From 2827cbbbe8a3cbd82fbcee3566d1b74155c18f13 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Sat, 29 Oct 2022 12:26:03 +0300 Subject: [PATCH] main : merge parallel example in main --- examples/CMakeLists.txt | 1 - examples/main/main.cpp | 24 +- examples/parallel/CMakeLists.txt | 3 - examples/parallel/README.md | 3 - examples/parallel/parallel.cpp | 432 ------------------------------- 5 files changed, 15 insertions(+), 448 deletions(-) delete mode 100644 examples/parallel/CMakeLists.txt delete mode 100644 examples/parallel/README.md delete mode 100644 examples/parallel/parallel.cpp diff --git a/examples/CMakeLists.txt b/examples/CMakeLists.txt index 629064e..0c1a43a 100644 --- a/examples/CMakeLists.txt +++ b/examples/CMakeLists.txt @@ -22,7 +22,6 @@ if (EMSCRIPTEN) add_subdirectory(whisper.wasm) else() add_subdirectory(main) - add_subdirectory(parallel) add_subdirectory(stream) add_subdirectory(bench) endif() diff --git a/examples/main/main.cpp b/examples/main/main.cpp index 1cf1d0a..f8d05ba 100644 --- a/examples/main/main.cpp +++ b/examples/main/main.cpp @@ -38,11 +38,12 @@ std::string to_timestamp(int64_t t, bool comma = false) { // command-line parameters struct whisper_params { - int32_t seed = -1; // RNG seed, not used currently - int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency()); - int32_t offset_t_ms = 0; - int32_t offset_n = 0; - int32_t max_context = -1; + int32_t seed = -1; // RNG seed, not used currently + int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency()); + int32_t n_processors = 1; + int32_t offset_t_ms = 0; + int32_t offset_n = 0; + int32_t max_context = -1; bool verbose = false; bool translate = false; @@ -74,6 +75,8 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) { params.seed = std::stoi(argv[++i]); } else if (arg == "-t" || arg == "--threads") { params.n_threads = std::stoi(argv[++i]); + } else if (arg == "-p" || arg == "--processors") { + params.n_processors = std::stoi(argv[++i]); } else if (arg == "-ot" || arg == "--offset-t") { params.offset_t_ms = std::stoi(argv[++i]); } else if (arg == "-on" || arg == "--offset-n") { @@ -128,6 +131,7 @@ void whisper_print_usage(int argc, char ** argv, const whisper_params & params) fprintf(stderr, " -h, --help show this help message and exit\n"); fprintf(stderr, " -s SEED, --seed SEED RNG seed (default: -1)\n"); fprintf(stderr, " -t N, --threads N number of threads to use during computation (default: %d)\n", params.n_threads); + fprintf(stderr, " -p N, --processors N number of processors to use during computation (default: %d)\n", params.n_processors); fprintf(stderr, " -ot N, --offset-t N time offset in milliseconds (default: %d)\n", params.offset_t_ms); fprintf(stderr, " -on N, --offset-n N segment index offset (default: %d)\n", params.offset_n); fprintf(stderr, " -mc N, --max-context N maximum number of text context tokens to store (default: max)\n"); @@ -350,7 +354,8 @@ int main(int argc, char ** argv) { // print system information { fprintf(stderr, "\n"); - fprintf(stderr, "system_info: n_threads = %d / %d | %s\n", params.n_threads, std::thread::hardware_concurrency(), whisper_print_system_info()); + fprintf(stderr, "system_info: n_threads = %d / %d | %s\n", + params.n_threads*params.n_processors, std::thread::hardware_concurrency(), whisper_print_system_info()); } // print some info about the processing @@ -363,8 +368,9 @@ int main(int argc, char ** argv) { fprintf(stderr, "%s: WARNING: model is not multilingual, ignoring language and translation options\n", __func__); } } - fprintf(stderr, "%s: processing '%s' (%d samples, %.1f sec), %d threads, lang = %s, task = %s, timestamps = %d ...\n", - __func__, fname_inp.c_str(), int(pcmf32.size()), float(pcmf32.size())/WHISPER_SAMPLE_RATE, params.n_threads, + fprintf(stderr, "%s: processing '%s' (%d samples, %.1f sec), %d threads, %d processors, lang = %s, task = %s, timestamps = %d ...