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.

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

Place histogram bucketing logic on CPU explicitly when using TPUStrategy #2883

Merged
merged 3 commits into from
Nov 5, 2019

Conversation

hongjunChoi
Copy link
Contributor

  • Motivation for features / changes
    Manually place bucketing logic in writing histogram to CPU when TPUStrategy is used.

  • Technical description of changes
    _buckets() generates a graph with dynamic shapes which are not supported on TPU. As so, place the function in outside compilation cluster when TPUStrategy is in scope.

@hongjunChoi hongjunChoi changed the title Update summary_v2.py Place histogram bucketing logic on CPU explicitly when using TPUStrategy Nov 4, 2019
@nfelt nfelt self-requested a review November 4, 2019 22:24
Copy link
Contributor

@nfelt nfelt left a comment

Choose a reason for hiding this comment

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

LGTM, but would you mind filing a GitHub issue for us to reconsider the fact that the format requires supporting dynamic shapes? It would be good ultimately to not need this special casing.

@nfelt nfelt merged commit f1443b3 into tensorflow:master Nov 5, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants