chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Commit Graph

  • f2d1b7c4ad Support sequence parallel and trivial dims in dump comparator (#19460) fzyzcjy 2026-02-27 08:11:15 +08:00
  • eb2ada3804 Support non-packed format when aligning tokens in dump comparator (#19459) fzyzcjy 2026-02-27 08:10:29 +08:00
  • e1e0cfd856 Use named tensors in dump comparator (#19458) fzyzcjy 2026-02-27 08:09:55 +08:00
  • eb0e905fc3 Update token layout and cleanup printer in dump comparator (#19457) fzyzcjy 2026-02-27 08:09:15 +08:00
  • 8ac64e1487 Support unifying axis ordering in dump comparator (#19456) fzyzcjy 2026-02-27 08:08:32 +08:00
  • 425d333ee3 Support token dim in arbitrary location in dump comparator (#19455) fzyzcjy 2026-02-27 08:07:38 +08:00
  • 5172c37845 [AMD] Use fused GEMM with FP8 cast for FP8 prefill (#19422) Thomas Wang 2026-02-27 07:14:52 +08:00
  • b79fa9ccd7 [CI] Add Claude skill for writing SGLang tests (#19448) sglang-bot 2026-02-26 13:37:54 -08:00
  • a1ef8e2cc0 [AMD] optimize Kimi K2.5 fused_moe_triton performance by tuning (#19228) RoyWang 2026-02-27 03:50:13 +08:00
  • 288300aafd [PD] Tiny code cleanup for prefill info registering (#19414) Shangming Cai 2026-02-27 01:35:26 +08:00
  • e4b708d3e9 [Spec V2] Support specV2 for mamba hybrid attention (#18808) zhangheng 2026-02-27 00:36:01 +08:00
  • 78d6674c45 [diffusion] feat: support hybrid parallelism for diffusers backend (#19405) DefTruth 2026-02-27 00:06:08 +08:00
  • 5939b8912a [MUSA][11/N] ci: add MUSA 4.3 kernel build and release pipeline (#18537) johnnycxm 2026-02-26 23:45:12 +08:00
  • e55e65535e [Bugfix] Add rids to the batch filtering for two batch overlap (#19418) Shangming Cai 2026-02-26 22:57:25 +08:00
  • 97f1fa5e6b [NPU] Fix disaggregation metadata buffer bootstrap_room_dtype for npu backend (#19423) Shangming Cai 2026-02-26 21:10:50 +08:00
  • 86eb80007e [NPU] support Kimi-K2.5 on NPU (#19331) khalilzhk 2026-02-26 20:41:44 +08:00
  • bdc1e46e5a [Qwen3.5] Qwen3.5-27B inference repeat bug fix (#19411) AlfredYong 2026-02-26 20:11:29 +08:00
  • 74c8e7b215 refactor(jit_kernel): reduce duplication and separate test code (#19323) Xiaoyu Zhang 2026-02-26 18:30:49 +08:00
  • a7152df2e3 [diffusion ] CLI: Fix typo in CLI usage doc string (#19316) Junhao Liu 2026-02-26 02:24:14 -08:00
  • 27fd014726 [PD] Add kv_cache_dtype consistency check for PD Disaggregation (#19407) Shangming Cai 2026-02-26 17:15:58 +08:00
  • e14fd4accb Fix nightly Mistral-Large-3 NVFP4 accuracy threshold (#19402) Alison Shao 2026-02-25 23:59:11 -08:00
  • d2b3c7fb14 [Tracing] update script for converting otel tracing data to perfetto format (#19396) Feng Su 2026-02-26 14:03:22 +08:00
  • de3d1e7669 [misc] use ORJSONResponse in http-server generate (#19191) Yilong Zhao 2026-02-25 21:26:25 -08:00
  • 0fd44ff342 Fix NSA CP positions mismatch in eagle NextN model (#19367) Alison Shao 2026-02-25 20:14:33 -08:00
  • 119c91cb8b Skip signal handler registration when not on main thread (#18752) Xinyu Zhang 2026-02-25 19:30:05 -08:00
  • 88ad3b894a [sgl-kernel][Feat][B200][2/N] Support MXFP8 Grouped GEMM in Blackwell (#14640) Qi Yuhang 2026-02-26 11:23:37 +08:00
  • b3202fe6d0 [PCG] fix piecewise cuda graph for Qwen3.5 (#19220) Minglei Zhu 2026-02-25 19:16:52 -08:00
  • a0a8f1473c [Benchmark] Fix generated_shared_prefix attribute naming and remove args dependency (#19363) Alison Shao 2026-02-25 18:45:54 -08:00
  • 6e82183f5a [Disagg] Route disagg prefill results through process_batch_result (#19364) sglang-bot 2026-02-25 18:38:39 -08:00
  • 914ed34757 update jit_kernel codeowners (#19385) Xiaoyu Zhang 2026-02-26 10:36:34 +08:00
  • 265eb56d44 Support multi-step alignment and pipeline integration in dump comparator (#19378) fzyzcjy 2026-02-26 10:23:22 +08:00
  • 4e843f1216 [DeepSeek-V3.2][JIT-kernel] Support nsa fuse store indexer k cache (#19148) Yuan Luo 2026-02-26 10:23:10 +08:00
  • f230967e65 [AMD] Fix ROCm Docker builds, update apache-tvm-ffi (#19359) Michael 2026-02-25 18:16:28 -08:00
  • f9a2f0398f Support token aligner planning and execution in dump comparator (#19377) fzyzcjy 2026-02-26 10:04:33 +08:00
  • d34d5aca07 Support loading token aligner data in dump comparator (#19376) fzyzcjy 2026-02-26 10:03:56 +08:00
  • e8dd14519d Add aligner entrypoint and bundle handler in dump comparator (#19375) fzyzcjy 2026-02-26 10:03:22 +08:00
  • 2ad475b4ed use flashinfer.sampling (#18696) pansicheng 2026-02-26 10:02:38 +08:00
  • 2739d7df62 Reorganize modules and pipeline in dump comparator (#19374) fzyzcjy 2026-02-26 10:00:13 +08:00
  • 508b8e3387 Handle warnings via sink for structured output and add pair in dump comparator (#19373) fzyzcjy 2026-02-26 09:59:15 +08:00
  • 46321ee70e Support dumping rid for correlation across passes in dump comparator (#19372) fzyzcjy 2026-02-26 09:57:57 +08:00
  • 7c9e8e2def [Re-land][jit kernel] Support per_token_group_quant_8bit jit kernel (#19140) Yuan Luo 2026-02-26 09:53:57 +08:00
  • beabaa8d37 [Kernel Slimming] Migrate marlin moe kernel to JIT (#19181) Linyu Wu 2026-02-26 09:05:13 +08:00
  • 350190487b Flashinfer MOE FP8 support for Mistral Large 3. (#15422) Daniel Cámpora 2026-02-26 00:00:37 +01:00
  • c60dcc40bb [Logging] Guard log_prefill_stats against idle batches in disagg prefill (#19361) Liangsheng Yin 2026-02-25 13:31:52 -08:00
  • 08957c88ea [Logging] Fix prefill side logging in pd disagg (#19350) YAMY 2026-02-25 12:42:18 -08:00
  • 306c552639 Revert "Fix HybridAttnBackend forward for linear attention" (#19356) Kangyan-Zhou 2026-02-25 11:49:50 -08:00
  • a0f3361023 Update (#19351) HAI 2026-02-25 11:37:07 -08:00
  • b2c46fc60b [AMD] Support Qwen3-Coder-Next on AMD platform (#18355) jacky.cheng 2026-02-26 03:06:22 +08:00
  • cc1ca61c81 fix: add --cuda-graph-max-bs to DSV3 FA3 FP8 KV cache test (#19307) Alison Shao 2026-02-25 10:27:31 -08:00
  • 0217e82a08 [diffusion] Clean code (#19325) Makcum888e 2026-02-25 21:16:03 +03:00
  • 2fb239450e Revert "bugfix: prioritize init_npu_backend to fix various initialization bugs" (#19343) Even Zhou 2026-02-25 23:04:30 +08:00
  • c7c4a1cbbd refactor linear attention backend (#18622) Yuhao Yang 2026-02-25 23:02:44 +08:00
  • 471acd98b9 [diffusion] logging: improve logging (#19312) Mick 2026-02-25 23:00:35 +08:00
  • 56891e46bc [Ascend ] Add qwen3.5 122B/35B/27B deployment examples on doc (#19339) Hexq0210 2026-02-25 21:09:22 +08:00
  • 59b9d1e86d [diffusion] improve: improve fuse_scale_shift_kernel with non-blocking op (#18710) Qingfu Wen 2026-02-25 21:04:20 +08:00
  • c144e55462 Fix HybridAttnBackend forward for linear attention (#19006) akhilg-nv 2026-02-25 05:02:37 -08:00
  • d38c0e537d fix(dense): fix Qwen3.5 dense model precision bug in TP_SIZE>1 (#19070) Zheng Li 2026-02-25 20:54:42 +08:00
  • cdc411160b [NPU] Fix a corner case where FusedMoE.top_k is not explicitly declared (#19287) Even Zhou 2026-02-25 20:49:59 +08:00
  • 9840cd3f68 [diffusion] chore: enable sequence shard for wan by default (#19311) Mick 2026-02-25 18:21:44 +08:00
  • 307a2b73ed [AMD] add testcases for Qwen3 235b Instruct 2507 models (#18805) mqhc2020 2026-02-25 18:17:52 +08:00
  • 60eeef7370 [AMD][with CI Fix] support two batch overlapping for mori ep (#19216) billishyahao 2026-02-25 18:14:08 +08:00
  • c4ef33862b [diffusion] fix: fix bugs to let LTX-2 pipeline support latest Sglang Args pipelines (#19295) GMI Xiao Jin 2026-02-25 01:30:36 -08:00
  • 671b595570 Fix trtllm_mha fp8 SWA KV index translation (#19107) Mohammad Miadh Angkad 2026-02-25 17:02:17 +08:00
  • a55f658835 [Misc] Normalize --host parameter to use plain hostname without scheme (#19309) Julian Huang 2026-02-25 16:37:24 +08:00
  • f75abb4521 [Fix][Qwen3.5] Fix KV cache slice transfer for GQA models with replicated KV heads (#19086) YAMY 2026-02-25 00:26:44 -08:00
  • d40cb2f725 [HiCache] Support heterogeneous tp for hicache storage (#18541) huangtingwei 2026-02-25 16:13:57 +08:00
  • 3d879c69e9 refactor: extract device-to-backend mapping into get_default_distributed_backend (#19202) Wang, Yi 2026-02-25 15:50:26 +08:00
  • d0bb140034 [NPU] bugfix for model Qwen3-Coder-Next at weight shape transpose for npu. (#18700) Hexq0210 2026-02-25 15:46:20 +08:00
  • a1b39c1c26 Perf/fuse mamba state scatter mtp verify (#18088) xutizhou 2026-02-25 15:40:55 +08:00
  • ab7071b545 [SKILL] Better claude skills for sgl-kernel and jit-kernel (#19302) Xiaoyu Zhang 2026-02-25 15:26:55 +08:00
  • 4a3a787f1e [Fix] Kimi K2.5 support pp (#18434) lw9527 2026-02-25 15:22:11 +08:00
  • 8d9ee6669e Fix comment for tp_rank calculation in dp_attention (#19306) Shangming Cai 2026-02-25 15:19:10 +08:00
  • 17b0affbdf [AMD] Support --enable-aiter-allreduce-fusion on AMD GPUs (#13747) Hubert Lu 2026-02-24 23:11:55 -08:00
  • 73fe389dd1 [Qwen3.5] Raise Exception when radix_cache and extra_buffer are enabled at the same time (#19169) YAMY 2026-02-24 23:04:37 -08:00
  • 4dba73087e [CI] grant ci permissions to sglang-npu-bot (#19308) Even Zhou 2026-02-25 14:55:23 +08:00
  • 76d5410e01 [Disagg] Fix decode querying unregistered dp_rank when prefill dp_size is 1 (#19305) Liangsheng Yin 2026-02-24 22:52:39 -08:00
  • 8bd644765f [AMD] Enable ROCm kvcache JIT path and add AMD CI coverage. (#18992) Hubert Lu 2026-02-24 22:15:05 -08:00
  • aff2f130ec [diffusion] CI: fix pr-test workflow file to include changes in jit kernels (#19285) Mick 2026-02-25 14:08:55 +08:00
  • ca09d71cf0 Fix nightly grok failure on rotary embedding import (#19192) Michael 2026-02-24 21:25:16 -08:00
  • 3fe93b5493 Updated benchmark guide (#19243) nvjullin 2026-02-25 13:11:17 +08:00
  • e138f7960a [AMD] Fix accuracy while using --enable-dp-attention (#19247) jacky.cheng 2026-02-25 12:50:28 +08:00
  • ab0f608788 [PD-Disagg] Fix bootstrap server race condition when prefill workers not yet registered (#19288) Liangsheng Yin 2026-02-24 20:22:16 -08:00
  • 539f772f54 [PD-Disagg] Fully support external DP dispatch w/ PD-disaggregation mode. (#19268) Liangsheng Yin 2026-02-24 19:58:01 -08:00
  • 241ee90164 [diffusion] chore: tiny fix pyproject.toml (#19256) Mick 2026-02-25 11:57:53 +08:00
  • 694924b878 Update .gitignore to remove '.claude/' (#19296) Xiaoyu Zhang 2026-02-25 11:51:50 +08:00
  • 23adb50751 [AMD] [GLM-5 Day 0] Add GLM-5 nightly test (#18911) Michael 2026-02-24 19:39:17 -08:00
  • 1dae3e32b9 update code owners for observability (#19294) Feng Su 2026-02-25 11:26:41 +08:00
  • f862214c67 ci: temporarily disable GB200 test stage (#19289) Alison Shao 2026-02-24 19:04:23 -08:00
  • 0fac2796b6 [PD-Disagg] Improve KVManager init across all backends (#19240) Shangming Cai 2026-02-25 10:37:09 +08:00
  • c0fdfd4b92 Delete mm.feature after decode phase (#17324) siyu 2026-02-25 10:13:03 +08:00
  • 9dff933164 [Kernel Slimming] Remove sgl-kernel AOT marlin kernels (#19241) Xiaoyu Zhang 2026-02-25 10:08:22 +08:00
  • 3b89302277 Refactor: observability code cleanup (#17862) Feng Su 2026-02-25 10:07:29 +08:00
  • 245430eaac Encoder Global Cache Manager (#16137) siyu 2026-02-25 10:05:43 +08:00
  • b7af58b9af Support replication axis in dump comparator (#19282) fzyzcjy 2026-02-25 09:48:43 +08:00
  • 2e2b18e870 Support context parallel zigzag reordering in dump comparator (#19281) fzyzcjy 2026-02-25 09:46:17 +08:00
  • 0de1f4b07b Support multi axis unsharding in dump comparator (#19280) fzyzcjy 2026-02-25 09:44:07 +08:00
  • 4bb678f28a Full test coverage in dump comparator (#19279) fzyzcjy 2026-02-25 09:43:29 +08:00
  • 94ca2ac5d7 Support TP unification and enhance tests in dump comparator (#19278) fzyzcjy 2026-02-25 09:42:56 +08:00
  • 39ba9b5ab5 Support simple unsharding in dumper comparator (#19277) fzyzcjy 2026-02-25 09:42:21 +08:00
  • 02ca107b2c Support dims annotation and enhance dump loader in dumper (#19276) fzyzcjy 2026-02-25 09:41:48 +08:00