Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 1646fea672 [Model] Molmo2: Enable quantized weight mapping for vision backbone (#32385) Kim Hee Su 2026-01-17 18:33:05 +09:00
• d3317bbba4 [Models] Lfm2Moe: minor name changes for resolving lora conflicts (#29063) Paul Pak 2026-01-16 23:12:55 -07:00
• 8e61425ee6 [CI] Implement uploading to PyPI and GitHub in the release pipeline, enable release image building for CUDA 13.0 (#31032) Shengqi Chen 2026-01-17 12:52:33 +08:00
• 2e7c89e708 Revert "[Attention][MLA] Make FLASHINFER_MLA the default MLA backen… (#32484) Matthew Bonanni 2026-01-16 23:42:39 -05:00
• 037a6487af apply _validate_input to MistralTokenizer token-id chat prompts (#32448) vanshil shah 2026-01-16 19:23:45 -08:00
• 5a3050a089 [Docs][Governance] Add @robertshaw2-redhat to lead maintainers group (#32498) Simon MoandClaude 2026-01-16 18:35:49 -08:00
• 484e22bc18 [TPU][Core] Enable Pipeline Parallelism on TPU backend (#28506) Chenyaaang 2026-01-16 15:29:20 -08:00
• ca21288080 [CI] Fix OOM in Hopper Fusion E2E Tests (H100) (#32489) Lucas Wilkinson 2026-01-16 14:27:16 -07:00
• 4c82b6fac7 [responsesAPI] allow tuning include_stop_str_in_output (#32383) Andrew XiaandAndrew Xia 2026-01-16 16:14:40 -05:00
• a884bc62d6 [LoRA] Update LoRA expand kernel heuristic (#32425) Xin Yang 2026-01-16 10:38:07 -08:00
• 7a1030431a Atomics Reduce Counting Optimization for SplitK Skinny GEMMs. (#29843) Hashem Hashemi 2026-01-16 09:45:04 -08:00
• 9fd918e510 [CI] Update deepgemm to newer version (#32479) Wentao Ye 2026-01-16 12:18:05 -05:00
• c9a533079c [EPLB][BugFix]Possible deadlock fix (#32418) Ilya MarkovandTyler Michael Smith 2026-01-16 15:11:01 +01:00
• 6ca4f400d8 [CI][AMD] Skip test_permute_cols since the kernel is not used and not built for ROCm (#32444) rasmith 2026-01-16 02:22:53 -06:00
• 180e981d56 [Chore] Replace swish with silu (#32459) Cyrus Leung 2026-01-16 16:22:45 +08:00
• b84c426a8c [ROCm][CI] Skip Qwen3-30B-A3B-MXFP4A16 Eval Test On Non-CUDA Platforms (#32460) Micah Williamson 2026-01-16 02:17:44 -06:00
• b66b0d6abb fix(rocm): Enable non-gated MoE (is_act_and_mul=False) support on ROCm (#32244) Rabi Mishra 2026-01-16 13:01:10 +05:30
• 03da3b52ef [Bugfix] Refactor to support DP parallel in R3 (#32306) Hongxin Xuandarlenxu 2026-01-16 15:13:58 +08:00
• 14ce524249 [CI] Breakup h200 tests (#30499) Lucas Wilkinson 2026-01-15 23:23:22 -07:00
• 4ae77dfd42 [Frontend][1/n] Make pooling entrypoints request schema consensus | CompletionRequest (#32395) wang.yuqi 2026-01-16 14:17:04 +08:00
• 73f635a75f [Bug] Add TPU backend option (#32438) XiongfeiWei 2026-01-15 21:17:12 -08:00
• 35bf5d08e8 [bugfix] Fix online serving crash when text type response_format is received (#26822) cjackalandj0shuajun 2026-01-16 13:23:54 +09:00
• 5de6dd0662 [Bugfix] [DeepSeek-V3.2] fix sparse_attn_indexer padding (#32175) KebeandLucas Wilkinson 2026-01-16 12:21:55 +09:00
• 709502558c [Model] Add Step3vl 10b (#32329) 4 people 2026-01-16 11:04:16 +08:00
• 46f8a982b1 [ROCm][CI] Enable AITER Unified Attention On ROCm For gpt-oss Test (#32431) Micah Williamson 2026-01-15 18:55:57 -06:00
