Skip to content

Don't preload tensorflow and torch#355

Merged
rosbo merged 1 commit intomasterfrom
sitecustomize
Oct 26, 2018
Merged

Don't preload tensorflow and torch#355
rosbo merged 1 commit intomasterfrom
sitecustomize

Conversation

@rosbo
Copy link
Contributor

@rosbo rosbo commented Oct 26, 2018

A little while back, we had issue with hiting the size limit for loading object in static TLS (thread local storage). A workaround was to preload those libraries. Newer versions of these libraries and their dependencies reduce the use of static TLS.

Removing the preloading of these libraries will make using pip install and the like in kernels easier.

I ran our test suite and the kernel reported by a user originally that was causing TLS issue and they all run: https://www.kaggle.com/kmader/library-overload

BUG=118388247

@rosbo rosbo requested a review from d1jang October 26, 2018 20:36
@d1jang d1jang requested a review from emzeq October 26, 2018 20:37
Copy link
Contributor

@emzeq emzeq left a comment

Choose a reason for hiding this comment

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

LGTM!

@rosbo rosbo merged commit 6694e89 into master Oct 26, 2018
@rosbo rosbo deleted the sitecustomize branch October 26, 2018 21:10
@sebbov
Copy link
Contributor

sebbov commented Oct 26, 2018

yay!

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants