Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 7790645b82 [diffusion] UX: replace deprecated ORJSONResponse with orjson_response (#21755) MickandClaude Opus 4.6 2026-03-31 21:41:33 +08:00
• 20d07c4384 Fix remote weight info nnode>1 and dp>1 (#17389) JD 2026-03-31 06:17:18 -07:00
• ca2b2130ba [PD] Tiny cleanup after KVReceiver refactor (#21760) Shangming Cai 2026-03-31 21:07:57 +08:00
• c7adca9992 Fix kimi-linear launch server error (#21752) Yuan Luoandluoyuan.luo 2026-03-31 21:07:08 +08:00
• dbc97456ad Enable evict swa with piecewise cuda graph (#21754) Ke Bao 2026-03-31 20:07:16 +08:00
• 91ab75b505 [CI] Fix ring test timeout (#21751) Ke Bao 2026-03-31 18:26:08 +08:00
• 4455d17619 [PD] Refactor Disagg Conn and Fix Hang with total_request/total_tokens Balancing (#21299) weireweireandWeiliangl User 2026-03-31 18:01:50 +08:00
• acd37d8701 [CI] Fix rerun-test suite detection to skip commented registrations (#21753) Ke Bao 2026-03-31 18:00:53 +08:00
• 6c03ae6fe2 [diffusion] fix: fix typo (#21746) R0CKSTAR 2026-03-31 17:51:46 +08:00
• a6a8b9b376 bugfix(model):fix deepstack index out of range error (#21727) xiaoqiandxiaoqi.31 2026-03-31 17:41:47 +08:00
• 2456889f98 Rename rerun-ut to rerun-test (#21747) Ke Bao 2026-03-31 17:31:55 +08:00
• 08d3c1a134 Fix disaggregation hybrid attention ci (#21745) Ke Bao 2026-03-31 16:22:54 +08:00
• d52757fe97 [CI]Remove msgm-en and mmlu tests which cause timeout (#21733) Baizhou Zhang 2026-03-31 01:10:05 -07:00
• 5628e908ae [AMD] Use tgemm.mm for MoEGate router gemm in deepseek_v2.py (#21657) Thomas Wang 2026-03-31 15:55:40 +08:00
• b4cb31f698 [NPU] fix conflict between empty_cache and use_mem_pool (#21507) xiazhahe 2026-03-31 15:37:33 +08:00
• dd9c9c1b8e Add explicit disable flag for FlashInfer allreduce fusion (#21446) Mohammad Miadh Angkad 2026-03-31 15:15:44 +08:00
• 68a4573627 [diffusion] fix: fix Flux.2 with tp(#21664) Yuhao Yang 2026-03-31 14:14:59 +08:00
• 8ba992411d [AMD] Fix CI multimodal-gen-test-1-gpu-amd for gen model (#21621) jacky.cheng 2026-03-31 14:02:20 +08:00
• 03e4f2858d [Perf]Remove H2D for Qwen3.5 SpecV2 (#20864) Jincong Chen 2026-03-31 11:54:58 +08:00
• 33e725b052 [Fix] Update supported custom_mem_pool types for mooncake (#21728) Lewisand百麒 2026-03-31 11:18:30 +08:00
• 505eb312ec Revert "DeepSeek-R1-0528-w4a8: DeepEP Low Latency Dispatch Adopts FP8 Communication" (#21719) Xiaoyu Zhang 2026-03-31 10:22:01 +08:00
• 9b6bee2d40 Fix human-eval CI install on 5090 runners (#21714) Alison ShaoandAlison Shao 2026-03-30 18:53:09 -07:00
• 4e480982fa [misc] multiprocess compilation to speed up test (#21483) DarkSharpness 2026-03-31 08:56:37 +08:00
• 3650bfb199 Remove flashinfer wheel cache cleanup that deletes other versions (#21711) Alison ShaoandAlison Shao 2026-03-30 16:47:04 -07:00
• 67c295b5f5 [AMD] fix performance regression issue when run gpt-oss with "--context-length 13824" (#21691) kk 2026-03-31 07:30:16 +08:00
• 4b8456e266 [AMD][MoRI] bump MoRI to v0.1.0 (#21673) jhchouuu 2026-03-31 05:44:11 +08:00
