Skip {% generation %} and {% endgeneration %} template handling

[alvarobartt]

What does this PR do?

This PR is just a tentative fix for both {% generation %} and {% endgeneration %} custom syntax introduced within some chat templates for returning a mask of the assistant generated tokens which is only useful during training and should indeed be ignored during inference, affecting models as e.g. microsoft/Phi-4-reasoning-plus.

Thanks to @Rocketknight1 for the explanation!

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline, Pull Request section?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the documentation guidelines, and here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

@Narsil

[Narsil]

LGTM

[Rocketknight1]

This will work fine in a lot of cases, but will fail to catch syntax like {%- generation %}! Because the tags can also include whitespace-stripping syntax like {-, it can be dangerous to just delete them instead of adding a null Jinja handler for generation.

[alvarobartt]

This will work fine in a lot of cases, but will fail to catch syntax like {%- generation %}! Because the tags can also include whitespace-stripping syntax like {-, it can be dangerous to just delete them instead of adding a null Jinja handler for generation.

Thanks! I'll make sure to update it to handle that properly rather than a raw replacement 👍