I am testing TensorRT 4.0 on linux 16 CUDA 9.0 and 9.2 with an SSD network. After reading image in I need to transform it to 300x300. I would like to use GPU for that. When I do I receive the error 48 from cuda_launch.
If I run the application without executing gpu code written by me it runs without errors.
In which direction shall I investigate?
Thank you for your help
Your compile settings don’t match the device you are attempting to run on.