 Yihao WangandClaude Fable 5
|
f5bed255c0
|
[diffusion] feat: support key masks on USPAttention's replicated-prefix path (#36735)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
2026-09-03 20:13:33 +08:00 |
|
|
Yihao WangandClaude Fable 5
|
14444c6a04
|
[diffusion] feat: add maybe_record_function profiler spans for request phases (#35922)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
2026-09-03 20:12:54 +08:00 |
|
|
|
2bb25dc18b
|
[Speculative Decoding] Add native UNO serving support (#37667)
Co-authored-by: drproduck <drproduck@MacBook-Air-2.local>
Co-authored-by: BBuf <1182563586@qq.com>
|
2026-09-03 20:08:41 +08:00 |
|
|
Cheng Wan
|
a11dba1a01
|
[Feature] Unified memory: support decode context parallelism for the trtllm_mla family (#37693)
|
2026-09-03 03:37:12 -07:00 |
|
|
Duyi-Wang
|
429ac2d82c
|
[AMD] Fix DSv4 draft extend taking the target compression path during prefill (#37713)
|
2026-09-03 02:46:49 -07:00 |
|
|
|
27b7a2dc3b
|
[Kimi K3] Rework skipped-think fix as opt-in force_nonempty_content with streaming coverage (#34187)
Co-authored-by: yhyang201 <yhyang201@gmail.com>
Co-authored-by: Yuhao Yang <47235274+yhyang201@users.noreply.github.com>
|
2026-09-03 17:37:28 +08:00 |
|
|
kkandwunhuang
|
a6001478f4
|
[AMD] Perf Kimi-K3 MoE optimization (#33838)
Co-authored-by: wunhuang <wunhuang@amd.com>
|
2026-09-03 02:28:28 -07:00 |
|
|
Xinyi SongandThomas Wang
|
7ed29eba80
|
[AMD] Fix FP4 indexer OOR (#37660)
Co-authored-by: Thomas Wang <1am9trash@gmail.com>
|
2026-09-03 01:49:46 -07:00 |
|
|
|
3bac084d4e
|
[Model] Add native IFM K2 Horizon serving support (#37654)
Co-authored-by: Xiaoyu Zhang <1182563586@qq.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
|
2026-09-03 16:39:43 +08:00 |
|
|
YC Yen-Ching TsengandPhil Li
|
1fb85053e7
|
[AMD][Diffusion] Migrate FlyDSL fused norm kernels to the v0.3.0 stable API (#36349)
Co-authored-by: Phil Li <haicli@amd.com>
|
2026-09-02 23:03:08 -07:00 |
|
|
xiaobochen-amdandZhang, Jiejing
|
030d7e7e9b
|
[ROCm] Define the DSA head-gate graph helpers on HIP (#37118)
Co-authored-by: Zhang, Jiejing <jiejing.zhang@amd.com>
|
2026-09-02 22:55:45 -07:00 |
|
|
Liangsheng Yin
|
54da74e83e
|
[Fix] Broadcast PP dynamic-chunk profiling failures so every rank disables together (#37675)
|
2026-09-02 21:09:04 -07:00 |
|
|
Khoa PhamandClaude Opus 5
|
cf3173aeb9
|
[Perf] Walk the radix tree by offset instead of re-slicing token storage (ported from #36507) (#37324)
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-03 11:13:03 +08:00 |
|
|
James Liu
|
4229088a48
|
feat(kernels): generalize persistent CuTe JIT cache (#33911)
|
2026-09-02 19:59:04 -07:00 |
|
|
Alex NailsandAlison Shao
|
28262c20df
|
[CI][RFC] Replace black-jupyter with ruff-format (#37210)
Co-authored-by: Alison Shao <a.shao@wustl.edu>
|
2026-09-02 19:46:08 -07:00 |
|
|
|
2641e427be
|
Xpu/weekly simple model enablement 2026 08 30 (#37193)
Co-authored-by: dayanandav <dayananda.vasantha.kumar@intel.com>
Co-authored-by: Girijala, Pavan Sivaram <pavan.sivaram.girijala@intel.com>
Co-authored-by: Cui, Lily <lily.cui@intel.com>
Co-authored-by: Juan Muneton <juan.muneton.gallego@intel.com>
