Repository navigation
fix: count literal special-token spellings in chunk text - #461
Open
jackiectl2 wants to merge 1 commit into
Open
jackiectl2 wants to merge 1 commit into
jackiectl2 wants to merge 1 commit into
Conversation
Signed-off-by: Tianlang (Jackie) Chen <305351710+jackiectl2@users.noreply.github.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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?
Encode corpus text with special-token rejection disabled in both token estimation and truncation. Literal strings such as <|endoftext|> in documents are ordinary content, not tokenizer control tokens.
Validation
3 regressions passed and fail on the base; Ruff checks passed. Tests use a real deterministic tiktoken encoder with a small byte vocabulary to remain offline.
The default cl100k_base vocabulary download was blocked by the environment; no embedding/model run was performed.
Checklist