@@ -41,6 +41,7 @@ AppCliArgs AppCliArgs::parse(int argc, char* *argv, bool requireMode) {
41
41
args.seed = (unsigned long long )time (nullptr );
42
42
args.chatTemplateType = TEMPLATE_UNKNOWN;
43
43
args.maxSeqLen = 0 ;
44
+ args.netTurbo = true ;
44
45
args.gpuIndex = -1 ;
45
46
int i = 1 ;
46
47
if (requireMode && argc > 1 ) {
@@ -108,6 +109,8 @@ AppCliArgs AppCliArgs::parse(int argc, char* *argv, bool requireMode) {
108
109
args.maxSeqLen = (unsigned int )atoi (value);
109
110
} else if (std::strcmp (name, " --gpu-index" ) == 0 ) {
110
111
args.gpuIndex = atoi (value);
112
+ } else if (std::strcmp (name, " --net-turbo" ) == 0 ) {
113
+ args.netTurbo = atoi (value) == 1 ;
111
114
} else {
112
115
throw std::runtime_error (" Unknown option: " + std::string (name));
113
116
}
@@ -252,7 +255,10 @@ void runInferenceApp(AppCliArgs *args, void (*handler)(AppInferenceContext *cont
252
255
253
256
if (network != nullptr ) {
254
257
network->resetStats ();
255
- network->setTurbo (true );
258
+ if (args->netTurbo ) {
259
+ network->setTurbo (true );
260
+ printf (" 🚁 Network is in non-blocking mode\n " );
261
+ }
256
262
}
257
263
258
264
AppInferenceContext context;
@@ -313,7 +319,7 @@ void runWorkerApp(AppCliArgs *args) {
313
319
if (inference.isFinished )
314
320
break ;
315
321
316
- if (!isTurboEnabled) {
322
+ if (args-> netTurbo && !isTurboEnabled) {
317
323
network->setTurbo (true );
318
324
isTurboEnabled = true ;
319
325
printf (" 🚁 Network is in non-blocking mode\n " );
0 commit comments