Skip to content

Conversation

@wangxicoding
Copy link
Contributor

PR types

Others

PR changes

Others

Describe

Compute interceptor 完成上下游消息依赖与消息发送。

@paddle-bot-old
Copy link

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

Copy link
Contributor

@FeixLiu FeixLiu left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM


private:
std::unordered_set<int64_t> upstream_deps_;
// FIXME(wangxi): if use step_ and max_steps_, how to restart step_ from 0
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is a message RESET, which was designed to reset the step_ to 0.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

那感觉还得加一个start的interceptor,发送开始信息,然后到了指定micro_step后,再发reset信息。

@FeixLiu
Copy link
Contributor

FeixLiu commented Nov 24, 2021

现在这么设计是不是停不下来呀,需要等那个step_的开发?outbuffer和inready都可以根据RESET消息重置吧?

@wangxicoding
Copy link
Contributor Author

现在这么设计是不是停不下来呀,需要等那个step_的开发?outbuffer和inready都可以根据RESET消息重置吧?

是的,停不下来,reset不能把interceptor的线程停下来,只有stop可以。

@wangxicoding wangxicoding merged commit be3b774 into PaddlePaddle:develop Nov 24, 2021
@wangxicoding wangxicoding deleted the complete_compute_interceptor branch November 24, 2021 02:46
Zjq9409 pushed a commit to Zjq9409/Paddle that referenced this pull request Dec 10, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants