Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 213027951a [AMD] Upgrade Aiter (#22264) Bingxu Chen 2026-04-11 09:40:43 +08:00
• 265696b176 chore: update CI test est_time values (#22565) 4 people 2026-04-10 18:15:02 -07:00
• b5e4ae7b1a fix: match est_time updates by backend, not just suite (#22563) Cheng Wan 2026-04-10 17:54:50 -07:00
• 75223c5404 [Diffusion][CI] Fix nunchaku unit test broken by #22365 (#22560) Alison ShaoandAlison Shao 2026-04-10 19:49:56 -05:00
• 0011d2aec0 fix: track est_time per suite instead of per backend (#22557) Cheng Wan 2026-04-10 16:58:40 -07:00
• b4a1d8fd71 [mem] Fix idle token_usage missing mamba_usage; add FIXME for naming (#22555) Liangsheng Yin 2026-04-10 16:20:33 -07:00
• 0af9166474 [tokenizer] improve non streaming request processing + some small fixes. (#20310) Alex Nails 2026-04-10 15:46:12 -07:00
• 451320596f [CI] Add GB200 nightly perf regression pipeline (#22461) Sahithi Chigurupati 2026-04-10 15:12:24 -07:00
• 3f39b3d811 feat: add weekly workflow to update CI test est_time values (#22545) Cheng WanandClaude Opus 4.6 2026-04-10 15:03:37 -07:00
• f7a1740101 [MUSA][9/N] Add FA3 attention backend support through MATE (MUSA AI Tensor Engine) (#22051) oriandzhiguo.qin 2026-04-11 05:18:39 +08:00
• 6af34b95b6 perf: precompute FA3 scheduler_metadata to eliminate per-layer prepare_varlen_num_blocks (#21104) Minglei Zhuandzminglei 2026-04-10 13:57:54 -07:00
• 4ace144fae feat: update ModelExpress metadata API to SourceIdentity-based schema (#21222) Zhongdongming Dai 2026-04-10 13:45:05 -07:00
• 6d8330bdb7 Update CI_PERMISSIONS.json (#22465) satyamk7054andSatyam Kumar 2026-04-10 13:43:50 -07:00
• 6d95602ea3 Reduce GPU memory for MoE parallel groups (#22515) Cheng WanandClaude Opus 4.6 2026-04-10 13:23:23 -07:00
• 059b287e25 Add offline auto-tuning for LoRA CSGMV kernel (#20391) satyamk7054andSatyam Kumar 2026-04-10 13:10:43 -07:00
• d8831355a3 Fix multi_layer_eagle_worker_v2 draft extend selection, add chain style multi layer mtp test (#22340) Qiaolin Yuand0xNullPath 2026-04-10 12:44:52 -07:00
• 7dbd0dd9f0 MiniMax-M2.5 - Support dp attention, dp reduce scatter, FP4 all gather, AR fusion in prepare_attn (#20067) Trevor Morris 2026-04-10 12:41:27 -07:00
• a937ec31be fix: server crash when stop_token_ids contains null (#22175) KrishnanPrash 2026-04-10 11:42:23 -07:00
• 5cb4ea1d4d perf: enable inductor combo_kernels for horizontal fusion (#21977) Jia GuoandClaude Opus 4.6 2026-04-10 10:01:14 -07:00
• 2ba94136ce [sgl] improve mamba_track_indices perf in specdec (#22380) Tarushii Goel 2026-04-10 09:39:53 -07:00
• f652135d52 [sgl] fix using symmetric memory issues for attention_tp (#22286) Bi Xue 2026-04-10 09:26:18 -07:00
• 8227187d47 [SKILL]: add component accuracy guidance to the diffusion add-model skill (#22460) Ratish P 2026-04-10 20:38:31 +05:30
• cf5ad12612 [diffusion][CI]: route multimodal component accuracy through run_suite (#21960) Ratish P 2026-04-10 20:36:03 +05:30
• 84194c25c1 [BugFix] fix the bug of minimax_m2.5 model that causes repeated outputs when using tp16 (#20967) kingkingleeljjandClaude Opus 4.6 2026-04-10 22:21:19 +08:00
• 1ff51555f2 [Diffusion] modelopt diffusion fp8 support for flux1/flux2 and wan2.2 (#22365) Xiaoyu Zhang 2026-04-10 20:56:57 +08:00
• 8ba9646044 Make GDN support non-continuous B/A Tensor input to fix the accuracy regression of Qwen3.5-27B (#22312) Yujun Dong 2026-04-10 18:58:13 +08:00
• 0668a7f51a [Perf] Remove two operations in gdn_backend extend verify path (#22444) Jincong Chen 2026-04-10 17:53:57 +08:00
• 1c76f322df [HiCache] Add CP support for HiCache (#20977) Shangming Cai 2026-04-10 17:52:51 +08:00
• 37107bee6f [Observability] Add pending token count to prefill log and get_load (#22480) Cheng WanandClaude Opus 4.6 2026-04-10 02:05:21 -07:00
• c554dc5c64 Add dedicated FlashInferCuteDslMoE layer for standard-path FP4 MoE (#21339) Lee Nau 2026-04-10 01:35:56 -07:00
• 7c6b5c095c [diffusion] fix: fix flux2 i2i accuracy (#22423) Mick 2026-04-10 16:16:51 +08:00
• 6cf7f210bf Add page_size to admission token budget check (#22495) Liangsheng Yin 2026-04-10 01:16:04 -07:00
• dd41764487 [AMD][HIP] NSA: bf16 passthrough from RMSNorm to eliminate FP8 dequantization (#22258) Jacob0226andClaude Opus 4.6 2026-04-10 16:08:32 +08:00
• f5fd5ab622 add whisper test (#22302) Yuhao Yang 2026-04-10 15:34:53 +08:00
• 2ab141547d [CPU] Add apply_routed_scaling_factor_on_output support for biased_grouped_topk fusion (#22413) jianan-guandMa Mingfei 2026-04-10 15:16:05 +08:00
• 599cce4d82 [Intel GPU] import flash_attn functions from sgl_kernel only (#22438) Polisetty V R K Jyothendra Varma 2026-04-10 12:40:00 +05:30
• 5ba7d4e523 [HiSparse]: Update HiSparse's user-guide (#22499) 3 people 2026-04-10 15:06:43 +08:00
• 18f41ac427 [Reland] DeepSeek-R1-0528-w4a8: DeepEP Low Latency Dispatch Adopts FP8 Communication (#22316) 3 people 2026-04-10 14:56:05 +08:00
• 0334d4b7e8 [sgl] Fix mamba tracking calculation in spec dec (#22239) Tarushii Goel 2026-04-09 23:46:16 -07:00
• 6d79c60995 [Lora] Lora kimi support (#22381) Ethan (Yusheng) Su 2026-04-09 22:31:53 -07:00
• 722e25a621 Fix SWA eviction boundary and page-align chunked prefill (#22470) Liangsheng Yin 2026-04-09 22:09:43 -07:00
• e77bfba24d Fix NCCL AllGather hanging issue for Qwen3 Next MTP (#22458) Ke Bao 2026-04-10 11:40:54 +08:00
• b853e2c41b [CI] Remove Slack notification from ci-auto-bisect workflow (#22483) Alison ShaoandAlison Shao 2026-04-09 22:32:09 -05:00
• 45b0182205 [CI] Update est_time for 64 tests based on actual elapsed times (#22305) 3 people 2026-04-09 22:31:37 -05:00
• 89553ff82b [Observability] Add Prometheus metrics endpoint for gRPC mode (#20801) Kangyan-ZhouandClaude Opus 4.6 2026-04-09 20:04:54 -07:00
• 42ffb168b3 [EPD][VLM] Support Kimi K25 EPD (#22269) LHXuuu 2026-04-10 10:58:42 +08:00
• 4644d28213 [sgl-kernel/cpu] fix build error on non-x86 platform (#22245) Yibo CaiandMa Mingfei 2026-04-10 09:58:07 +08:00
• 3aaaf53f59 [Docker] Fix CI docker target after Dockerfile restructure (#22478) ishandhanani 2026-04-09 20:53:42 -05:00
• d283808457 [AMD] Replace triton rotary_emb with aiter rotary_emb for Wan2.2 denoise (#22422) jacky.cheng 2026-04-10 09:21:02 +08:00
