Merge pull request ggerganov#2 from bobqianic/patch

Patch
jettoblack · Jan 15, 2024 · 7047d32 · 7047d32
2 parents 6648641 + c8528a7
commit 7047d32
Show file tree

Hide file tree

Showing 6 changed files with 108 additions and 27 deletions.
diff --git a/Makefile b/Makefile
@@ -364,19 +364,19 @@ server: examples/server/server.cpp $(SRC_COMMON) $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) examples/server/server.cpp $(SRC_COMMON) $(WHISPER_OBJ) -o server $(LDFLAGS)
 
 stream: examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o stream $(CC_SDL) $(LDFLAGS)
+	$(CXX) $(CXXFLAGS) examples/stream/stream.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o stream $(CC_SDL) $(LDFLAGS)
 
-command: examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
+command: examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o command $(CC_SDL) $(LDFLAGS)
 
 lsp: examples/lsp/lsp.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) examples/lsp/lsp.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o lsp $(CC_SDL) $(LDFLAGS)
 
 talk: examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk $(CC_SDL) $(LDFLAGS)
+	$(CXX) $(CXXFLAGS) examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk $(CC_SDL) $(LDFLAGS)
 
 talk-llama: examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk-llama $(CC_SDL) $(LDFLAGS)
+	$(CXX) $(CXXFLAGS) examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk-llama $(CC_SDL) $(LDFLAGS)
 
 #
 # Audio samples

diff --git a/examples/command/command.cpp b/examples/command/command.cpp
@@ -8,6 +8,7 @@
 
 #include "common-sdl.h"
 #include "common.h"
+#include "console.h"
 #include "whisper.h"
 #include "grammar-parser.h"
 
@@ -59,9 +60,9 @@ struct whisper_params {
     std::string grammar;
 };
 
-void whisper_print_usage(int argc, char ** argv, const whisper_params & params);
+void whisper_print_usage(int argc, const char ** argv, const whisper_params & params);
 
-bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
+bool whisper_params_parse(int argc, const char ** argv, whisper_params & params) {
     for (int i = 1; i < argc; i++) {
         std::string arg = argv[i];
 
@@ -100,7 +101,7 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
     return true;
 }
 
-void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & params) {
+void whisper_print_usage(int /*argc*/, const char ** argv, const whisper_params & params) {
     fprintf(stderr, "\n");
     fprintf(stderr, "usage: %s [options]\n", argv[0]);
     fprintf(stderr, "\n");
@@ -151,7 +152,6 @@ std::string transcribe(
 
     wparams.print_progress   = false;
     wparams.print_special    = params.print_special;
-    wparams.print_realtime   = false;
     wparams.print_timestamps = !params.no_timestamps;
     wparams.translate        = params.translate;
     wparams.no_context       = true;
@@ -356,7 +356,6 @@ int process_command_list(struct whisper_context * ctx, audio_async &audio, const
 
             wparams.print_progress   = false;
             wparams.print_special    = params.print_special;
-            wparams.print_realtime   = false;
             wparams.print_timestamps = !params.no_timestamps;
             wparams.translate        = params.translate;
             wparams.no_context       = true;
@@ -678,7 +677,7 @@ int process_general_transcription(struct whisper_context * ctx, audio_async & au
     return 0;
 }
 
-int main(int argc, char ** argv) {
+int run(int argc, const char ** argv) {
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {
@@ -773,3 +772,23 @@ int main(int argc, char ** argv) {
 
     return ret_val;
 }
+
+#if _WIN32
+int wmain(int argc, const wchar_t ** argv_UTF16LE) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    std::vector<std::string> buffer(argc);
+    std::vector<const char*> argv_UTF8(argc);
+    for (int i = 0; i < argc; ++i) {
+        buffer[i] = console::UTF16toUTF8(argv_UTF16LE[i]);
+        argv_UTF8[i] = buffer[i].c_str();
+    }
+    return run(argc, argv_UTF8.data());
+}
+#else
+int main(int argc, const char ** argv_UTF8) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    return run(argc, argv_UTF8);
+}
+#endif
diff --git a/examples/lsp/lsp.cpp b/examples/lsp/lsp.cpp
@@ -171,7 +171,6 @@ json unguided_transcription(struct whisper_context * ctx, audio_async &audio, js
     }
     wparams.print_progress   = false;
     wparams.print_special    = params.print_special;
-    wparams.print_realtime   = false;
     wparams.print_timestamps = false;
     wparams.translate        = params.translate;
     wparams.no_context       = jparams.value("no_context", true);
@@ -210,7 +209,6 @@ json guided_transcription(struct whisper_context * ctx, audio_async &audio, cons
 
