FastDeploy icon indicating copy to clipboard operation
FastDeploy copied to clipboard

服务端支持的模型列表是只能用哪些模型,比如有一个变化检测的模型p2vnet,怎么使用fastdeploy部署工具?

Open funny000 opened this issue 1 year ago • 3 comments

funny000 avatar Feb 27 '24 03:02 funny000

可以支持其他模型,服务端其实是先依赖本地推理引擎支持端到端。 如果你要支持检测模型在服务端部署:

  1. 在检测模型里支持本地端到端推理 : 比如yolov5的前后处理 https://github.com/PaddlePaddle/FastDeploy/blob/develop/fastdeploy/vision/detection/contrib/yolov5/preprocessor.cc
  2. 参照其他检测模型的服务化支持,配置对应的config(也可能不要修改): https://github.com/PaddlePaddle/FastDeploy/tree/develop/examples/vision/detection/paddledetection/serving

比如服务化的预处理其实调用的是本地推理的预处理接口: https://github.com/PaddlePaddle/FastDeploy/blob/cfd80e95a12cd1f21e2dd76f566c8106e03b8c5d/examples/vision/detection/paddledetection/serving/models/preprocess/1/model.py#L95C29-L95C42

heliqi avatar Feb 27 '24 07:02 heliqi

这是本地推理集成新模型的详细文档: https://github.com/PaddlePaddle/FastDeploy/blob/develop/docs/cn/faq/develop_a_new_model.md

heliqi avatar Feb 27 '24 07:02 heliqi

好的,我先看看,谢谢

funny000 avatar Feb 27 '24 09:02 funny000