• bcf2333cd6 [CI] Fix LM Eval Large Models (H100) (#32423) Matthew Bonanni 2026-01-15 19:52:49 -05:00
• 83239ff19a Add thread_n=64 support to Marlin MoE (#32360) Michael Goin 2026-01-15 19:45:44 -05:00
• c277fbdf31 [Feat] Support non-gated MoE with Marlin, NVFP4 CUTLASS, FP8, INT8, compressed-tensors (#32257) 4 people 2026-01-16 02:15:05 +02:00
• aca5c51487 [Refactor] Remove unused file (#32422) Wentao Ye 2026-01-15 17:59:38 -05:00
• 31c29257c8 [MoE Refactor][17/N] Apply Refactor to Bf16 (#31827) Yongye ZhuandRobert Shaw 2026-01-15 12:53:40 -08:00
• 8c11001ba2 [ROCM] DSfp4 mla projection gemms weight dynamic quantization (#32238) Aleksandr MalyshevandAleksandr Malyshev 2026-01-15 12:13:08 -08:00
• bd292be0c0 [BugFix] Python file source reading can fail on UnicodeDecodeError (#32416) Richard Zou 2026-01-15 15:01:41 -05:00
• 41c544f78a [ROCm] [CI] [Release] Rocm wheel pipeline with sccache (#32264) TJian 2026-01-16 02:56:18 +08:00
• 1be5a73571 [UX] Use kv_offloading_backend=native by default (#32421) Michael Goin 2026-01-15 13:55:11 -05:00
• c36ba69bda [BugFix] Fix assert x_s.shape[-1] == x_q.shape[-1] // group_shape[1] in Blackwell Quantized MoE Test (#32362) Lucas Wilkinsonandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2026-01-15 11:19:12 -07:00
• 047413375c [Attention][AMD] Make flash-attn optional (#30361) Matthias Gehre 2026-01-15 18:18:24 +01:00
• 74e4bb1c5a fixing podman build issue (#32131) 3 people 2026-01-15 09:07:08 -08:00
• b34474bf2c [Feature] Support async scheduling + PP (#32359) Wentao Ye 2026-01-15 12:06:23 -05:00
• 6218034dd7 [Model Runner V2] Support FlashInfer backend & Fix CUDA Graph bug [1/2] (#32348) Woosuk Kwon 2026-01-15 08:59:23 -08:00
• 77c16df31d [ROCm][Bugfix] Disable hip sampler to fix deepseek's accuracy issue on ROCm (#32413) Pleaplusone 2026-01-16 00:35:47 +08:00
• 130d6c9514 [ROCm][Perf] Enable shuffle kv cache layout and assembly paged attention kernel for AiterFlashAttentionBackend (#29887) Pleaplusone 2026-01-15 23:29:53 +08:00
• 361dfdc9d8 [Quant] Support MXFP4 W4A16 for compressed-tensors MoE models (#32285) Dipika SikkaandMichael Goin 2026-01-15 10:25:55 -05:00
• 8ebfacaa75 [Attention][MLA] Make FLASHINFER_MLA the default MLA backend on Blackwell, and TRTLLM the default prefill (#32339) Matthew BonanniandWentao Ye 2026-01-15 09:49:57 -05:00
• b89275d018 [ROCm] Improve error handling while loading quantized model on gfx120… (#31715) brian033andTJian 2026-01-15 20:16:00 +08:00
• 28459785ff [3/N] Group together media-related code (#32406) Cyrus Leung 2026-01-15 19:52:12 +08:00
• 8853a50af2 [CI][BugFix][AMD][FP8] Fix test_rms_norm so it runs correctly on ROCm (#32372) rasmithandRandall Smith 2026-01-15 05:05:54 -06:00
• c5891b5430 [ROCM] Add ROCm image build to release pipeline (#31995) Douglas LehrandDoug Lehr 2026-01-15 05:01:40 -06:00
• 707b44cc28 [Refactor] [11/N] to simplify the mcp architecture (#32396) Chauncey 2026-01-15 18:49:31 +08:00
• 3a4e10c847 [Benchmark] [Feature] add vllm bench sweep startup command (#32337) rongfu.leng 2026-01-15 17:25:46 +08:00
• cbbae38f93 [2/N] Move cache factories to MM registry (#32382) Cyrus Leung 2026-01-15 17:02:30 +08:00
• cdba4c74b3 [Model] Avoid token selection in SigLIP pooling head (#32389) Cyrus Leung 2026-01-15 17:01:59 +08:00
• a52d1396a7 fix: avoid crash on zero-arg tool calls in glm4 parser (#32321) seeksky 2026-01-15 16:45:59 +08:00
• 1e584823f8 [Bugfix] Strengthen the check of X-data-parallel-rank in Hybrid LB mode (#32314) dtc 2026-01-15 16:31:16 +08:00
• 4c1c501a7e [Refactor] [10/N] to simplify the vLLM openai completion serving architecture (#32369) Chauncey 2026-01-15 15:41:34 +08:00
• ae1eba6a9a [ROCm][CI] Pin transformers 4.57.3 to fix jina test failures (#32350) Andreas Karatzas 2026-01-15 01:19:34 -06:00
• e9ec2a72d8 [Bugfix] Fix stale common_attn_metadata.max_seq_len in speculative decoding with Eagle (#32312) Ofir Zafrir 2026-01-15 08:39:37 +02:00
• 2c9b4cf5bf [BugFix] Fix DeepSeek-V3.1 + DeepGEMM incompatible scale shapes (#32361) Lucas WilkinsonandEldar Kurtić 2026-01-14 23:32:22 -07:00
• 9d7ae3fcdb [code clean] remove duplicate check (#32376) Ning Xie 2026-01-15 13:29:34 +08:00
• 3c2685645e [CI][AMD][Quantization][BugFix] Fix fp8 max in quant_utils.py and update test_fp8_quant.::test_static_fp8_quant_group_2d to use correct fp8 dtype and adjust atol/rtol (#32201) rasmith 2026-01-14 23:04:34 -06:00
• 773d7073ae [ROCm][CI] Disable async scheduling on ROCm for test_structured_output[meta-llama/Meta-Llama-3.1-8B-Instruct-xgrammar-auto-speculative_config9] (#32355) Micah Williamson 2026-01-14 22:53:43 -06:00
• edadca109c [Bugfix] Add CpuCommunicator.dispatch and combine to fix DP+MoE inference (#31867) kzwrime 2026-01-15 12:50:48 +08:00
• d86fc23bdd [Misc] Remove redundant line (#32366) Li Wang 2026-01-15 12:29:56 +08:00
• 375e5984fe Support configure skip_special_tokens in openai response api (#32345) Shiyan Deng 2026-01-14 20:07:26 -08:00
• 19b251fe3d Fix optional parameter parsing in MiniMax M2 tool parser #32278 (#32342) baonudesifeizhai 2026-01-14 23:05:48 -05:00
• 15422ed3f7 [CI/Build][Hardware][AMD] Fix v1/shutdown (#31997) Ryan Rock 2026-01-14 22:01:42 -06:00
• 8471b27df9 [compile] raise on compile_size implicit padding (#32343) dolpm 2026-01-14 12:46:56 -08:00
• 66652e8082 [BugFix] Assign page_size_padded when unifying kv cache spec. (#32283) Lumosis 2026-01-14 12:10:01 -08:00
• e27078ea80 [Bugfix][ROCm][performance] Resolve the performance regression issue of the Qwen3-Next-80B-A3B-Thinking under rocm_atten (#32336) vllmellm 2026-01-15 03:32:48 +08:00
• d084e9fca7 [MODEL] Fix handling of multiple channels for gpt-oss with speculative decoding (#26291) Aleksandr Samarinandsouthfreebird 2026-01-14 21:20:52 +03:00
• 3a612322eb [CI] Move rixl/ucx from Dockerfile.rocm_base to Dockerfile.rocm (#32295) qli88 2026-01-14 10:53:36 -06:00
• 9ea07b41da [1/N] Reorganize multimodal processing code (#32327) Cyrus Leung 2026-01-14 23:25:31 +08:00
• 552b262936 rename tokenize serving api request id prefix to tokenize (#32328) Ning Xie 2026-01-14 22:52:20 +08:00
• 00e6402d56 [Frontend] track responsesAPI server_load (#32323) Chauncey 2026-01-14 20:00:37 +08:00
• ce0946249d [Misc] Make mem utils can be reused by other platforms (#32322) Shanshan Shen 2026-01-14 19:46:01 +08:00
• 3f28174c6a [Frontend] Standardize use of create_error_response (#32319) Cyrus Leung 2026-01-14 19:22:26 +08:00
• 769d0629e1 [Refactor] [9/N] to simplify the vLLM openai translations serving ar chitecture (#32313) Chauncey 2026-01-14 18:20:58 +08:00
• 90db5b31e4 [Refactor] Move top-level dummy data generation to registry (#32310) Cyrus Leung 2026-01-14 18:17:46 +08:00
• b8199f6049 [Model] Re-implement Qwen3Omni Audio Encoder (#32167) Roger Wang 2026-01-13 23:40:30 -08:00
• 7e6f123810 Add Molmo2 multimodal model support (#30997) sangho.leeandIsotr0py 2026-01-13 23:33:09 -08:00
• 9312a6c03a [Refactor] [8/N] to simplify the vLLM openai responsesapi_serving architecture (#32260) Chauncey 2026-01-14 15:26:24 +08:00
• 6388b50058 [Docs] Add docs about OOT Quantization Plugins (#32035) Michael Goin 2026-01-14 02:25:45 -05:00
• 048bb59728 AMD CI Test - unskip moe_sum test and moe_align_block_size tests (#32039) Hongxia Yang 2026-01-14 02:25:10 -05:00
• 7933638051 [misc] Remove is_torch_equal_or_newer(2.4) cases (#32296) Angela Yi 2026-01-13 23:22:07 -08:00
• 6b176095e3 [Build] Relax anthropic version pin from ==0.71.0 to >=0.71.0 (#32289) DavidandMichael Goin 2026-01-14 02:21:39 -05:00
• 9d0d7f48d5 [ROCm][CI] Handle missing vision_config in Isaac model attention patch (#32281) Andreas Karatzas 2026-01-14 01:21:26 -06:00
• 50632adc58 Consolidate Intel Quantization Toolkit Integration in vLLM (#31716) Yi Liu 2026-01-14 15:11:30 +08:00
• 6fa6e7ef0c [ROCm][CI] Disable Async Scheduling For Qwen3-Next-80B-A3B-Instruct MTP Async EPLB Accuracy Test (#32275) Micah Williamson 2026-01-13 23:29:42 -06:00
• 90c0836902 [Model Runner V2] Refactor Sampler (#32245) Woosuk Kwon 2026-01-13 17:58:12 -08:00
• 8ef50d9a6b [Kernel][Performance] Enable smaller Scaling Factor tiling for NVFP4 small-batch decoding (#30885) Roberto L. Castro 2026-01-14 00:22:53 +01:00
• 2a60ac91d0 [Improvement] Persist CUDA compat libraries paths to prevent reset on apt-get (#30784) emricksini-h 2026-01-13 23:35:05 +01:00
• 9e65bb4ef4 Add mergify label job for "bug" in PR titles (#31980) Michael Goin 2026-01-13 17:28:19 -05:00
• 0db574b185 [Build] Add scripts for cherry-picking and trigger build (#32282) Simon MoandCursor Agent 2026-01-13 13:21:05 -08:00
• 2f4a71daf2 [Misc] Add In-Container restart capability through supervisord for sagemaker entrypoint (#28502) HappyAmazonian 2026-01-13 13:06:10 -08:00
• 69f8a0ea37 fix(rocm): Use refresh_env_variables() for rocm_aiter_ops in test_moe (#31711) Rabi Mishra 2026-01-14 00:41:54 +05:30
• f28125d87b [Perf] Optimize grouped topk kernel, 1.2%~2% E2E Throughput improvement (#32058) Wentao Ye 2026-01-13 13:58:18 -05:00
• 46f8c6b725 Fix CUDA 13 wheel installation doc (#32276) Dmitry Tokarev 2026-01-13 13:48:37 -05:00
• af54d2e2d0 [responseAPI] support partial message generation (#32100) 3 people 2026-01-13 13:41:26 -05:00
• 6beef12b9b [EPLB][Cleanup] Remove is_async_enabled from EplbModelState (#32050) Sage Moore 2026-01-13 10:19:03 -08:00
• ab74b2a27a [Trivial] Remove duplicate enable_mfu_metrics (#32246) Mark McLoughlin 2026-01-13 17:09:23 +00:00
• 2263d44b68 [4/N][Attention] Move MLA common to model_executor (#32060) Matthew BonanniandLuka Govedič 2026-01-13 12:08:45 -05:00