Allow BPE to treat special tokens as one token #939
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
|
/gcbrun |
mattdangerw
requested changes
Mar 30, 2023
| ) | ||
| self.assertAllEqual(call_output, expected) | ||
|
|
||
| def test_tokenize_with_special_tokens(self): |
There was a problem hiding this comment.
have you tested detokenization? That is probably worth checking out
There was a problem hiding this comment.
yea, it's actually not affected by this PR.
Here the special token has to be present inside the vocab, such as <|endoftext|>.
| if special_tokens: | ||
|
|
||
| def build_alt(i): | ||
| return " " + "I" * i + "IGuessIamAValidWord" + "d" * i |
There was a problem hiding this comment.
heh this feels a little too janky :)
What if we...
- Move this function out (don't nest it) for readability.
- Take in the original special_token.
- Strip all splittable
\s\p{L}\p{N}from the original special token. - Add a unique prefix or suffix.
f"Ĵ{i}"or something like that.
That is nice in that it won't expand the length of these characters much, and if someone is debugging intermediate state, it will look a little more readable.
|
/gcbrun |
|
/gcbrun |
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Outdated suggestions cannot be applied.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.
There are some cases that BPE needs to treat special tokens as one token, e.g.,
"<|endoftext|>"in GPT2.