Commit Graph

Select branches

Hide Pull Requests

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• c05eb856f7 [CI] Fix nightly test failures (#34523) Liangsheng Yin 2026-08-12 15:40:20 -07:00
• 2b4381956f fix(glm5.2): restrict MoE weights to local PP layers (#33793) binbin Deng 2026-08-13 03:32:08 +08:00
• 4f883636a2 [Perf] FlashInfer MLA: remove blocking D2H in spec-decode plan (#27689) elvischenv 2026-08-13 03:29:03 +08:00
• 3974b00359 Add bit-exact unified radix cache KL test for hybrid SWA + mamba (#34607) Ke Bao 2026-08-13 02:15:36 +08:00
• 773faf992d Reserve multimodal runtime allocations and keep padded inputs aligned (#34141) 5 people 2026-08-12 11:04:07 -07:00
• e6250c7c70 docs: update Qwen3.8 disaggregated serving configs (#34601) YAMY 2026-08-12 11:02:35 -07:00
• 126010c373 [misc] update CI_PERMISSIONS.json (#34610) Ke Bao 2026-08-13 01:58:21 +08:00
• b501311fa1 [Kimi K3] Fuse MLA gate projection into QKV-A GEMM (#33623) 3 people 2026-08-13 00:21:34 +08:00
• 8549cce11b [BugFix] Fix race in c128 prefill plan kernel on ragged extend (#32467) yiheng 2026-08-13 00:15:41 +08:00
• d21eefc94f [Docs] Rename Qwen3.8-Max-DSpark to Qwen3.8-2.4T-A95B-DSpark (#34590) Xinyuan Tong 2026-08-12 23:25:48 +08:00
• ad47dde65c [diffusion] optimize: fuse cosmos qk norm, rope, and kv packing (#34275) Mick 2026-08-12 23:18:16 +08:00
• f28bc5a6de docs(cookbook): add BF16 recipes to Nemotron 3.5 Lightning (#34573) Yichi Zhang 2026-08-12 23:15:34 +08:00
• 8e7c07fae7 [Docs] Add Qwen3.8 cookbook (#34587) zijiexiaandClaude Opus 5 2026-08-12 08:07:46 -07:00
• 197832bcf5 [Kimi-K3][NPU] Support Kimi-K3 on NPU (#33465) McZyWu 2026-08-12 21:20:14 +08:00
• dc5f6c4883 [diffusion] optimize: stream and parallelize bit-exact video output saves (#34564) Mick 2026-08-12 20:03:37 +08:00
• 9701cc138c [diffusion] optimize: optimize bit-exact h3 reference video ingress (#34563) Mick 2026-08-12 20:02:49 +08:00
• 9deb6952af [AMD][DI][CI] Add GLM-5.2 MXFP4 wide-EP16 2P1D nightly recipes (#34476) Zhaoyi Li 2026-08-12 05:53:14 -05:00
• b5d1453ed2 [AMD] GLM 5.2 MXFP4 SGLANG COOKBOOK (#34379) ajith-sirra-amdandgiovanniguastiamd 2026-08-12 16:09:25 +05:30
• 00bdafe944 [AMD][CI] Swap the AMD PR gate to ROCm 7.2 and demote ROCm 7.0 to a daily shadow (#34204) 3 people 2026-08-12 18:31:03 +08:00
• b3bffef70a [diffusion] UX: suppress noisy worker startup warnings (#34512) Mick 2026-08-12 17:53:45 +08:00
• 644d55ebfa [diffusion] feat: support native and peft minimax h3 loras (#34359) Mick 2026-08-12 17:52:40 +08:00
• 00e57d74f0 Bump FlashInfer to 0.6.17 and remove Kimi K3 workarounds (#33997) Mohammad Miadh Angkad 2026-08-12 17:17:26 +08:00
• 2d76d537e5 feat: support deterministic FA4 for GLM-4.7-Flash (#33945) Yuzhen Zhou 2026-08-12 01:57:34 -07:00
• 1f008dc226 [Diffusion][LTX-2] Allocate AdaLN outputs from one contiguous slab (#34508) Xiaoyu Zhang 2026-08-12 16:28:27 +08:00
• daae3acb36 [Diffusion][Z-Image] Tune native QK RMSNorm launch for SM103 (#34507) Xiaoyu Zhang 2026-08-12 16:27:44 +08:00
• 4827061247 [Diffusion] Make weight-only FP8 dequant cache torch.compile-safe (#34506) Xiaoyu Zhang 2026-08-12 16:26:31 +08:00
