {
|
||||
|
||||
// whisper init
|
||||
|
||||
struct whisper_context * ctx_wsp = whisper_init_from_file(params.model_wsp.c_str());
|
||||
struct whisper_context_params cparams;
|
||||
cparams.use_gpu = params.use_gpu;
|
||||
|
||||
struct whisper_context * ctx_wsp = whisper_init_from_file_with_params(params.model_wsp.c_str(), cparams);
|
||||
|
||||
// llama init
|
||||
|
||||
@ -269,6 +276,9 @@ int main(int argc, char ** argv) {
|
||||
lcparams.seed = 1;
|
||||
lcparams.f16_kv = true;
|
||||
lcparams.n_threads = params.n_threads;
|
||||
if (!params.use_gpu) {
|
||||
lcparams.n_gpu_layers = 0;
|
||||
}
|
||||
|
||||
struct llama_context * ctx_llama = llama_new_context_with_model(model_llama, lcparams);
|
||||
|
||||
|
||||
Reference in New Issue
Block a user