Co-authored-by: Gao, Pengfei <pengfei.gao@intel.com>
|
2026-09-03 09:35:59 +08:00 |
|
|
Liangsheng Yin
|
a522c8a4b6
|
[misc] Extract PP dynamic chunk sizing into a DynamicChunkSizer scheduler component (#37674)
|
2026-09-02 18:35:43 -07:00 |
|
|
Mick
|
0dd66def7c
|
[chore] harden checkpoint quantization metadata parsing (#36922)
|
2026-09-03 09:27:22 +08:00 |
|
|
|
fbf909b460
|
[Fix] Alpha-channel images and tool-result media ordering (port of #36507) (#37320)
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
|
2026-09-02 18:23:06 -07:00 |
|
|
Lee NauandYangmin Li
|
6e41f1ad29
|
[Fix] Preserve FP32 in SM107 MXFP8 fallback (#37489)
Co-authored-by: Yangmin Li <yangminl@nvidia.com>
|
2026-09-02 17:55:14 -07:00 |
|
|
Shiyan DengandLianmin Zheng
|
80e8302d03
|
Add SGLANG_CRASH_ON_JIT_COMPILE to forbid on-the-fly JIT compilation (#36615)
Signed-off-by: Shiyan Deng <dsy842974287@meta.com>
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>
|
2026-09-02 17:26:25 -07:00 |
|
|
|
87d60a2229
|
Improve CUDA graph and speculative execution output handling (#37329)
Co-authored-by: jiayisuse <jiayisuse@fb.com>
Co-authored-by: Yinghai Lu <yinghai@meta.com>
Co-authored-by: Hao Zhang <zhisbug@users.noreply.github.com>
Co-authored-by: Yichao Fu <yichaofu@meta.com>
|
2026-09-02 17:25:27 -07:00 |
|
|
Alison Shao
|
db1eb48651
|
[CI] Graceful teardown for the PD and HiSparse server fixtures (#37485)
|
2026-09-02 17:22:18 -07:00 |
|
|
|
ff04a00d73
|
Reduce tokenizer overhead and offload CUDA VMM publication (#37330)
Co-authored-by: Shiyan Deng <dsy842974287@meta.com>
Co-authored-by: Yinghai Lu <yinghai@meta.com>
|
2026-09-02 17:21:08 -07:00 |
|
|
Oguz Ulgen
|
f15748d965
|
[bench] Support real-traffic replay with early-stop-aware steady-state metrics in bench_one_batch_server (#37469)
|
2026-09-02 17:20:25 -07:00 |
|
|
Liangsheng Yin
|
5c46ce37f5
|
[Fix] Apply the attention-CP broadcast result in PP dynamic-chunk profiling (#37669)
|
2026-09-02 17:20:07 -07:00 |
|
|
Byron HsuandByronHsu
|
046cdaabaa
|
[Sampling] Capture masks from sampler support (#36630)
Co-authored-by: ByronHsu <ByronHsu@users.noreply.github.com>
|
2026-09-02 17:13:28 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
5ddca6819e
|
Fix unified SWA: size a non-owner's v2p by the id space it must address (#37560)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-09-02 16:56:18 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
5a1275a519
|
Converge the two SWA predicates, and stop conditioning the capture sink on the pool (#37550)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-09-02 16:55:46 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
d9848b9ecd
|
Build the unified read stream directly, without the page-table rectangle (#37512)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-09-02 16:55:16 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
18d5ffb42a
|
Size the unified read-table grid from bs, and fuse the allocator's tombstone scatters (#37511)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-09-02 16:54:26 -07:00 |
|
|
|
c05f8ae830
|
[PD] Optimize paged allocator free-list release (#37146)
