Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 043f13171f [Performance] Optimize NSA Indexer K/S Buffer Access with Fused Triton Kernels (#13812) JohnsonmsandJohnsonms 2025-12-02 18:53:06 -08:00
• f764c6910d [diffusion] feat: support distilled vae generic (#14195) 3 people 2025-12-02 21:27:31 -05:00
• 922054079c [Doc] Update DeepSeek-V3.2 document (#14321) Baizhou Zhang 2025-12-02 18:19:39 -08:00
• 7d1a130cde Refactor custom allreduce logics (#13710) 4 people 2025-12-03 09:20:05 +08:00
• 7ae368efde chore: bump SGLang version to 0.5.6 (#14316) sglang-botandsglang-bot 2025-12-02 17:17:13 -08:00
• c4e20caded ci: Add zyzshishui to CI permissions (#14324) sunxxunsandroot 2025-12-02 20:00:31 -05:00
• 5dad1ff1b6 [model-gateway] add workflow for external model providers (#14323) Simo Lin 2025-12-02 16:47:06 -08:00
• ca52ed425f Clean up imports and move files (#14317) Lianmin Zheng 2025-12-02 16:31:54 -08:00
• 5c03aa3e9d Adding section for scheduled PR test runs on main (#14309) Douglas Yang 2025-12-02 14:52:44 -08:00
• 253be18e52 Fix nonetype error for ci failure monitor (#14319) Douglas Yang 2025-12-02 14:24:25 -08:00
• 084b06e79d Add /rerun-stage slash command to rerun specific PR test stages (#14262) alisonshao 2025-12-02 14:23:38 -08:00
• fc6fb550ec [Minor] update docs on CI (#14315) Lianmin Zheng 2025-12-02 12:06:54 -08:00
• 7c38eca1e4 feat: DeepSeek new v3.2 encoding (#14249) Eva20150932-atlascloudandXinyuan Tong 2025-12-02 19:41:05 +00:00
• 427b08e24d Init TBO with dp_padded batch (#11423) 3 people 2025-12-03 02:34:26 +08:00
• 0141ca370f Revert PR #14044: Restore separate memory pool for piecewise CUDA graph (#14278) alisonshao 2025-12-02 09:53:16 -08:00
• 25a6be4930 Fix duplicate download log messages in multi-process environment (#14299) alisonshao 2025-12-02 09:33:18 -08:00
• df1f31241b [sgl-kernel] fix runtime error while preloading CUDA runtime (#13089) Antonin Vidon 2025-12-02 10:03:51 -05:00
• c5947ecd85 Opt moe align block size kernel (#14133) Xiaoyu Zhang 2025-12-02 19:13:55 +08:00
• 9530b76630 [diffusion] refactor: simplify DmdDenoisingStage (#14269) Mick 2025-12-02 18:59:40 +08:00
• 3067b3f050 [diffusion] chore: improve model info registration and searching strategy (#14281) 3 people 2025-12-02 18:28:59 +08:00
• e0ec42c710 [CI] Fix 4-GPU test timeout by using 3 partitions (#14287) alisonshao 2025-12-02 00:47:26 -08:00
• 9c9d7091cb Update CODEOWNERS for multimodal_gen (#14286) Mick 2025-12-02 16:15:19 +08:00
• 51a86ce69e [model-gateway] change rust package name to sgl-model-gateway instead (#14283) Simo Lin 2025-12-02 00:06:54 -08:00
• 64092c8b55 [Auto Sync] Rename is_hybrid to is_hybrid_swa (#14252) 4 people 2025-12-01 23:24:24 -08:00
• 63b9300f00 chore: bump sgl-kernel version to 0.3.18.post2 (#14244) sglang-bot 2025-12-01 23:14:12 -08:00
• 21ec99beff [VLM][Doc] Document for VLM DP Encoder (#14279) Yuan Luoandluoyuan.luo 2025-12-02 15:08:37 +08:00
• 383689e3ad [model-gateway] fix version output (#14276) Simo Lin 2025-12-01 22:51:57 -08:00
• 236a7c2370 fix trtllm mla spec (#13738) b8zhongandBrayden Zhong 2025-12-01 22:16:25 -08:00
• 3dabd609fb Optimize topk sigmoid in minimax_m2 (#14047) Roger Youngandxuebi 2025-12-02 14:07:12 +08:00
• 73df525382 [model-gateway] include smg version command in py binding (#14274) Simo Lin 2025-12-01 21:21:02 -08:00
• e6420100ee sync attention doc and ep doc to doctree (#14257) b8zhongandBrayden Zhong 2025-12-01 21:15:22 -08:00
• c9e2090101 fix: Support PP for Mistral Small 3.1 (#14254) Kevin Li 2025-12-01 21:04:14 -08:00
• 106df4eac5 Fix mrope_positions size when req is retracted (#13700) kun-llflandXuchun Shang 2025-12-02 11:38:20 +08:00
• 427a19b643 Remove cargo config also in .zshenv (#14267) Liangsheng Yin 2025-12-02 11:31:53 +08:00
