Commit Graph
10627 Commits
Author SHA1 Message Date
heziiop 235a488c87 [NPU] ascend fuseep use moe ep group (#32040) 2026-07-23 14:07:14 +08:00
silencejade 09071be105 [NPU] [FIX] Fix performance degradation of Qwen3.5-397B-A17B (#32130) 2026-07-23 14:05:10 +08:00
Xiaoyu ZhangandClaude Opus 4.8 11b0e5c5ad [Kernel] Classification cleanup: unify _jit_ naming, drop empty/model groups, add elementwise (RFC #29630) (#32148)
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-23 13:47:02 +08:00
Liangsheng Yin 1b63155efe [Perf] Skip blocks past per-request live length in full-width Triton kernels (#32109) 2026-07-22 22:00:25 -07:00
Xiaoyu ZhangandClaude Opus 4.8 2d1a7be8c4 [Kernel] Reclassify kernel tests by ops group + move helpers out of the package (RFC #29630) (#32128)
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-23 12:18:27 +08:00
jiayisunxandMa Mingfei a2935ce329 [XPU][GDN] add XPU path for causal_conv1d_fn and causal_conv1d_update (#31250)
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
2026-07-23 10:50:20 +08:00
Артем Савкин 108182cb81 [Bugfix] [NPU] Fix w4a8 MoE performance degradation (#32113) 2026-07-23 10:31:32 +08:00
zhaozx-cn 926530b0cf [NPU]remove duplicate code (#31863) 2026-07-23 10:24:31 +08:00
huangtingwei 86e1bb584d [HiCache] Add model-aware key isolation to Mooncake Store (#31920) 2026-07-23 10:20:13 +08:00
YanbingJiang 9a7ac3ecef Fix unnecessary gather/scatter on CPU for non-contiguous Mamba statepool (#31754) 2026-07-23 09:08:45 +08:00
Xiaoyu ZhangandClaude Opus 4.8 99f636a86f [Kernel] RFC #29630 finale: retire sglang.jit_kernel into sglang.kernels (#32072)
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-23 08:35:09 +08:00
danielafrimiandDaniel Afrimi 24da0e51b6 Warn on small Mamba chunked prefill size (#30938)
Co-authored-by: Daniel Afrimi <dafrimi@aws-dfw-cs-001-login-01.cm.cluster>
2026-07-22 15:09:53 -07:00
Mohammad Miadh AngkadandBrayden Zhong 0c29c8fece Bump FlashInfer to 0.6.15.post1 (#31927)
Co-authored-by: Brayden Zhong <b8zhong@uwaterloo.ca>
2026-07-22 14:21:59 -07:00
c20c48b8fd Add 'anyOf' schema support for qwen3_coder tool call parser (#30832)
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
Co-authored-by: Liangsheng Yin <hnyls2002@gmail.com>
Co-authored-by: Alex Nails <alex.nails@radixark.ai>
2026-07-22 14:14:47 -07:00
Brayden Zhong e7511141ea Support CuteDSL GEMM BF16 on SM100 on by default when allowed by heuristic (#30567) 2026-07-22 14:13:12 -07:00
Liangsheng Yin 98cc8d91cf [Fix] Evict only the KV shortfall in evict_from_tree_cache (#32016) 2026-07-22 12:41:06 -07:00
Liangsheng Yin 5c6a29b3c3 [Fix] Unify pinned host pool release on graceful shutdown (#32029) 2026-07-22 12:29:34 -07:00
Cheng Wan f5dcbe8f14 Revert RuntimeContext config-namespace reads/roles (#31813–#31817) (#32100) 2026-07-22 11:52:41 -07:00
Ke Bao 40ac197fb2 Fix get_server_args import lint error (#32096) 2026-07-23 00:09:25 +08:00
Raghavendra Vedula 4eaa5ca651 Treat partial_json_parser AssertionError as incomplete JSON (#31975) 2026-07-22 23:25:25 +08:00
Raghavendra Vedula a9497e8d73 Guard min_new_tokens penalizer against None eos_token_id (#31973) 2026-07-22 23:24:25 +08:00
Chengze Fan 40b2119b23 [AMD] Cache AITER expert mask across decode (#31889) 2026-07-22 07:50:23 -07:00
e8e765b9d6 [AMD] Add fused all-reduce RMSNorm per-group quant for Qwen3.5 FP8 (#24651)
Co-authored-by: jacky.cheng <yichiche@amd.com>
Co-authored-by: yctseng0211 <yctseng@amd.com>
Co-authored-by: HAI <hixiao@gmail.com>
2026-07-22 07:33:03 -07:00
Mohammad Miadh Angkad b855efd9e6 Fix Inkling kernel imports after migration (#32076) 2026-07-22 22:09:01 +08:00
0a6d1930c3 [Attention Backend] Add HPC-Ops attention backend (#30540)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Halcyon <56064364+VAthree@users.noreply.github.com>
2026-07-22 22:06:22 +08:00
Fan Kunand范坤 004df6b520 [FIX] Prevent Lightning Attention extra-buffer mamba state corruption (#29973)
