Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• c1df359b44 Add XPU profiler activity support in benchmark code (#12981) Kalyan Kumarandgemini-code-assist[bot] 2026-03-05 12:52:56 +05:30
• 2bdd89a6cd [Kernel Slimming] Migrate NVFP4 kernels to JIT (#19437) Mohammad Miadh Angkad 2026-03-05 15:22:28 +08:00
• 1bbfed0539 [misc] add env for http keep alive timeout (#19847) Yilong Zhao 2026-03-04 22:00:51 -08:00
• 86c5617787 [BUG]: fix prevent illegal memory access in Mamba SSM tracking during EAGLE speculative verification (#19415) Chenxi LiandConnorLi96 2026-03-04 21:13:21 -08:00
• 10c65df48a [Bug] Fix lora tp bug on H200 (#19769) Baizhou Zhang 2026-03-04 20:11:02 -08:00
• feda2b11c4 [AMD] Add AWQ AMD CI coverage and quantization platform compatibility docs (#19550) Bruce Changlong Xu 2026-03-04 19:50:55 -08:00
• 0e6a64712a [bugfix] Fix PPMissingLayer AttributeError when Using PP (#19804) Xinyi Song 2026-03-04 19:48:15 -08:00
• 198381d9ce Add SSL/TLS support for HTTP and gRPC servers (#18973) Kangyan-Zhouand 2026-03-04 19:27:16 -08:00
• 9c11a7ae40 [diffusion] fix: fix the frame interpolation testcase in CI regarding number of frames (#19659) Junhao LiuandMick 2026-03-04 19:21:53 -08:00
• fc53307ce9 [diffusion] hardware: SiluAndMul/RMSNorm/LayerNorm MUSA implementations (custom ops, 12/N) (#18583) R0CKSTARandQingfu Wen 2026-03-05 11:10:57 +08:00
• 9795b4cd5b [Diffusion] Open t5 encoder parallel folding for wan2.2 and mova video (#18493) Xiaoyu ZhangandMick 2026-03-05 10:18:00 +08:00
• e555a6c171 [feat] Enhance lora_update_weight_from_tensor for RL training (#19314) Ethan (Yusheng) Su 2026-03-04 18:10:42 -08:00
• d8427d0156 [NPU][CI] Cache pytorch dependency in ci (#19754) monkeyLoveding 2026-03-05 09:24:37 +08:00
• 0eb64c1e72 [smg] Extract tokenizer_path from /model_info into discovered labels (#19905) Kangyan-ZhouandClaude Opus 4.6 2026-03-04 17:12:20 -08:00
• 861d78635f [CI] remove itl testing due to unstable networking (#19904) Liangsheng Yin 2026-03-04 16:32:17 -08:00
• 43bdee703e Fix Fp8 MTP layer a2a backend without EP. (#18515) Shu WangandBaizhou Zhang 2026-03-04 18:28:10 -06:00
• 33c92732f4 [Triton] Use dynamic loop bound in alloc_extend_kernel (#19898) Liangsheng Yin 2026-03-04 16:15:58 -08:00
• a710b7d791 [Sarvam] Add inference support for Sarvam MoE LLMs (#18938) rakeshandgemini-code-assist[bot] 2026-03-05 04:58:00 +05:30
• 376dfb03f7 Fix issue 19717 by making qo_indptr uniform strided instead of packed (#19807) kpham-sgl 2026-03-04 15:27:10 -08:00
• 28c931e1a5 feat: Priority-based scheduling optimization (including default priority, preemption toggle, priority-based metrics, etc.) (#17026) zhuxinjie-nzandhnyls2002 2026-03-05 06:52:08 +08:00
• 9457c049e1 [Qwen3.5] Enable MTP spec_v2 and add test for nvidia/Qwen3.5-397B-A17B-NVFP4 (#19391) hlu1 2026-03-04 14:01:25 -08:00
• 0ee9d3c8e9 fix(grpc): send last chunk before completion during streaming (#19895) Chang Su 2026-03-04 13:21:21 -08:00
• 329817e262 [AMD] Move get_global_server_args import out of CUDA-only block to fix NameError on AMD (#19866) Bingxu Chen 2026-03-05 02:23:42 +08:00
