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

Eager: CPU Performance/Operation Overheads #14130

Closed
asimshankar opened this Issue Oct 31, 2017 · 2 comments

Comments

Projects
None yet
4 participants
@asimshankar
Member

asimshankar commented Oct 31, 2017

(This applies only when eager execution has been enabled via tfe.enable_eager_execution())

Eager execution re-uses most of the same Python code used for constructing TensorFlow graphs. Many of these paths have not been optimized for part of the critical path of computation. As a result, the CPU overheads of executing Python code for every operation are higher than we’d like.

Consequently, the performance of eager execution on models with many small computations, or models executed on CPU may be dominated by these overheads.

Overheads are measured using microbenchmarks such as in benchmarks_test.py and model-level benchmarks such as those used for ResNet50 and the PTB RNN

@tensorflowbutler

This comment has been minimized.

Member

tensorflowbutler commented Oct 23, 2018

Nagging Assignee @akshaym: It has been 195 days with no activity and this issue has an assignee. Please update the label and/or status accordingly.

@asimshankar

This comment has been minimized.

Member

asimshankar commented Oct 23, 2018

Closing this out as many many improvements have been made and others are coming too.
Not sure about the value of this issue since it isn't very specific (yes, I know I filed this :))

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