 Augusto Yao
|
ea8f4e9f3f
|
[feature] implement dcp for deepseek_v2 (#14194)
|
2026-06-25 15:15:04 -07:00 |
|
|
jiayisunx
|
d717e926c1
|
fix(runner): prevent eager token buffer under-allocation (#28894)
|
2026-06-25 14:37:19 +08:00 |
|
|
Liangsheng Yin
|
c7734e6871
|
[Spec] Dissolve EagleDraftInputV2Mixin so spec-info dataclasses hold data only (#29220)
|
2026-06-24 18:08:30 -07:00 |
|
|
Jiajun LiandClaude Opus 4.8
|
e4bf0043fe
|
[fix] Skip routed expert capture for draft model under spec v2 (#26980)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-06-24 16:56:07 -07:00 |
|
|
Lianmin ZhengandYinghai Lu
|
7e63feee6f
|
Add scheduler metrics extension hooks (#29207)
Co-authored-by: Yinghai Lu <yinghai@meta.com>
|
2026-06-24 15:50:05 -07:00 |
|
|
Lianmin ZhengandCheng Wan
|
d6aacd2801
|
Handle input-embed-only batches in eager runner (#29121)
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
|
2026-06-24 10:33:52 -07:00 |
|
|
cctryandcctry
|
76db6c9d9e
|
[Refactor] Share CUDA graph memory pool across prefill and decode (#28973)
Co-authored-by: cctry <cctry@fb.com>
|
2026-06-24 10:18:20 -07:00 |
|
|
|
b2c8f7a22e
|
[AMD] Support triton backend decode context parallel for Qwen3.5 (#25090)
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
Co-authored-by: Kangyan-Zhou <zky314343421@gmail.com>
Co-authored-by: Khoa Pham <khoa.pham@radixark.ai>
Co-authored-by: Hubert Lu <55214931+hubertlu-tw@users.noreply.github.com>
Co-authored-by: zhengyao <zayao@amd.com>
|
2026-06-23 17:35:56 -07:00 |
|
|
Lianmin Zheng
|
ecab3f322e
|
Revert "Improve MFU metrics for prefill and verify timing" (#29079)
|
2026-06-23 15:46:22 -07:00 |
|
|
Trevor Morris
|
f74a1722e6
|
[NVIDIA] Support TF32 matmul to improve MiniMax gate gemm performance (#22744)
|
2026-06-23 14:54:54 -07:00 |
|
|
YAMY
|
93015a9e6b
|
fix(runner): autotune flashinfer MoE on a decode-shaped buffer (#29069)
|
2026-06-23 13:31:47 -07:00 |
|
|
Lianmin ZhengandPranjal Shankhdhar
|
b60185c41c
|
Improve MFU metrics for prefill and verify timing (#29000)
Co-authored-by: Pranjal Shankhdhar <pranjal.ssh@gmail.com>
|
2026-06-23 12:26:56 -07:00 |
|
|
Cheng Wan
|
c4376aaa88
|
[Refactor] Remove dead out_cache_loc_swa buffers (#28968)
|
2026-06-23 00:03:30 -07:00 |
|
|
Lianmin Zheng
|
ad9723af03
|
Clean up CUDA graph capture logs (#28937)
|
2026-06-22 06:15:26 -07:00 |
|
|
Yuwei An
|
2ce32366a0
|
[Fix][BCG][Spec] Restore EAGLE prefill plumbing dropped by #23906 (#28870)
|
2026-06-22 01:54:54 -07:00 |
|
|
Liangsheng Yinandthanhhao98
|
106d2930a6
|
[core] Gate the overlap WAR barrier on forward reads to recover decode throughput (#28363)
Co-authored-by: thanhhao98 <31717833+thanhhao98@users.noreply.github.com>
|
2026-06-22 00:40:56 -07:00 |
|
|
cctryandcctry
|
0c065671c9
|
[Spec] Redo: split init_backends; account draft weights in --mem-fraction-static (#28855)
Co-authored-by: cctry <cctry@fb.com>
|
2026-06-21 20:45:16 -07:00 |
|
|
Mohammad Miadh Angkad
|
643ee748c6
|
[PP] Pass DSA topk through PP warmup proxy buffers (#28785)
|
2026-06-21 23:55:39 +08:00 |
|
|
Lianmin Zheng
|
7942d546d1
|
Revert "[Spec] Split init_backends; account draft weights in --mem-fraction-static" (#28841)
|
2026-06-21 07:52:35 -07:00 |
|
|
cctry
|
9691a29fe0
|
[Spec] Split init_backends; account draft weights in --mem-fraction-static (#28683)
|
2026-06-21 01:22:26 -07:00 |
|
|
cctry
|
6d4ca9bc54
|
Cap SWA pool sizing with chunk cache (#28755)
|
2026-06-21 01:06:59 -07:00 |
|
|
|
f42ec350b4
|
[mtp] add rejection sampling for speculative decoding (#26312)
Co-authored-by: lyc508653 <lyc508653@alibaba-inc.com>
Co-authored-by: Qiaolin-Yu <liin1211@outlook.com>
Co-authored-by: Huiqiang Jiang <30883354+iofu728@users.noreply.github.com>
Co-authored-by: Yi Zhang <25844240+yizhang2077@users.noreply.github.com>
Co-authored-by: Yizhong Cao <114661107+cao1zhg@users.noreply.github.com>
|
2026-06-20 15:10:42 -07:00 |
|
|
Lianmin Zheng
|
fe428dd845
|
Clean up startup log noise (#28807)
|
2026-06-20 15:02:52 -07:00 |
|
|
Cheng Wan
|
6fdcb9934c
|
fix(runner): size eager static buffers for prefill budget and MLP-sync autotune (#28677)
|
