Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 3f0ba021fc [AMD] Improve openai/gpt-oss performance (#21020) 5 people 2026-03-21 14:16:47 +08:00
• 67cad3e69e Revert "Support CuteDSL mm_fp4 backend" (#21077) Baizhou Zhang 2026-03-20 22:47:47 -07:00
• c076968c52 [CI] Remove obsolete AOT-only jit-kernel benchmarks after sgl-kernel 4.0 (#21075) Xiaoyu Zhang 2026-03-21 13:40:42 +08:00
• 5f3393c04c Fix deepseek-v32-fp4 b200 ci (#21072) Baizhou Zhang 2026-03-20 22:28:40 -07:00
• 9614271ae4 [Test] Fix flaky multi-instance memory occupation test (#21074) Liangsheng Yin 2026-03-20 22:19:03 -07:00
• 427442304a Add CI maintenance mode gate to nightly-test-nvidia.yml (#21073) Lianmin Zheng 2026-03-20 22:07:34 -07:00
• 50404e0d1f ci: refactor CUDA dependency install script (#21017) Lianmin Zheng 2026-03-20 21:36:58 -07:00
• d5473f87c4 Add CI maintenance mode gate to pr-test.yml (#21069) Lianmin Zheng 2026-03-20 21:27:38 -07:00
• 048d90e165 Revert "[AMD] Add MoE weights and scales padding" (#21067) Alison Shao 2026-03-20 20:26:17 -07:00
• a75e74b534 Add revocation of permissions section (#21066) Lianmin Zheng 2026-03-20 20:22:24 -07:00
• 3798a8c88d docs: add out-of-tree model integration guide (#21050) 3 people 2026-03-20 20:07:46 -07:00
• 6c91590e1b [HiCache] refactor: hicache normalization flow and compatibility checks (#19669) shuwenn 2026-03-21 09:38:44 +08:00
• 9419453713 [AMD] Add MoE weights and scales padding (#18684) mqhc2020 2026-03-21 05:55:09 +08:00
• f97c09dac1 [AMD] Enable aiter unified attention for non-SWA models (Qwen3-VL) (#20897) YC Yen-Ching Tsengandwunhuang 2026-03-21 03:07:41 +08:00
• 576e397b6e [Test] Add unit tests for srt/sampling (#20891) Zijun Gao 2026-03-20 10:38:48 -05:00
• 146700db68 Add e2e demo test in dump comparator (#21031) fzyzcjy 2026-03-20 22:41:01 +08:00
• 6703cc4484 Enhance output formatting in dump comparator (#21029) fzyzcjy 2026-03-20 22:04:50 +08:00
• fdbcb8156e Refactor dp_utils to use ParallelAxis enum in dump comparator (#21028) fzyzcjy 2026-03-20 22:04:20 +08:00
• 154395ab7d Support s≡t dimension name equivalence in dump comparator (#21027) fzyzcjy 2026-03-20 22:03:34 +08:00
• cc22601d28 Validate replicated axes orthogonality in dump comparator (#21026) fzyzcjy 2026-03-20 22:02:40 +08:00
• 2f01950a0e Support jointly-determined axes inference in dump comparator (#21025) fzyzcjy 2026-03-20 22:01:26 +08:00
• ecd7e40d20 Support dependent axis auto-resolution in dump comparator (#21024) fzyzcjy 2026-03-20 21:56:39 +08:00
• 2d7a262ca3 ci: rename 1/2-gpu-runner labels to 1/2-gpu-h100 (#21008) Lianmin Zheng 2026-03-20 06:04:15 -07:00
• 104b10f70a refactor: consolidate is_in_ci (jit_kernel, sgl-kernel benchmarks, tests) (#21009) Lianmin Zheng 2026-03-20 05:55:36 -07:00
• 9fbe6800aa [NPU] [Diffusion] Update CI performance baseline for Wan2.2-T2V-A14B-Diffusers-w8a8 (#20997) Артем Савкин 2026-03-20 15:54:12 +03:00
• f41832795e Add compile-time 256-bit vector guard for pre-Blackwell (#19794) xingsy97 2026-03-20 18:25:12 +08:00
• 2dd9196079 [JIT Kernel][Feature] Support JIT custom all reduce (rewrite as v2) (#19880) DarkSharpnessandXiaoyu Zhang 2026-03-20 18:24:07 +08:00
• 2099943a49 Fix scale_step_k computation in the fp8_kernel (#20819) Muqi LiandXiaoyu Zhang 2026-03-20 18:09:31 +08:00
• ec01ef9092 Fix torch.compile/dynamo crash with Qwen3 QK-norm in piecewise CUDA g… (#19818) Jia GuoandClaude Opus 4.6 2026-03-20 03:05:09 -07:00
• c1da420799 ci: run Stage A CUDA tests as stage-a-test-small-1-gpu on 5090 (#20988) Lianmin Zheng 2026-03-20 02:55:16 -07:00
• fa89d152c0 [diffusion] CI: fix hunyuan3d JIT cache (#20773) Prozac614anddaiweitao 2026-03-20 17:51:55 +08:00
• a0a4dae67f Revert "Fix DeepSeek V32 FP4 test" (#21003) Lianmin Zheng 2026-03-20 02:19:28 -07:00
• 112b628227 Replace _resolve_future_token_ids with JIT kernel + platform dispatch (#20976) Lianmin ZhengandClaude Opus 4.6 2026-03-20 01:47:03 -07:00
• cf60c5bd15 [CI] Fix jit_kernel benchmark ci (#20990) Xiaoyu Zhang 2026-03-20 16:40:20 +08:00
• c82d20d48e Fix DeepSeek V32 FP4 test (#20984) Baizhou Zhang 2026-03-20 01:04:32 -07:00
• 26f709e97d misc: make prefill-delayer compatible with multiple types of mem pool (#20979) Yilong Zhao 2026-03-20 00:05:53 -07:00
• 193bbf9b66 chore(ci): remove deprecated CI Monitor workflow (#20993) Lianmin Zheng 2026-03-20 00:05:22 -07:00
• cea79530f8 ci: merge CPU default and stage-a-cpu-only suites (#20992) Lianmin Zheng 2026-03-20 00:04:55 -07:00
• 95327458ee misc: add BatchTokenizerReq hook into dp controller (#20981) Yilong Zhao 2026-03-19 23:59:53 -07:00
• 3ebe6e9894 ci: remove redundant RUNNER_LABELS from pr-test.yml (#20991) Lianmin Zheng 2026-03-19 23:59:47 -07:00
• 712a48c5d2 ci: move metrics scripts under scripts/ci/utils (#20986) Lianmin Zheng 2026-03-19 23:47:57 -07:00
• 0be88ac44c ci: rename pr-test workflow_call input ref to git_ref Lianmin Zheng 2026-03-19 23:45:06 -07:00
• 46a76af97b [Bugifx] qwen3 rope parameter compatibility (#20931) lviyandXinyuan Tong 2026-03-20 13:22:01 +08:00
• 87549f8f0b perf(mamba): use Triton conv1d for non-contiguous input to avoid .contiguous() copy (#20469) Jia Guo 2026-03-19 19:38:46 -07:00
• db995fba47 perf(kimi_linear): replace einops rearrange with native torch ops in Kimi-Linear KDA path (#20396) Vedant V Jhaveri 2026-03-19 19:38:12 -07:00
• fa0d8f6629 perf: avoid unnecessary gpu-cpu sync in eagle_info (#20266) ehuaaandroot 2026-03-20 10:37:29 +08:00
• 3d749c49ca [JIT Kernel] Fix NVFP4 multi-arch compilation failure (#20874) Mohammad Miadh Angkad 2026-03-20 10:30:04 +08:00
• 6ad78b008b Consolidate streaming session tests (4 files → 3, fewer server starts) (#20961) Liangsheng Yin 2026-03-19 19:13:28 -07:00
• 22e378af86 Fix result writer in tuning_block_wise_kernel.py, and add FP8 kernel config for L40 (#20368) cs-cat 2026-03-20 09:28:54 +08:00
• d9794ef9f7 [Qwen3-Next] Fuse Qwen3-Next GDN's qkvz_proj and ba_proj (#19321) Yuan Luoandluoyuan.luo 2026-03-20 09:25:29 +08:00
