[WIP] Add a dedicated tokenizer for byte level transformers #36216
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
|
I've written tests but I'm not sure where I should place them. Where do you want them, if anywhere? @ArthurZucker |
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Outdated suggestions cannot be applied.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.
What does this PR do?
This is a simple implementation of a byte tokenizer.
It is useful for models like the recent BLT from Meta.
It accepts all the encoding schemes from the built-in
encodefunction of the standard library.In particular, "UTF-32-BE" is useful since it allows to use fixed size patching for the bytes.
The class doesn't use a vocabulary, but I still kept the usual methods for compatibility.
Worth noting is that the special tokens use the legacy codepoints built-in Unicode.
For example the mask token is
"\u001a", which is the "substitute" codepoint.See the Unicode table.
Fixes #36202