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

Temporary fix to get nightly green again #696

Merged
merged 1 commit into from
Jan 27, 2023

Conversation

mattdangerw
Copy link
Member

Our segfault appears to do with import order. Importing sentencepiece first, everything is ok, importing tf first, the trainer breaks.

This should be a safe change to get CI green while we look for a longer term solution.

Copy link
Contributor

@chenmoneygithub chenmoneygithub left a comment

Choose a reason for hiding this comment

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

thx!

Copy link
Contributor

@jbischof jbischof left a comment

Choose a reason for hiding this comment

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

Short fix is good. Thanks!

I guess this means that someone running a SP-based tokenizer will be prompted to install the lib separately?

@mattdangerw
Copy link
Member Author

I guess this means that someone running a SP-based tokenizer will be prompted to install the lib separately?

We actually don't need the sentencepiece package for tokenization, that is handled by tf-text.

We need sentencepiece for training a "sentencepiece proto", which happens both in our unit testing and the compute_sentence_piece_proto utility. If an end users attempts to call compute_sentence_piece_proto without the package installed, they get an error with instructions on installation.

@mattdangerw mattdangerw merged commit 9c5f850 into keras-team:master Jan 27, 2023
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