• daf697afda [AMD] Add SGLANG_DISAGGREGATION_NUM_PRE_ALLOCATE_REQS env var for configurable KV transfer overlap (#20410) Zhai FeiyueandHaiShaw 2026-03-31 05:37:16 +08:00
• d6029de6ad [Bugfix][NPU] Skip FRACTAL_NZ format for MoE weights with unaligned dimensions (#21209) 3 people 2026-03-31 01:52:17 +05:30
• 4a9ffc3ab6 fix nemotron capture for non attention layers (#21436) Vedant V Jhaveri 2026-03-30 12:50:49 -07:00
• ad064c2f4e [GLM-V and GLM-4.7] Cast to FP32 before gate projection for GLM model. (#21660) Yuxuan Zhang 2026-03-31 03:25:27 +08:00
• a20d12ae96 [diffusion][doc]: add ring sp performance benchmark page (#20998) yuefeng Wuandgemini-code-assist[bot] 2026-03-31 01:26:05 +08:00
• f4b0e9c64a [diffusion] [NPU] support ring attention on NPU with FA (#21383) Makcum888e 2026-03-30 20:10:55 +03:00
• 752d260c77 [NPU][diffusion]: support parallel decoding of qwen-image (#20757) GXINand高鑫 2026-03-31 01:03:24 +08:00
• ba6d54d0f0 [NPU] GLM-5 optimize with fused kernels (#18617) cen121212 2026-03-30 22:48:15 +08:00
• 7119d59747 DeepSeek-R1-0528-w4a8: DeepEP Low Latency Dispatch Adopts FP8 Communication (#14162) xieminghe1andundefined 2026-03-30 22:27:28 +08:00
• 673ffb3116 [NPU] fix eagle3 accept rate (#21255) heziiop 2026-03-30 21:58:25 +08:00
• c5c58c3349 [NPU][Diffusion] fix sp modulate for qwen-image-edit (#20974) GXINand高鑫 2026-03-30 21:18:48 +08:00
• 0a1fb42869 [diffusion] CI: relax pr-test threshold (#21682) Mick 2026-03-30 20:23:46 +08:00
• b76730701b [diffusion] feat: enhance overlay mechanism (#21648) Mick 2026-03-30 19:45:34 +08:00
• 1d6424d5ad fix: Mistral Small 4 fails to start due to config/weight format mismatch (#21620) 4 people 2026-03-30 16:57:35 +08:00
• b246269444 fix mamba cache leak when adder fails to add a matched req. (#21404) strgrb 2026-03-30 16:45:49 +08:00
• 62a63eeff7 [Fix] Fix weight_loader property assignment for qwen3-next FP8 models (#21662) Baizhou ZhangandClaude Opus 4.6 2026-03-30 01:35:59 -07:00
• e6071e60c0 [AMD] Support AMD MXFP4 Qwen3.5-397B-A17B model (#21234) Hubert Lu 2026-03-30 01:14:18 -07:00
• 965f03cdc2 [NPU] Update DeepSeek-V3.2 model deployment instructions in documentation (#21468) Michelle Wuandwuxue 2026-03-30 15:51:42 +08:00
• b9a68c304e [AMD] Fused rope kv store (#21315) kkandwunhuang 2026-03-30 15:05:41 +08:00
• af62bd9486 [CPU] Implement MXFP4 Gemm kernels for intel AMX to support GPT OSS series. (#14385) Ma Mingfei 2026-03-30 14:44:12 +08:00
• ed01e1d5d6 [CPU] add kernel apply_rotary_pos_emb_cpu for Qwen3-VL and Qwen3-Omni (#13121) blzhengandMa Mingfei 2026-03-30 14:43:46 +08:00
• 6da8f5f69e fix topk softmax performance issue (#14702) Ma Mingfei 2026-03-30 14:43:16 +08:00
• c32ee48886 MFU metrics in Prometheus (#19395) Aishwarya Ramasethu 2026-03-30 02:40:06 -04:00
• 1a4b383fac [CI] [FlashInfer v0.6.7] Use offline quantized checkpoint for MXFP8 Gemm tests (#21625) Ziang Li 2026-03-29 22:47:46 -07:00
• 5b19c9a05d [Doc] Update tips for developer new-comers (#21659) Baizhou Zhangandgemini-code-assist[bot] 2026-03-29 22:40:36 -07:00
• f0303fd07e [Intel GPU] Enable DeepSeek R1 inference on XPU (#18461) Polisetty V R K Jyothendra Varma 2026-03-30 11:05:59 +05:30
