reformat.cpp::executeCutensor::388 checkMacros.cpp::catchCudaError::202

[leslie27ch]

Hello, I'm really at a loss here and would greatly appreciate your guidance:
I trained an epoch 2 model (using CUDA 11.3), performed PTQ quantization, and converted it to ONNX. When I built and ran it (using CUDA 11.6), everything was normal, and the result visualization was also fine.
Then, I used the epoch 6 model, and without changing anything else, I followed the same steps, but when I ran it (using CUDA 11.6), I encountered the following error. it lools error happened during transfusionhead forward。