Commit Graph

Select branches

Hide Pull Requests

cch/megamoe-rc-v0.1

Mono Color

• ae4be601c2 Fix CI when sgl-kernel is changed but srt is not changed (#10515) fzyzcjy 2025-09-16 17:49:54 +08:00
• 9b876889b7 Update CUTLASS. Refine KernelSchedule for fp8 (grouped) gemm. (#10491) Qi Yuhang 2025-09-16 17:47:37 +08:00
• c0c6f543e4 chore: upgrade sgl-kernel 0.3.10 (#10500) Yineng Zhang 2025-09-16 02:00:53 -07:00
• edd6a07bc0 Minor fix lint introduced by #10466 (#10507) Shangming Cai 2025-09-16 16:38:25 +08:00
• b6dd4bcb81 feat: update support for qwen3next model (#10466) cao1zhg 2025-09-16 16:09:56 +08:00
• b2435be682 Cache the result of is_blackwell platform check (#10498) b8zhong 2025-09-15 22:30:28 -07:00
• 5fe39e85a2 [router] fix router manager and router init in server (#10499) Chang Su 2025-09-15 22:23:26 -07:00
• fa5d0bf6a5 Remove wrong imports from sglang.python (#10493) Liangsheng Yin 2025-09-16 13:12:21 +08:00
• 16e9335998 [router] add router db connector for responses api (#10487) Simo Lin 2025-09-16 01:04:56 -04:00
• f1c692f6f8 Add Logprobs unit test with a loose threshold (#10230) Night 2025-09-16 01:04:40 -04:00
• 80572c8345 [ModelOpt] Respect kv_cache_quant_algo in ModelOpt checkpoints (#10336) brayden-hai 2025-09-15 20:16:49 -07:00
• 4bb08f6e07 [Hicache] Evaluate Per-Round Metrics in Multiturn Bench (#10203) ykwd 2025-09-16 10:34:40 +08:00
• ec272dda9c Temporay work-around for rocm 7.0.0 alpha with enabling data-parallel issue (#10434) kk 2025-09-16 10:08:04 +08:00
• a220c14f81 fix crash of DeepSeek-V3 update_weights_from_disk (#8863) scut-cbq 2025-09-16 09:45:15 +08:00
• 35ef3f2902 [router] fix worker registration in multi model mode (#10486) Chang Su 2025-09-15 18:05:00 -07:00
• 31fb19a0a2 [Auto Sync] Update registry.py (20250915) (#10484) Lianmin Zheng 2025-09-15 17:34:28 -07:00
• 3f41b48c40 [2/2] Introduce Chunked-SGMV kernels and corresponding LoRA backend for improved performance (#10286) Lifu Huang 2025-09-15 16:04:03 -07:00
• 2689f0bf02 [router] multi model registration fix (#10481) Chang Su 2025-09-15 15:22:21 -07:00
• 5207424014 chore: bump v0.3.10 sgl-kernel (#10478) Yineng Zhang 2025-09-15 15:20:09 -07:00
• c3c26f76b3 [Env] minimal version for organizing envs (#10479) Liangsheng Yin 2025-09-16 03:51:25 +08:00
• 2cf811a9da Fix --dataset-path in bench_one_batch_server (#10475) Liangsheng Yin 2025-09-16 02:55:02 +08:00
• 3b25dc127a [1/2] Speed up trtllm_mla attention backend (>10% e2e) (#10473) fzyzcjy 2025-09-16 02:53:21 +08:00
• 5c08d7d21d fix: resolve sgl-kernel ut (#10476) Yineng Zhang 2025-09-15 11:42:48 -07:00
• a45d9a4ee8 model: support solar (#8189) Praneth Paruchuri 2025-09-15 23:51:13 +05:30
• 28c79dc84a fix: gpt-oss streaming dropping normal content when tools are provided but not used (#9657) Jonas 2025-09-15 20:02:32 +02:00
• 1fcccda4b2 fix(internvl): fix accuracy issue of normalization (#10375) Kevin Tuan 2025-09-16 01:56:01 +08:00
• 79acec4fe7 [CI] Fix runner for sgl-kernel (#9887) Sahithi Chigurupati 2025-09-15 10:55:48 -07:00
• b1721edbac [PD metrics] Add latency Histogram metrics of each stage for generate requests (#8710) Yingchun Lai 2025-09-16 01:52:49 +08:00
• 57234d0c9c [bugfix] fix typo (#10471) Jiayi Yan 2025-09-15 22:29:20 +08:00
• b93acd7020 [router] minor code clean up in server startup (#10470) Chang Su 2025-09-15 07:28:25 -07:00
• 86a32bb5cd chore: bump v0.5.3rc0 (#10468) Yineng Zhang 2025-09-15 03:55:18 -07:00
• 5afd036533 feat: support pip install sglang (#10465) Yineng Zhang 2025-09-15 03:09:17 -07:00
• 059c13de5c Fix trtllm_moe wrong correction bias (#10440) fzyzcjy 2025-09-15 16:02:05 +08:00
• 50dc0c1e9c Run tests based on labels (#10456) Lianmin Zheng 2025-09-15 00:29:20 -07:00
• 76becc1dbc Add rtx5880 moe triton (#10439) Jimmy_L 2025-09-15 15:12:10 +08:00
• 2a37b24d23 [HotFix]: Hot fix import path in 3fs_bench_client.py (#10463) hzh0425 2025-09-15 14:45:46 +08:00
• f73aae0bfc Update GITHUB_TOKEN secret for documentation push (#10458) Lianmin Zheng 2025-09-14 21:59:13 -07:00
• 69b35793a0 [router] fix logger ordering git ctx (#10457) Chang Su 2025-09-14 21:37:21 -07:00
• 957482c8f2 [router] add dependency for router (#10401) ooapex 2025-09-15 12:14:14 +08:00
• 3795b6a43f fix(server_args): Skip chunked_prefill_size validation when disaggregation mode is decode (#10358) Jimmy 2025-09-15 12:13:35 +08:00
• 7eccbe992d [router] fix service discovery and mcp ut (#10449) Simo Lin 2025-09-15 00:07:23 -04:00
• 0549f21c60 fix: fix max_new_tokens uninitialized error (#9343) Mick 2025-09-15 12:06:55 +08:00
• b354e3c90d [CI] Fix token key in label-pr.yml workflow (#10452) Lianmin Zheng 2025-09-14 20:45:53 -07:00
• 65e6f48ce4 Update permissions in label-pr.yml (#10450) Lianmin Zheng 2025-09-14 20:41:43 -07:00
• 0ec580a86c Fix label PR (#10445) Lianmin Zheng 2025-09-14 20:33:09 -07:00
• 0b14159fc4 Add reasoning examples for GPT-OSS in Markdown examples (#9626) Vincent Zhong 2025-09-14 23:27:40 -04:00
• 1489cd6c02 [docs / oneliner] update mmmu docs instruction (#9768) Vincent Zhong 2025-09-14 23:26:39 -04:00
• fc2c3a3d8e metrics: support customer labels specified in request header (#10143) Yingchun Lai 2025-09-15 11:00:08 +08:00
• 8f6a175803 Fix label pr for ci (#10441) Lianmin Zheng 2025-09-14 19:48:06 -07:00
• 010181388c Tiny fix wrong naming (#10437) fzyzcjy 2025-09-15 10:24:41 +08:00
• 4844fac91d Refactor TopK to ensure readability and extensibility (#9338) Cheng Wan 2025-09-14 19:16:25 -07:00
• b7d385e812 automatically label pr for ci (#10435) Lianmin Zheng 2025-09-14 19:13:11 -07:00
• 305c9e8c2d [4/N]DP refactor: support watching mode get_load and shortest queue strategy (#10201) Liangsheng Yin 2025-09-15 10:06:08 +08:00
• ca63f075b7 Revert "Fix FA4 import cause moe_fused_gate output be illegal memory" (#10432) fzyzcjy 2025-09-15 10:03:27 +08:00
• f9ee6ae17a [router]: Add Embedding routing logic (#10129) Jintao Zhang 2025-09-15 09:44:35 +08:00
• dcee42c200 feat: add dsv3 fp4 cutlass moe etp ut (#10433) Yineng Zhang 2025-09-14 18:44:09 -07:00
• 258d02c86d Fix correction bias undefined behavior for nvfp4 models (#10426) fzyzcjy 2025-09-15 09:41:09 +08:00
• 60d7beda6b Add split tile size for Triton attention (#10425) Ke Bao 2025-09-15 08:35:49 +08:00
• 2f8ba6fe82 [Fix] MoE: fix w8a8_fp8 MoE and add tests to cover this code path (#10429) Cheng Wan 2025-09-14 17:34:28 -07:00
• 7ce6c10eb6 fix: enable cu124 and cu128 build on main push (#10431) Yineng Zhang 2025-09-14 16:19:35 -07:00
• 55025b9282 fix: use latest flashinfer (#10428) Yineng Zhang 2025-09-14 15:07:14 -07:00
• 4c21b09074 [Feature] Sglang Tracing: Fine-Grained Tracking for Request Latency - Part 1 (#9962) Feng Su 2025-09-15 02:08:02 +08:00
• 165abeebca Typo: in --enable-custom-logit-processor: agree with cli arg (#10076) 艾力可 2025-09-14 11:27:09 +02:00
• 21ca4c3afa [PD metrics] Fix some uncompleted PD related metrics (#8627) Yingchun Lai 2025-09-14 17:26:58 +08:00
• e3cf812f7d Fix sgl-kernel + srt CI (#10419) fzyzcjy 2025-09-14 16:44:47 +08:00
• 4da5533682 Support profile args in Engine API (#6539) fzyzcjy 2025-09-14 16:21:10 +08:00
• ac964d2e58 Support global scale in addition to per expert scale for cutedsl moe (#10270) fzyzcjy 2025-09-14 16:17:00 +08:00
