I run nvcr.io/nvidia/tritonserver:23.01-py3 docker image with the following command
docker run --gpus=0 --rm -it --net=host -v ${PWD}/models:/models nvcr.io/nvidia/tritonserver:23.01-py3 tritonserver --model-repository=/models
i was compiled yolov8n.pt to engine format, by nvidia-tensorrt 8.4.1.5 and in model folder: . ├── yolov8 │ ├── 1 │ │ └── model.plan │ └── config.pbtxt
and i received this: I0629 08:04:00.307095 1 tensorrt.cc:211] TRITONBACKEND_ModelInitialize: yolov8 (version 1) I0629 08:04:01.044353 1 logging.cc:49] Loaded engine size: 169 MiB E0629 08:04:01.239643 1 logging.cc:43] 1: [stdArchiveReader.cpp::StdArchiveReader::32] Error Code 1: Serialization (Serialization assertion magicTagRead == kMAGIC_TAG failed.Magic tag does not match) E0629 08:04:01.250591 1 logging.cc:43] 4: [runtime.cpp::deserializeCudaEngine::66] Error Code 4: Internal Error (Engine deserialization failed.) I0629 08:04:01.273851 1 tensorrt.cc:237] TRITONBACKEND_ModelFinalize: delete model state
humm, may be model.plan was broke when i try to compile it, how to solve that problem.
I am not sure whether this works
yolo export model=./yolov8n.pt format=onnx simplify=True opset=12
trtexec.exe --onnx=yolov8n.pt --saveEngine=yolov8n.engine --workspace=3000 --fp16