• 1f930cd23d [diffusion] CI: add testcase-wise retry mechanism (#14261) Mick 2025-12-02 11:06:12 +08:00
• 11ce05163d Fix NIXL exception message (#14172) Kartik Ramesh 2025-12-01 20:39:45 -06:00
• cd4151abc7 [model-gateway] add audio and moderation in model card (#14263) Simo Lin 2025-12-01 18:36:38 -08:00
• 8fe8b63576 Revert "Try to remove wrong logic about max total token in spec decoding" (#14259) Stefan He 2025-12-01 18:18:03 -08:00
• 8a7b1b8301 [Docs] Update CI docs (#14260) Lianmin Zheng 2025-12-01 18:15:03 -08:00
• 26aebf83d3 [VLM] Support Piecewise CUDA Graph for Qwen3-Omni-MOE (#14222) Yuan Luoandluoyuan.luo 2025-12-02 10:12:10 +08:00
• 3ab8ae6847 [diffusion] fix: fix Flux.2 condition image resize (#14232) Mickandgemini-code-assist[bot] 2025-12-02 10:05:44 +08:00
• 03888b9de5 [Minor] Upgrade cutedsl version in Dockerfile (#13968) Baizhou Zhang 2025-12-01 17:15:26 -08:00
• 796d82b107 [Auto Sync] Add max_total_num_tokens metric: Update scheduler_metrics_mixin.py, collector.py (20251202) (#14256) 3 people 2025-12-01 16:34:34 -08:00
• 1da59e8304 [Auto Sync] optionally disable fake register in Update fp8_kernel.py (20251202) (#14255) 3 people 2025-12-01 16:11:12 -08:00
• 1d66a14c2e [model-gateway] Add e2e tests of streaming events and tool choice for response api (#13880) Xinyue ZhangandSimo Lin 2025-12-01 15:27:12 -08:00
• 02af51e4fc Support fp4 fp8 non gated moe (#13794) 3 people 2025-12-02 01:26:28 +02:00
• eb5008846a [CI] Fix test_deepep_large.py (#14247) Baizhou Zhang 2025-12-01 15:18:48 -08:00
• 079b173853 Fix a distributed initialization error (#13843) Zhiyu 2025-12-01 15:10:05 -08:00
• 57f933fd7d [model-gateway] Migrate Worker trait to model-aware methods (#14250) Simo Lin 2025-12-01 14:32:09 -08:00
• 1f2b84d28d Fix NSA Bug in Centralize NSA Dispatch Logic (#14245) YAMY 2025-12-01 13:18:18 -08:00
• e7d6027e4a [model-gateway] add ModelCard support to WorkerMetadata (#14243) Simo Lin 2025-12-01 13:04:09 -08:00
• 07821352fb Revert "Skip weight loading in deepgemm compilation" (#14241) ishandhanani 2025-12-01 12:59:09 -08:00
• edbeaf3b88 [MM][style] rename inputs_embeds to input_embeds for consistency (#14240) Byron Hsu 2025-12-01 11:36:51 -08:00
• d9dca28247 Update pr-test.yml to fix unknown job name deepep-8-gpu Kangyan-Zhou 2025-12-01 10:02:29 -08:00
• 9325f94534 [model-gateway] add ModelCard and ProviderType for model configuration (#14237) Simo Lin 2025-12-01 09:47:51 -08:00
• 41b7aab848 Disable Deepep 8 GPU tests (#14152) Kangyan-Zhou 2025-12-01 09:37:38 -08:00
• 491f4fe8e1 Modify git tag for DeepGemm in sgl-kernel. (#14179) Sulfur6-L8972 2025-12-02 01:01:22 +08:00
• 34035d8cd9 chore: bump sgl-kernel version to 0.3.18.post2 (#14229) sglang-bot 2025-12-01 08:50:40 -08:00
• 92ca629582 [model-gateway] add ModelType bitflags and Endpoint enum for worker (#14230) Simo Lin 2025-12-01 08:14:44 -08:00
• ec92d7f14e [model-gateway] fix v1/models response format to be oai compatible (#13693) Chang SuandSimo Lin 2025-12-01 07:20:38 -08:00
• 79b389da6b [model-gateway] refactor oai router 1/n (#14228) Simo Lin 2025-12-01 07:16:35 -08:00
• 3de09aadbc Add new moe wna16 marlin gemm (#14122) Xiaoyu Zhang 2025-12-01 23:07:53 +08:00
• 2e8f54e61e [spec-overlap] bugfix for pd disaggregation and npu (#14088) liupeng374andEven Zhou 2025-12-01 22:58:20 +08:00
• e55731b6e8 [model-gateway] Avoid logging MCP connection token (#13887) Wenyi Xu 2025-12-01 20:44:31 +08:00
• 45264554f3 Super tiny fix typo (#14219) fzyzcjy 2025-12-01 20:19:17 +08:00
• c4293f59ac Change PR test schedule to run every 6 hours (#14218) Lianmin Zheng 2025-12-01 03:50:58 -08:00
• a2423052f6 Add cuda event based on waiting value (#14214) Liangsheng Yin 2025-12-01 18:51:44 +08:00
• bc3d2a85af [Minor] update docs (#14212) Lianmin Zheng 2025-12-01 02:33:58 -08:00