• 42f4b7276c Revert "feat(mm)(grpc): compute M-RoPE positions for preprocessed VL inputs" (#20956) Baizhou Zhang 2026-03-19 18:03:04 -07:00
• 8d22efdeea [Tiny Fix] Fix IS_BLACKWELL env var empty string warning in rerun-ut workflow (#20957) Liangsheng Yin 2026-03-19 17:24:35 -07:00
• 2b53e660de Simplify streaming session logprob handling (#20955) Liangsheng Yin 2026-03-19 17:09:40 -07:00
• 63c38aba5e Fix token leak with logprob_start_len=0 in streaming sessions (#20557) Leon Gao 2026-03-19 15:37:27 -07:00
• 62361e8fd5 Temporarily adjust the scheulde for pr-test.yml to 12 hours (#20951) Kangyan-Zhou 2026-03-19 14:48:10 -07:00
• b42b9f6e1a Support CuteDSL mm_fp4 backend (#18801) Brayden Zhong 2026-03-19 17:20:01 -04:00
• d8ece7fb22 [Tiny Fix] Filter lru related warning with pcg (#20940) Yuwei An 2026-03-19 13:20:49 -07:00
• 0949b138af Simplify server startup output (#20885) Lianmin ZhengandClaude Opus 4.6 2026-03-19 13:11:37 -07:00
• a02cff7f2b [Fix] Patch is_flash_attn_2_available for flash-attn-4 in VLM input format test (#20946) Xinyuan Tong 2026-03-19 20:00:51 +00:00
• c562e0d13b [feat] Enhance Kimi-K2/K2.5 function call and reasoning detection (#19552) 3 people 2026-03-20 03:57:57 +08:00
• 2ee7f41e25 [AMD] add mori ep normal tbo unittest (#20941) billishyahao 2026-03-20 02:52:25 +08:00
• 9e629d31fd [AMD] CI - Fix AMD CI (multimodal test, move flaky test to non-deterministic group) (#20815) YC Yen-Ching Tseng 2026-03-20 02:50:19 +08:00
• 3b89c6d3ea simplify and improve pr-test.yml (#20911) Lianmin Zheng 2026-03-19 11:01:06 -07:00
• 29ced9c162 [UX] Suppress noisy httpx/httpcore INFO logs (#20944) Mohammad Miadh Angkad 2026-03-20 01:58:41 +08:00
• 319bb4974c [Fix] RayEngine multi-node: co-locate rank0 scheduler with Engine and fix CUDA device setting (#20722) Xinyu Zhang 2026-03-19 10:27:16 -07:00
• 274581fb77 Add support for more batch sizes in cpu_graph_runner (#13881) Cao E 2026-03-20 00:50:56 +08:00
• 4c52b7fcc6 [CI] Improve PP consistency check success rate (#20838) Shangming Cai 2026-03-19 19:12:50 +08:00
• 671bd266c1 [AMD] CI - update amd nightly cronjob (#20928) YC Yen-Ching TsengandBingxu Chen 2026-03-19 18:10:17 +08:00
• c8f0122acf Fix gpu-fault issue when run deepseek-r1 and enable dp (#20841) kkandwunhuang 2026-03-19 17:36:12 +08:00
• cb8105fe28 [sgl-kernel][6/7]Support Expert Specialization Grouped GEMM (#15471) Qi Yuhang 2026-03-19 15:39:52 +08:00
• 574572b21b [BugFix] bug fix for DeepSeek eagle3 in Attn-DP mode (#20492) khalilzhk 2026-03-19 14:48:46 +08:00
• fd05532da1 Add logging for BootstrapServer for CI diagnosis (#20844) Shangming Cai 2026-03-19 14:42:12 +08:00
• a98b456c70 [CPU] Add frontend support for Gemma (#12590) blzheng 2026-03-19 14:02:26 +08:00
• 8d4fcf2f7b [CPU] Fix MoE layer support for DeepSeek-OCR models (#12555) jianan-gu 2026-03-19 13:57:55 +08:00
• 85fe8c6793 [AMD] Use aiter_dsv3_router_gemm kernel if number of experts <= 256. (#18451) Matti Varjokallio 2026-03-19 07:40:48 +02:00
