ANSCENTER/ANSCORE
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix AMD by changing from GetTensorData<T>() to GetTensorMutableData<T>()

Browse Source
This commit is contained in:
Tuan Nghia Nguyen
2026-04-28 13:25:02 +10:00
parent f4b74c837e
commit dcf974c35c
18 changed files with 359 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
engines/ONNXEngine/ONNXEngine.cpp
View File

@@ -819,7 +819,10 @@ namespace ANSCENTER {
input_values_handler.clear();
input_values_handler.shrink_to_fit();
const float* vals = output_tensors[0].GetTensorData<float>();
// GetTensorMutableData on DirectML — the const GetTensorData
// triggers a per-call host-readable mapping that on AMD DML
// exhausts a staging-buffer pool after ~8 calls and hangs.
const float* vals = output_tensors[0].GetTensorMutableData<float>();
const unsigned int hidden_dim =
static_cast<unsigned int>(output_node_dims.at(0).at(1));
@@ -1377,7 +1380,9 @@ namespace ANSCENTER {
Ort::RunOptions{ nullptr },
in_names, &input_tensor, 1, out_names, 1);
out_result = post_processing(outputs[0].GetTensorData<float>());
// GetTensorMutableData (not GetTensorData) — see comment in this
// file's other output-read sites; const GetTensorData hangs on AMD DML.
out_result = post_processing(outputs[0].GetTensorMutableData<float>());
}
Ort::Value MOVINET::transform(const cv::Mat& mat)