🐯 fix: use_liger_kernel with IterableDataset

[jue-jue-zi]

What does this PR do?

When using load_dataset(.., streaming=True) to load the training set as an IterableDataset, the dataset.column_names may be None.

If use_liger_kernel is enabled at the same time, SFTTrainer will attempt to remove unused columns during data preparation:

collator_expected_keys = {"input_ids", "seq_lengths", "completion_mask", "assistant_masks"}
dataset.select_columns(collator_expected_keys.intersection(dataset.column_names))

Since dataset.column_names is None in this case, this operation causes a training error.

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? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes?
• 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.

[qgallouedec]

cool, thanks!

[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.