Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
set batch default value
雾聪930fe72f4352
diff --git a/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cpp b/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cppindex bb032e605..c252bc7a0 100644--- a/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cpp+++ b/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cpp@@ -144,7 +144,7 @@ int main(int argc, char *argv[])TCLAP::ValueArg<std::int32_t> thread_num("", THREAD_NUM, "multi-thread num for rtf", false, 1, "int32_t");TCLAP::ValueArg<std::string> hotword("", HOTWORD, "the hotword file, one hotword perline, Format: Hotword Weight (could be: 阿里巴巴 20)", false, "", "string");TCLAP::SwitchArg use_gpu("", INFER_GPU, "Whether to use GPU for inference, default is false", false);- TCLAP::ValueArg<std::int32_t> batch_size("", BATCHSIZE, "batch_size for ASR model when using GPU", false, 1, "int32_t");+ TCLAP::ValueArg<std::int32_t> batch_size("", BATCHSIZE, "batch_size for ASR model when using GPU", false, 4, "int32_t");cmd.add(model_dir);cmd.add(quantize);diff --git a/runtime/onnxruntime/bin/funasr-onnx-offline.cpp b/runtime/onnxruntime/bin/funasr-onnx-offline.cppindex f0234c89e..764c58158 100644--- a/runtime/onnxruntime/bin/funasr-onnx-offline.cpp+++ b/runtime/onnxruntime/bin/funasr-onnx-offline.cpp@@ -60,7 +60,7 @@ int main(int argc, char** argv)TCLAP::ValueArg<std::int32_t> audio_fs("", AUDIO_FS, "the sample rate of audio", false, 16000, "int32_t");TCLAP::ValueArg<std::string> hotword("", HOTWORD, "the hotword file, one hotword perline, Format: Hotword Weight (could be: 阿里巴巴 20)", false, "", "string");TCLAP::SwitchArg use_gpu("", INFER_GPU, "Whether to use GPU for inference, default is false", false);- TCLAP::ValueArg<std::int32_t> batch_size("", BATCHSIZE, "batch_size for ASR model when using GPU", false, 5, "int32_t");+ TCLAP::ValueArg<std::int32_t> batch_size("", BATCHSIZE, "batch_size for ASR model when using GPU", false, 4, "int32_t");cmd.add(model_dir);cmd.add(quantize);diff --git a/runtime/websocket/bin/funasr-wss-server.cpp b/runtime/websocket/bin/funasr-wss-server.cppindex f395e593b..aee202480 100644--- a/runtime/websocket/bin/funasr-wss-server.cpp+++ b/runtime/websocket/bin/funasr-wss-server.cpp@@ -126,7 +126,7 @@ int main(int argc, char* argv[]) {TCLAP::ValueArg<std::int32_t> fst_inc_wts("", FST_INC_WTS,"the fst hotwords incremental bias", false, 20, "int32_t");TCLAP::SwitchArg use_gpu("", INFER_GPU, "Whether to use GPU, default is false", false);- TCLAP::ValueArg<std::int32_t> batch_size("", BATCHSIZE, "batch_size for ASR model when using GPU", false, 5, "int32_t");+ TCLAP::ValueArg<std::int32_t> batch_size("", BATCHSIZE, "batch_size for ASR model when using GPU", false, 4, "int32_t");// add filecmd.add(hotword);