Skip to content

[Bugfix] Fix Incremental Detokenization with tokenizers == 0.22.0 #24159

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
vllm-bot merged 5 commits into from
Sep 4, 2025
+3 −2
Merged

[Bugfix] Fix Incremental Detokenization with tokenizers == 0.22.0 #24159

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4adf155
fix tokenizers update
faaany Sep 3, 2025
f3dadb9
Update vllm/v1/engine/detokenizer.py
faaany Sep 3, 2025
f8dbaf6
Merge branch 'main' into fix_tokenizers_update
faaany Sep 3, 2025
095f441
Merge branch 'main' into fix_tokenizers_update
faaany Sep 4, 2025
c447eab
Merge branch 'main' into fix_tokenizers_update
faaany Sep 4, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions vllm/v1/engine/detokenizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -234,7 +234,7 @@ def _protected_step(self, next_token_id: int) -> Optional[str]:
try:
token = self.stream.step(self.tokenizer, next_token_id)
except Exception as e:
if str(e) != INVALID_PREFIX_ERR_MSG:
if INVALID_PREFIX_ERR_MSG not in str(e):
Comment thread
faaany marked this conversation as resolved.
Outdated
raise e
# Recover from edge case where tokenizer can produce non-monotonic,
# invalid UTF-8 output, which breaks the internal state of
Expand All @@ -243,7 +243,8 @@ def _protected_step(self, next_token_id: int) -> Optional[str]:
logger.warning(
"Encountered invalid prefix detokenization error"
" for request %s, resetting decode stream.", self.request_id)
self.stream = DecodeStream(self.skip_special_tokens)
self.stream = DecodeStream(
skip_special_tokens=self.skip_special_tokens)

@yma11 yma11 Sep 3, 2025

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this change necessary? DecodeStream introduce more args?

@faaany faaany Sep 3, 2025 •
edited
Loading

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes, as can be seen from: https://github.com/huggingface/tokenizers/pull/1856/files#diff-780be0b9b76e7260ed2be2249d17c4a879b7e2e98e9f30f26dc9c65501f775d1R672. Otherwise, we would get an error that ids should not be Boolean type.

token = self.stream.step(self.tokenizer, next_token_id)
return token

Expand Down
Loading