Co-authored-by: 范坤 <fankun@U-Q0542J2D-0225.local>
2026-07-22 22:01:29 +08:00
ishandhananiandConnor Carpenter 21065bc862 feat: add native gRPC sidecar module launcher (#31076)
Signed-off-by: Ishan Dhanani <ishandhanani@gmail.com>
Signed-off-by: Connor Carpenter <connorc@nvidia.com>
Co-authored-by: Connor Carpenter <connorc@nvidia.com>
2026-07-22 06:39:08 -07:00
Xiaoyu ZhangandClaude Opus 4.8 74338e94f1 [Kernel] Phase 4 batch-3: migrate tangled JIT subsystems + new groups into kernels.ops (RFC #29630) (#32045)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-22 21:15:03 +08:00
b13abfdedf Fix LongCat n-gram token-table crashes on padded batches (#31312)
Co-authored-by: whn09 <whn09@users.noreply.github.com>
Co-authored-by: Xiaoyu Zhang <1182563586@qq.com>
2026-07-22 20:24:21 +08:00
Xiaoyu ZhangandClaude Opus 4.8 977ea336cd [Kernel] Phase 4 batch-2: migrate JIT operator groups into kernels.ops (no shims) (RFC #29630) (#32015)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-22 17:49:51 +08:00
McZyWu 9456cef279 [NPU]fix rl update weights 'Parameter' object has no attribute 'weight_LOADER' (#31796) 2026-07-22 16:34:59 +08:00
Cheng Wan 745b2ca45c config: read parallel config leaves via get_parallel() (#31816) 2026-07-22 01:18:50 -07:00
Cheng Wan 602b546a4e config: load-time declarations write the config bags (#31815) 2026-07-22 01:18:31 -07:00
Cheng Wan 11a4c2d057 config: read resolved config via namespace accessors (#31814) 2026-07-22 01:18:05 -07:00
Cheng Wan 1d0a6ee178 runtime_context: record the publishing process role (#31813) 2026-07-22 01:17:42 -07:00
Cheng Wan e1479cc966 config: route runtime config adjustments through the namespace bags (#31812) 2026-07-22 01:17:22 -07:00
Cheng Wan 97e2c0c4ee config: make ServerArgs read-only with a single audited mutation entry (#31811) 2026-07-22 01:16:55 -07:00
Cheng Wan 09688d58bc runtime_context: add resolved-config namespace bags and accessors (#31810) 2026-07-22 01:16:24 -07:00
Cheng Wan 1a19f2b50f config: annotate ServerArgs fields with their runtime-config namespace (#31809) 2026-07-22 01:15:11 -07:00
Yuan Luoandluoyuan.luo ae2bc3321e [KDA] Fix mixed exponent bases in Triton chunk prefill (#31904)
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
2026-07-22 16:03:29 +08:00
Xun Sun ddaf430e6c [Elastic EP] Centralize Mooncake PG configuration (#31708) 2026-07-22 14:52:37 +08:00
Vladimir Serov 0a3cd26b28 LoRA: Ascend: Update ascend LoRA backend to support new kernels (#15912) 2026-07-22 14:11:40 +08:00
Xiaoyu ZhangandClaude Opus 4.8 246b3c3eaf [Kernel] Phase 3+4: move JIT infra + operator groups into sglang.kernels (RFC #29630) (#31666)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-22 11:28:27 +08:00
xdtbynd d708969f68 [bugfix][NPU] Fix startup bug in olmoe 1b 7b (#31782) 2026-07-22 10:51:15 +08:00
Jinyan Yi 394b0dd13a [NPU] Update non-vit vision part for cumulative seqlen (#31867) 2026-07-22 10:41:48 +08:00
93cb9a5482 [CI] Point diffusion CI writes to sgl-project/ci-data-diffusion (#31983)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Alison Shao <a.shao@wustl.edu>
2026-07-21 19:15:15 -07:00
3217b7e3ce fix(hicache): support staged write-back for asymmetric MHA (#30981)
Co-authored-by: 晟海 <huangtingwei.htw@antgroup.com>
Co-authored-by: Zhangheng <hzh0425@apache.org>
2026-07-22 09:58:49 +08:00
Jialin OuyangandAlex Nails 4597dd4d88 Fix reward/classification models broken by load_weights v2 dispatch (#28671) (#31988)
Co-authored-by: Alex Nails <alex.nails@radixark.ai>
2026-07-21 18:47:49 -07:00
ashwini rathi 14c0a31829 [Bugfix] Fix Cohere2MoeConfig import crash from huggingface_hub @strict (#31769) 2026-07-22 08:56:30 +08:00
DarkSharpnessandClaude Fable 5 8bb0d8d005 [JIT] Trait-driven per_token_group_quant: unify the quant kernel family (flat + masked) (#30924)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-07-22 08:46:52 +08:00