Dataloader now shuffles the shards and documents within #52
Conversation
This file contains 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
lukasugar
reviewed
Jul 4, 2024
fraserlove
force-pushed
the
dataloader-shuffle
branch
from
dd75feb
to
faaed08
Compare
|
@lukasugar I have been training multiply model and have event customized the model. I was seeing if you would be interest in working together on a new model. I have a FastTokenizer with 128K vocab and new special tokens. If you are interested I can share more information in a PM |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Added a function to separate a shard into documents, shuffle the documents, and then concatenate the documents back to a single shard. Individual shards are also shuffled. This removes the unusual periodicity present within the fineweb-edu dataset and results in a smoother training loss trajectory (see image below).