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
add batch for bins
雾聪b58334974427
diff --git a/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cpp b/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cppindex 8afa77a05..4ff216a97 100644--- a/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cpp+++ b/runtime/onnxruntime/bin/funasr-onnx-offline-rtf.cpp@@ -144,6 +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("", "batch-size", "batch_size for ASR model when using GPU", false, 1, "int32_t");cmd.add(model_dir);cmd.add(quantize);@@ -163,6 +164,7 @@ int main(int argc, char *argv[])cmd.add(audio_fs);cmd.add(thread_num);cmd.add(use_gpu);+ cmd.add(batch_size);cmd.parse(argc, argv);std::map<std::string, std::string> model_path;@@ -181,7 +183,8 @@ int main(int argc, char *argv[])struct timeval start, end;gettimeofday(&start, nullptr);bool use_gpu_ = use_gpu.getValue();- FUNASR_HANDLE asr_handle=FunOfflineInit(model_path, 1, use_gpu_);+ int batch_size_ = batch_size.getValue();+ FUNASR_HANDLE asr_handle=FunOfflineInit(model_path, 1, use_gpu_, batch_size_);if (!asr_handle){diff --git a/runtime/onnxruntime/bin/funasr-onnx-offline.cpp b/runtime/onnxruntime/bin/funasr-onnx-offline.cppindex 7472932cc..c8b0d4b2a 100644--- a/runtime/onnxruntime/bin/funasr-onnx-offline.cpp+++ b/runtime/onnxruntime/bin/funasr-onnx-offline.cpp@@ -60,6 +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("", "batch-size", "batch_size for ASR model when using GPU", false, 5, "int32_t");cmd.add(model_dir);cmd.add(quantize);@@ -78,6 +79,7 @@ int main(int argc, char** argv)cmd.add(audio_fs);cmd.add(hotword);cmd.add(use_gpu);+ cmd.add(batch_size);cmd.parse(argc, argv);std::map<std::string, std::string> model_path;@@ -96,7 +98,8 @@ int main(int argc, char** argv)gettimeofday(&start, nullptr);int thread_num = 1;bool use_gpu_ = use_gpu.getValue();- FUNASR_HANDLE asr_hanlde=FunOfflineInit(model_path, thread_num, use_gpu_);+ int batch_size_ = batch_size.getValue();+ FUNASR_HANDLE asr_hanlde=FunOfflineInit(model_path, thread_num, use_gpu_, batch_size_);if (!asr_hanlde){