Conversation
Store compressed indexer pools in packed token KV with per-request tails, and transfer the complete hybrid runtime state across PD stages. Use native 512-wide MLA and vLLM Top512 selection as an indexer method. Share mHC kernels, update packed KDA tuning, and fix GLM tool-call parsing. Validation: 118 targeted tests passed on H200; Python and shell syntax checks and git diff --check passed.
Pack only valid K-pool pages and schedule each decode or MTP query with its actual GPU length. Preserve the contiguous prefill path and the 1M CUDA Graph capacity. Validation: 233 unit tests passed, including exact logits/top-k scores, MTP widths 1/3/6, empty HOLD rows, and changing lengths in 1M Graph replay. Black and Flake8 passed.
Pass activation limits through DeepEP prefill and decode, and enable GLM expert parallelism with complete shared expert outputs.
Keep global shared request slots while using the local DP capacity for GPU state allocation, scheduling, and CUDA Graph capture. Preserve global capacity for diverse mode and cross-DP prompt cache fetch. Validated with 69 targeted tests and DP8/EP8/MTP2 inference at 32 concurrent requests and 1M context.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
No description provided.