Spaces:
Running
Running
Andreas Lubbe
commited on
cli : add --suppress_nst support (#2664)
Browse files- examples/cli/cli.cpp +5 -0
examples/cli/cli.cpp
CHANGED
|
@@ -71,6 +71,7 @@ struct whisper_params {
|
|
| 71 |
bool log_score = false;
|
| 72 |
bool use_gpu = true;
|
| 73 |
bool flash_attn = false;
|
|
|
|
| 74 |
|
| 75 |
std::string language = "en";
|
| 76 |
std::string prompt;
|
|
@@ -170,6 +171,7 @@ static bool whisper_params_parse(int argc, char ** argv, whisper_params & params
|
|
| 170 |
else if (arg == "-ls" || arg == "--log-score") { params.log_score = true; }
|
| 171 |
else if (arg == "-ng" || arg == "--no-gpu") { params.use_gpu = false; }
|
| 172 |
else if (arg == "-fa" || arg == "--flash-attn") { params.flash_attn = true; }
|
|
|
|
| 173 |
else if ( arg == "--suppress-regex") { params.suppress_regex = argv[++i]; }
|
| 174 |
else if ( arg == "--grammar") { params.grammar = argv[++i]; }
|
| 175 |
else if ( arg == "--grammar-rule") { params.grammar_rule = argv[++i]; }
|
|
@@ -237,6 +239,7 @@ static void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params
|
|
| 237 |
fprintf(stderr, " -ls, --log-score [%-7s] log best decoder scores of tokens\n", params.log_score?"true":"false");
|
| 238 |
fprintf(stderr, " -ng, --no-gpu [%-7s] disable GPU\n", params.use_gpu ? "false" : "true");
|
| 239 |
fprintf(stderr, " -fa, --flash-attn [%-7s] flash attention\n", params.flash_attn ? "true" : "false");
|
|
|
|
| 240 |
fprintf(stderr, " --suppress-regex REGEX [%-7s] regular expression matching tokens to suppress\n", params.suppress_regex.c_str());
|
| 241 |
fprintf(stderr, " --grammar GRAMMAR [%-7s] GBNF grammar to guide decoding\n", params.grammar.c_str());
|
| 242 |
fprintf(stderr, " --grammar-rule RULE [%-7s] top-level GBNF grammar rule name\n", params.grammar_rule.c_str());
|
|
@@ -1128,6 +1131,8 @@ int main(int argc, char ** argv) {
|
|
| 1128 |
|
| 1129 |
wparams.no_timestamps = params.no_timestamps;
|
| 1130 |
|
|
|
|
|
|
|
| 1131 |
whisper_print_user_data user_data = { ¶ms, &pcmf32s, 0 };
|
| 1132 |
|
| 1133 |
const auto & grammar_parsed = params.grammar_parsed;
|
|
|
|
| 71 |
bool log_score = false;
|
| 72 |
bool use_gpu = true;
|
| 73 |
bool flash_attn = false;
|
| 74 |
+
bool suppress_nst = false;
|
| 75 |
|
| 76 |
std::string language = "en";
|
| 77 |
std::string prompt;
|
|
|
|
| 171 |
else if (arg == "-ls" || arg == "--log-score") { params.log_score = true; }
|
| 172 |
else if (arg == "-ng" || arg == "--no-gpu") { params.use_gpu = false; }
|
| 173 |
else if (arg == "-fa" || arg == "--flash-attn") { params.flash_attn = true; }
|
| 174 |
+
else if (arg == "-sns" || arg == "--suppress-nst") { params.suppress_nst = true; }
|
| 175 |
else if ( arg == "--suppress-regex") { params.suppress_regex = argv[++i]; }
|
| 176 |
else if ( arg == "--grammar") { params.grammar = argv[++i]; }
|
| 177 |
else if ( arg == "--grammar-rule") { params.grammar_rule = argv[++i]; }
|
|
|
|
| 239 |
fprintf(stderr, " -ls, --log-score [%-7s] log best decoder scores of tokens\n", params.log_score?"true":"false");
|
| 240 |
fprintf(stderr, " -ng, --no-gpu [%-7s] disable GPU\n", params.use_gpu ? "false" : "true");
|
| 241 |
fprintf(stderr, " -fa, --flash-attn [%-7s] flash attention\n", params.flash_attn ? "true" : "false");
|
| 242 |
+
fprintf(stderr, " -sns, --suppress-nst [%-7s] suppress non-speech tokens\n", params.suppress_nst ? "true" : "false");
|
| 243 |
fprintf(stderr, " --suppress-regex REGEX [%-7s] regular expression matching tokens to suppress\n", params.suppress_regex.c_str());
|
| 244 |
fprintf(stderr, " --grammar GRAMMAR [%-7s] GBNF grammar to guide decoding\n", params.grammar.c_str());
|
| 245 |
fprintf(stderr, " --grammar-rule RULE [%-7s] top-level GBNF grammar rule name\n", params.grammar_rule.c_str());
|
|
|
|
| 1131 |
|
| 1132 |
wparams.no_timestamps = params.no_timestamps;
|
| 1133 |
|
| 1134 |
+
wparams.suppress_nst = params.suppress_nst;
|
| 1135 |
+
|
| 1136 |
whisper_print_user_data user_data = { ¶ms, &pcmf32s, 0 };
|
| 1137 |
|
| 1138 |
const auto & grammar_parsed = params.grammar_parsed;
|