Repository navigation
fix: count literal special tokens like <|endoftext|> as plain text - #129
Merged
Merged
Conversation
Repos that vendor tokenizer code (e.g. CLIP) contain the literal string '<|endoftext|>'. tiktoken rejects it by default, so count_tokens raised and module clustering aborted before any LLM call.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
Some repos contain the literal string
<|endoftext|>in source code (e.g. vendored CLIP tokenizers).count_tokenscallstiktoken'sencode, which raisesValueError: Encountered text corresponding to disallowed special token '<|endoftext|>'by default. This aborts module clustering before any LLM call (seen onFsoft-AIC/SBDC).Fix
Pass
disallowed_special=()incount_tokensso such strings are counted as normal text. This only affects token counting, not what is sent to the model.