Co-authored-by: wangwenming.41 <wangwenming.41@jd.com>
Co-authored-by: hnyls2002 <lsyincs@gmail.com>
Co-authored-by: Liangsheng Yin <hnyls2002@gmail.com>
|
2026-09-02 16:51:37 -07:00 |
|
|
Mohammad Miadh Angkad
|
718bd39fe0
|
[Fix] DP attention: correct the decode->extend prefix off-by-one (#37505)
|
2026-09-02 16:14:16 -07:00 |
|
|
paulzhang-tmandQiaolin-Yu
|
3fa6b86504
|
[Spec] Publish the final multi-layer EAGLE shared-read event (#36752)
Co-authored-by: Qiaolin-Yu <liin1211@outlook.com>
|
2026-09-02 15:57:50 -07:00 |
|
|
eeechoandBaizhou Zhang
|
19c30dff56
|
[SM120] DeepSeek-V4: DeepGEMM paged-MQA indexer +FP4 MoE+ page-split (#29927)
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
|
2026-09-02 15:49:59 -07:00 |
|
|
YAMY
|
982aa8acfc
|
[Bugfix] Load Qwen3.5 MTP embedding under PP (#37471)
|
2026-09-02 15:19:40 -07:00 |
|
|
YAMY
|
3c9cea8f10
|
[EAGLE] Prune draft-extend logits to selected rows (#35546)
|
2026-09-02 15:10:08 -07:00 |
|
|
YAMY
|
fe45af1e6f
|
perf(gdn): select ReplaySSM verify loop unrolling by shape (#36970)
|
2026-09-02 15:07:49 -07:00 |
|
|
cctry
|
3a855b050a
|
fix(disagg): poll receivers during decode preallocation (#37483)
|
2026-09-02 14:26:40 -07:00 |
|
|
cctry
|
ad6e830858
|
[Bugfix] Key CUDA graph dedup signatures on kernel function identity (#37657)
|
2026-09-02 14:24:31 -07:00 |
|
|
Liangsheng Yin
|
19c7679e9e
|
[mem_cache] Make free_swa sync-free on page_size == 1 (#36723)
|
2026-09-02 14:18:22 -07:00 |
|
|
JoeandBBuf
|
acea43079f
|
Fix native MoE handling of noncontiguous top-k IDs (#36407)
Co-authored-by: BBuf <1182563586@qq.com>
|
2026-09-02 14:16:19 -07:00 |
|
|
|
f8cbf000f4
|
[AMD] Enable FP4 indexer for Deepseek V4 (#37353)
Co-authored-by: 1am9trash <1am9trash@gmail.com>
Co-authored-by: AMD-yanfeiwang <256076023+AMD-yanfeiwang@users.noreply.github.com>
Co-authored-by: Thomas Wang <thomawan@amd.com>
|
2026-09-02 09:45:08 -07:00 |
|
|
Mick
|
f6aed6ec53
|
[diffusion] doc: rewrite stale diffusion compatibility matrix (#36987)
|
2026-09-02 23:42:54 +08:00 |
|
|
Kevin MiandClaude Fable 5
|
f586654518
|
[diffusion] feat: support FastH3 (4-step VSA-distilled MiniMax-H3) with a VSA-H3 attention backend (#37480)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
2026-09-02 21:39:54 +08:00 |
|
|
Liangsheng Yin
|
862a909a08
|
[Fix] Lock PP dynamic-chunk profiling requests before releasing through the tree cache (#37509)
|
2026-09-02 02:09:01 -07:00 |
|
|
Liangsheng Yin
|
ebfd8c60e5
|
[CI] Install sgl-eval from PyPI through the test extra (#37504)
|
2026-09-02 01:45:25 -07:00 |
|
|
Shuwen WangandClaude Opus 5
|
fe3d4b9bbb
|
fix: restore missing get_component_forced_attn_backend import in minimax_h3 (#37566)
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-02 15:58:31 +08:00 |
|
|
|
4b329482e8
|
[diffusion] feat: support cube sparse attention for minimax h3 (#34893)
Co-authored-by: zhenaozhenfu <zhenaozhenfu@minimaxi.com>
Co-authored-by: Reynor <reynor@minimaxi.com>
|
2026-09-02 15:21:33 +08:00 |
|
|
Mick
|
9175590aa0
|
[diffusion] refactor: admit explicit attention backends by capability (#37441)
|
2026-09-02 15:17:57 +08:00 |
|