When I install CUDA12.1.2 toolkit in Ubuntu20.04 with a GPU Geforce GTX16750, in the instruction, the nvidia driver version is 530.30.02-1. But at the last two steps, the deiver version will automatically update to driver 535.54.03 and cuda12.2, whichwill cause incompatible with tensorRT v8.6.1. Because I have install an ecosystem with Tensorflow 12.5.0, and will install cuDNN8.9.7.29, TensorRT 8.6.1, Python3.11.7 to run in a virtual environment, can you help to help me to resolve this compatibility problem. If there is any Tensor RT version which is compatible with nvida driver 535.54.03 and cuda 12.2.
There is no CUDA 12.1.2 toolkit, see here.
Using driver 535.54.03 instead of 530.30.02 will not introduce any compatibility problem.
But if I use driver 535.54.03, the cuda version will be cuda 12.2, but tensorRT 8.6.1 is not compatible with cuda 12.2, it is compatible to cuda12.1 instead. Or, which tensor RT version will be compatible with the combination if I use driver 535.40.03, and cuda 12.2.