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

昇腾310 npu 适配 paddle 然后用paddleocr测试,npu用起来,但是性能好差,是哪里需要优化吗? #70962

Open
hecheng64 opened this issue Jan 23, 2025 · 4 comments
Assignees
Labels
heco Hardware Ecosystem status/new-issue 新建 type/question 用户提问

Comments

@hecheng64
Copy link

hecheng64 commented Jan 23, 2025

昇腾310 npu 适配 paddle 然后用paddleocr测试,npu用起来,但是性能好差,是哪里需要优化吗?

@hecheng64
Copy link
Author

@onecatcn

@onecatcn onecatcn added the heco Hardware Ecosystem label Jan 23, 2025
@onecatcn
Copy link
Contributor

我们当前的模型维护都是基于paddlex,建议使用paddlex
https://paddlepaddle.github.io/PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/OCR.html#21

@hecheng64 hecheng64 changed the title 为啥import paddlehub 昇腾npu会报错 昇腾310 npu 适配 paddle 然后用paddleocr测试,npu用起来,但是性能好差,是哪里需要优化吗? Jan 23, 2025
@hecheng64
Copy link
Author

@onecatcn 昇腾310 npu 适配 paddle 然后用paddleocr测试,npu用起来,但是性能好差,是哪里需要优化吗?

@onecatcn
Copy link
Contributor

onecatcn commented Jan 23, 2025

@onecatcn 昇腾310 npu 适配 paddle 然后用paddleocr测试,npu用起来,但是性能好差,是哪里需要优化吗?

因为框架的基础适配方案是基于算子,如果没有特别优化过,推理就是跑小算子,没有性能优化。
我们已经在测试ppocr在昇腾卡(训练卡和推理卡)上基于paddle2onnx使用OM的推理方案了,那个方案的性能比较好。目前还有Paddlex的部分代码工作要完成,计划今年3月开源出来。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
heco Hardware Ecosystem status/new-issue 新建 type/question 用户提问
Projects
None yet
Development

No branches or pull requests

3 participants