Description
When converting an onnx model (SSD300 from pytorch) to tensorrt with trtexec, I get the following error :
[01/27/2025-13:33:41] [V] [TRT] Running: NMSCastFusion on NonMaxSuppression_4152_4645
[01/27/2025-13:33:41] [E] Error[2]: [myelinBuilderUtils.cpp::getMyelinSupportTypeNoConstraints::959] Error Code 2: Internal Error (Assertion !scopedOp failed. DeviceToShapeHostNode should not have been inserted into a scope)
[01/27/2025-13:33:41] [E] Engine could not be created from network
[01/27/2025-13:33:41] [E] Building engine failed
[01/27/2025-13:33:41] [E] Failed to create engine from model or file.
[01/27/2025-13:33:41] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v100600] [b26] # trtexec --onnx=./models/ssd.onnx --saveEngine=./models/ssd.engine --verbose
Environment
TensorRT Version: 10.6.0.26-1+cuda12.6
GPU Type: jetson orin nano
CUDA Version: V12.6.77
CUDNN Version: 9.5.0
Operating System + Version: Ubuntu 22.04
Python Version (if applicable): 3.10.12
PyTorch Version (if applicable): torch-2.5.0-cp310-cp310-linux_aarch64