September 12, 2018 – At GTC Japan, NVIDIA announced the latest version of the TensorRT high-performance deep learning inference optimizer and runtime.
TensorRT 5 provides support for the new Turing architecture, new optimizations and INT8 APIs that achieves up to 40x faster inference over CPU-only platforms.
This latest version dramatically speeds up inference of recommenders, neural machine translation, speech, and natural language processing apps.
TensorRT 5 Highlights:
- Speeds up inference by 40x over CPUs for models such as translation using mixed precision on Turing Tensor Cores
- Optimizes inference models with new INT8 APIs
- Supports Xavier-based NVIDIA Drive platforms and the NVIDIA DLA accelerator for FP16
TensorRT 5 will be available to members of the NVIDIA Developer Program.
The TensorRT inference server is a containerized microservice that maximizes GPU utilization and runs multiple models from different frameworks concurrently on a node. It leverages Docker and Kubernetes to integrate seamlessly into DevOps architectures.