     wparams.print_progress   = false;
     wparams.print_special    = params.print_special;
-    wparams.print_realtime   = false;
     wparams.print_timestamps = false;
     wparams.translate        = params.translate;
     wparams.no_context       = true;

diff --git a/examples/stream/stream.cpp b/examples/stream/stream.cpp
@@ -4,6 +4,7 @@
 //
 #include "common-sdl.h"
 #include "common.h"
+#include "console.h"
 #include "whisper.h"
 
 #include <cassert>
@@ -56,9 +57,9 @@ struct whisper_params {
     std::string fname_out;
 };
 
-void whisper_print_usage(int argc, char ** argv, const whisper_params & params);
+void whisper_print_usage(int argc, const char ** argv, const whisper_params & params);
 
-bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
+bool whisper_params_parse(int argc, const char ** argv, whisper_params & params) {
     for (int i = 1; i < argc; i++) {
         std::string arg = argv[i];
 
@@ -97,7 +98,7 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
     return true;
 }
 
-void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & params) {
+void whisper_print_usage(int /*argc*/, const char ** argv, const whisper_params & params) {
     fprintf(stderr, "\n");
     fprintf(stderr, "usage: %s [options]\n", argv[0]);
     fprintf(stderr, "\n");
@@ -126,7 +127,7 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
     fprintf(stderr, "\n");
 }
 
-int main(int argc, char ** argv) {
+int run(int argc, const char ** argv) {
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {
@@ -213,7 +214,11 @@ int main(int argc, char ** argv) {
 
     std::ofstream fout;
     if (params.fname_out.length() > 0) {
+    #if _WIN32
+        fout.open(console::UTF8toUTF16(params.fname_out));
+    #else
         fout.open(params.fname_out);
+    #endif
         if (!fout.is_open()) {
             fprintf(stderr, "%s: failed to open output file '%s'!\n", __func__, params.fname_out.c_str());
             return 1;
@@ -314,7 +319,6 @@ int main(int argc, char ** argv) {
 
             wparams.print_progress   = false;
             wparams.print_special    = params.print_special;
-            wparams.print_realtime   = false;
             wparams.print_timestamps = !params.no_timestamps;
             wparams.translate        = params.translate;
             wparams.single_segment   = !use_vad;
@@ -431,3 +435,23 @@ int main(int argc, char ** argv) {
 
     return 0;
 }
+
+#if _WIN32
+int wmain(int argc, const wchar_t ** argv_UTF16LE) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    std::vector<std::string> buffer(argc);
+    std::vector<const char*> argv_UTF8(argc);
+    for (int i = 0; i < argc; ++i) {
+        buffer[i] = console::UTF16toUTF8(argv_UTF16LE[i]);
+        argv_UTF8[i] = buffer[i].c_str();
+    }
+    return run(argc, argv_UTF8.data());
+}
+#else
+int main(int argc, const char ** argv_UTF8) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    return run(argc, argv_UTF8);
+}
+#endif
diff --git a/examples/talk-llama/talk-llama.cpp b/examples/talk-llama/talk-llama.cpp
@@ -3,6 +3,7 @@
 
 #include "common-sdl.h"
 #include "common.h"
+#include "console.h"
 #include "whisper.h"
 #include "llama.h"
 
@@ -77,9 +78,9 @@ struct whisper_params {
     std::string path_session = "";       // path to file for saving/loading model eval state
 };
 
-void whisper_print_usage(int argc, char ** argv, const whisper_params & params);
+void whisper_print_usage(int argc, const char ** argv, const whisper_params & params);
 
-bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
+bool whisper_params_parse(int argc, const char ** argv, whisper_params & params) {
     for (int i = 1; i < argc; i++) {
         std::string arg = argv[i];
 
@@ -127,7 +128,7 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
     return true;
 }
 
-void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & params) {
+void whisper_print_usage(int /*argc*/, const char ** argv, const whisper_params & params) {
     fprintf(stderr, "\n");
     fprintf(stderr, "usage: %s [options]\n", argv[0]);
     fprintf(stderr, "\n");
@@ -180,7 +181,6 @@ std::string transcribe(
 
     wparams.print_progress   = false;
     wparams.print_special    = params.print_special;
-    wparams.print_realtime   = false;
     wparams.print_timestamps = !params.no_timestamps;
     wparams.translate        = params.translate;
     wparams.no_context       = true;
@@ -247,7 +247,7 @@ The transcript only includes text, it does not include markup like HTML and Mark
 {1}{4} Blue
 {0}{4})";
 
-int main(int argc, char ** argv) {
+int run(int argc, const char ** argv) {
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {
@@ -708,3 +708,23 @@ int main(int argc, char ** argv) {
 
     return 0;
 }
+
+#if _WIN32
+int wmain(int argc, const wchar_t ** argv_UTF16LE) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    std::vector<std::string> buffer(argc);
+    std::vector<const char*> argv_UTF8(argc);
+    for (int i = 0; i < argc; ++i) {
+        buffer[i] = console::UTF16toUTF8(argv_UTF16LE[i]);
+        argv_UTF8[i] = buffer[i].c_str();
+    }
+    return run(argc, argv_UTF8.data());
+}
+#else
+int main(int argc, const char ** argv_UTF8) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    return run(argc, argv_UTF8);
+}
+#endif
diff --git a/examples/talk/talk.cpp b/examples/talk/talk.cpp
@@ -3,6 +3,7 @@
 
 #include "common-sdl.h"
 #include "common.h"
+#include "console.h"
 #include "whisper.h"
 #include "gpt-2.h"
 
@@ -41,9 +42,9 @@ struct whisper_params {
     std::string fname_out;
 };
 
-void whisper_print_usage(int argc, char ** argv, const whisper_params & params);
+void whisper_print_usage(int argc, const char ** argv, const whisper_params & params);
 
-bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
+bool whisper_params_parse(int argc, const char ** argv, whisper_params & params) {
     for (int i = 1; i < argc; i++) {
         std::string arg = argv[i];
 
@@ -79,7 +80,7 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
     return true;
 }
 
-void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & params) {
+void whisper_print_usage(int /*argc*/, const char ** argv, const whisper_params & params) {
     fprintf(stderr, "\n");
     fprintf(stderr, "usage: %s [options]\n", argv[0]);
     fprintf(stderr, "\n");
@@ -116,7 +117,6 @@ std::string transcribe(whisper_context * ctx, const whisper_params & params, con
 
     wparams.print_progress   = false;
     wparams.print_special    = params.print_special;
-    wparams.print_realtime   = false;
     wparams.print_timestamps = !params.no_timestamps;
     wparams.translate        = params.translate;
     wparams.no_context       = true;
@@ -170,7 +170,7 @@ Here is how {0} (A) continues the dialogue:
 
 A:)";
 
-int main(int argc, char ** argv) {
+int run(int argc, const char ** argv) {
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {
@@ -373,3 +373,23 @@ int main(int argc, char ** argv) {
 
     return 0;
 }
+
+#if _WIN32
+int wmain(int argc, const wchar_t ** argv_UTF16LE) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    std::vector<std::string> buffer(argc);
+    std::vector<const char*> argv_UTF8(argc);
+    for (int i = 0; i < argc; ++i) {
+        buffer[i] = console::UTF16toUTF8(argv_UTF16LE[i]);
+        argv_UTF8[i] = buffer[i].c_str();
+    }
+    return run(argc, argv_UTF8.data());
+}
+#else
+int main(int argc, const char ** argv_UTF8) {
+    console::init(true, true);
+    atexit([]() { console::cleanup(); });
+    return run(argc, argv_UTF8);
+}
+#endif