Repository navigation
Conversation
An empty tokenized prompt (e.g. "prompt": "" with a tokenizer that adds no BOS) reached BatchGenerator.insert_segments, which raises ValueError inside the generation thread. The thread ended, and every later request failed. Reject the request with 400 before it is inserted; the server keeps serving. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
An empty tokenized prompt crashes the batched server.
"prompt": ""with a tokenizer that adds no BOS token.BatchGenerator.insert_segments, which raisesValueError("Sequence ... has an empty prompt.")inside the generation thread.This PR:
max_tokens: 0;ValueErrors raised while setting up generation.#1513 / #1909 track the general "keep the generation loop alive" problem. This is the narrow fix for one trigger.
Test:
test_empty_prompt_is_rejected_and_server_survives. It sends an empty prompt (expects 400), then a normal request (expects 200). It fails onmainand passes with this change; all 44 tests intests/test_server.pypass.🤖 Generated with Claude Code