• 5638d40f3a [nvidia] Gemma4 nvfp4 fix (#22079) Shu Wang 2026-04-09 19:44:34 -05:00
• cebd9c2a1e [sgl] add ability to return logprobs in MultiLayerEagleWorkerV2 (#22241) Tarushii Goel 2026-04-09 16:20:55 -07:00
• aa103eab8d [Docker] Optimize Dockerfile for BuildKit layer caching (#22160) ishandhanani 2026-04-09 17:34:57 -05:00
• c3833ba929 Enable DFLASH support for additional model backends (#22358) Mohammad Miadh AngkadandDavid Wang 2026-04-10 05:36:12 +08:00
• 28ef6de091 [Lora] Lora quat info re-factor and support deepseekv3 mla lora (#22323) Ethan (Yusheng) Su 2026-04-09 14:19:58 -07:00
• 60acdc31f2 [Fix] Fix several bugs on DSA models (#22430) Baizhou Zhang 2026-04-09 12:46:23 -07:00
• 606aa11ea8 [DSA] Enable all reduce fusion for DSA models (#22390) Baizhou Zhang 2026-04-09 12:42:44 -07:00
• 9d905efa2c [Docker] Fix Trivy CVEs, cubin download 403s, and kernels command order (#22322) Kangyan-ZhouandClaude Opus 4.6 2026-04-09 12:26:22 -07:00
• 8eb235ab51 fix: do not strip whitespace from GLM tool call values (#20543) Lawrence WuandKangyan-Zhou 2026-04-09 11:14:15 -07:00
• 8b991d98a1 [feature] asr: add chunk-based streaming ASR for Qwen3-ASR (#22089) Lishan H 2026-04-09 13:49:03 -04:00
• 021bd77c91 Add skills for debugging hanging issues (#22463) Ke Bao 2026-04-10 01:37:44 +08:00
• 1df9f4e2f6 [AMD] Add prealloc token env for mori-ep (#22329) billishyahao 2026-04-10 00:34:35 +08:00
• 8216b921a1 Add MLX profiling to bench_one_batch.py (#22159) Jonah Bernard 2026-04-09 08:45:21 -04:00
• 7603b226ce Upgrade sglang-torch-profiler-analysis SKILLS (#22440) Xiaoyu Zhang 2026-04-09 18:23:03 +08:00
• 9fed58805f [Doc] Clarify SWA HybridSWAPoolConfigurator comments on all-SWA vs hybrid semantics (#22443) Liangsheng Yin 2026-04-09 03:02:16 -07:00
• 8a67fb20ea [Speculative] Support penalty for spec v2 overlap scheduling (#22049) YMbmzy 2026-04-09 16:59:04 +08:00
• 628df31d08 [AMD] Use aiter CK layernorm2d for LayerNorm to reduce NSA indexer kernel launches (#22424) Thomas Wang 2026-04-09 16:55:29 +08:00
• 57ffc55fb6 feat: [1/2] [DeepEP] Fuse shared expert into MoE dispatch under EP (#20089) 3 people 2026-04-09 16:48:28 +08:00
• 7965573eb4 fix issues for npu docs (#22307) amote-i 2026-04-09 16:27:34 +08:00
• 8ec0934f8f [NPU]add Qwen3-32b and Qwen3-8b low latency md (#22429) Liwansi 2026-04-09 16:18:34 +08:00
• 19bbaeb3ee [HiSparse]: Add HiSpares-DSA Model's nightly CI (#22425) 3 people 2026-04-09 16:00:55 +08:00
• a64905a7b8 [CICD] [prefill-only] Consolidate prefill-only model E2E tests (#22405) Sundara Raman Ramachandran 2026-04-09 00:54:34 -07:00
• 9709192ce9 [diffusion] feat: support FLUX.2-small-decoder (#22414) Mick 2026-04-09 15:53:14 +08:00
• 8ff01d6841 [Test] Add CPU unit tests for MemoryPoolConfigurator (#22420) Liangsheng YinandKe Bao 2026-04-09 00:39:19 -07:00
• de441ac6bb [core] Introduce MemoryPoolConfigurator class hierarchy (#22389) Liangsheng Yin 2026-04-09 00:29:19 -07:00
• b9c316917b fix AttributeError: 'LazyValue' object has no attribute 'keys' in eplb_manager.py for qwen3 moe (#21822) Evgueni Petrov 2026-04-09 14:13:29 +07:00
