TensorRT 6.0 (6.0.1.8 – libvinfer v6) and 7.0 (7.0.0.11 – libvinfer v7) are now available on all EECS Research Servers and Managed Linux Desktops. EECS users can load the desired versions through the relevant environment modules.
NVIDIA TensorRT is an SDK for high-performance deep learning inference. It includes a deep learning inference optimizer and runtime that delivers low latency and high-throughput for deep learning inference applications.
The following table shows the various combinations for the versions available in the EECS Infrastructure:
| CUDA version | CuDNN version | |||||
| 9.0 | 10.0 | 10.1 | 10.2 | |||
| TensorRT version | 5 | ✓ | ✓ | ✓ | x | 7.5 |
| 6 | ✓ | ✓ | ✓ | ✓ | 7.6 | |
| 7 | ✓ | ✓ | X | ✓ | 7.6 | |