[Bugfix][MRV2][Spec Decode] Reject draft slots that were never proposed (#58784)

qizixi committed 23 days ago

[Core] Rework scheduler `skipped_waiting` queue (#58947)

Nick Hill committed 23 days ago

[Mypy] Fix mypy typing for Voxtral and vision models (#58251)

Ashraf Bhuiyan committed 23 days ago

[Bugfix][Scheduler] Preserve logprobs across streaming continuations (#57447)

0xsensei committed 23 days ago

[Perf][MRV2] Allow FULL decode graphs for one-token prompt tails (#58400)

Nick Hill committed 23 days ago

[Model Runner V2][Spec Decode] Support spec decode with draft model (#43091)

Xiaoshuang Wang committed 23 days ago

[Bugfix] Fix resumable request + async scheduling handoff race (#58259)

yzong-rh committed 23 days ago

[Bugfix] Don't sync-police or retry FlashInfer all-reduce workspace creation in eager mode (#58498)

Kevin H. Luu committed 23 days ago

[ROCm][MLA] Add an AITER ASM round-robin decode route for DCP multi-token verify (#56861)

xiaohuguo2023 committed 23 days ago

[ROCm][CI] Increase timeout for Entrypoints Unit (#59076)

djramic committed 23 days ago

[ROCm][CI] Pin OpenTelemetry to LMCache's cap in the ROCm images (#59051)

Rohan Potdar committed 23 days ago

[CI][Bugfix] Relax packed_qk_rope_ correctness test to one ULP (#59008)

stefankoncarevic committed 23 days ago

[Quant] Use canonical N-first weight format for CT WNA16 MoE (#52798)

HDCharles committed 23 days ago

[KVConnector][NIXL] Support packed MLA KV layouts in pipeline-parallel push prefill (#50499)

qizixi committed 23 days ago

[Bugfix][Kimi-K3] Refresh DSpark context KV cache pointers after the KV cache is re-bound (#58814)

Oxana Korzh committed 23 days ago

[ROCm]Keep LMCache OpenTelemetry on the image's 1.40 stack (#59056)

Micah Williamson committed 23 days ago

[ROCm][Perf] Replace torch.topk in DSA candidate block selection (#58208)

Fangzhou Ai committed 23 days ago

[LoRA] Support variable num_labels for sequence classification (#57766)

linitra24 committed 23 days ago

[Feature] Add fixed-token prefill scoring (#54335)

aoshen02 committed 23 days ago

[ROCm][Perf] Kimi-K3 Enable sharded latent MoE up-projection under EP (#54956)

xaguilar-amd committed 23 days ago