Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix Pybind memory leak #23105

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
+3 −2
Open

Fix Pybind memory leak #23105

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions onnxruntime/python/onnxruntime_pybind_mlvalue.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -280,7 +280,7 @@ void DmlToCpuMemCpy(void* dst, const void* src, size_t num_bytes) {
uint32_t readback_heap_size = gsl::narrow_cast<uint32_t>(sizeof(readback_heap));
ORT_THROW_IF_FAILED(d3d12_device->GetPrivateData(dml_readback_heap_guid, &readback_heap_size, &readback_heap));

// ReadbackFromGpu already syncs with the CPU and waits for the copy to be completed, so we don't need to sync after
// ReadbackFromGpu already syncs with the CPU and waits for the copy to be completed, so we dont need to sync after
// this call
readback_heap->ReadbackFromGpu(
gsl::make_span(static_cast<std::byte*>(dst), num_bytes),
Expand Down Expand Up @@ -428,7 +428,7 @@ MLDataType NumpyTypeToOnnxRuntimeTensorType(int numpy_type) {
// Special, not a C type expands to enum value of 16
{NPY_FLOAT16, DataTypeImpl::GetType<MLFloat16>()},
{NPY_DOUBLE, DataTypeImpl::GetType<double>()},
// We don't want to use size specific types such
// We dont want to use size specific types such
// as NPY_INT32 bc they are not enums but hash defines
// which may map into other enums and may conflict with other entries here
// also NPY docs define these sizes as platform specific, thus we
Expand Down Expand Up @@ -581,6 +581,7 @@ static void CopyDataToTensor(PyArrayObject* darray, int npy_type, Tensor& tensor
for (int i = 0; i < total_items; ++i, src += item_size) {
// Python unicode strings are assumed to be USC-4. Strings are stored as UTF-8.
PyObject* item = PyArray_GETITEM(darray, src);
UniqueDecRefPtr<PyObject> itemGuard(item, DecRefFn<PyObject>());
PyObject* pStr = PyObject_Str(item);
UniqueDecRefPtr<PyObject> strGuard(pStr, DecRefFn<PyObject>());
dst[i] = py::reinterpret_borrow<py::str>(pStr);
Expand Down
Loading