• d8ab41dce5 [Fix] Handle pre-release tags in nightly wheel version parsing (#21656) Kangyan-ZhouandClaude Opus 4.6 2026-03-29 22:29:40 -07:00
• 90bdc3192b Update sponsorship details in README.md (#21658) Ying Sheng 2026-03-29 21:42:59 -07:00
• db5d9eb8ce [diffusion] CI: fix dashboard chart (nightly) display issues (#21653) 3 people 2026-03-30 12:02:01 +08:00
• 9b4dd27478 [Fix] Fix Qwen3.5 MoE model loading and Mamba cache sharding in PP mode (#21448) Feng Suandzhangxiaolei123456 2026-03-30 11:57:26 +08:00
• c06ca1526c Fix circular reference in CustomTestCase.__init_subclass__ (#21650) Liangsheng Yinandwan4ch 2026-03-29 20:38:12 -07:00
• afb32d7622 README: coding agent sponsorship for long-term contributors (#21642) Lianmin Zheng 2026-03-29 16:02:51 -07:00
• 9f7792415a Clean up TokenizerManager: remove dead code and improve rid validation (#21639) Lianmin ZhengandClaude Opus 4.6 2026-03-29 15:12:49 -07:00
• f3970b17ef [Cleanup] Remove unused BatchMultimodalOutput and BatchMultimodalDecodeReq (#21640) Lianmin ZhengandClaude Opus 4.6 2026-03-29 14:54:25 -07:00
• 1d9c8e8c9e Simplify routed experts test and move base64 encoding to tokenizer manager (#21634) Lianmin ZhengandClaude Opus 4.6 2026-03-29 12:44:01 -07:00
• 2acdda1d85 [Fix] Remove redundant allreduce fusion block and skip TP=1 (#20621) Mohammad Miadh Angkad 2026-03-30 03:30:40 +08:00
• bda94fc779 [Fix] SGLANG_USE_CUDA_IPC_TRANSPORT=1 and SGLANG_ENABLE_MM_SPLITTING=1 do not work at the same time. (#19915) wili 2026-03-30 01:15:26 +08:00
• d2440dcf58 [VLM] perf: optimize CUDA IPC for multimodal transfer by caching IPC pool handles (#21418) saatwiknagpal 2026-03-29 09:20:38 -07:00
• 5bb9ca0e63 [Feature] Optimizations for JPEG input on NVIDIA GPU (#19749) wili 2026-03-30 00:06:14 +08:00
• 42c46e6334 [sgl] disable piecewise cuda graph when a model doesn't have layers (#21565) Bi Xue 2026-03-29 08:04:20 -07:00
• aa9177152e fix cuda graph capturing error in sm120 mxfp8 triton path (#19835) Hanlin Bi 2026-03-29 04:59:24 -04:00
• fec9961a1f Clean up _wait_for_scheduler_ready implementation (#21626) Liangsheng Yin 2026-03-29 01:02:33 -07:00
• c34593f951 [HiCache] fix: graceful shutdown of pending async tasks in bench_mix.py (#20276) shuwenn 2026-03-29 15:46:32 +08:00
• d2fa8d67ba Wrap IPv6 addresses in gRPC, bench_serving, and log messages (#21236) 3 people 2026-03-29 00:36:31 -07:00
• 18074e25dc fix: scheduler launch hang when non-current rank dies (#20287) shuwenn 2026-03-29 15:28:45 +08:00
• 22e4733ab9 Add subprocess liveness monitor to detect scheduler crashes (#18582) 3 people 2026-03-29 15:09:13 +08:00
• 35f5a0ff35 [CI] Lossen test_return_routed_experts threshold (#21270) Junrong Lin 2026-03-29 13:04:53 +08:00
• 9d64a82173 feat(ci): add GB300 nightly benchmark test suites (#21487) Kangyan-ZhouandClaude Opus 4.6 2026-03-28 21:54:03 -07:00
• 166e9090ee [CI] Skip flaky elastic EP test (#21619) Shangming Cai 2026-03-29 12:50:40 +08:00