• 1b76eb9361 [Doc] Update version references and add automation (#18409) Mohammad Miadh Angkad 2026-03-05 01:51:46 +08:00
• 44208d2adf [vlm][minicpm] support input formats of processor output and embedding (#19614) Ken J 2026-03-04 09:11:12 -08:00
• c03deb8175 Fix disagg PD bootstrap and KV transfer metrics (#19009) Kangyan-ZhouandClaude Opus 4.6 2026-03-04 09:08:10 -08:00
• 34c19a32c1 fix flaky test for test_kda_kernels (#19864) strgrb 2026-03-04 22:47:29 +08:00
• 738ebfd330 KDA: fuse qkv conv and support stride for fused_sigmoid_gating_delta_rule_update_kernel (#19506) strgrb 2026-03-04 22:45:53 +08:00
• 6910c1b281 [Feature][NPU]: add runtime support for GPTQ-quantized MoE models (#16364) 3 people 2026-03-04 21:02:19 +08:00
• c2b66d320d [HiCache] Add an env var to control transfer engine reuse (#19867) Shangming Cai 2026-03-04 20:36:32 +08:00
• e33e833d11 update model names (#19870) amote-i 2026-03-04 19:27:37 +08:00
• 88cfa6c11d [NPU]Releasing redundant memory of w13_weight and nz when the ascend_fuseep feature is enabled (#19813) chenxu214 2026-03-04 19:26:29 +08:00
• 17119a697d Optimization: Reduce the number of D2H operations (#19424) sky 2026-03-04 16:32:42 +08:00
• f07d668ba1 Remove flashinfer version argument from cu13 docker release workflow (#19862) Baizhou Zhang 2026-03-04 00:26:06 -08:00
• 52dcade4aa Fix flashinfer bump workflow (#19855) Baizhou Zhang 2026-03-04 00:09:58 -08:00
• 78ddf05afd [Fix] Install tomli in flashinfer bumping workflow (#19841) Baizhou Zhang 2026-03-03 23:12:22 -08:00
• 09fa012ba7 Fix /health regression from early prebound socket listen (#19805) Mohammad Miadh Angkad 2026-03-04 15:00:46 +08:00
• 115f879958 Helios: Real Real-Time Long Video Generation Model (#19782) Yuhao Yang 2026-03-04 14:58:04 +08:00
• c287d9b645 chore: add flashinfer version bump workflow (#19837) Baizhou ZhangandClaude 2026-03-03 22:53:34 -08:00
• 562c3ff2d0 [Feature] implement the standard multi-layer MTP for step3p5 (#18564) 3 people 2026-03-04 14:48:53 +08:00
• e9b5706545 [diffusion] feat: support torch compile for diffusers backend (#19673) DefTruth 2026-03-04 14:08:45 +08:00
• c6850ac30c [AMD] Fix Qwen3-Coder-Next: Add missing k_scale/v_scale args to extend_attention_fwd in aiter_backend (#19736) Michael 2026-03-03 22:01:08 -08:00
• 5972f97f11 Remove naive rotary forward overriding. (#19263) Jue Wang 2026-03-04 00:50:40 -05:00
• ac1f07487a Fix triton alloc extend kernel (#19780) ybyang 2026-03-04 13:01:16 +08:00
• 73bf2c5bdc [sgl]add pin_mem to remove cpu->gpu copy sync point (#19795) 3 people 2026-03-03 21:00:51 -08:00
• b7f7df7ee6 [NSA] Fix line-too-long lint in can_nsa_prefill_cp_round_robin_split (#19829) sglang-bot 2026-03-03 20:34:22 -08:00
• ca44aa25af Fix dp_attention crash when dp_size < tp_size in warmup dummy run (#19760) Yuhao Yang 2026-03-04 11:43:13 +08:00
• da9dcbc906 [diffusion] fix: fix corrupted image editing outputs in Multi-GPU SP mode for FLUX.2-klein models (#19454) Ruihang Li 2026-03-04 11:35:46 +08:00
• 6851613b93 [Bugfix] For cp: Fixed hang problem in prefix cache and kvcache support fp8 in-seq-split mode (#19656) Baidu-AIAKandvincent 2026-03-04 11:19:46 +08:00
