env: Tesla v100, Driver Version: 440.82, CUDA Version: 10.2
docker: nvcr.io/nvidia/tritonserver:20.03-py3, opencv-3.2.0
customlib: yolov3 plugin, image_preprocess, image_postprocess
discribe: the server regularly works well at beginning, but after several hours, it crushed, how can I handle this problem?
anyone can help ?