vLLM/vllm/model_executor at ae0350c0de721f99d9779bc06605da27282f8ed7 - vLLM

iHore/vLLM

Fork 0

Files

T

History

MOVZXandHinata ae0350c0de

[Spec Decode][MRV2] RecoverSSM for Qwen GDN and PLE short conv (PR #58863 )

Intake of open upstream PR vllm-project/vllm#58863. Active only with
VLLM_USE_V2_MODEL_RUNNER=1; inert otherwise.

Speculative decoding on Gated DeltaNet snapshotted the full recurrent
state after every draft position and reserved one Mamba block per draft
position. On a 24 GB card that is reserved VRAM we never use, plus
per-step copy traffic.

RecoverSSM, the protocol Kimi-K3 KDA already uses, replaces this:
verify runs from a read-only checkpoint and stores a small per-token
replay record [correction (V), normalized key (K), gate (1)] in fp32;
commit replays only the accepted tokens and writes the recurrent state
once, plus the block-boundary state in align mode. The PLE dilated
short conv compacts its accepted window with the same plan.

Conflict in vllm/models/qwen4_exp/nvidia/mtp.py: our get_top_tokens()
work put LocalArgmaxMixin on Qwen4ExpMTP, the PR adds SupportsReplaySSM
to the same base list. Kept both.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>

2026-09-29 13:39:09 +07:00

determinism

[Bugfix] batch_invariant: keep non-AllReduce collectives enabled on NCCL >= 2.31 (#58179 )

2026-09-22 14:44:00 -04:00

hw_agnostic

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212 )

2026-09-16 10:57:24 -07:00

kernels

[Bugfix] Revert "[ROCm][Perf] Replace torch.topk in DSA candidate block selection" (#59125 )

2026-09-28 20:11:48 -07:00

layers

[Spec Decode][MRV2] RecoverSSM for Qwen GDN and PLE short conv (PR #58863 )