[Bugfix] Catch and log invalid token ids in detokenizer #24351
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
Signed-off-by: Nick Hill <nhill@redhat.com>
njhill
requested review from
WoosukKwon,
alexm-redhat,
comaniac,
robertgshaw2-redhat and
ywang96
as code owners
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces a targeted bugfix to handle a rare OverflowError within the detokenizer. The change correctly isolates the OverflowError, logs the exception with the problematic token ID for future diagnostics, and ensures the system can continue processing by treating the failed token as None. This is a robust and appropriate way to handle an intermittent, hard-to-reproduce issue. The implementation is sound and effectively insulates the system from this specific failure mode.
DarkLight1337
approved these changes
Sep 6, 2025
eicherseiji
pushed a commit
to eicherseiji/vllm
that referenced
this pull request
Sep 9, 2025
…#24351) Signed-off-by: Nick Hill <nhill@redhat.com>
skyloevil
pushed a commit
to skyloevil/vllm
that referenced
this pull request
Sep 13, 2025
…#24351) Signed-off-by: Nick Hill <nhill@redhat.com>
FeiDaLI
pushed a commit
to FeiDaLI/vllm
that referenced
this pull request
Sep 25, 2025
…#24351) Signed-off-by: Nick Hill <nhill@redhat.com>
Closed
1 task
njhill
added a commit
to njhill/vllm
that referenced
this pull request
Oct 8, 2025
This is an update to the "workaround" added in vllm-project#24351. That PR insulates against occasional negative token ids that can be produced occasionally, though we still don't know the root cause. With the update to tokenizers 0.22.1, this error manifests as a TypeError rather than an OverflowError, so the patch needs to be updated to account for this. Signed-off-by: Nick Hill <nhill@redhat.com>
xuebwang-amd
pushed a commit
to xuebwang-amd/vllm
that referenced
this pull request
Oct 10, 2025
…#24351) Signed-off-by: Nick Hill <nhill@redhat.com> Signed-off-by: xuebwang-amd <xuebwang@amd.com>
xuebwang-amd
pushed a commit
to xuebwang-amd/vllm
that referenced
this pull request
Oct 24, 2025
…#24351) Signed-off-by: Nick Hill <nhill@redhat.com> Signed-off-by: xuebwang-amd <xuebwang@amd.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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 is a token overflow issue with quantized models that happens occasionally: #21951
The exact cause is not yet clear and it appears to be difficult to reproduce.
This change just insulates against the error, skipping the problem token but still logging the exception. It also prints the token id in question (suspected to be negative) which should help with further diagnosis.