推理server 三 torchserve示例
kserve在用torchserve时,需要按照这个文档中指示,将torchserve重新build一下:https://github.com/pytorch/serve/tree/master/kubernetes/kfserving
docker run -it --network=host -v /home/test/models_torch:/mnt/models pytorch/torchserve-kfs:latest --model_name cifar10 --model_class_name Net --model_dir /mnt/models/cifar10 --http_port 8080 --workers 1 (workers设多了,可能会出现cuda初始化错误)
py-input.json:
posted on 2021-11-09 11:25 MissSimple 阅读(176) 评论(0) 编辑 收藏 举报