The latest version of the TRT-LLM has some issues with the CUDA 13. We need to make sure this won't cause any issues for the release.
Either TRT-LLM team should fix it or we need to find a TRT-LLM version that works with the CUDA 13.
Alternatively, we should check with the automation team if we can use a different version of the CUDA.