• 84ce7502cf [Diffusion][MiniMax H3] Extend exact QKNorm+RoPE rounding to SM103 (#34505) Xiaoyu Zhang 2026-08-12 16:25:24 +08:00
• 45f7063335 [Diffusion] Tune QK head LayerNorm for SM103 (#34503) Xiaoyu Zhang 2026-08-12 16:24:18 +08:00
• a2e88279c2 metrics: don't clock-rebase unset time sentinels in ReqTimeStats deserialization (#34335) Sam ShleiferandClaude Fable 5 2026-08-11 23:27:47 -07:00
• 793a11dd27 [AMD] Publish an undated tag for the miles nightly images (#34521) Xinyu Jiang 2026-08-12 02:27:23 -04:00
• 0dab252ffc Fix DFlash sliding attention causality defaults (#34524) Mohammad Miadh Angkad 2026-08-12 14:05:55 +08:00
• 687967c70d Fix tokenizer warning filtering for processors (#34500) Xinyuan Tong 2026-08-12 13:26:44 +08:00
• 9d2d737ebf [Benchmark] Remove 22 unmaintained benchmarks (#34520) Liangsheng Yin 2026-08-11 22:24:43 -07:00
• cac3269305 [XPU] Fix NemotronH (hybrid mamba2) launch on --device xpu (#32227) Juan MunetonandMa Mingfei 2026-08-11 22:23:23 -07:00
• 256981ce16 [CI] Align rerun-test environment with the test stages (#34195) Liangsheng Yin 2026-08-11 21:37:10 -07:00
• 22e4b3a81f [Diffusion] Avoid slow cuBLASLt GELU epilogue on SM120 (#34350) Xiaoyu Zhang 2026-08-12 12:12:00 +08:00
• a9a355774a [diffusion] feat: support dynamically cpu offload components (#34391) Mick 2026-08-12 11:38:27 +08:00
• 5899674504 [Fix] Make DeepSeek-V4 reasoning and tool-call streaming parsing chunk-invariant (#34458) +313 people 2026-08-11 20:03:28 -07:00
• 2be9773a21 [diffusion] doc: update cosmos3 edge and distilled cookbook (#34497) Mick 2026-08-12 10:51:11 +08:00
• 4aff4b1822 [Diffusion] Improve bit-exact fusion fallback diagnostics (#34412) Xiaoyu Zhang 2026-08-12 10:42:48 +08:00
• 3f9d184833 [Diffusion] Tune QK head LayerNorm for SM120 (#34349) Xiaoyu Zhang 2026-08-12 10:41:54 +08:00
• f7b6800b22 [diffusion] model: support cosmos3 edge and distilled (#31590) 3 people 2026-08-12 04:29:40 +02:00
• 81c88da1ab [diffusion] fix: nightly diffusion benchmark passes the retired --warmup flag (#34423) MickandClaude Opus 5 2026-08-12 10:25:14 +08:00
• 30cb848d4b [Bugfix][DSA] Fix num_splits "(b+1)" crash on prefill-CP speculative decode (#34443) NVShreyasandClaude Opus 4.8 2026-08-11 19:13:49 -07:00
• a53d3636ce [diffusion][model] Add native SANA-Video T2V support (#32921) Xiaoyu Zhang 2026-08-12 10:07:24 +08:00
• 93e9db5eb8 [Fix][Qwen]: fused shared-expert detection PP-safe protection (#34447) zijiec 2026-08-11 21:50:05 -04:00
• 37c631ef23 [diffusion] Ideogram-4: fuse Qwen3-style RoPE and SwiGLU silu-mul (denoise -5.1% H100 / -4.7% H200, bit-exact) (#34314) Xiaoyu ZhangandClaude Fable 5 2026-08-12 09:19:35 +08:00
• b1b8ce715b [Diffusion][MiniMax H3] Fix SM120 QKNorm+RoPE rounding (#34347) Xiaoyu Zhang 2026-08-12 09:18:16 +08:00
• a2d723820e [diffusion] fix: fix model-driven dit layerwise offload auto policy (#34401) Mick 2026-08-12 09:08:31 +08:00
• 1ce515a53d Refocus LoRA tests on regression coverage (#34464) Yanbin Jiang 2026-08-11 17:37:52 -07:00
