Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

请问下使用fastdeploy部署paddleOCRv3的性能问题 #2570

Open
ouerum opened this issue Dec 4, 2024 · 4 comments
Open

请问下使用fastdeploy部署paddleOCRv3的性能问题 #2570

ouerum opened this issue Dec 4, 2024 · 4 comments
Assignees

Comments

@ouerum
Copy link

ouerum commented Dec 4, 2024

按照官方文档部署ppOCRv3 fastdeploy服务,使用jmeter压测性能一直卡在10QPS左右,修改instance_group里面的count数值也没有提高并发。

  1. 推理后端使用tensorrt,精度是trt_fp8。
  2. 从原有的矩阵输入改成了图片base64输入。

参考文档连接:https://github.com/PaddlePaddle/FastDeploy/blob/develop/examples/vision/ocr/PP-OCR/serving/fastdeploy_serving/README.md

部署环境
【GPU】NVIDIA T4
【docker镜像】fastdeploy:1.0.1-gpu-cuda11.4-trt8.4-21.10

@ouerum
Copy link
Author

ouerum commented Dec 4, 2024

这是使用的配置文件 pp_ocr.tar.gz

另外请问下官方有测试过fastdeploy服务化部署ppocr的性能,以及有什么优化手段?

@Jiang-Jia-Jun
Copy link
Collaborator

这个需要自行debug看下是不是卡在了预处理这些cpu处理环节

@ouerum
Copy link
Author

ouerum commented Dec 4, 2024 via email

@Jiang-Jia-Jun
Copy link
Collaborator

这有可能是检测出来的框过多,每个框都需要crop出来,后处理耗时就会比较久。 看是否可以根据检测的置信度做一些过滤,减少后处理需要处理的框个数

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants