Commit Graph

Select branches

Hide Pull Requests

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 1e80d938ff [HiCache]fix draft host pool allocator type (#29823) huangtingwei 2026-07-01 19:31:16 +08:00
• 79f334b1aa [diffusion] CI: add 5090 job (#29791) Mick 2026-07-01 19:09:43 +08:00
• 7d9de81cda feat(hicache): Use NIXL path-mode (#27060) Lukas Humbel 2026-07-01 11:59:45 +02:00
• 07ca24372b Add fused EH norm for DeepSeek NextN (#29667) Mohammad Miadh Angkad 2026-07-01 17:46:21 +08:00
• 8ee200972e [fix] Add support for flashinfer MOE A2A to Qwen3 BF16 model path (#26255) Daniel Stokes 2026-07-01 20:59:54 +12:00
• 677a11bfa9 [Doc] Tiny update dsv4 doc (#29827) Baizhou Zhang 2026-07-01 01:47:35 -07:00
• 5134dcdcab [Intel XPU] Initially add nightly GSM8K accuracy tests for Llama-3.1-8B (TP=2) and Qwen3-32B (TP=4) (#28908) ashwini rathiandSinghal, Shubham 2026-07-01 13:57:24 +05:30
• 548f505cc5 [AMD] Cover DeepSeek-R1 MXFP4 TP4 MTP nightly CI (#29290) Bingxu Chen 2026-07-01 16:21:17 +08:00
• 69ce72020a [AMD] Register ltx2_ada_values JIT kernel test for AMD nightly CI (#29693) Michael 2026-07-01 01:13:55 -07:00
• 81a472efcc [AMD] Update ROCm AITER pin to 9127c94 (#29816) Bingxu Chen 2026-07-01 16:09:11 +08:00
• 5e1ccd9320 [HiCache] Optimize HiCache hash generation with bulk token byte conversion (#28287) huangtingwei 2026-07-01 15:44:23 +08:00
• df0dfbaa45 [Kernel] Strengthen kernel shape coverage (#29636) 4 people 2026-07-01 15:44:20 +08:00
• 8205aa3603 chore: clean diffusion dead code (#29789) Xiaoyu Zhang 2026-07-01 15:42:16 +08:00
• 308d89e042 [AMD] Split qwen3.5 triton DCP test into its own nightly job (#29409) YC Yen-Ching Tseng 2026-07-01 15:39:39 +08:00
• 00e12cebb4 [Fix]: Defer DSA MLA CP KV gather for fp8 trtllm prefill in PD mode (#29161) Shijin Zhang 2026-07-01 15:20:12 +08:00
• 721350656d [NPU] Fix glm 4.6v (#29381) zhaozx-cn 2026-07-01 15:13:48 +08:00
• 40594bd381 [passthrough] engine: zstd request-body decompression + header overrides (#29684) Jialin Ouyang 2026-06-30 23:48:56 -07:00
• 47ae1241d3 [CI] Relax ModelOpt NVFP4 diffusion consistency thresholds (#29767) Xiaoyu Zhang 2026-07-01 14:45:08 +08:00
• fcb9f229b3 [KDA-Pilot] Add LTX2 QKNorm split-RoPE CUDA fast path (#29708) Xiaoyu Zhang 2026-07-01 14:42:07 +08:00
• 41779d56fd fix: make write_token dynamic (#29271) Chang Min Bark 2026-07-01 01:55:57 -04:00
• 546d6e23ad fix(mlx): default prefill_aware_swa=False on MlxModelRunnerStub (#29681) Tai An 2026-06-30 22:55:43 -07:00
• 3cdc2415b1 [MLX] Fix step-bounded profiling for bench tools on Apple Silicon (#29217) Lijuan Tang 2026-06-30 22:55:29 -07:00
• a7390b17f8 [NPU]Modify --lora-backend & --moe-runner-backend description. (#29793) qinsir5522 2026-07-01 11:15:46 +08:00
• 5de66d7fbc [EPD][BugFix] Fix encoder health check with global cache TP (#29296) 4 people 2026-07-01 11:04:03 +08:00
• 56f22cd520 Optimize C128 state pool allocation using request state pool (#28612) zhangxiaolei 2026-07-01 10:11:29 +08:00
• 5b76f55d90 [Fix]: Inline H2D during CUDA graph capture to avoid stream isolation in Offloader (#29166) Shijin Zhang 2026-07-01 08:49:58 +08:00
• a01afdd526 Support real draft tokens to simulated acceptance (#29645) weireweireandweireweire 2026-07-01 08:22:05 +08:00
• 53c61bd5e6 [bug2] skip swa recovery on locked full kv (#29352) 3 people 2026-06-30 16:07:06 -07:00
• 99b8f36cb1 Skip custom all-reduce v2 CUDA graph capture with torch memory saver. (#27948) Yuzhen Zhou 2026-06-30 14:53:29 -07:00
• bb98629157 docs(cookbook): add AMD MI300X/MI325X/MI355X support for GLM-5.2 (#28471) 4 people 2026-06-30 14:08:51 -07:00
• 081a01c37d docs: sync LMSYS SGLang blog cards (#29307) sglang-botandsglang-bot 2026-06-30 13:48:10 -07:00
• e721259079 typo-fix: correct mooncake package name to "mooncake-transfer-engine" (#25153) TzZtzt 2026-07-01 02:54:36 +08:00
• 97fc4dfd73 [Doc]Checking and modifying Markdown formatting issues and link validity (#28586) 3 people 2026-07-01 01:27:38 +08:00
• 1c2523b719 [bug1] keep full kv when swa skips leaf data (#29351) Feng Yaoandispobock 2026-06-30 09:51:18 -07:00
• c70cc96905 [AMD] Stop rocm720 pr auto-runs (#29768) YC Yen-Ching Tsengandbingxche 2026-06-30 23:58:25 +08:00
• f2756f53f3 [AMD] Update AMD local registry address (#29765) YC Yen-Ching Tsengandbingxche 2026-06-30 23:40:00 +08:00
• cf36dca6d4 [JIT Kernel] Triton moe fused gate (#25835) 5 people 2026-06-30 23:26:18 +08:00
• a531d81c19 [diffusion][cache-dit] support Krea-2 + run-driven has_separate_cfg (#29688) Yihao WangandClaude Opus 4.8 2026-06-30 08:22:05 -07:00
• c6a7c98ae4 [NPU]GLM-4.7-Flash optimize with fused kernels (#29509) Estrella-xx 2026-06-30 19:22:19 +08:00
• 2f730e299f Disable dsr1 prefill cudagraphs by default (#28053) nvjullin 2026-06-30 16:34:44 +08:00
• a5e6dd3767 [AMD] Implement QuarkW4A8MXFp4MoE to support amd/gpt-oss-120b-w-mxfp4-a-fp8 (#27204) Stanley Winata 2026-06-30 01:24:22 -07:00
• 89620b9169 [NPU] Support DeepSeek V4 Flash MTP on Ascend (#28980) 5 people 2026-06-30 16:22:13 +08:00
• 4b4b4af583 Budget EAGLE/STANDALONE draft KV pool in SWA pool configurators (#29622) cctryandcctry 2026-06-30 00:58:33 -07:00
• ff51acd67b [NPU] Qwen3-VL-8B use split_qkv_rmsnorm_rope for extend (#29627) silencejade 2026-06-30 15:55:50 +08:00
• bae78a44da [Deps] Bump transformers to 5.12.1 (#29393) Mohammad Miadh Angkad 2026-06-30 15:54:01 +08:00
• f920a37da4 [AMD] Copy decode result on forward_stream instead of copy_stream (#29642) Thomas Wang 2026-06-30 14:53:48 +08:00
• b6fceaa789 [DSA] Use cos_sin_cache for DSA indexer fusion (#29613) Mohammad Miadh Angkad 2026-06-30 14:49:00 +08:00
• dbb51c46ac [CI] Migrate JIT tests missed by #29066 to runner_config registration (#29715) 3 people 2026-06-29 23:21:14 -07:00
• dad0120c57 [HiSparse]: Skip flaky hisparse-nixl ci (#29710) Zhangheng 2026-06-30 14:04:34 +08:00
• cc238270b1 Update GLM tests to 5.2 and delete redundant tests (#29686) Baizhou Zhang 2026-06-29 23:04:26 -07:00
• b6907d9664 docker: install dynamo nightly in the dev image for rapid iteration/testing (#29634) 3 people 2026-06-30 07:58:19 +02:00
• 54e71506b3 [AMD][DSV4] Remove per-batch D2H syncs in MTP to avoid bubbles between 2 batches (#29420) 3 people 2026-06-30 13:39:59 +08:00
• f41d455f37 [Model] Add HrmTextForCausalLM (Hierarchical Reasoning Model - Text) (#27887) yifei wuandXinyuan Tong 2026-06-30 13:32:32 +08:00
• 7d9116d9d7 Fix test_type_based_dispatcher after TokenizedGenerateReqInput field changes (#29376) YC Yen-Ching Tseng 2026-06-30 13:22:22 +08:00
• 7da4e30d9b [AMD] Register fused_metadata_copy JIT kernel test for AMD nightly CI (#29671) Michael 2026-06-29 21:26:11 -07:00
• 6bf15aa2b9 [AMD]: Enable NIXL PD disaggregation for ROCm(1/n) (#28348) Zhaoyi Liandbingxche 2026-06-29 23:23:09 -05:00
• 2068ae7eec [diffusion] chore: document VAE decode parallel group axes (#29364) Mick 2026-06-30 12:14:55 +08:00
• 3add35e26d [Diffusion] Reuse shared AlignedVector and tidy jit_kernel/diffusion (#29664) Xiaoyu ZhangandClaude Opus 4.8 2026-06-30 11:38:22 +08:00
• 8dbf04fc56 Add new Intel SGLang members into CI_permission list (#29700) Zaili Wang 2026-06-30 11:06:28 +08:00
• 3a72d02415 [DSA] Optimize DSA CUDA graph replay metadata generation (#29499) Mohammad Miadh Angkad 2026-06-30 10:53:23 +08:00
• 25b6051c70 [diffusion] warmup: default to model sampling resolution (declare Z-Image default) (#29519) Mick 2026-06-30 10:32:11 +08:00
• 3e16be2122 [CI] Migrate JIT tests to runner config registration (#29066) 3 people 2026-06-29 19:12:14 -07:00
• bc8b3ab1f5 [Kernel] Add SM90 Q8KV8 FP8 Sparse MLA Prefill JIT Kernel with Tests and Benchmark (#25751) Ho-Ren (Jack) Chuang 2026-06-29 18:00:11 -07:00
• a6bc432fd8 [Perf][Kernel] Fuse SiLU+Mul into NVFP4 Expert Quantization for CUTLASS MoE (#18612) Ho-Ren (Jack) Chuang 2026-06-29 16:51:01 -07:00
• f0bf96390b [SM120] Add FlashInfer sparse MLA decode for DSv4-Flash (#27455) eeechoandAliceChenyy 2026-06-30 07:27:16 +08:00
• b8c25bfaa7 [NVIDIA] Support flashinfer a2a with flashinfer_trtllm_routed moe (#22394) Trevor Morris 2026-06-29 16:23:58 -07:00
• 6bdecb8206 [DeepSeek V3] Reland: run routed experts on main stream in dual-stream MoE (#29463) Khoa PhamandClaude Opus 4.7 2026-06-29 15:05:01 -07:00
• 45314a9fcb [spec] Fix index_share_for_mtp_iteration being a no-op in EAGLE MTP draft (#29654) Xinyuan Tong 2026-06-29 22:52:23 +01:00
• fc96edd297 feat(mem_cache): page-major (layer-major within a page) KV/state layout (#29533) Cheng Wanandlch1475369 2026-06-29 14:49:54 -07:00
• 6c018eb4d1 Fix bounded checkpoint prefetching and buffered drop-cache handling (#29156) Mohammad Miadh Angkad 2026-06-30 05:49:16 +08:00
• 5556631789 [Speculative Decoding] Validate vocabulary compatibility in STANDALONE mode (#23838) 3 people 2026-06-30 00:45:12 +03:00
• 5106b42cbd [cookbook] GLM-5.2 NVFP4 B300: TP8 recipe + 3 strategies (#29557) zijiexiaandClaude Opus 4.8 2026-06-29 14:42:19 -07:00
• 98d0e702c3 [GLM-5] Tune the threshold of router GEMM (#29470) Brayden Zhong 2026-06-29 14:19:24 -07:00
• 11b7ed7c9e [Docs] Add --prerelease=allow so uv installs the latest sglang (#29676) zijiexiaandClaude Opus 4.8 2026-06-29 14:13:34 -07:00
• 74a197af9d docs: add B200 NVFP4 recipes + benchmarks to GLM-5.2 cookbook (#29674) zijiexiaandClaude Opus 4.8 2026-06-29 14:06:30 -07:00
• a5c45a12bb CUDA graph executable dedup via cudaGraphExecUpdate (#29625) cctryandcctry 2026-06-29 13:56:33 -07:00
• f480c5f1f9 [Spec] Frozen-KV MTP: delay target KV binding to pool init + reset stale draft out_cache_loc (#29616) 5 people 2026-06-29 11:57:10 -07:00
• b0be644133 [diffusion] feat: keep image-model auxiliary components resident under auto memory policy (#29649) Mick 2026-06-30 01:25:37 +08:00
• e6c15f76f3 [optimize] fix swa eviction boundary for unfinished inserts (#29350) Feng Yao 2026-06-29 09:19:15 -07:00
• 473a278dd1 model: support nvidia/LocateAnything-3B (#28958) Jyothirmai KottuandXinyuan Tong 2026-06-29 09:16:42 -07:00
• 5169df70f6 [AMD] Sgl-data mount opt-in (#29661) YC Yen-Ching Tsengandbingxche 2026-06-29 22:11:37 +08:00
• bb7d3440b5 bugfix revise interface get cpu copy for npu mem pool to align with gpu (#29146) McZyWu 2026-06-29 19:30:14 +08:00
• 489017b3d6 [NPU] [DOC] Update deterministic inference feature support status to A2, A3 (#29632) amote-i 2026-06-29 19:17:06 +08:00
• 93926b9a95 Add Qwen3 MoE tests for PP compatibility with CP and DP (#29640) Shangming Cai 2026-06-29 18:52:03 +08:00
• a2b5ce2ed1 Add stochastic rounding for FP16 Mamba SSM cache (#26929) danielafrimiandDaniel Afrimi 2026-06-29 11:47:09 +03:00
• d5133e925b [NPU][Bugfix] Add scoring_func for mimo_v2 (#29493) iridiumineandiridiumine 2026-06-29 16:34:54 +08:00
• 3b1b512a9a Fix disaggregation receiver ZMQ cleanup (#29570) bef0rewind 2026-06-29 16:22:29 +08:00
• 91cf159696 [PP] bugfix: include CP size in PP rank offset (#29571) SSSunzt 2026-06-29 16:01:41 +08:00
• c7b9b92d9a Fix CI caused by https://github.com/sgl-project/sglang/pull/29576 (#29596) Brayden ZhongandBrayden Zhong 2026-06-28 23:46:18 -07:00
• c0d45cda29 [Spec] Add DFLASH basic sanity CI test (#29338) Liangsheng Yin 2026-06-28 23:45:05 -07:00
• 62d7929b3a [NPU][Bugfix] Accept in_capture in Ascend replay metadata (#29598) AndyLi429 2026-06-29 14:37:45 +08:00
• cd91dd0757 fix test_weight_checker_comparator assertion and ue8m0 scale unpack (#29623) Yueming Yuan 2026-06-28 23:36:37 -07:00
• f85cc94d82 [NPU] Qwen3-VL-30B use split_qkv_rmsnorm_rope for extend (#29505) silencejade 2026-06-29 14:35:19 +08:00
• 38d4ffcd86 [cookbook] drop redundant serve flags (GLM-5.2) + fix M3 page-size note (#28731) Xinyuan Tong 2026-06-29 06:34:13 +01:00
• bb74ed4a8d Replace hasattr with isinstance in SHM feature helpers (#29549) Lianmin Zheng 2026-06-28 21:31:28 -07:00
• f76e707f59 Clean up follow-ups for eagle hidden dim clean up (#29546) Lianmin Zheng 2026-06-28 21:30:53 -07:00
• 2260e612f6 [NPU] update best practicce docs from testcase (#29492) jianzhao-xu 2026-06-29 11:30:27 +08:00
• 06fd2efedd [NPU][Bugfix] Fix a ModelSlim loading failure (#29029) Peng Xingchen 2026-06-29 09:53:01 +08:00
• be17475a1b [Spec] Publish DFLASH verify read-done event for fine-grained WAR barrier (#29541) Liangsheng Yin 2026-06-28 18:51:38 -07:00
• e1ca92a7fd [weight checker] refactor: add precision branch; allow ULP quant err; used chunked compare (#28974) Yueming Yuan 2026-06-28 18:44:05 -07:00