I am working on a project that involves object detection. I’m using tensorflow 1.7 for this, and according to their documentation tensorRT can be used to optimize tensorflow’s performance, however I could not find a version of tensorRT for Nvidia’s CUDA toolkit 9.1. Is it possible to use a version of tensorRT with CUDA 9.1 or do I have to downgrade to CUDA 9.0 to use it? How much would tensorRT improve the speed of the object detection or would it work just as well using only CUDA 9.1?
I’m currently running Ubuntu 16.04 using a Nvidia GTX 960M graphics card (as I am only making a prototype for our final object detection solution). If there is any more system information that people need, I would be happy to oblige.