• ef6bfc1197 [AMD] Add GLM-5.1-FP8 nightly accuracy and performance benchmarks for MI30x and MI35x (#22336) Michael 2026-04-08 22:57:43 -07:00
• e379befbac Add symmetric debug mode to print stack trace of comm ops with unregistered tensors (#18569) Nicolas Castet 2026-04-09 00:34:58 -05:00
• 6b96f8341d [AMD] Fix multimodal diffusion test crash on ROCm by falling back to SDPA (#22335) Bingxu Chen 2026-04-09 13:32:49 +08:00
• 30b738d3a6 [SKILL] add torch profiler analysis workflow (#22353) Xiaoyu Zhang 2026-04-09 12:53:48 +08:00
• edfddda192 Move runai model loader test to nightly suite (#22418) Liangsheng Yin 2026-04-08 21:39:32 -07:00
• 355fcbcc17 [diffusion] fix: fix cache dit refresh none mask (#22374) Mick 2026-04-09 11:58:24 +08:00
• 6838a23226 [Feature] Add token embedding overrides for sparse embedding replacement (#20960) jsheng_LinkedinandClaude Opus 4.6 2026-04-08 20:51:36 -07:00
• a69be2e866 [Feature] Support eagle3 for qwen3-vl (#22230) Kurkur 2026-04-09 11:45:36 +08:00
• ddc8ef1038 Lazy import flash_attention_v4 to avoid loading flash_attn.cute at startup (#22306) Lianmin Zheng 2026-04-08 20:40:25 -07:00
• f127d67823 [Spec][Ngram] Misc enhance support for multiple SAMs (#22294) Khoa Pham 2026-04-08 19:56:23 -07:00
• c431b11d8b [CI] Use UV to improve pip install speed (#22029) tfhddd 2026-04-09 09:18:32 +08:00
• 1b7c33a5b7 [diffusion] rl: revamp rollout Log-Prob support with SDE/CPS for RL post-training (#21204) Kangrui DuandMikukuOvO 2026-04-08 18:00:00 -07:00
• 2c4e113dd7 [CI] Fast-fail on lint check failure in check-stage-health (#22400) Liangsheng Yin 2026-04-08 17:17:07 -07:00
• 46c2b77627 [CI] Add GLM-5.1 nightly tests and update Qwen3.5 model (#22399) Kangyan-ZhouandClaude Opus 4.6 2026-04-08 17:04:57 -07:00
• cf27b11498 [CI] Increase stage-c-test-4-gpu-b200 partitions from 4 to 5 (#22395) Alison ShaoandAlison Shao 2026-04-08 16:36:27 -07:00
• 614dd1d4f5 [CI] Add alexnails to CI_PERMISSIONS.json (#22391) Alison ShaoandAlison Shao 2026-04-08 16:22:05 -07:00
• cc8ea08b8b [CI] Replace upload/download-artifact with job outputs in release-docker-runtime (#22388) Kangyan-ZhouandClaude Opus 4.6 2026-04-08 16:13:33 -07:00
• 1e3f6ebea6 [core] Extract pool sizing logic to pool_configurator.py (#22384) Liangsheng Yin 2026-04-08 16:13:21 -07:00
• e41647f52b [CI] Add pre-commit hook to validate test/registered/ files have CI registry (#22308) 3 people 2026-04-08 15:59:15 -07:00
• 5e1a9834f9 Update ci permission (#22387) Baizhou Zhang 2026-04-08 15:42:06 -07:00
• 4e5b8cb041 Fix get_version_tag.py to handle dot-separated post versions (#22385) Baizhou ZhangandClaude Opus 4.6 2026-04-08 15:18:22 -07:00
• df3275bd6c chore: bump flashinfer version to 0.6.7.post3 (#22382) sglang-botandsglang-bot 2026-04-08 14:49:45 -07:00
• c89afaea7c Fix hybrid_linear_attn_backend crash with ngram speculation (#20739) 3 people 2026-04-09 03:52:07 +08:00
• c26b8b4a4b [GDN] Remove FlashInfer GDN decode + no_buffer guard and default to FlashInfer on SM100+ (#21861) YAMY 2026-04-08 11:59:15 -07:00
• db30a63a13 [sgl-kernel] support > 1024 experts in moe_align_block_size kernel (#21610) Kurt Shuster 2026-04-08 14:45:13 -04:00