fix(pd_vllm): prefill request hygiene, ignore_eos support, usage in its own chunk - #55
Merged
Merged
Conversation
…efill request Supports streaming clients against a PD deployment: the prefill request now drops stream_options and max_completion_tokens from the client body, since the first contradicts the stream=False we force (vLLM rejects the pair with a 400 during body parsing) and the second overrides our max_tokens=1. Verified with `vllm bench serve --backend openai-chat`, which sends both unconditionally and previously failed every request.
mzmssg
approved these changes
Aug 5, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two fixes for serving OpenAI-compatible clients over a PD deployment.
The prefill request now drops
stream_optionsandmax_completion_tokensfrom the client body: the first contradicts thestream=Falsewe force (vLLM rejects the pair with a 400 during body parsing) and the second overrides ourmax_tokens=1. Verified withvllm bench serve --backend openai-chat, which sends both unconditionally and previously failed every request.ignore_eosis now forwarded to the decode engine and honoured by the MLA/NSA adapter, streaming usage moves into its own trailing chunk carryingtotal_tokens(the shape vLLM and the OpenAI API emit), and thetransformers/tokenizerspins are relaxed to>=.