• c7c03ec53b [NVIDIA] Add flashinfer MNNVL backend for allreduce only (#30700) Shu Wang 2026-08-11 18:46:46 -05:00
• 983dfd6a9a [Misc] Sanitize the structure of environ.py (#34472) Baizhou Zhang 2026-08-11 16:31:33 -07:00
• 857910bd35 docs(semianalysis): Update Qwen3.5 B200 NVFP4 MTP config (#34357) Faradawn Yang 2026-08-11 16:12:17 -07:00
• d82a1d4802 [XPU] Pad MoE expert weight row stride to avoid L3 aliasing (#33905) 3 people 2026-08-11 15:54:13 -07:00
• fde9ad2531 [Feature] Add Muse Glimmer model support (#34262) 7 people 2026-08-11 15:41:52 -07:00
• 9c1517df4a Raise PD zmq per-context socket cap via SGLANG_DISAGGREGATION_ZMQ_MAX_SOCKETS (#34450) Hanming Lu 2026-08-11 15:27:32 -07:00
• 7fb6e61b95 Fix CUDA 13.0 VMM handle type compatibility (#34431) Mohammad Miadh Angkad 2026-08-12 06:16:34 +08:00
• 9ced8d0981 Optimize FP32 LM head for bf16/fp16 (#32370) ilyasher-harmonic 2026-08-11 15:14:50 -07:00
• 2c07ca5e8d [Fix] Allow flashinfer_sparse_mla DSA backend for HiSparse on SM120 FP8 KV (#33075) gongwei1027 2026-08-12 06:05:04 +08:00
• 59450c4f18 docs(cookbook): Kimi-K3 — drop --enable-symm-mem from the GB cells (#34444) Douglas YangandClaude Opus 5 2026-08-11 12:23:41 -07:00
• d5d41d07ed [Docs] Add Ling-3.0-tiny INT4 recipes (#34395) Xinyuan Tong 2026-08-12 03:16:03 +08:00
• 93c1bff1d4 Fix default dtype restoration after model loader errors (#34440) 3 people 2026-08-11 11:33:04 -07:00
• c58953d90a O(1) slot allocation in ReqToTokenPool.alloc() (#32208) Dmitrii SergeevandZhiqiang Xie 2026-08-11 20:26:05 +02:00
• aadb9720fe fix: route scheduler aborts to multi-tokenizer workers (#33940) Jeremy Zhang 2026-08-12 02:15:16 +08:00
• f9153df62e Add bit-exact hicache logprob-consistency test (#34356) Ke Bao 2026-08-12 01:54:45 +08:00
• 8f3d3a31f4 [HiCache] Fix Mamba track-boundary bookkeeping under overlap scheduling (#29792) huangtingweiandHanming Lu 2026-08-12 00:36:48 +08:00
• a3bd7d9401 Bump CuTeDSL to 4.6.2 (#34372) Mohammad Miadh Angkad 2026-08-12 00:35:35 +08:00
• 8267d76c2c [VLM] replace deprecated image processor use_fast (#34175) Mick 2026-08-12 00:14:07 +08:00
• b20c375c10 Fix flaky decode cache-hit check in Inkling test (#34405) Ke Bao 2026-08-11 21:47:01 +08:00
• f148eb6e6e Add Hunyuan3 On Ascend Doc (#30223) Jinyan Yi 2026-08-11 21:20:30 +08:00
• a50ab9cec7 [npu] [bugfix] Fix HiCache MHA backup for NPU (#34341) gjsheu 2026-08-11 21:09:45 +08:00
• 3add7e19ff Add NVIDIA Nemotron 3.5 Lightning cookbook (#33481) Faradawn YangandRyan Stewart 2026-08-11 06:00:57 -07:00
• 2d193077f7 [JIT Kernel] Migrate per-token FP8 quantization from AOT to JIT (#34257) Mohammad Miadh Angkad 2026-08-11 20:40:40 +08:00
• a0a76e4485 [DSV4] perf: Enable alt stream during BCG prefill (#29070) 4 people 2026-08-11 14:10:19 +02:00
• 546965fc72 [diffusion] LTX-2: mount the bit-exact fused modulate at the 8 bare adaLN sites (ltx23-one-stage denoise -2.8% H100 / -2.6% H200) (#34315) Xiaoyu ZhangandClaude Fable 5 2026-08-11 18:23:08 +08:00
• dd8c5849af [diffusion] refactor: move dit execution capabilities to runtime models (#34249) Mick 2026-08-11 18:20:43 +08:00
