-
Test device:
- Xeon(R) Gold 6326, 512GB RAM, A30
- Container: Pytorch 22.11 (tensorRT: 8.5.1.7)
- NVIDIA Driver Version: 510.
- CUDA Version: 11.6
-
Source:
- modify_onnx.py: modify input and output of onnx model
- build_engine.py: build engine with fp32, fp16, int8 (must provide calib images)
- image_batcher.py: for calibrate with int8, must provide preprocess for in model in that script
- test_running_time: test .trt running time