Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 5819ca8944 [Docs] Add reo analytics (#33957) Simon Mo 2026-02-05 17:42:22 -08:00
• 79028d4388 [Perf] Disable clean_logits in deepgemm fp8_mqa_logits kernel (#33568) Xin Yang 2026-02-05 17:34:00 -08:00
• 325ab6b0a8 [Feature] OTEL tracing during loading (#31162) emricksini-h 2026-02-06 01:59:28 +01:00
• 91a07ff618 [Bugfix] Fix DeepSeek v3.2 tokenizer outputting None issue (#33832) Wei Zhao 2026-02-05 18:50:49 -05:00
• d5c4800112 Adds padding and perf improvements to wvSplitK_fp8 (#33527) Hashem Hashemi 2026-02-05 14:16:02 -08:00
• 42d5d705f9 [Minor] Sort safetensors files to ensure deterministic loading order (#33491) Lumosisandmgoin 2026-02-05 14:05:09 -08:00
• 116880a5a0 [Bugfix] Make MM batching more robust (#33817) Cyrus Leung 2026-02-06 04:40:58 +08:00
• 4145e50d85 [Bugfix] Fix DSV3.2 NVFP4 (#33932) Matthew Bonanni 2026-02-05 14:22:19 -05:00
• 20f5d185a6 [Misc] Rename translations to speech_to_text for OAI serving component (#33904) Nicolò Lucchesi 2026-02-05 20:16:52 +01:00
• 1887acca9e Fix tokenizer test for renamed attr on Transformers v5 (#33902) Harry Mellor 2026-02-05 19:16:20 +00:00
• 92e7562a99 [Bugfix] Suppress non-TTY color output on the process name part of the log (#29714) Tsukasa OI 2026-02-06 03:47:09 +09:00
• 87d0d17ab5 [Models] Consolidate Deepseek-OCR2 processor (#33909) Isotr0py 2026-02-06 02:29:20 +08:00
• a57c8228ff [Moe Refactor] Make Inplace Flag for FusedMoEModularKernel part of the constructor (#33375) bnellnmandRobert Shaw 2026-02-05 13:07:18 -05:00
• 1ee95841bd [Bugfix] Fix swapped engine_ids in NIXL Llama 4 local attention path (#33795) zackyoray 2026-02-05 19:51:58 +02:00
• 7d8c6804e2 [Misc] Add debug logs (#33931) Nicolò Lucchesi 2026-02-05 18:42:40 +01:00
• af3162d3aa [Spec Decode] Unified Parallel Drafting (#32887) Benjamin Chislett 2026-02-05 12:37:18 -05:00
• 5b2a9422f0 [BugFix] Fix LoRA Fp8 (#33879) danisereb 2026-02-05 19:25:55 +02:00
• c1858b7ec8 [Feat][RL][1/2] Native Weight Syncing API: NCCL (#31943) Aaron HaoandSumanthRH 2026-02-05 09:13:23 -08:00
• 82914d2ae8 [Bugfix] Fix step3p5 parser when using mtp (#33690) Mario Hong 2026-02-06 00:04:04 +08:00
• 81a90e5277 [Docs] Add bart-plugin to docs (#33905) Nicolò Lucchesi 2026-02-05 13:20:25 +01:00
• 1c3a221d3b [Bugfix] Fix corner case of sparse embedding (#33886) wang.yuqi 2026-02-05 18:51:22 +08:00
• 7bd42e609d [Refactor] Clean up input preprocessing (#33687) Cyrus Leung 2026-02-05 18:43:42 +08:00
• a2522839d8 [Bugfix] Fix Kimi-K2.5 NVFP4 checkpoints weight loading (#33876) Isotr0py 2026-02-05 18:29:54 +08:00
• 59a5cb387a [perf] Integrate flashinfer concat_mla_k (#31171) jiahanc 2026-02-05 18:23:11 +08:00
• 8322d4e47f Enable Cross layers KV cache layout at NIXL Connector V2 (#33339) 4 people 2026-02-05 12:17:02 +02:00
• 3e472e81f9 [ROCm][Bugfix][CI] Fix hybrid models and their tests (Mamba/Jamba/Bamba) (#32710) Andreas KaratzasandMatthew Wong 2026-02-05 04:01:23 -06:00