• d815d00248 Tiny call cudaProfilerStart only on first rank in node (#14211) fzyzcjy 2025-12-01 18:18:45 +08:00
• fa9021b21f fix: Increase FlashInfer workspace size for Qwen3VL models (#14173) Xiaoyu Zhang 2025-12-01 17:54:23 +08:00
• 9c80072845 Add peak output tokens per second in bench_serving (#14165) Xiaoyu Zhangandgithub-actions[bot] 2025-12-01 17:47:54 +08:00
• 630a693081 [VLM] Boost Memory Pool based CUDA IPC (#14123) Yuan Luoandluoyuan.luo 2025-12-01 17:17:46 +08:00
• 7ce8faae28 [diffusion] refactor: remove hard-code of instanceof on PipelineConfig (#14186) Mickandgemini-code-assist[bot] 2025-12-01 16:35:34 +08:00
• de153cf76a Fix speculative decoding error when retracting (#14180) fzyzcjy 2025-12-01 15:30:13 +08:00
• f4a0c5c76b Try to remove wrong logic about max total token in spec decoding (#14167) fzyzcjy 2025-12-01 15:29:58 +08:00
• 0f8e53947d [Piecewise] Use same global graph memory pool as the main cuda graph … (#14044) 3 people 2025-11-30 23:04:10 -08:00
• e8ba5a668c Support profiling only prefill or decode without the other (#14182) fzyzcjy 2025-12-01 14:46:30 +08:00
• a2960bdd6b Super tiny allow millisecond precision in logging (#14183) fzyzcjy 2025-12-01 14:46:09 +08:00
• 487c8d4df3 Tiny add several args to bench serving (#14181) fzyzcjy 2025-12-01 14:45:47 +08:00
• f87b8eab23 Tiny fix transform_scale_ue8m0 wrong output in some scenarios (#14003) fzyzcjy 2025-12-01 14:45:27 +08:00
• e8542db558 [piecewise] move piecewise_cuda_graph_runner init to model_runner initialize (#14034) 3 people 2025-11-30 22:16:04 -08:00
• 6df1e8d628 [Auto Sync] Update backend.py (20251130) (#14153) 4 people 2025-11-30 22:15:02 -08:00
• 4addb60274 Pull Request Instructions: RL and Training Framework Integrations (#14187) Richard Chenandzhaochenyang20 2025-12-01 00:58:54 -05:00
• bd0e690857 [Feature] Enable PTPC FP8 for compressed tensors moe (aiter kernel) (#12181) qichu-yun 2025-12-01 13:54:28 +08:00
• 0825d7f4c6 [piecewise] Refactor VLM to support input embed buffer and remove external embedder hack (#14155) Byron Hsu 2025-11-30 21:43:09 -08:00
• 0b9dbea593 [diffusion] chore: improve z-image (#14104) Yuhao Yang 2025-12-01 12:26:17 +08:00
• 982db4ebac Feat: GLM-4.6 supports shared experts fusion (#13873) 3 people 2025-12-01 11:33:18 +08:00
• f5f3a5d98c [PD] Support json file configuration for Transfer Engine (#14059) Teng MaandShangming Cai 2025-12-01 10:47:33 +08:00
• f138ae5789 [model-gateway] support VL models in router (#14140) ooapex 2025-12-01 09:50:22 +08:00
• decb48965d [DeepSeekV3.2] Enable pure TP & Partial DP Attention (#13646) YAMY 2025-11-30 15:59:23 -08:00
• c72f0756d2 Fix: fix flashmla fp8 kv cache acc error (#13841) Fan Yinandybyang 2025-12-01 05:38:19 +08:00
• f1115cf58d Revert "[Minor]Raise Error when deepep num dispatch token per rank is smaller than cuda graph bs" (#14171) Baizhou Zhang 2025-11-30 13:49:46 -07:00
• 412160f4c1 [sgl-kernel] fix b200 kernel ci (#13907) Fan YinandHydraQYH 2025-12-01 02:15:37 +08:00
• 7b03cc6482 [Minor]Raise Error when deepep num dispatch token per rank is smaller than cuda graph bs (#14065) Baizhou Zhang 2025-11-30 11:11:42 -07:00
• 9872a677b4 [ci]fix deepep import error on H20 action (#14166) Hank Han 2025-12-01 01:42:57 +08:00
• c15c864b6f Fix LMCache unit test and init bug (#14005) Dongjoo Seo 2025-11-30 07:57:32 -08:00
• dc7bdc7329 bugfix[schedule]: Excessive preemption occurs when preempting running requests to schedule new prefill requests. (#12494) PiteXChen 2025-11-30 22:29:26 +08:00
• 0a9d64530d Support grammar + spec + reasoning (#14163) Liangsheng Yin 2025-11-30 21:19:57 +08:00
• 340c613ab5 Support numactl bind for CPU and memory before process starts (#14156) fzyzcjy 2025-11-30 17:00:33 +08:00
• 36b729c2b8 Implement profiler v2 and fix stage mixture bug (#14148) fzyzcjy 2025-11-30 16:59:52 +08:00