\n", + __func__, fname_inp.c_str(), int(pcmf32.size()), float(pcmf32.size())/WHISPER_SAMPLE_RATE, + params.n_threads, params.n_processors, params.language.c_str(), params.translate ? "translate" : "transcribe", params.no_timestamps ? 0 : 1); @@ -393,7 +399,7 @@ int main(int argc, char ** argv) { wparams.new_segment_callback_user_data = ¶ms; } - if (whisper_full(ctx, wparams, pcmf32.data(), pcmf32.size()) != 0) { + if (whisper_full_parallel(ctx, wparams, pcmf32.data(), pcmf32.size(), params.n_processors) != 0) { fprintf(stderr, "%s: failed to process audio\n", argv[0]); return 8; } diff --git a/examples/parallel/CMakeLists.txt b/examples/parallel/CMakeLists.txt deleted file mode 100644 index 153c422..0000000 --- a/examples/parallel/CMakeLists.txt +++ /dev/null @@ -1,3 +0,0 @@ -set(TARGET parallel) -add_executable(${TARGET} parallel.cpp) -target_link_libraries(${TARGET} PRIVATE whisper ${CMAKE_THREAD_LIBS_INIT}) diff --git a/examples/parallel/README.md b/examples/parallel/README.md deleted file mode 100644 index 919537b..0000000 --- a/examples/parallel/README.md +++ /dev/null @@ -1,3 +0,0 @@ -# parallel - -TODO diff --git a/examples/parallel/parallel.cpp b/examples/parallel/parallel.cpp deleted file mode 100644 index df4a58a..0000000 --- a/examples/parallel/parallel.cpp +++ /dev/null @@ -1,432 +0,0 @@ -#include "whisper.h" - -// third-party utilities -// use your favorite implementations -#define DR_WAV_IMPLEMENTATION -#include "dr_wav.h" - -#include -#include -#include -#include -#include -#include - -// Terminal color map. 10 colors grouped in ranges [0.0, 0.1, ..., 0.9] -// Lowest is red, middle is yellow, highest is green. -const std::vector k_colors = { - "\033[38;5;196m", "\033[38;5;202m", "\033[38;5;208m", "\033[38;5;214m", "\033[38;5;220m", - "\033[38;5;226m", "\033[38;5;190m", "\033[38;5;154m", "\033[38;5;118m", "\033[38;5;82m", -}; - -// 500 -> 00:05.000 -// 6000 -> 01:00.000 -std::string to_timestamp(int64_t t, bool comma = false) { - int64_t msec = t * 10; - int64_t hr = msec / (1000 * 60 * 60); - msec = msec - hr * (1000 * 60 * 60); - int64_t min = msec / (1000 * 60); - msec = msec - min * (1000 * 60); - int64_t sec = msec / 1000; - msec = msec - sec * 1000; - - char buf[32]; - snprintf(buf, sizeof(buf), "%02d:%02d:%02d%s%03d", (int) hr, (int) min, (int) sec, comma ? "," : ".", (int) msec); - - return std::string(buf); -} - -// command-line parameters -struct whisper_params { - int32_t seed = -1; // RNG seed, not used currently - int32_t n_threads = std::max(std::min(4, (int32_t) std::thread::hardware_concurrency()) / 2, 1); - int32_t n_processors = 2; - int32_t offset_t_ms = 0; - int32_t offset_n = 0; - int32_t max_context = -1; - - bool verbose = false; - bool translate = false; - bool output_txt = false; - bool output_vtt = false; - bool output_srt = false; - bool print_special_tokens = false; - bool print_colors = false; - bool no_timestamps = false; - - std::string language = "en"; - std::string model = "models/ggml-base.en.bin"; - - std::vector fname_inp = {}; -}; - -void whisper_print_usage(int argc, char ** argv, const whisper_params & params); - -bool whisper_params_parse(int argc, char ** argv, whisper_params & params) { - for (int i = 1; i < argc; i++) { - std::string arg = argv[i]; - - if (arg[0] != '-') { - params.fname_inp.push_back(arg); - continue; - } - - if (arg == "-s" || arg == "--seed") { - params.seed = std::stoi(argv[++i]); - } else if (arg == "-t" || arg == "--threads") { - params.n_threads = std::stoi(argv[++i]); - } else if (arg == "-p" || arg == "--processors") { - params.n_processors = std::stoi(argv[++i]); - } else if (arg == "-ot" || arg == "--offset-t") { - params.offset_t_ms = std::stoi(argv[++i]); - } else if (arg == "-on" || arg == "--offset-n") { - params.offset_n = std::stoi(argv[++i]); - } else if (arg == "-mc" || arg == "--max-context") { - params.max_context = std::stoi(argv[++i]); - } else if (arg == "-v" || arg == "--verbose") { - params.verbose = true; - } else if (arg == "--translate") { - params.translate = true; - } else if (arg == "-l" || arg == "--language") { - params.language = argv[++i]; - if (whisper_lang_id(params.language.c_str()) == -1) { - fprintf(stderr, "error: unknown language '%s'\n", params.language.c_str()); - whisper_print_usage(argc, argv, params); - exit(0); - } - } else if (arg == "-otxt" || arg == "--output-txt") { - params.output_txt = true; - } else if (arg == "-ovtt" || arg == "--output-vtt") { - params.output_vtt = true; - } else if (arg == "-osrt" || arg == "--output-srt") { - params.output_srt = true; - } else if (arg == "-ps" || arg == "--print_special") { - params.print_special_tokens = true; - } else if (arg == "-pc" || arg == "--print_colors") { - params.print_colors = true; - } else if (arg == "-nt" || arg == "--no_timestamps") { - params.no_timestamps = true; - } else if (arg == "-m" || arg == "--model") { - params.model = argv[++i]; - } else if (arg == "-f" || arg == "--file") { - params.fname_inp.push_back(argv[++i]); - } else if (arg == "-h" || arg == "--help") { - whisper_print_usage(argc, argv, params); - exit(0); - } else { - fprintf(stderr, "error: unknown argument: %s\n", arg.c_str()); - whisper_print_usage(argc, argv, params); - exit(0); - } - } - - return true; -} - -void whisper_print_usage(int argc, char ** argv, const whisper_params & params) { - fprintf(stderr, "\n"); - fprintf(stderr, "usage: %s [options] file0.wav file1.wav ...\n", argv[0]); - fprintf(stderr, "\n"); - fprintf(stderr, "options:\n"); - fprintf(stderr, " -h, --help show this help message and exit\n"); - fprintf(stderr, " -s SEED, --seed SEED RNG seed (default: -1)\n"); - fprintf(stderr, " -t N, --threads N number of threads to use during computation (default: %d)\n", params.n_threads); - fprintf(stderr, " -p N, --processors N number of processors to use during computation (default: %d)\n", params.n_processors); - fprintf(stderr, " -ot N, --offset-t N time offset in milliseconds (default: %d)\n", params.offset_t_ms); - fprintf(stderr, " -on N, --offset-n N segment index offset (default: %d)\n", params.offset_n); - fprintf(stderr, " -mc N, --max-context N maximum number of text context tokens to store (default: max)\n"); - fprintf(stderr, " -v, --verbose verbose output\n"); - fprintf(stderr, " --translate translate from source language to english\n"); - fprintf(stderr, " -otxt, --output-txt output result in a text file\n"); - fprintf(stderr, " -ovtt, --output-vtt output result in a vtt file\n"); - fprintf(stderr, " -osrt, --output-srt output result in a srt file\n"); - fprintf(stderr, " -ps, --print_special print special tokens\n"); - fprintf(stderr, " -pc, --print_colors print colors\n"); - fprintf(stderr, " -nt, --no_timestamps do not print timestamps\n"); - fprintf(stderr, " -l LANG, --language LANG spoken language (default: %s)\n", params.language.c_str()); - fprintf(stderr, " -m FNAME, --model FNAME model path (default: %s)\n", params.model.c_str()); - fprintf(stderr, " -f FNAME, --file FNAME input WAV file path\n"); - fprintf(stderr, "\n"); -} - -void whisper_print_segment_callback(struct whisper_context * ctx, void * user_data) { - const whisper_params & params = *(whisper_params *) user_data; - - const int n_segments = whisper_full_n_segments(ctx); - - // print the last segment - const int i = n_segments - 1; - if (i == 0) { - printf("\n"); - } - - if (params.no_timestamps) { - if (params.print_colors) { - for (int j = 0; j < whisper_full_n_tokens(ctx, i); ++j) { - if (params.print_special_tokens == false) { - const whisper_token