2026-06-19 13:23:26 -07:00 |
|
|
Cheng Wan
|
2aa7b58aa7
|
refactor(runner): reuse a prepared static buffer for every dummy run (#28740)
|
2026-06-19 13:16:07 -07:00 |
|
|
Cheng WanandClaude Opus 4.8
|
856b0dc74b
|
refactor(runner): move kernel warmup into the shared runner lifecycle (warmup()) (#28739)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-06-19 13:13:13 -07:00 |
|
|
Cheng WanandClaude Opus 4.8
|
d705a91de1
|
refactor(runner): add EagerRunner, own the eager path, polymorphic dispatch (#28386)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-06-19 13:05:43 -07:00 |
|
|
|
c436a8161a
|
[AMD] Enable HiSparse on ROCm (#26639)
Co-authored-by: clintg6 <7388379+clintg6@users.noreply.github.com>
Co-authored-by: HAI <hixiao@gmail.com>
|
2026-06-19 11:59:45 -07:00 |
|
|
Mohammad Miadh Angkad
|
88c261c3f3
|
Fix IndexCache PP topk handoff (#28532)
|
2026-06-19 23:39:09 +08:00 |
|
|
Oguz Ulgen
|
3af991fb3e
|
[AMD] Make breakable CUDA graph run on ROCm/HIP (#28173)
|
2026-06-19 07:16:00 -07:00 |
|
|
Cheng Wan
|
a6db86d535
|
refactor(runner): split BaseRunner (shared) from BaseCudaGraphRunner (#28385)
|
2026-06-19 02:04:15 -07:00 |
|
|
Cheng Wan
|
1c6331cbd6
|
refactor(runner): rename runner replay/load/can_run for the shared surface (#28384)
|
2026-06-19 01:45:04 -07:00 |
|
|
Lucia Fang
|
1c8551169d
|
Add opt-in CUDA-graph capture-trace export (#28551)
|
2026-06-18 18:51:14 -07:00 |
|
|
Baizhou Zhang
|
e3026ef016
|
[3/N][CP] Implement zigzag CP strategy (#28421)
|
2026-06-18 15:10:30 -07:00 |
|
|
Cheng Wan
|
f83e4d5968
|
refactor(runner): unify eager-forward DP/MLP-sync padding into one helper (#28383)
|
2026-06-18 13:41:30 -07:00 |
|
|
Qiaolin Yu
|
2411737244
|
[spec decoding] fully overlap spec decoding for hybrid linear attention backend (#28579)
|
2026-06-18 13:31:12 -07:00 |
|
|
Cheng Wan
|
67db2ac3e7
|
refactor(runner): unify pp_proxy_tensors forward kwarg into one helper (#28382)
|
2026-06-18 02:23:47 -07:00 |
|
|
cctry
|
fa71064147
|
fix: speculative draft worker clobbering target attention backend (#28559)
|
2026-06-18 01:21:14 -07:00 |
|
|
|
9b10821c8e
|
[NPU] Add Ascend NPU support for DeepSeek-V4 (#25144)
Co-authored-by: khalil2ji3mp6 <khalilzhk@gmail.com>
Co-authored-by: randgun <kelonlu@163.com>
Co-authored-by: t00937989 <tanlei33@huawei.com>
|
2026-06-18 15:30:25 +08:00 |
|
|
Yuan Luoandluoyuan.luo
|
3b9db3a1f0
|
[Mamba][GDN] Deduplicate spec conv-window intermediate cache via sliding window layout (#28302)
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
|
2026-06-17 20:42:31 -07:00 |
|
|
Cheng Wan
|
53318911ca
|
Add get_parallel(): a structured accessor for parallel-topology state (#28567)
|
2026-06-17 20:23:43 -07:00 |
|
|
Qiaolin Yu
|
753aa89a83
|
[spec decoding] fix mrope_positions in draft extend (#28464)
|
2026-06-17 13:25:31 -07:00 |
|
|
Ryan Zzzandzhujunyu
|
8fd1694dd2
|
Deepseek v4: support mixed dtype compression states (#27277)
Co-authored-by: zhujunyu <zhujunyu.666@bytedance.com>
|
2026-06-17 00:56:41 -07:00 |
|
|
Brayden ZhongandBrayden Zhong
|
b8a73bfba0
|
Call Flashinfer mm_fp8 for per-tensor FP8 GEMMs on SM100 (#28333)
Co-authored-by: Brayden Zhong <brayden@radixark.ai>
|
2026-06-16 20:50:03 -07:00 |
|
|
Jason Mancuso
|
c0a6c3ce66
|
Fix circular import when sglang.srt.model_executor.runner_backend is imported first (#28002)
|
2026-06-16 12:04:34 -07:00 |
|
|
feliang-gitandxutizhou
|
92b42c8d8a
|
LPLB: linear-programming load balancer for MoE expert parallelism (#24515)
Co-authored-by: xutizhou <xutingz@nvidia.com>
|
2026-06-16 10:19:42 -07:00 |
|
|
Shu Wang
|
32685874f3
|
Reenable MNNVL backend for FlashInfer allreduce fusion (#23402)
|
2026-06-15 20:19:15 -07:00 |
|
|
zhangxiaolei
|
063ab89ac1
|
DeepSeek-V4 Online Compress support MTP (#26471)
|
2026-06-15 19:56:07 -07:00 |
|
|
Lianmin Zheng
|
f18d38d040
|
Revert "[AMD][Quantization] Online MXFP4 quantization 2/N - FP8 to MXFP4 requantization on AMD GPUs" (#28213)
|
2026-06-14 13:34:06 -07:00 |
|
|
Liangsheng Yin
|
bb48405c31
|
Unify NVTX annotation helpers and split the enable gate per subsystem (#28165)
|
2026-06-14 00:04:59 -07:00 |
|