• 038914b7c8 [Refactor] Move task outside of PoolingParams.verify (#33796) Cyrus Leungandwang.yuqi 2026-02-05 17:33:11 +08:00
• d2f4a71cd5 [Bugfix] Kimi-K2 grouped_topk usage for Flashinfer monolithic kernels. (#33858) Pavani Majety 2026-02-05 01:32:10 -08:00
• 2abd97592f [KV Connector][Metrics] Do not count local prefix cache hits in connector queries (#30522) Mark McLoughlin 2026-02-05 07:57:27 +00:00
• 6abb0454ad [Perf] Optimize the performance of structured output + reasoning (#33557) Chauncey 2026-02-05 15:45:29 +08:00
• db6f71d4c9 [CI/Build] Fix CPU CI test case title (#33870) Li, Jiang 2026-02-05 15:07:14 +08:00
• fd03538bf9 [CPU][BugFix] Allow w8a8 oneDNN quantized matmul to support 3D inputs (#33727) Fadi Arafeh 2026-02-05 06:26:09 +00:00
• 1f70313e59 [Bugfix] Fix ScoreMultiModalParam multi-document scoring returning single result (#33837) Andreas Karatzasandwang.yuqi 2026-02-05 00:17:00 -06:00
• 07daee132b [CI/Build] Parallelize CPU CI tests (#33778) Li, Jiang 2026-02-05 13:53:48 +08:00
• 9595afda18 [2/N] move responses/serving _make_response_output_items logic to parser (#33281) Andrew XiaandAndrew Xia 2026-02-05 00:46:15 -05:00
• c1395f72cd [CI][AMD][BugFix] Ensure VLLM_ROCM_USE_AITER is set so test_rocm_aiter_topk.py can run correctly (#33840) rasmith 2026-02-04 23:05:48 -06:00
• 007b183d74 [docs] fix unintentional misspellings (#33863) rinbaro 2026-02-04 20:50:59 -08:00
• add9f1fbd9 [Minor] Include StreamingInput in inputs package (#33856) Nick Hill 2026-02-04 20:38:20 -08:00
• e3bf79ffa0 Revert "[Attention][FA3] Update FA3 to include new swizzle optimization" (#33841) Luka Govedič 2026-02-04 22:54:27 -05:00
• fb1270f1f8 [CI][Bugfix]: return McpCall for built-in MCP tools in non-streaming mode (#32762) Andreas Karatzas 2026-02-04 21:14:06 -06:00
• 72bb24e2db [release] Minor fixes to release annotation (#33849) Kevin H. Luu 2026-02-04 18:07:35 -08:00
• a7be77beef [Bugfix] fix DeepSeek R1 with CUTLASS MLA Broken on B200 (#33637) Chauncey 2026-02-05 09:28:36 +08:00
• bbe0574d8e [Bugfix] Disable TRTLLM attention when KV transfer is enabled (#33192) zhanqiuhu 2026-02-04 19:49:18 -05:00
• 4d9513537d [CI][torch.compile] Reduce e2e fusion test time (#33293) Luka Govedič 2026-02-04 19:09:03 -05:00
• 439afa4eea feat: Add ColBERT late interaction model support (#33686) 3 people 2026-02-05 01:05:13 +01:00
• fa4e0fb028 [Core] Don't schedule spec tokens with prefill chunks (#33652) Nick Hill 2026-02-04 15:40:22 -08:00
• ce498a6d61 Change the type signature of MixtureOfExperts.expert_weights to MutableSequence[Sequence[Tensor]] (#33573) Sage MooreandRobert Shaw 2026-02-04 14:02:46 -08:00
• 9f14c9224d Revert "[torch.compile] Significantly speed up cold start times" (#33820) Richard Zou 2026-02-04 13:59:59 -08:00
• 535de06cb1 [Model] Add transcription support for Qwen3-Omni (#29828) Muhammad HashmiandNickLucche 2026-02-04 13:17:47 -08:00
• 4292c90a2a [Bugfix] Support RotaryEmbedding CustomOp for gpt-oss (#33800) Simon Danielsson 2026-02-04 21:17:41 +01:00
• 6e98f6d8b6 Implement zero-copy GQA for multimodal and CPU (#33732) Taeksang Kim 2026-02-05 05:11:39 +09:00
