wenet-e2e
Segmentation fault during MNN Model export
I get a "segmentation fault" when I try to convert the ONNX model in MNN model:
MNNConvert -f ONNX --modelFile voxceleb_resnet34.onnx --MNNModel model.mnn --bizCode MNN
The device support i8sdot:0, support fp16:0, support i8mm: 0
Start to Convert Other Model Format To MNN Model..., target version: 2.8
[13:09:33] /home/user/MNN-976d1d7c0f916ea8a7acc3d31352789590f00b18/tools/converter/source/onnx/onnxConverter.cpp:46: ONNX Model ir version: 7
[13:09:33] /home/user/MNN-976d1d7c0f916ea8a7acc3d31352789590f00b18/tools/converter/source/onnx/onnxConverter.cpp:47: ONNX Model opset version: 14
Start to Optimize the MNN Net...
Segmentation fault
It happens both with and without --saveStaticModel. The model used is the pretrained model available, but I tried exporting a model using the updated export_onnx.py script and the problem persists. The MNN version and flags are the same as defined in the CMakeFiles.txt.
Still segmentation fault. If I try to export the model using export_mnn.py it doesn't find the executable "MNNConvert", since it calls MNNConvert using os.system.
How can I compile MNN to recreate the same environment of the test? I'm starting to think that I'm not compiling it properly.
Still segmentation fault. If I try to export the model using export_mnn.py it doesn't find the executable "MNNConvert", since it calls MNNConvert using os.system.
How can I compile MNN to recreate the same environment of the test? I'm starting to think that I'm not compiling it properly.
try: pip install MNN
I managed to convert the model without a segmentation fault happening. I installed mnn with pip install MNN as you said and then instead of using MMNConvert I executed the python version:
python -m mnnconvert -f ONNX --modelFile model.onnx --MNNModel model.mnn --bizCode MNN