Skip to content

在线C++多线程服务,CPU预测性能如何优化,有哪些建议 #17449

@sshilei

Description

@sshilei

我现在使用的 baidu/paddlepaddle/paddle@paddle_1-0-15-1_PD_B, 我的模型是简单的CBOW, C++预测时间约为20ms,不太符合预期,请问有哪些预测时间优化建议。
代码及config如下:

image

image

Metadata

Metadata

Labels

预测原名Inference,包含Capi预测问题等

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions