Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 43a6acfb7d [Model] fix ernie45 reasoning_parser (#27973) CSWYF3634076 2025-11-04 15:16:46 +08:00
• 58279c60b5 [KV Connector] Make KVCacheConfig an explicit constructor argument (#27887) Mark McLoughlin 2025-11-04 07:00:49 +00:00
• 2f84ae1f27 [CI/Build] Update LM Eval Version in AMD CI (#27944) Zhewen Li 2025-11-03 22:36:40 -08:00
• f32cbc9a0c [CPU]Improve dynamic 4bit moe performance (#27240) xiangze-arm 2025-11-04 14:33:23 +08:00
• 7e4be74104 [Bug] Batch invariant: Fix flash attn MLA RuntimeError: scheduler_metadata must have shape (metadata_size) (#27884) Wentao Ye 2025-11-04 01:05:55 -05:00
• 380ba6816d [Metrics] Enable sleep state metric outside of dev mode (#27867) Mark McLoughlin 2025-11-04 04:35:36 +00:00
• 14a125a06d [NIXL][XPU] Pin NIXL version to 0.7.0 (#27849) liuzhenwei 2025-11-04 11:28:35 +08:00
• c02fccdbd2 [Refactor] Lazy import tool_parser (#27974) Chauncey 2025-11-04 10:10:10 +08:00
• 6ddae74054 [LoRA] Lora shrink swizzle (#27694) li2haipengandJee Jee Li 2025-11-03 17:30:20 -08:00
• b13a447546 [Bugfix][ROCm] Fix ViT rotary embeddings for torch.compile compatibility on ROCm (#27748) vllmellm 2025-11-04 09:12:19 +08:00
• 7956b0c0bc Remove the tpu docker image nightly build. (#27997) QiliangCui 2025-11-03 16:35:54 -08:00
• 3758757377 [Bugfix] Fix MoE Routing Simulation (#28002) Tyler Michael Smith 2025-11-03 17:26:49 -05:00
• ccd3e55e51 [Bugfix][plugin] fla crash on plugin (#27322) Hank_ 2025-11-04 05:27:03 +08:00
• 01baefe674 Add TP parameter to attention tests (#27683) Matthew Bonanni 2025-11-03 16:04:40 -05:00
• 786030721e [Docs] add runai_streamer_sharded to LoadConfig (#27937) Ning Xie 2025-11-04 04:35:16 +08:00
• 145c00a4d3 [Bugfix] change FlashMLA reorder_batch_threshold (#27777) Matthew Bonanni 2025-11-03 15:17:10 -05:00
• 55011aef24 [Bugfix][Qwen][Multimodal] Move Qwen2_5_vl sdpa to custom op and reenable compile (#27764) Lucas Kabela 2025-11-03 11:12:15 -08:00
• a4398fbb5e [Feature][Benchmarks] Support inf burstiness (#26941) Sophie du Couédic 2025-11-03 19:33:17 +01:00
• 2c19d96777 [Spec Decode] Integrate Suffix Decoding from Arctic Inference (#25784) Aurick QiaoandAurick Qiao 2025-11-03 09:23:31 -08:00
• 4bc400f47e [CI/Testing] Add basic single node dual batch overlap test (#27235) Lucas Wilkinson 2025-11-04 02:00:46 +09:00
• cac4c10ef0 [BUG] Make 'binary' default option for saving torch compile artifacts when using standalone_compile (#27616) ahao-anyscale 2025-11-03 08:13:51 -08:00
• f7d2946e99 [Bugfix] Skip gs:// model paths for speculator detection (#27846) pwschuurman 2025-11-03 06:31:03 -08:00
• 294c805f1d Early exit for MoE LoRA kernels (#27131) gnovackandJee Jee Li 2025-11-03 04:22:17 -08:00
• 40b69e33e7 [Model] Add PaddleOCR-VL Model Support (#27758) 3 people 2025-11-03 19:04:22 +08:00
• 32257297dd [CI/Build] Remove the flaky gpt-oss lora test (#27966) Jee Jee Li 2025-11-03 16:50:06 +08:00
• ba464e6ae2 Add ORCA endpoint load metrics support (#24905) Misha Efimov 2025-11-03 03:21:31 -05:00
• 7f4bdadb92 [XPU]Refine Dockerfile.xpu, avoid oneccl dependency issue (#27964) Kunshang Ji 2025-11-03 15:36:59 +08:00
• cec7c28833 [Bugfix] Padded Eagle Specdec with Chunked Prefill (#26263) Rémi DelacourtandBenjamin Chislett 2025-11-03 08:22:46 +01:00
• 18961c5ea6 [Hybrid] Pass kernel block size to builders (#27753) Thomas Parnell 2025-11-03 06:48:03 +01:00
• 470ad118b6 [Frontend] Align finish_reason when tool is called with OpenAI (#25054) Sungyoon JeongandChauncey 2025-11-03 13:21:18 +09:00
• 1bf43ae35d [BugFix][LoRA] use adapter_id instead of id field of lora_request (#27728) Biswa Panda 2025-11-02 18:08:08 -08:00
• 0ce743f4e1 Fix(llm): Abort orphaned requests when llm.chat() batch fails Fixes #26081 (#27420) Vensen 2025-11-03 00:24:01 +08:00
• 6c317a656e [Misc] Provide Siglip2 chat template (#27939) Cyrus Leung 2025-11-02 21:42:38 +08:00
• 00b31a36a2 [V1] [Hybrid] Mamba1 Automatic Prefix Caching (#26377) Asaf Joseph Gardin 2025-11-02 14:16:23 +02:00
• 73444b7b56 Performance fix MistralTokenizer: cache special ids and tokens (#27925) Julien DenizeandPatrick von Platen 2025-11-02 09:48:33 +01:00
• 853a8eb53b [Bugfix] Fix Qwen Omni audio inference (#27920) Cyrus Leung 2025-11-02 13:06:05 +08:00
• 758ea2e980 [CI/Build] Fix flaky test_transcription_validation.py::test_basic_audio_gemma (#27924) Ben Browning 2025-11-01 23:45:02 -04:00
• 685c99ee77 [KV offload] Offloading connector async scheduling support (#27648) Yue ZhangandNick Hill 2025-11-02 05:08:56 +08:00
• 1e88fb751b Adds anthropic /v1/messages endpoint to openai api_server (#27882) Benjamin Bartels 2025-11-01 19:45:42 +00:00
• c2ed069b32 [BugFix] Fix mixed penalties batch with async scheduling (#27910) Nick Hill 2025-11-01 10:51:24 -07:00
• af6e19f50f [Core][TPU] Support TPU Data Parallalism (#27365) wenxindongwork 2025-11-01 11:14:44 -06:00
• 99d69af9ec [Bugfix] Python 3.10 compatibility for Self (#27918) Cyrus Leung 2025-11-01 23:28:54 +08:00
• d811b442d3 [Bugfix] DeepSeek V3.2 MTP metadata & CUDA graph issues (#26779) Haco 2025-11-01 22:52:43 +08:00
• 30a14b034f [V0 deprecation] Remove VLLM_USE_V1 usage in platform and v1 module (#27798) wangxiyuanandHarry Mellor 2025-11-01 18:17:45 +08:00
• 799ce45cc1 [Docs] Mock all imports for docs (#27873) Harry Mellor 2025-11-01 10:02:23 +00:00
• 2c0c7c39bd feat(benchmarks): support HF model names in multi-turn benchmark (#27850) ai-jz 2025-11-01 01:04:52 -07:00
• e675118849 [Add] cmdline argument parsing for KV cache offloading modules (#27621) Yihua ChengandHarry Mellor 2025-11-01 00:17:07 -07:00
• e2347dbf58 [Bugfix] [Model] Missing MRoPE function definition from KeyeForConditionalGeneration (#27895) TJian 2025-10-31 22:45:23 -07:00
• 879a06579e [CI/Build] Bump transformers version (#27528) 3 people 2025-11-01 13:11:07 +08:00
• 29de3cdee4 Adding SplitK in fused_moe_lora kernel (#27818) yugong333andJee Jee Li 2025-10-31 21:55:46 -07:00
• 7e2729b57e [Multimodal][XPU]Enable vision attn backend for xpu platform (#27525) 4 people 2025-11-01 12:45:02 +08:00
• 3a5de7d2d6 [Bugfix] Fix KDA output (#27905) Jee Jee Li 2025-11-01 11:54:36 +08:00
• bc4486d609 [Kernel] Enable FusedMoEModularKernel support bias (#27754) Jee Jee Li 2025-11-01 10:05:12 +08:00
• 0cdbe7b744 [Core] Async scheduling + structured outputs compatibility (#26866) Nick Hill 2025-10-31 17:35:04 -07:00
• df334868ca [Hybrid] A simpler algorithm to find kernel_block_size (#26476) Chen Zhang 2025-10-31 14:30:28 -07:00
• 0e0a638c3b Batch invariance doc (#27839) 3 people 2025-10-31 17:22:19 -04:00
• f29aeb5a25 Add FLASHINFER_MLA to test_mla_backends and add B200 CI run (#27663) Matthew Bonanni 2025-10-31 14:12:19 -04:00
• 5e8862e9e0 [Feature] Pydantic validation for scheduler.py and structured_outputs.py (#26519) Vinay R DamodaranandHarry Mellor 2025-10-31 11:05:50 -07:00
• 9e5bd3076e [Cleanup] Remove no-longer-used SpeculativeConfig.enable_chunked_prefill (#27826) Nick Hill 2025-10-31 10:57:45 -07:00
• fc16f1c477 Flashinfer_CUTLASS_MOE fuses quantization for TP (#27223) Shu Wang 2025-10-31 10:54:29 -07:00
• bc306fe5e9 fix incorrect type annotation in KimiMLP (#27885) ZiTian Zhao 2025-11-01 01:38:02 +08:00
• 103a468bbf [bugfix] Missing cached item in beam search (#27874) Chenguang ZhengandCyrus Leung 2025-11-01 01:34:27 +08:00
• 70bfbd7b16 Docs update tpu install instructions (#27824) 3 people 2025-10-31 13:29:55 -04:00
• d6517be3cd [Bugfix] Missing NIXL metadata for handshake initialization if instance spans multi-node (#26338) GuanLuoandNicolò Lucchesi 2025-11-01 01:16:00 +08:00
• 7e06c40e63 [Bugfix] Fix broken MRoPE for GLM-4.1V/GLM-4.5V (#27860) Isotr0py 2025-11-01 01:04:51 +08:00
• 675704ac01 [Bugfix] Allow 64-bit integer values for LoRA IDs to avoid overflow/truncation (#27876) Madeesh Kannan 2025-10-31 17:58:42 +01:00
• 0384aa7150 [CI/Build] Add gpt-oss LoRA test (#27870) Jee Jee Li 2025-10-31 22:17:21 +08:00
• 3857eb8725 [Perf] Decouple torch op from GDA to leverage torch.compile (#27871) Jiangyun Zhu 2025-10-31 21:35:52 +08:00
• 933cdea440 [BugFix] Don’t compute reorder threshold when there are no attention groups (#27861) Huamin Li 2025-10-31 04:36:18 -07:00
• 3933f18a5e [Bugfix] Avoid too small block m/n for FlexAttention kernel option (#27853) Isotr0py 2025-10-31 19:33:12 +08:00
• e5ef4dfc11 [Kimi-Linear] Correct prefixes and add compatibility to AWQ quants (#27834) toncaoandtoncao 2025-10-31 16:36:37 +07:00
• 36960501d3 [Hardware][Powerpc] Fix VLLM_CPU_OMP_THREADS_BIND="auto" low CPU utilization for Power (#27734) Akash kaothalkarandAkash Kaothalkar 2025-10-31 13:15:26 +05:30
• b2e65cb4a7 [benchmark] Make request IDs unique across clients by default (#27723) Seiji Eicher 2025-10-30 19:40:35 -05:00
• 2bf0bcc1fc [CI Test] Add Scheduled Integration Test (#27765) Wentao Ye 2025-10-30 20:29:26 -04:00
• 697f507a8e [CI/Build][Intel] Enable performance benchmarks for Intel Gaudi 3 (#26919) Jakub Sochacki 2025-10-31 00:57:22 +01:00
• d5d2a0fe74 [Misc] Make all tool scripts executable (#27831) Matthew Bonanni 2025-10-30 19:46:02 -04:00
• c9791f1813 [BugFix] Fix broken import in initialize_ray_cluster() (#27838) Nick Hill 2025-10-30 16:26:13 -07:00
• e7acb20076 [Feature] Batch invariant torch.compile (#27660) Paul ZhangandWentao Ye 2025-10-30 16:11:29 -04:00
• 4b68c4a55b [Core][Perf] Only invoke save_new_computed_blocks when computed blocks are not empty (#27799) Jialin Ouyang 2025-10-30 12:47:30 -07:00
• a8141fa649 [Refactor] Remove VLLM_DEEPEP_LOW_LATENCY_ALLOW_NVLINK (#27750) Wentao Ye 2025-10-30 15:32:39 -04:00
• 4917002523 [Fix] Skip record_sleep_state logic in PrometheusStatsLogger if not in dev mode (#27789) Sumanth R Hegde 2025-10-30 12:26:27 -07:00
• a2981c4272 [EP/DP][API Server] Enable DP-aware routing in OpenAI API requests (#24945) cong-metaandCong Chen 2025-10-30 12:10:16 -07:00
• 4574d48bab [Core][Bookkeeping] Update cu_num_accepted_tokens for all req_index (#27629) Jialin Ouyang 2025-10-30 11:52:36 -07:00
• ab98f6556f [Bugfix] Fix 2 precommit issues - (mamba_block_size, kv_cache_config) (#27811) Tyler Michael SmithandNick Hill 2025-10-30 14:52:18 -04:00
• 2918c1b49c [Model] Use the same fused_moe configs for all H200 devices (#23642) Roger Meier 2025-10-31 01:36:56 +08:00
• 1004205795 [MTP] Refactor mtp predictor to avoid d2h operation (#27643) Mengqing Cao 2025-10-31 01:27:39 +08:00
• ba33e8830d Reapply "Install pre-built xformers-0.0.32.post2 built with pt-2.9.0" (#27768) Huy Do 2025-10-30 10:22:30 -07:00
• 33a0ea5f32 [Docs] add Shanghai Meetup - 2025/10 (#27545) 3 people 2025-10-31 01:33:13 +09:00
• 60f76baa66 [Misc] Replace CUDA_VISIBLE_DEVICES in DP with torch.cuda.set_device for device selection on cuda-like devices (#27564) Ilya MarkovandTyler Michael Smith 2025-10-30 16:41:44 +01:00
• e5e076cad7 [BugFix] Stopgap - Flashinfer Autotuner + GPT-OSS + DP/TP (#27762) Varun Sundar RabindranathandVarun Sundar Rabindranath 2025-10-30 11:24:31 -04:00
• eebf00cb0c [Bugfix][CPU] Fix MRoPE dispatch on the CPU backend (#27800) Li, Jiang 2025-10-30 23:12:05 +08:00
• 9956aae4ea [Model][Ouro] Support Ouro Model (#27794) 4 people 2025-10-30 22:34:41 +08:00
• 0fe0140408 [KV offload] Enable CPU KV offload on CUDA alike Platforms (#27770) Zhewen Li 2025-10-30 07:10:29 -07:00
• 4e68cc9b6a [Model] Introduce Kimi Linear to vLLM (#27809) Zhiyuan Li 2025-10-30 21:02:27 +08:00
• 1994de99ea [CI Failure] Fix test_kv_cache_model_load_and_run (#27717) Huamin Li 2025-10-30 05:27:53 -07:00
• 4464723f22 [Frontend][Doc][5/N] Improve all pooling task | Polish encode (pooling) api & Document. (#25524) 3 people 2025-10-30 20:13:05 +08:00
• 74374386e2 [Bugfix] Improve GPU validation logging in Ray fallback scenarios (#25775) Sairam Pillai 2025-10-30 17:27:59 +05:30
• c01f6e525f [CI] Fix mypy for vllm/v1/core and vllm/v1/engine (#27108) Wentao Ye 2025-10-30 07:32:17 -04:00
• c7d2a554ba [CI Failure] fix test_default_mm_loras (#27795) Huamin Li 2025-10-30 03:13:03 -07:00
• af826e0820 [V0 deprecation] Remove VLLM_USE_V1 usage in config module (#27784) wangxiyuan 2025-10-30 17:42:49 +08:00