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.

Sign up for GitHub

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

Jump to bottom

summarize_grads_and_vars and debug_summaries are very slow #91

Closed
hnyu opened this issue Jul 4, 2019 · 5 comments
Closed

summarize_grads_and_vars and debug_summaries are very slow #91

hnyu opened this issue Jul 4, 2019 · 5 comments

Comments

@hnyu
Copy link
Collaborator

hnyu commented Jul 4, 2019 •
edited
Loading

In the PPO humanoid example, these two summaries will take roughly 3 seconds per training update while the training update itself is only about 2 seconds.
@emailweixu
Copy link
Contributor

We can adjust the summary interval. We need to have something to measure how much time is spent on environment.step, predict, training, summary, and evaluation so that we can make right adjustment.

@hnyu hnyu mentioned this issue Jul 13, 2019
Closed
@hnyu
Copy link
Collaborator Author

hnyu commented Jul 13, 2019

Is it possible to optimize the summary code?

@witwolf
Copy link
Contributor

witwolf commented Jul 16, 2019 •
edited
Loading

summary interval do not much help , it only reduce the frequency of write to file , computation for data (particular histogram data ..) , serialization for metadata etc has no reduction， and this part can be optimized at least

@witwolf witwolf mentioned this issue Jul 16, 2019
Merged
@hnyu
Copy link
Collaborator Author

hnyu commented Jul 17, 2019

Close this issue for #124

@hnyu hnyu closed this as completed Jul 17, 2019
@emailweixu
Copy link
Contributor

#120 also helps alleviate this issue

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@witwolf @emailweixu @hnyu