TensorRT版本性能差异测试ResNet50.onnx
模型-耗时ms/张 | TRT7.1.3.4 | TRT5.1.5.0 | 版本性能提升 |
---|---|---|---|
ResNet50-b1-fp32-Nvidia2080Ti | 2.0444 | 2.3022 | 11.20% |
ResNet50-b4-fp32 | 0.6491 | 1.1063 | 41.32% |
ResNet50-b8-fp32 | 0.3533 | 0.8849 | 60.07% |
ResNet50-b16-fp32 | 0.2037 | 0.7502 | 72.85% |
ResNet50-b1-int8 | 0.7206 | 1.0161 | 29.08% |
ResNet50-b4-int8 | 0.2754 | 0.3471 | 20.64% |
ResNet50-b8-int8 | 0.1635 | 0.2030 | 19.48% |
ResNet50-b16-int8 | 0.1083 | 0.1505 | 28.02% |
Batch性能提升(相对于b1) | 68.23%/82.72%/90.04% 61.78%/77.31%/84.97% |
51.95%/61.56%/67.41% 65.84%/80.02%/85.19% |