DataCollatorForLanguageModeling warning error fixed

[mjaliz]

What does this PR do?

This PR fixes a bug in the DataCollatorForLanguageModeling class that causes a TypeError when instantiating the data collator with whole_word_mask=True.

Problem

When users try to instantiate DataCollatorForLanguageModeling with whole_word_mask=True, they encounter the following error:

TypeError: category must be a Warning subclass, not 'str'
This error occurs in the __post_init__ method at line 724-727, where warnings.warn() is called incorrectly with two separate string arguments. The second string is being interpreted as the category parameter, which must be a Warning subclass (like UserWarning), not a string.
Solution
The fix combines the two warning message strings into a single message and explicitly passes UserWarning as the category parameter, following the pattern used elsewhere in the same file (e.g., line 719-722). Before:
warnings.warn(
    "Random token replacement is not supported with whole word masking.",
    "Setting mask_replace_prob to 1.",
)
After:
warnings.warn(
    "Random token replacement is not supported with whole word masking. "
    "Setting mask_replace_prob to 1.",
    UserWarning,
)
This ensures the data collator can be instantiated correctly when using whole word masking for masked language modeling tasks.
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](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section?
 Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? 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](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation).
 Did you write any new necessary tests?
Who can review?
@SunMarc @ArthurZucker - This is a small bug fix in the data collator that prevents instantiation with whole word masking enabled.
Note: Discovered and fixed by @mjaliz
y

[Rocketknight1]

Yes, thank you for the fix! The extra comma made the second line appear to be the warning category, so we crashed in this case instead of throwing the warning correctly.

[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.

[mjaliz]

Thank you so much for your attention.