Skip to content

[Trainer] Undo #29896#30129

Merged
NielsRogge merged 4 commits intohuggingface:mainfrom
NielsRogge:undo_29896
Apr 9, 2024
Merged

[Trainer] Undo #29896#30129
NielsRogge merged 4 commits intohuggingface:mainfrom
NielsRogge:undo_29896

Conversation

@NielsRogge
Copy link
Copy Markdown
Collaborator

What does this PR do?

This PR reverts #29896 as a better way to handle it is #30102, but that one requires a huge amount of code updates + more extensive offline discussion as it's quite a big change to something people use a lot.

Therefore, a better fix for #29790 is simply this:

default_collator = (
            DataCollatorWithPadding(tokenizer)
            if tokenizer is not None and isinstance(tokenizer, PreTrainedTokenizerBase)
            else default_data_collator
        )

@HuggingFaceDocBuilderDev
Copy link
Copy Markdown

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
Copy Markdown
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.

Thanks! Just a comment on separating out the revert and additional change

@NielsRogge
Copy link
Copy Markdown
Collaborator Author

Ok opened a separate PR for this

Copy link
Copy Markdown
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.

Thanks for handling this!

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