Repository navigation
Fix vLLM routing capture and add GB10 validation - #139
Merged
WestWaters merged 1 commit intoOct 6, 2026
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fix FlashInfer routing accounting, including one-token prompt continuations that use a decode kernel. Keep the scheduler's phase flags without changing kernel dispatch. Unsupported observations stay in an unknown bucket.
Added code, reasoning and general-text workloads at two context lengths. Serial and batched runs on one GB10 matched 21,504 prompt tokens and 744 decode-feedback tokens in every router layer, with no recording errors or unknown tokens.
Serial outputs matched two runs after removing the recorder. Five batched profiles matched both controls. Short-code controls differed from each other, so batched output stability remains unresolved.
45 focused tests pass. Includes a CPU-only test job and aggregate results, not raw prompts, outputs or machine details. The runtime was a local vLLM development build, not a stock release. This is routing validation, not a quality, speed or multi-node claim.
Details:
notes/vllm-routing-capture.mdandnotes/vllm-routing-profiles.md.I used AI assistance and reviewed and tested the changes.