How to use.
save_model.py
참고 test_model.py
참고하여 inference 동작 체크 필요ensemble
, core model(.pt file)
, preprocessing
, postprocessing
conda-pack -n ${conda_env_name} -o pre_env.tar.gz
와 같은 방식으로 conda environment 패키징 docker pull nvcr.io/nvidia/tritonserver:21.10-py3
./triton
)docker run --gpus='"device=0"' -it --rm --shm-size=8g -p 8005:8000 -v ${MODEL_FOLDER_PATH}:/model_dir nvcr.io/nvidia/tritonserver:21.10 tritonserver --model-repository=/model_dir --strict-model-config=false --model-control-mode=poll --repository-poll-secs=10 --backend-config=tensorflow,version=2 --log-verbose=1
client.py
참고