• fa46e2bd40 Support offloading in fp8 (#9948) fzyzcjy 2025-09-14 16:14:28 +08:00
• b047b553c2 [2/2] Speed up prefill mla attention concat (#10157) fzyzcjy 2025-09-14 16:12:04 +08:00
• a0f844ed5a Let sgl-kernel changes be tested on srt (#10313) fzyzcjy 2025-09-14 16:09:17 +08:00
• 2df532ef20 Fix the global scale fix does not support EPLB and improve enabling condition (#10369) fzyzcjy 2025-09-14 16:07:47 +08:00
• abea9250da Auto determine sgl kernel version in blackwell CI (#10318) fzyzcjy 2025-09-14 16:06:30 +08:00
• b3c977622f Add h200 fused moe config for Qwen3-Next (#10404) Ximingwang-09 2025-09-14 14:48:46 +08:00
• b8347b40b1 Add self.capture_aux_hidden_states For GLM-4.5V (#10228) Yuxuan Zhang 2025-09-14 14:31:55 +08:00
• 72dfa96aeb Fix cutlass moe accuracy drop caused by attention UB from DP padding mode (#10414) fzyzcjy 2025-09-14 13:29:09 +08:00
• 05b01ef4da fix duplicated logger in eager_utils (#10410) lijin23 2025-09-14 13:25:40 +08:00
• 55a6e644b0 [Hack] Add pd-disaggregation decode polling interval (#10411) Liangsheng Yin 2025-09-14 10:18:23 +08:00
• 6897e06b69 Remove repeatedly lists adding in init_incremental_detokenization (#10412) Liangsheng Yin 2025-09-14 10:05:52 +08:00
• a360511d7b [Generative Score API] Scoring(Prefill-only) optimizations. (#9748) Sundara Raman Ramachandran 2025-09-13 10:57:06 -07:00
• 94d0f656fb [Performance] Dynamic Batch Tokenizer (#9382) Sundara Raman Ramachandran 2025-09-13 10:56:04 -07:00
• eca59f96c3 Update ROCm docker image to add sgl-router support (#10406) kk 2025-09-14 01:52:18 +08:00
• 9752861002 [Fix] Support qwen3-next MTP+DP (#10392) Binyao Jiang 2025-09-13 02:45:04 -07:00
• 297d374510 support qwen3_next blackwell (#10403) Yi Zhang 2025-09-13 17:18:26 +08:00
• 31e9d3a5aa [Fix] Init mamba related memory pools with torch.zeros (#10400) Binyao Jiang 2025-09-12 23:16:48 -07:00
• 6f4676ef85 fix: tool parse in large streaming chunk beginning with normal content (#10397) Xinyuan Tong 2025-09-12 22:29:35 -07:00
• c9ec4cae5b Fix the style of sgl kernel (#10398) Lianmin Zheng 2025-09-12 22:20:21 -07:00
• 99757cc3e6 fix probs name which without temp scaling name (#9984) narutolhy 2025-09-12 21:19:57 -07:00
• cdddab056c [Auto Sync] Update xgrammar_backend.py (20250913) (#10395) Lianmin Zheng 2025-09-12 17:46:56 -07:00
• 7c5a0a1b77 [router] add not implemented functions for multi model trait (#10394) Simo Lin 2025-09-12 19:44:18 -04:00
• 49f169d53e [HiCache] doc: update deployment in readme (#10332) Teng Ma 2025-09-13 07:35:37 +08:00
• 7fce2fd91a [HiCache] fix mooncake config in different tp size (#10377) Teng Ma 2025-09-13 07:34:23 +08:00
• 16cd550c85 Support Qwen3-Next on Ascend NPU (#10379) Even Zhou 2025-09-13 07:31:37 +08:00
• d5e2a37414 Benchmark: Support API_KEY without 'bearer' (#10380) Muqi Li 2025-09-13 07:29:04 +08:00
• 366043db8e [router] Add get and cancel method for response api (#10387) Keyang Ru 2025-09-12 16:19:38 -07:00
• 2f173ea074 [router] allow one router to support different model families and serving mode (#10244) Simo Lin 2025-09-12 19:18:27 -04:00
• 321fecab74 Add sentencepiece to project dependencies (#10386) Mohammad Miadh Angkad 2025-09-13 07:02:54 +08:00
• 9d775b1a2d feat: add deepseek v3 fp4 ut (#10391) Yineng Zhang 2025-09-12 15:43:29 -07:00
• 78b7465cad Fix GPU fault issue when run dsv3 with dp mode and enable torch-compile (#10361) kk 2025-09-13 06:05:51 +08:00
• 07bcad7fb7 [bug] fix router ci syntax error (#10390) Simo Lin 2025-09-12 17:39:15 -04:00
• 98adac8ef7 fix: exclude protobuf generated code (#10388) Yineng Zhang 2025-09-12 13:52:09 -07:00