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

Compiling graphs #85

Closed
mortendahl opened this issue Sep 3, 2018 · 6 comments
Closed

Compiling graphs #85

mortendahl opened this issue Sep 3, 2018 · 6 comments
Assignees
@justin1121 @mortendahl
Labels
research Research heavy work of more theoretical nature

Comments

@mortendahl
Copy link
Member

mortendahl commented Sep 3, 2018 •
edited

Look into whether this gives any significant performance boosts:
@mortendahl
Copy link
Member Author

this might be an issue: "the resulting pruned subgraph cannot contain Placeholder or Variable nodes"

@mortendahl mortendahl added the research Research heavy work of more theoretical nature label Sep 4, 2018
@morgangiraud
Copy link
Contributor

tfcompile is made to be used for inference purpose, so as long as we can freeze a graph to remove variables, we're fine with the Variable limitation.

About the placeholders, I guess you can always tfcompile from the first operation after placeholders , no?

@mortendahl
Copy link
Member Author

mortendahl commented Sep 5, 2018 •
edited

I think you can still have placeholders in the compiled graph, no? Ah no, "feeds" seems to be different than placeholders!

@mortendahl
Copy link
Member Author

next step could be to simply get a compiled version of TensorFlow with XLA support running

@mortendahl mortendahl added the prioritized A label denoting that the issue has been prioritized and is waiting for someone to tackle it. label Dec 14, 2018
This was referenced Dec 14, 2018
Open
Open
@mortendahl mortendahl removed the prioritized A label denoting that the issue has been prioritized and is waiting for someone to tackle it. label Dec 20, 2018
@justin1121
Copy link
Member

Further investigation has gone into using XLA to compile graphs and doesn't really seem beneficial at this point.

@justin1121
Copy link
Member

See this comment https://github.com/mortendahl/tf-encrypted/issues/229#issuecomment-462420840.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@justin1121 justin1121

@mortendahl mortendahl

Labels
research Research heavy work of more theoretical nature
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@morgangiraud @justin1121 @mortendahl