← PROJECTS

TensorRT Global Performance Tuner TRIAL

The public implementation is testable, but gains need reproduction on local models.

Next stepCompare accuracy and throughput on one production ONNX model.

Judgment reviewed 2026-10-24 · Basis: NVIDIA opens accuracy-aware tuning for TensorRT build routes

RSS

Availability & adoption

availability
open-source trtexec implementation; Python support planned for Torch-TensorRT 2.15

Timeline