Technology News
NVIDIA Announces Global Performance Tuner in TensorRT 11.1 for Performance Optimization
According to NVIDIA Developer. The new Global Performance Tuner feature introduces customizable build routes that expand the solution space and validate candidates against accuracy and performance.
VERIFIEDVerified · 1 sources
Illustration generated by TigerIQ from the story topic.
KEY TAKEAWAYS
- • The Global Performance Tuner is released in TensorRT 11.1 to provide customizable build routes.
- • TensorRT builds inference engines through interdependent scheduling decisions across graph transformations and quantization.
- • A mathematical framework models compilation using a builder algorithm and a searching algorithm.
Global Performance Tuner Feature
The Global Performance Tuner, released in TensorRT 11.1 , introduces customizable build routes that expand the solution space and validate each candidate against end-to-end accuracy and performance.
Compilation and Scheduling
NVIDIA TensorRT builds inference engines by making interdependent scheduling decisions across graph transformations, fusion, implementation selection, memory placement, and quantization.
SOURCES & METHODOLOGY
- 1. developer.nvidia.comOpen source