• 071f0f1e9d [diffusion] ERNIE-Image: fuse rotate-half RoPE + GELU-mul and hoist rope cos/sin (denoise -16.2% H100 / -12.7% H200, bit-exact) (#34306) Xiaoyu ZhangandClaude Fable 5 2026-08-11 18:18:37 +08:00
• ba3dc16401 [diffusion] weight-only FP8: dequantize linear weights once at first use (Ideogram-4 denoise -18.8% H200 / -7.8% H100, bit-exact) (#34305) Xiaoyu ZhangandClaude Fable 5 2026-08-11 18:15:26 +08:00
• 5469faec45 HiSparse: shared-index (IndexShare) plan-then-IO swap-in prefetch (#34329) Zhiqiang XieandTingwei Huang 2026-08-11 01:58:28 -07:00
• 396722e490 [NPU] Disable failed test cases (#34377) 3 people 2026-08-11 16:38:15 +08:00
• 6f3fe13a9c [AMD] Install AITER's pinned Triton wheel in the ROCm 7.2 image (#34364) Bingxu ChenandYC Yen-Ching Tseng 2026-08-11 16:36:45 +08:00
• d8a61c26a6 [CI] Add a scheduled workflow to close stale PRs (#34380) Liangsheng Yin 2026-08-11 00:54:02 -07:00
• e74ea5b1d7 [ROCm/gfx95] Fix fp8 per-channel attention for Kimi-K2.7-code-mxfp4 o… (#31105) 3 people 2026-08-11 15:51:57 +08:00
• 1c06c160f9 [Docs] Add Ling-3.0-flash INT4 and MXFP4 recipes (#34363) Xinyuan Tong 2026-08-11 15:35:27 +08:00
• 9d4be40124 Fix DSpark + DeepSeek V4 prefill CP compatibility (#33865) ybyang 2026-08-11 14:26:29 +08:00
• afa2d5570b [Diffusion] Add online FP8 support for Krea-2 (#34136) Tanmay patil 2026-08-11 11:42:05 +05:30
• 667e18d99d [PD] Support pipeline-parallel prefill with Mooncake staging buffer (#33807) YAMYandShangming Cai 2026-08-10 22:53:17 -07:00
• 13aeb91b6e [Fix] Update multimodal CUDA VMM helper import (#34358) Xinyuan Tong 2026-08-11 13:41:21 +08:00
• b3c02cbce7 [perf] Collapse the DP attention scheduler sync to a single D2H copy (#34338) Liangsheng YinandAlex Nails 2026-08-10 21:52:29 -07:00
• 585c3c6816 [Refactor] Split the FlashInfer autotune dummy-run flag from the LM-head policy (#34336) Liangsheng Yin 2026-08-10 21:20:05 -07:00
• a58fa0388e [Fix] Correct W4AFP8 DeepEP scaling and mode-specific dtypes (#33669) EchO 2026-08-11 11:21:42 +08:00
• 704808ed27 [MiniMax-H3] SubBlock: training-free block-sparse attention for the DiT (#34148) forrestlandClaude Opus 5 2026-08-11 11:11:11 +08:00
• dd20826e0a [AMD] Preserve the AITER expert mask across torch_memory_saver pause/resume (#34220) Zhiyao Jiang 2026-08-10 22:59:15 -04:00
• b498f46271 [CI] Add MegaMoE runner compatibility alias (#34326) Baizhou Zhang 2026-08-10 19:47:38 -07:00
• 8d050dd880 [AMD][CI] Run MI300 8-GPU stage-C shards two at a time (#34324) Michael 2026-08-10 18:49:08 -07:00
• 5af5183351 test: isolate metal profiler tests from ambient SGLANG_USE_MLX (#34300) Jae B. 2026-08-10 21:44:16 -04:00
• aeab1de1de [diffusion] optimization: support cuda graph for Pi-0.5 prefix encoding (#34256) Mick 2026-08-11 09:14:53 +08:00
• df986c4d5e Consolidate CUDA VMM allocation helpers (#34199) cctry 2026-08-10 18:11:11 -07:00
• ba5183fe10 [diffusion] UX: fix CI server warmup progress logging (#34301) Mick 2026-08-11 09:09:32 +08:00
• 418975ba64 [EPD] feat: pipeline owner-only multimodal preprocessing (#34206) Mick 2026-08-11 09:08:05 +08:00