• ba6b501f3a Clean up detokenizer and remove dead multimodal_gen code (#21588) Lianmin ZhengandClaude Opus 4.6 2026-03-28 21:44:40 -07:00
• 516cff97a3 [Diffusion] Align diffusion benchmark skill presets with nightly comparison cases (#21616) Xiaoyu Zhangandgemini-code-assist[bot] 2026-03-29 12:12:17 +08:00
• 343a7ac652 [GDN] Fuse GDN kkt + solve_tril into one kernel (#21411) Yuan Luoandluoyuan.luo 2026-03-29 12:02:07 +08:00
• c86f6c2831 [AMD] Add peft>=0.18.0 to diffusion_hip deps for transformers 5.x compat for AMD diffusion model (#21442) jacky.chengandHaiShaw 2026-03-29 11:28:05 +08:00
• 4e69f14b95 fix bench_serving sglang backend to support image dataset (#21294) Yuhao Yang 2026-03-29 10:02:11 +08:00
• 3ab9afd653 fix: piecewise_cuda_graph get correct qo_indptr (#21452) eigenandAvery Huang 2026-03-28 18:57:29 -04:00
• efebcab43e Support skip-softmax attention (#19089) Shu Wang 2026-03-28 17:55:48 -05:00
• 40ff652862 Skip ci for .md files (#21482) Артем Савкин 2026-03-28 23:00:17 +03:00
• 27071e0a43 [NPU] Update quantization&CI documentation (#21100) Артем СавкинandTamir Baydasov 2026-03-28 21:42:21 +03:00
• ced69c9f84 feat: enable CUDA graph and timestamp for the whisper model(#21190) Xinyuan Tong 2026-03-28 17:46:03 +00:00
• 57cf4790ca [VLM] Optimize ShmPointerMMData for multi-pickle safety and deferred unwrap (#21465) Yuhao Yang 2026-03-28 23:11:12 +08:00
• fc9de157f9 [diffusion] feat: support overlay model materialization (#21600) Mick 2026-03-28 23:02:38 +08:00
• ee15c104ef [CI] hot-fix ci lint (#21608) Yuan Luo 2026-03-28 21:32:39 +08:00
• 627e162335 [diffusion] fix: fix Flux2-Klein prompt tokenization length to 512 and add regression coverage (#21407) Aditya Sharma 2026-03-28 14:58:02 +05:30
• edd4d54023 [Clean] Remove deprecated environs (#21536) Baizhou Zhang 2026-03-28 00:35:44 -07:00
• 7078e385ea [AMD] Add GLM-4.7-FP8 accuracy CI test for MI35x (#21534) Jacob0226andClaude Opus 4.6 2026-03-28 15:28:56 +08:00
• 6ef4318ec0 [CI] Move v32 cp test to deepep running suite (#21585) Baizhou Zhang 2026-03-27 22:49:06 -07:00
• 402628e560 Patch transformers is_base_mistral in CI to avoid HF 429 rate limiting (#21586) Liangsheng Yin 2026-03-27 22:19:36 -07:00
• 33cca495ae [CI] Replace upload/download-artifact with job outputs in release-docker workflow (#21579) Kangyan-ZhouandClaude Opus 4.6 2026-03-27 22:12:55 -07:00
• daf02bde33 Fix Piecewise CUDA Graph crash with -enable-mixed-chunk (#20441) Jianyingandjianyingzhu 2026-03-28 12:56:21 +08:00
• 19b1f75186 Fix HFRunner hang when subprocess dies during init (#21582) Liangsheng Yin 2026-03-27 21:22:42 -07:00
• 5ef56682b8 reduce CPU peak memory in multimodal tensor hashing (#21123) Yuhao Yang 2026-03-28 11:09:16 +08:00
• 588320262e Update CODEOWNERS for transformers.py and docs (#21555) Adarsh ShirawalmathandLianmin Zheng 2026-03-28 08:37:53 +05:30
• 9fa7b974fd [diffusion] chore: remove redundant identity preprocess_text functions(#20633) Fengyuan YuandFengyuan Yu 2026-03-28 10:07:30 +08:00
• e570ca96f6 [diffusion] refactor: Unify TeaCacheParams and WanTeaCacheParams (#20706) Eitan TurokandMick 2026-03-28 04:51:44 +03:00