id = whisper_full_get_token_id(ctx, i, j); - if (id >= whisper_token_eot(ctx)) { - continue; - } - } - - const char * text = whisper_full_get_token_text(ctx, i, j); - const float p = whisper_full_get_token_p (ctx, i, j); - - const int col = std::max(0, std::min((int) k_colors.size(), (int) (std::pow(p, 3)*float(k_colors.size())))); - - printf("%s%s%s", k_colors[col].c_str(), text, "\033[0m"); - } - } else { - const char * text = whisper_full_get_segment_text(ctx, i); - printf("%s", text); - } - fflush(stdout); - } else { - const int64_t t0 = whisper_full_get_segment_t0(ctx, i); - const int64_t t1 = whisper_full_get_segment_t1(ctx, i); - - if (params.print_colors) { - printf("[%s --> %s] ", to_timestamp(t0).c_str(), to_timestamp(t1).c_str()); - for (int j = 0; j < whisper_full_n_tokens(ctx, i); ++j) { - if (params.print_special_tokens == false) { - const whisper_token id = whisper_full_get_token_id(ctx, i, j); - if (id >= whisper_token_eot(ctx)) { - continue; - } - } - - const char * text = whisper_full_get_token_text(ctx, i, j); - const float p = whisper_full_get_token_p (ctx, i, j); - - const int col = std::max(0, std::min((int) k_colors.size(), (int) (std::pow(p, 3)*float(k_colors.size())))); - - printf("%s%s%s", k_colors[col].c_str(), text, "\033[0m"); - } - printf("\n"); - } else { - const char * text = whisper_full_get_segment_text(ctx, i); - - printf("[%s --> %s] %s\n", to_timestamp(t0).c_str(), to_timestamp(t1).c_str(), text); - } - } -} - -bool output_txt(struct whisper_context * ctx, const char * fname) { - std::ofstream fout(fname); - if (!fout.is_open()) { - fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname); - return false; - } - - fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname); - - const int n_segments = whisper_full_n_segments(ctx); - for (int i = 0; i < n_segments; ++i) { - const char * text = whisper_full_get_segment_text(ctx, i); - fout << text; - } - - return true; -} - -bool output_vtt(struct whisper_context * ctx, const char * fname) { - std::ofstream fout(fname); - if (!fout.is_open()) { - fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname); - return 9; - } - - fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname); - - fout << "WEBVTT\n\n"; - - const int n_segments = whisper_full_n_segments(ctx); - for (int i = 0; i < n_segments; ++i) { - const char * text = whisper_full_get_segment_text(ctx, i); - const int64_t t0 = whisper_full_get_segment_t0(ctx, i); - const int64_t t1 = whisper_full_get_segment_t1(ctx, i); - - fout << to_timestamp(t0) << " --> " << to_timestamp(t1) << "\n"; - fout << text << "\n\n"; - } - - return true; -} - -bool output_srt(struct whisper_context * ctx, const char * fname, const whisper_params & params) { - std::ofstream fout(fname); - if (!fout.is_open()) { - fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname); - return false; - } - - fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname); - - const int n_segments = whisper_full_n_segments(ctx); - for (int i = 0; i < n_segments; ++i) { - const char * text = whisper_full_get_segment_text(ctx, i); - const int64_t t0 = whisper_full_get_segment_t0(ctx, i); - const int64_t t1 = whisper_full_get_segment_t1(ctx, i); - - fout << i + 1 + params.offset_n << "\n"; - fout << to_timestamp(t0, true) << " --> " << to_timestamp(t1, true) << "\n"; - fout << text << "\n\n"; - } - - return true; -} - -int main(int argc, char ** argv) { - whisper_params params; - - if (whisper_params_parse(argc, argv, params) == false) { - return 1; - } - - if (params.seed < 0) { - params.seed = time(NULL); - } - - if (params.fname_inp.empty()) { - fprintf(stderr, "error: no input files specified\n"); - whisper_print_usage(argc, argv, params); - return 2; - } - - // whisper init - - struct whisper_context * ctx = whisper_init(params.model.c_str()); - - if (ctx == nullptr) { - fprintf(stderr, "error: failed to initialize whisper context\n"); - return 3; - } - - for (int f = 0; f < (int) params.fname_inp.size(); ++f) { - const auto fname_inp = params.fname_inp[f]; - - // WAV input - std::vector pcmf32; - { - drwav wav; - if (!drwav_init_file(&wav, fname_inp.c_str(), NULL)) { - fprintf(stderr, "%s: failed to open WAV file '%s' - check your input\n", argv[0], fname_inp.c_str()); - whisper_print_usage(argc, argv, {}); - return 4; - } - - if (wav.channels != 1 && wav.channels != 2) { - fprintf(stderr, "%s: WAV file '%s' must be mono or stereo\n", argv[0], fname_inp.c_str()); - return 5; - } - - if (wav.sampleRate != WHISPER_SAMPLE_RATE) { - fprintf(stderr, "%s: WAV file '%s' must be 16 kHz\n", argv[0], fname_inp.c_str()); - return 6; - } - - if (wav.bitsPerSample != 16) { - fprintf(stderr, "%s: WAV file '%s' must be 16-bit\n", argv[0], fname_inp.c_str()); - return 7; - } - - int n = wav.totalPCMFrameCount; - - std::vector pcm16; - pcm16.resize(n*wav.channels); - drwav_read_pcm_frames_s16(&wav, n, pcm16.data()); - drwav_uninit(&wav); - - // convert to mono, float - pcmf32.resize(n); - if (wav.channels == 1) { - for (int i = 0; i < n; i++) { - pcmf32[i] = float(pcm16[i])/32768.0f; - } - } else { - for (int i = 0; i < n; i++) { - pcmf32[i] = float(pcm16[2*i] + pcm16[2*i + 1])/65536.0f; - } - } - } - - // print system information - { - fprintf(stderr, "\n"); - fprintf(stderr, "system_info: n_threads = %d / %d | %s\n", params.n_threads, std::thread::hardware_concurrency(), whisper_print_system_info()); - } - - // print some info about the processing - { - fprintf(stderr, "\n"); - if (!whisper_is_multilingual(ctx)) { - if (params.language != "en" || params.translate) { - params.language = "en"; - params.translate = false; - fprintf(stderr, "%s: WARNING: model is not multilingual, ignoring language and translation options\n", __func__); - } - } - fprintf(stderr, "%s: processing '%s' (%d samples, %.1f sec), %d threads, %d processors, lang = %s, task = %s, timestamps = %d ...\n", - __func__, fname_inp.c_str(), int(pcmf32.size()), float(pcmf32.size())/WHISPER_SAMPLE_RATE, - params.n_threads, params.n_processors, - params.language.c_str(), - params.translate ? "translate" : "transcribe", - params.no_timestamps ? 0 : 1); - - fprintf(stderr, "\n"); - } - - - // run the inference - { - whisper_full_params wparams = whisper_full_default_params(WHISPER_SAMPLING_GREEDY); - - wparams.print_realtime = false; - wparams.print_progress = false; - wparams.print_timestamps = !params.no_timestamps; - wparams.print_special_tokens = params.print_special_tokens; - wparams.translate = params.translate; - wparams.language = params.language.c_str(); - wparams.n_threads = params.n_threads; - wparams.n_max_text_ctx = params.max_context >= 0 ? params.max_context : wparams.n_max_text_ctx; - wparams.offset_ms = params.offset_t_ms; - - // this callback is called on each new segment - if (!wparams.print_realtime) { - wparams.new_segment_callback = whisper_print_segment_callback; - wparams.new_segment_callback_user_data = ¶ms; - } - - if (whisper_full_parallel(ctx, wparams, pcmf32.data(), pcmf32.size(), params.n_processors) != 0) { - fprintf(stderr, "%s: failed to process audio\n", argv[0]); - return 8; - } - - printf("\n"); - - // output to text file - if (params.output_txt) { - const auto fname_txt = fname_inp + ".txt"; - output_txt(ctx, fname_txt.c_str()); - } - - // output to VTT file - if (params.output_vtt) { - const auto fname_vtt = fname_inp + ".vtt"; - output_vtt(ctx, fname_vtt.c_str()); - } - - // output to SRT file - if (params.output_srt) { - const auto fname_srt = fname_inp + ".srt"; - output_srt(ctx, fname_srt.c_str(), params); - } - } - } - - whisper_print_timings(ctx); - whisper_free(ctx); - - return 0; -}