• 88290690f0 Add kpham-sgl into CI Permission list (#19819) kpham-sgl 2026-03-03 18:55:10 -08:00
• 4348976f80 [Diffusion] Refactor diffusion benchmark/profile skill to reuse diffusion-perf skill and clarify profiling trigger (#19783) Xiaoyu Zhang 2026-03-04 10:54:42 +08:00
• 82e7139c06 [VLM] Support cos sin cache for Ernie4.5-VL (#19743) Yuan Luoandluoyuan.luo 2026-03-04 10:54:23 +08:00
• 525d046990 [AMD] CI - new runner label for MI325 8gpu (#19815) YC Tseng 2026-03-04 05:45:23 +03:00
• 115e9a1acd [Diffusion] Delete useless _ulysses_input_split func (#19786) Xiaoyu Zhang 2026-03-04 10:45:11 +08:00
• ee5ccde0ad support fused_moe_triton and moe_sum_all_reduce kernel fusion[reduce … (#19672) xieminghe1andundefined 2026-03-04 10:30:33 +08:00
• d22c6a3847 fix: Properly return abort error for streaming requests if the abort is triggered by scheduler (#19357) Charles Chen 2026-03-03 17:18:15 -08:00
• eb6bcc5c86 [CI] Register test_quant_config_parsing.py in CI suite (#19809) Alison ShaoandAlison Shao 2026-03-03 16:53:31 -08:00
• e2af840c3d Various SM120 improvements (#19721) Brayden Zhong 2026-03-03 19:46:13 -05:00
• a69b943356 [SGLang-Diffusion] Add offline throughput benchmark script for multi-modal models (#18154) 4 people 2026-03-03 16:39:46 -08:00
• c18cff4f90 [CI] Add DeepGEMM warmup to stage-c-test-deepep-4-gpu (#19806) Alison ShaoandAlison Shao 2026-03-03 16:18:12 -08:00
• 441045a7bf [AMD] Fix EAGLE3 speculative decoding with aiter attention backend (#19362) Hubert Lu 2026-03-03 16:12:13 -08:00
• e6411ba315 Increase max_concurrent_jobs in job queue (#19797) Eric Zhang 2026-03-03 15:54:51 -08:00
• c7ffbf25e9 [CI] Fix rerun-ut workflow: add DeepEP install, RDMA env, Blackwell detection (#19803) Liangsheng Yin 2026-03-03 15:17:16 -08:00
• 753da27535 [Bugfix] fix parse_lscpu_topology bug (#18520) Jiayi Yan 2026-03-04 07:15:36 +08:00
• 069c7e4188 Fix CI failures (#19303) Cao E 2026-03-04 07:04:44 +08:00
• b8c71f895e Add tuned triton==3.5.1 h200 tp2, tp4 for qwen 3 next (#15948) Yi Zhong 2026-03-03 17:47:19 -05:00
• 0c760c4cd7 Add tuned triton==3.5.1 b200 tp2, tp4 for qwen 3 next (#15917) Yi Zhong 2026-03-03 17:47:05 -05:00
• fb37c0a400 [args] Add Expert Parallelism Argument To SRT Runner (#18492) Jonah BernardandQiaolin Yu 2026-03-03 17:16:35 -05:00
• f7897def96 [Feature] Improve weight loading log (#18651) Praneth Paruchuriandgemini-code-assist[bot] 2026-03-04 03:46:13 +05:30
• 9305f0e58d Support triton_kernels for GPT-OSS on SM120 (#19718) Brayden Zhongandamittell 1388680+amittell@users.noreply.github.com 2026-03-03 17:14:01 -05:00
• 1135e214b3 [CI] support /rerun-ut command in slash handler (#19800) Liangsheng Yin 2026-03-03 14:10:49 -08:00
• 5b2e2750b5 Enable XQA for SM90 and SM120 (#17115) Sam (Kesen Li)andXiaowei Wang 2026-03-04 06:09:44 +08:00
• dc92f88a21 Enhance bench_multiturn.py with OpenAI API support and richer metrics (#19724) Kangyan-ZhouandClaude Opus 4.6 2026-03-03 13:48:04 -08:00
• f749802402 [Score API][18132] return token usage in Score API response (#18381) Guy Stone 2026-03-03 16:45:35 -05:00
• b0f26698f5 feat(benchmark script): add similar to vllm --ready-check-timeout-sec parameter (#15466) almaslof 2026-03-03 23:44:38 +02:00
• ac2819c81f Fix assertion tolerance for bf16 precision in triton attention UT (#17461) Rahul Vijayaraghavan 2026-03-04 03:13:58 +05:30
• 85ab6a7f54 cli: Add lazy imports and fail-fast config validation (RFC #9853) (#19368) Karthik Korallaandhnyls2002 2026-03-03 16:03:49 -05:00
• d6ac5f23cc [Docs] Add GDN attention backends matrix documentation (#19755) zwang86andZeyu Wang 2026-03-03 13:00:34 -08:00
• cedb86a950 Feature:Reserve HTTP server port before model loading to immediately detect port conflicts instead of failing after several minutes of model loading. (#17754) xrwang8 2026-03-04 03:59:18 +08:00
• 2e1b9e2547 Fix routed_dp_rank boundary validation (#19762) 3 people 2026-03-04 03:55:15 +08:00
• 05f68e1230 [AMD] Fix the hipDeviceGetName issue in ROCm based docker images (#19440) Hubert Luandbingxche 2026-03-03 11:42:48 -08:00
• 85f7a0aa30 feat: support Kimi K2.5 for Eagle3 (#19689) 5 people 2026-03-04 02:41:15 +08:00
• 95e4a25b17 [fix typo]: funtion -> function (1 line change) (#19790) SoluMilken 2026-03-04 00:57:02 +08:00
• daabfe7e4c [hotfix] fix apply function name compressed_tensors_w4a4_mxint4_moe (#19713) Tamir Baydasov 2026-03-03 16:50:37 +03:00
• c6377bbbca feat(gdn): add FlashInfer K-last SSM layout support for GDN prefill and decode for Hopper (#18361) 6 people 2026-03-03 20:30:48 +08:00
• d939e26585 [model gateway][0/N] router EPD support: add encoder grpc server backend support (#16552) 3 people 2026-03-03 19:38:15 +08:00
• facde4c6d3 [PD] Enable all CP ranks for KVCache transfer (#19765) Shangming Cai 2026-03-03 19:35:21 +08:00
• 365ca1edb5 [NPU] bugs fix: fix a condition bug when using speculative inference on Qwen3 and Qwen3 moe (#19532) shengzhaotian 2026-03-03 17:59:25 +08:00
• 666caaf9ce [Tool Call] Stream DeepSeek-V3.2 function call parameters in JSON format. (#16091) Muqi LiandHuixxi 2026-03-03 17:46:29 +08:00
• 4c95953b77 Fix/nemotron mtp quantaized (#19433) Shaun Kotek 2026-03-03 11:07:46 +02:00
• af0d35b224 Fix: Reject requests with a duplicate request ID which can cause server crash/hang (#19035) 3 people 2026-03-03 00:33:25 -08:00
• 6af0448cc9 [Bugfix] Catch errors when DeepSeek-V3.2 generates malformed JSON (#18174) Muqi Li 2026-03-03 16:10:07 +08:00
• 7a2d3df96f Apply default stream to priority 0 in scheduling. (#16438) Liangsheng Yin 2026-03-03 00:05:27 -08:00
• 07b8d763ef feat: Add FP8 KV cache support for Triton attention backend (#18882) Zack Yu 2026-03-02 23:38:34 -08:00
• 62480ebb1b [SGLang-Diffusion] Fix custom op fake impl missing eps default for torch.compile (#19725) 赵晨阳andXiaoyu Zhang 2026-03-02 23:24:36 -08:00
• 3c01b44700 [Fix] NPU deepep hccl buffer and fix IPC safe check (#17804) 1StepForever 2026-03-03 14:56:06 +08:00
• 43a9249e7c CI: update CI_PERMISSIONS.json (#19744) Mick 2026-03-03 14:34:41 +08:00
• dbf1247fe0 Add KimiK2Detector with tool interruption support (#19696) Xinyuan Tong 2026-03-03 06:04:49 +00:00
• 63003a39cf [BUG] Support tuple hidden_states from fused MXFP4/FP8 quantization (#19643) Yuzhen Zhou 2026-03-02 20:39:06 -08:00
• 0abb9f4176 Piecewise Cuda Graph Docs (#19738) Yuwei AnandWenyao Gao 2026-03-02 19:51:17 -08:00