Product
TensorRT
NVIDIA推出的高性能深度学习推理优化器与运行时，自2015年发布，支持算子融合、精度校准（FP32/FP16/INT8/FP8）、内核自动调优、动态形状，是GPU推理部署的事实标准工具链
Timeline (last 90 days)
Aug 26
对于工程部署，ONNX Runtime、TensorRT等推理框架对Windows的支持已相当完善
Unverified50%
Aug 4
TensorRT optimizes inference performance through layer fusion, precision calibration, and automatic kernel tuning
Unverified90%
Aug 4
TensorRT only supports the NVIDIA ecosystem
Unverified85%