• 126cd5cfae gpt-oss decode performance optimization (#20392) kkandwunhuang 2026-03-19 13:30:03 +08:00
• cd22aa27a9 [CPU] Add FP8 Bmm support (#9744) blzhengandFan Yin 2026-03-19 13:19:48 +08:00
• c2b01bd2fc [CPU] fix bug in AVX512 implementation of flash_attn_softmax (#20220) blzhengandWu, Chunyuan 2026-03-19 13:18:47 +08:00
• 687d9eb66f [CPU] Optimize image preprocessor performance for Qwen2VLImageProcessorFast (#15168) Ma Mingfei 2026-03-19 13:18:15 +08:00
• 62d7454976 optimize conv3d used in patch embedding (#16040) Ma Mingfei 2026-03-19 13:17:53 +08:00
• cbea9f6909 [CPU] improve numa memory binding (#19666) blzhengandgemini-code-assist[bot] 2026-03-19 13:15:50 +08:00
• 2f4babe32b [CPU] support LayerNorm with 3D shape (#15075) Zaili WangandMa Mingfei 2026-03-19 13:15:24 +08:00
• dc6aa26ce9 [CPU] Add mrope kernel for Qwen3-vl (#12531) blzhengandMa Mingfei 2026-03-19 13:12:48 +08:00
• 4052b53227 fix scheduler for non-cuda devices and disable piecewise cuda graph f… (#19992) Juan MunetonandKangyan-Zhou 2026-03-18 21:54:19 -07:00
• f85455ab24 [Bugfix] fix qwen3vl hang when --mm-enable-dp-encoder is enable (#20759) Ling Zhang 2026-03-19 12:51:39 +08:00
• 7f6f1a3ab1 [LoRA][II] Add fused MOE LoRA Triton kernel and tests (#19711) Ethan (Yusheng) Su 2026-03-18 19:58:14 -07:00
• 7553b7dcb0 chore: extract diffusion_common in python/pyproject_other.toml (#20803) R0CKSTAR 2026-03-19 10:39:16 +08:00
• eea9e19c13 fix lint introduced in #20708 (#20886) Qiaolin Yu 2026-03-18 15:38:52 -07:00
• 0d23a461a0 feat(mm)(grpc): compute M-RoPE positions for preprocessed VL inputs (#19973) Chang Su 2026-03-18 15:34:50 -07:00
• 8b9482e665 fix(dp-attn): consistent overlap disable decision across DP ranks (#20853) Liangsheng Yin 2026-03-18 15:16:39 -07:00
• 4e8829e4cd Replace topk_ids with curr_topk_ids in fused_moe.py (#20302) maocheng23 2026-03-18 14:57:05 -07:00
• a3196d08b8 [MiniMax M2] Fix KV cache scale loading (#20870) Chad VoegeleandClaude Sonnet 4.6 2026-03-18 16:54:43 -05:00
• 6b8a6545b2 Add Mistral Small 4 (Pixtral) support (#20708) 4 people 2026-03-18 21:15:32 +00:00
• df1d046de2 Add packed_modules_mapping for MiniMax-M2 (#19995) Trevor Morris 2026-03-18 14:10:01 -07:00
• d1e95af282 Upgrade transformers==5.3.0 (#17784) 4 people 2026-03-18 20:50:43 +00:00
• e5750a572c Support TP for lora lm_head layer (#18511) Bruce WuandEthan Su 2026-03-18 13:48:03 -07:00
• 8f0f36c64b [1/2] Add ModelExpress coordination for remote instance weight loading - matching TP (#19920) 3 people 2026-03-18 15:38:32 -05:00
• 6cca5b9b97 [NPU] [BUGFIX] Test ascend memory consumption.py fix (#17995) Артем Савкин 2026-03-18 22:47:43 +03:00
• 46a392658e Refine RL & Post-Training description in README (#20877) Lianmin ZhengandClaude Opus 4.6 2026-03-18 12:43:42 -07:00
• c7a71740a5 [NPU][diffusion] npu support enable_torch_compile for torchair backend on diffusion models (#20687) Yaochen Han 2026-03-19 03:40:35 +08:00