Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

fix onnxruntime memoryleak when load model (#2108)

locasxec494d8fd18c2
diff --git a/runtime/onnxruntime/src/fsmn-vad.cpp b/runtime/onnxruntime/src/fsmn-vad.cpp
index 42ce83b45..e34767009 100644
--- a/runtime/onnxruntime/src/fsmn-vad.cpp
+++ b/runtime/onnxruntime/src/fsmn-vad.cpp
@@ -60,18 +60,19 @@ void FsmnVad::ReadModel(const char* vad_model) {
LOG(ERROR) << "Error when load vad onnx model: " << e.what();
exit(-1);
}
- GetInputOutputInfo(vad_session_, &vad_in_names_, &vad_out_names_);
+ GetInputOutputInfo(vad_session_, &vad_in_names_, &vad_out_names_, &vad_allocator);
}
void FsmnVad::GetInputOutputInfo(
const std::shared_ptr<Ort::Session> &session,
- std::vector<const char *> *in_names, std::vector<const char *> *out_names) {
+ std::vector<const char *> *in_names, std::vector<const char *> *out_names,
+ Ort::AllocatorWithDefaultOptions *allocator) {
Ort::AllocatorWithDefaultOptions allocator;
// Input info
int num_nodes = session->GetInputCount();
in_names->resize(num_nodes);
for (int i = 0; i < num_nodes; ++i) {
- std::unique_ptr<char, Ort::detail::AllocatedFree> name = session->GetInputNameAllocated(i, allocator);
+ std::unique_ptr<char, Ort::detail::AllocatedFree> name = session->GetInputNameAllocated(i, *allocator);
Ort::TypeInfo type_info = session->GetInputTypeInfo(i);
auto tensor_info = type_info.GetTensorTypeAndShapeInfo();
ONNXTensorElementDataType type = tensor_info.GetElementType();
@@ -90,7 +91,7 @@ void FsmnVad::GetInputOutputInfo(
num_nodes = session->GetOutputCount();
out_names->resize(num_nodes);
for (int i = 0; i < num_nodes; ++i) {
- std::unique_ptr<char, Ort::detail::AllocatedFree> name = session->GetOutputNameAllocated(i, allocator);
+ std::unique_ptr<char, Ort::detail::AllocatedFree> name = session->GetOutputNameAllocated(i, *allocator);
Ort::TypeInfo type_info = session->GetOutputTypeInfo(i);
auto tensor_info = type_info.GetTensorTypeAndShapeInfo();
ONNXTensorElementDataType type = tensor_info.GetElementType();
@@ -310,6 +311,8 @@ void FsmnVad::Test() {
}
FsmnVad::~FsmnVad() {
+ for (auto vad_in_name_item : vad_in_names_) vad_allocator.Free((void*)vad_in_name_item);
+ for (auto vad_out_name_item : vad_out_names_) vad_allocator.Free((void*)vad_out_name_item);
}
FsmnVad::FsmnVad():env_(ORT_LOGGING_LEVEL_ERROR, ""),session_options_{} {
diff --git a/runtime/onnxruntime/src/fsmn-vad.h b/runtime/onnxruntime/src/fsmn-vad.h
index f06a9651b..5443a3ba4 100644
--- a/runtime/onnxruntime/src/fsmn-vad.h
+++ b/runtime/onnxruntime/src/fsmn-vad.h
@@ -34,6 +34,7 @@ public:
std::shared_ptr<Ort::Session> vad_session_ = nullptr;
Ort::Env env_;
Ort::SessionOptions session_options_;
+ Ort::AllocatorWithDefaultOptions vad_allocator;
std::vector<const char *> vad_in_names_;
std::vector<const char *> vad_out_names_;
std::vector<std::vector<float>> in_cache_;
@@ -56,7 +57,8 @@ private:
static void GetInputOutputInfo(
const std::shared_ptr<Ort::Session> &session,
- std::vector<const char *> *in_names, std::vector<const char *> *out_names);
+ std::vector<const char *> *in_names, std::vector<const char *> *out_names,
+ Ort::AllocatorWithDefaultOptions *allocator);
void FbankKaldi(float sample_rate, std::vector<std::vector<float>> &vad_feats,
std::vector<float> &waves);