feat: Enable Packing for pretokenised dataset #468
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dushyantbehl
requested review from
Ssukriti,
aluu317,
anhuong,
fabianlim and
kmehant
as code owners
|
Thanks for making a pull request! 😃 |
dushyantbehl
force-pushed
the
packing-for-pretokenised
branch
from
d6bcee1 to
4c3f1c3
Compare
2 tasks
dushyantbehl
force-pushed
the
packing-for-pretokenised
branch
from
4c3f1c3 to
d0e0e20
Compare
Collaborator
|
5 tasks
dushyantbehl
force-pushed
the
packing-for-pretokenised
branch
5 times, most recently
from
6d4e771 to
73b81c1
Compare
Remove skip_prepare_dataset flag as we want trainer to pack the dataset if tokenized which is done via prepare_dataset Signed-off-by: Dushyant Behl <[email protected]>
dushyantbehl
force-pushed
the
packing-for-pretokenised
branch
from
73b81c1 to
5030dba
Compare
dushyantbehl
changed the title
kmehant
reviewed
Mar 20, 2025
Comment on lines
-399
to
-402
| dataset_kwargs = {} | ||
| if is_tokenized_dataset: | ||
| dataset_kwargs["skip_prepare_dataset"] = True | ||
|
|
Collaborator
There was a problem hiding this comment.
Any reason we dont want to skip anymore?
Collaborator
Author
There was a problem hiding this comment.
yes...added explanation here #468 (comment)
We need to remove this check to call prepare_dataset and and enable packing for the pretokenized dataset.
Collaborator
There was a problem hiding this comment.
Just checked, looks like it checks for tokenized with trl and skips prepare.
Collaborator
Author
There was a problem hiding this comment.
This line is what we need - https://github.com/huggingface/trl/blob/a34987956cd5bf08ed7501da2510b9404bede695/trl/trainer/sft_trainer.py#L467
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description of the change
This PR adds support for packing for pretokenized datasets which was added in
transformers>=4.46Adds
DataCollatorForSeqtoSeqfor usage with pretokenized dataset when packing is enabled withpadding=FalseRemoves
skip_prepare_datasetas its needed for enabling packing on a tokenized dataset.Related issue number
How to verify the PR
Was the PR tested