Skip to content

Conversation

@gante
Copy link
Contributor

@gante gante commented Sep 18, 2024

What does this PR do?

Fixes failing CI on rag, since #32564.

From diving into the error, #32564 did not cause the error, but rather the bug it fixes probably exposed this existing configuration issue.

@gante gante requested a review from LysandreJik September 18, 2024 18:37
question_encoder_tester = DPRModelTester(self)
dpr_config_and_inputs = question_encoder_tester.prepare_config_and_inputs()
generator_tester = T5ModelTester(self, vocab_size=1100)
generator_tester = T5ModelTester(self, vocab_size=1101)
Copy link
Contributor Author

@gante gante Sep 18, 2024

Choose a reason for hiding this comment

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

If we go check the test tokenizer, the last expanded token has index 1100 (pad token), so the vocab size has to be 1101 otherwise we will get indexing errors (causing the red CI)

Prior to #32564 the same was happening. I'm assuming the fix in #32564 exposed this issue.

@HuggingFaceDocBuilderDev

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

Copy link
Contributor

@amyeroberts amyeroberts left a comment

Choose a reason for hiding this comment

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

Giving a quick approval as @LysandreJik is in SF. Thanks for fixing so quickly!

@gante gante merged commit f3b3810 into huggingface:main Sep 19, 2024
@gante gante deleted the fix_rag branch September 19, 2024 10:55
BernardZach pushed a commit to BernardZach/transformers that referenced this pull request Dec 5, 2024
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