• 2f6d17cb2f [rocm][ray] Fix: Unify Ray device visibility handling across CUDA and ROCm (#33308) kourosh hakhamaneshi 2026-02-04 10:09:14 -08:00
• 192ad4648b [Bugfix] Fix interns1-pro initialization and PP (#33793) Isotr0py 2026-02-05 01:54:45 +08:00
• 0e92298622 [Misc] Delay deprecation of CommonAttentionMetadata properties (#33801) Lucas Wilkinson 2026-02-04 09:41:57 -07:00
• 87d9a26166 [Bugfix] Fix ubatch wrapper num_tokens calculate (#33694) jiangkuaixue123 2026-02-05 00:41:45 +08:00
• 80f921ba4b [Bugfix] Fix normalize still being passed to PoolerConfig (#33794) Cyrus Leung 2026-02-04 23:56:02 +08:00
• 711edaf0d0 [Perf] Optimize spec decoding + async scheduling, 1.5% Throughput improvement (#33612) Wentao YeandNick Hill 2026-02-04 09:34:32 -05:00
• 1d367a738e [Bugfix][ROCm] Include float8_e4m3fnuz in NCCL Dtype Dispatching (#33713) Micah Williamson 2026-02-04 07:36:29 -06:00
• 32a02c7ca2 Apply #33621 to main (#33758) 3 people 2026-02-04 21:35:39 +08:00
• f67ee8b859 [Perf] Optimize chat completion streaming performance (#33782) Chauncey 2026-02-04 20:30:36 +08:00
• e57ef99b40 [Model] Apply #32631 for recent models (#33785) Cyrus Leung 2026-02-04 20:23:01 +08:00
• f8516a1ab9 [Bugfix][Model] Fix audio-in-video support for Qwen2.5-Omni and Qwen3-Omni (#33605) Yueqian LinandRoger Wang 2026-02-04 07:15:29 -05:00
• 824058076c [PERF] Change GDN Attention State Layout from [N, HV, K, V] to [N, HV, V, K] (#33291) Vadim Gimpelson 2026-02-04 15:20:52 +04:00
• 8e32690869 [KV Connector][BugFix] scheduler: Delay freeing blocks of aborted async loads (#32255) Or Ozeri 2026-02-04 13:16:34 +02:00
• a208439537 [compile] Remove runner type from ignored caching factor list. (#33712) Zhengxu Chen 2026-02-04 05:56:45 -05:00
• bcd2f74c0d [compile] Clean up AOT compile bypass on evaluate_guards. (#33578) Zhengxu Chen 2026-02-04 05:12:53 -05:00
• f79f777803 [XPU][2/N] add support unquantized moe support for xpu (#33659) Kunshang Ji 2026-02-04 18:12:25 +08:00
• 4c8d1bf361 use ORJSONResponse when available to improve the efficiency of request process (#33548) Augusto Yao 2026-02-04 18:04:11 +08:00
• 061da6bcf7 [XPU] remove common path warning log (#33769) Kunshang Ji 2026-02-04 16:40:17 +08:00
• 4403e3ed4c [Metrics] Add labeled prompt token metrics for P/D disaggregation (#33290) zhanqiuhu 2026-02-04 02:46:48 -05:00
• 08e094997e [Hardware][AMD][CI] Refactor AMD tests to properly use BuildKite parallelism (#32745) Matt 2026-02-04 00:51:33 -06:00
• d88a1df699 [Deprecation] Deprecate profiling envs (#33722) Wentao Ye 2026-02-04 00:58:21 -05:00
• 90d74ebaa4 [Deprecation] Remove _get_data_parser in MM processor (#33757) Cyrus Leung 2026-02-04 13:51:52 +08:00
• 45f8fd6f97 [Feature] Enable TRITON_ATTN for Batch Invariance (#33688) Frank Wang 2026-02-03 21:27:34 -08:00
• 5e1e0a0fbd [Refactor] Remove unused dead code (#33718) Wentao Ye 2026-02-04 00:25:11 -05:00
• eb5ed20743 [Bugfix] Define router_logits_dtype for remaining MoE models (#33737) Michael Goin 2026-02-04 00:24:14 -05:00
• 2647163674 Save startup benchmark results as a list of values (#33629) Huy Do 2026-02-03 20:37:51 -08:00
• 9fb27dd3b3 [MM] Align the prefix of MMEncoderAttention with Attention (#33750) Shanshan Shen 2026-02-04 12:07:30 +08:00
• 4dffc5e044 [CPU] Split attention dispatch by head_dim alignment (#32161) R3hankhan 2026-02-04 09:07:15 +05:30
• e1bf04b6c2 [1/N] Initial Implementation of Parser for ResponsesAPI (#32712) Andrew XiaandAndrew Xia 2026-02-03 21:59:03 -05:00
• 02080179a3 [Bugfix] Fix torchrun PP broadcast deadlock with async scheduling (#33701) Isotr0py 2026-02-04 10:17:37 +08:00
• 1b8fe6f7c4 [Frontend][4/n] Make pooling entrypoints request schema consensus | ScoreRequest (#33060) wang.yuqi 2026-02-04 09:48:40 +08:00
• 52ee21021a [BugFix][Spec Decoding] Fix negative accepted tokens metric crash (#33729) Nick Hill 2026-02-03 15:34:41 -08:00
• 655efb3e69 [Dependency] Remove comments of ray in dependency files (#33351) Wentao Ye 2026-02-03 18:30:47 -05:00
• bd8da29a66 [Bugfix] Fix sparse MLA metadata building (#33579) Matthew Bonanni 2026-02-03 18:29:48 -05:00
• 2a99c5a6c8 [Bugfix] Disable TRTLLM FP8 MoE if router_logits_dtype==float32 and routing_method!=DeepSeekV3 (#33613) Michael Goin 2026-02-03 16:26:51 -05:00
• 3f7662d650 [Voxtral Realtime] Change name (#33716) Patrick von Platen 2026-02-03 22:03:28 +01:00
• a372f3f40a [MISC] Fix Tensor Parallelism for Quantized Mamba Models with n_groups=1 (#33257) Vadim Gimpelson 2026-02-04 00:10:31 +04:00
• 61e632aea1 Turn @config into a dataclass_transform (#31541) Harry Mellor 2026-02-03 17:40:59 +00:00
• b1bb18de8d [torch.compile] Significantly speed up cold start times (#33641) Richard Zou 2026-02-03 09:12:11 -08:00
• 2267cb1cfd [Attention][FA3] Update FA3 to include new swizzle optimization (#23465) Lucas Wilkinson 2026-02-03 09:08:47 -07:00
• 0d6ccf68fa [P/D] rework mooncake connector and introduce its bootstrap server (#31034) dtcandNicolò Lucchesi 2026-02-04 00:08:25 +08:00
• 18e7cbbb15 [Bugfix] Fix startup hang for Granite Speech (#33699) Cyrus Leung 2026-02-03 23:57:56 +08:00
• f0d5251715 [Voxtral models] Skip warm-up to skip confusing error message in warm-up (#33576) Patrick von PlatenandCyrus Leung 2026-02-03 16:22:34 +01:00
• 5c4f2dd6ef [MM] Pass prefix parameter to MMEncoderAttention (#33674) Shanshan Shen 2026-02-03 22:47:41 +08:00
• f3d8a34671 [Bugfix] Do not add extra \n for image-only cases when constructing multimodal text prompts. (#33647) wang.yuqi 2026-02-03 22:43:47 +08:00
• 4bc913aeec Feat/add nemotron nano v3 tests (#33345) shaharmor98 2026-02-03 15:52:49 +02:00
• fbb3cf6981 [Bugfix][Async][Connector] avoid vllm-side double free during async scheduling + request abort + async KV cache transfer (#33377) Kuntai Du 2026-02-03 21:50:15 +08:00
• 2df2b3499d Document NixlConnector backend selection via kv_connector_extra_config (#33552) Krish Gupta 2026-02-03 19:19:59 +05:30
• 2a8d84e66d Fix Gemma3n audio encoder for Transformers v5 (#33673) Harry Mellor 2026-02-03 13:49:49 +00:00