Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 4e2c7caf2d [Bugfix] Add regression test for MoE quant_config under torch.compile (#34335) Matthias Gehre 2026-02-20 06:27:26 +01:00
• d9e62c03eb [Quark] Fix MoE fp8 activation scale handling on mi300 (#34386) Bowen Bao 2026-02-19 21:27:14 -08:00
• a1a2d79442 [ci] Use the right tag for CPU arm64 image (#34915) Kevin H. Luu 2026-02-19 19:59:15 -08:00
• ac900c89bb [Refactor] Implement output type check in LLM (#34794) Cyrus Leung 2026-02-20 11:57:55 +08:00
• 76df6072ff [Core] Fix state names in pause_scheduler() (#34840) Mark McLoughlin 2026-02-20 01:21:46 +00:00
• 16f24e8797 [CI] Add GPT-OSS Eval job for H100 (#34359) Michael Goin 2026-02-19 20:14:54 -05:00
• 40b2f1c3d9 [Model Runner V2] Minor CPU optimizations (#34856) Nick Hill 2026-02-19 16:05:37 -08:00
• 648951a9c3 [Bugfix] Fix benchmark_fused_collective crash on CustomOp init (#34665) 3 people 2026-02-19 16:01:00 -08:00
• f72061a19a [UX] More descriptive reasons in is_supported_config for MoE (#34908) Michael Goin 2026-02-19 18:20:52 -05:00
• 662205d34e [Bugfix] Fix Basic Models Test (#34818) Matthew BonanniandRobert Shaw 2026-02-19 17:49:07 -05:00
• 4fb8beefaa [Bugfix] Fix cutlass fp8 kernel on hopper for Qwen3.5 (#34914) Roger Wang 2026-02-19 13:34:55 -08:00
• 304319c4ed Change targets for AMD build in the "CI" pipeline (#34918) Alexei-V-Ivanov-AMD 2026-02-19 15:26:53 -06:00
• c683d11c94 [Refactor] Deprecate head_first for chunk_gated_delta_rule (#34263) Wentao Ye 2026-02-19 13:23:49 -05:00
• 3eff45d793 Revert "[NemotronH] Do not force router to run in fp32 (#34582)" (#34808) roikoren755andMichael Goin 2026-02-19 19:47:05 +02:00
• 4685a630a2 [Model Bash][DeepSeekR1] Remove Shared Expert Clone (#34344) Robert ShawandRobert Shaw 2026-02-19 10:56:14 -05:00
• ee1d25f199 [Llama4,Quantization] Simplify and generalize logic for Q/K permutations in quantized self-attn layers (#34471) 3 people 2026-02-19 16:55:41 +01:00
• 6fff24f30f [Bugfix] Qwen3.5 kv-scale weight remapping (#34719) Linda 2026-02-19 13:13:37 +01:00
• 23210a911e [CI/Build] Try to make beam search test less flaky (#34885) Cyrus Leung 2026-02-19 19:16:58 +08:00
• 1391378861 [Bugfix] Fix edge case in UUID data parsing (#34884) Cyrus Leung 2026-02-19 18:24:30 +08:00
• f6220f9877 [ROCm][Test] Fix beam search determinism failures from batch-size-dependent FP divergence and removed wrong marker (#34878) Andreas Karatzas 2026-02-19 02:25:26 -06:00
• 2df2bb27b0 [ROCm][CI] Removing all blocking labels from MI355 until stable infra (#34879) Andreas Karatzas 2026-02-19 01:53:08 -06:00
• f75b61a9e9 [Voxtral Realtime] Fix engine crash on empty multimodal embeddings (#34862) Tal NirandClaude Opus 4.6 2026-02-19 02:21:47 -05:00
• 7f51e93864 [Bug] Fix DeepSeek V3 weight loading caused by incorrect prefix (#34876) Wei Zhao 2026-02-19 02:20:30 -05:00
• 4611af1663 [Bugfix] Add Quant Config to Llava Next Projector (#34847) Alex Brooks 2026-02-19 00:18:23 -07:00
• ad5aa6bd9f fix(docs): fix typos in comments and docstrings (#34836) Manrique Vargas 2026-02-19 02:17:41 -05:00
• 9681068cf9 [Frontend] Fix reasoning_tokens for text-based parsers in Responses API (#33513) Jaeyeon Kim(김재연) 2026-02-19 08:16:41 +01:00
• b6101d384d Deprecate test-pipeline.yaml (#34864) Kevin H. Luu 2026-02-18 18:15:27 -08:00
• 5fcb0cdd68 [Model Runner V2] Use FP32 for Gumbel Noise (#34854) Woosuk Kwon 2026-02-18 17:07:37 -08:00
• c878b43b64 [Model Runner V2] Remove unnecessary copies in PW CUDA graph capture (#34849) Woosuk Kwon 2026-02-18 15:52:50 -08:00
• 2b84ac669c [CI][AMD][BugFix] Use torch.testing.assert_close instead of assert torch.allclose in test_rocm_skinny_gemms.py (#34181) rasmith 2026-02-18 17:10:19 -06:00
• 11d3976b88 [Model Runner V2] support piecewise & mixed cudagraph (#32771) zhrrr 2026-02-19 07:03:17 +08:00
• 40da9625a1 [MoE Refactor] Convert mxfp4 marlin into modular kernel format (#34588) Yongye ZhuandRobert Shaw 2026-02-18 17:37:14 -05:00
• 8d9babd4de Fix empty tool_call_id in Anthropic messages API tool result conversion (#34745) 3 people 2026-02-18 17:31:59 -05:00
• e99ba957ec [BUG] Fixing Weight Sync unit test (#34841) Aaron Hao 2026-02-18 14:20:10 -08:00
• 64ac1395e8 [Docs] Clean up speculators docs (#34065) Kyle Sayers 2026-02-18 16:48:11 -05:00
• 61cf087680 [Bugfix] Fix lora tests (#34834) Cyrus LeungandMichael Goin 2026-02-19 05:22:31 +08:00
• 847a57cd12 [Bugfix][MoE Kernel] Fix incorrect routing selection for models without expert groups (e.g., MiniMax-M2.1) (#34673) 3 people 2026-02-18 13:03:24 -08:00
• fcd6ac97ed [CI][AMD][BugFix] Skip tests in test_unquantized_backend_selection that should not run on ROCm (#34655) rasmith 2026-02-18 14:00:40 -06:00
• 95be2a7f22 [Model Runner V2] Minor simplification for DCP (#34786) Woosuk Kwon 2026-02-18 11:04:53 -08:00
• 0e60c925cf [Bugfix] Remove assert causing hipErrorStreamCaptureUnsupported (#34455) Jaden Mathias 2026-02-18 13:54:54 -05:00
• d7ff22204a [Misc] Add mooncake-transfer-engine to kv_connectors requirements (#34826) Teng Ma 2026-02-19 02:26:24 +08:00
• c0bd8b13da [Bugfix] Redo Qwen3.5/Qwen3-Next GDN projector fusion (#34697) Isotr0pyandJJJYmmm 2026-02-19 01:46:53 +08:00
• caeb887bf6 [Bugfix] Fix NVFP4 TRTLLM MoE non-gated support; add gsm8k for Nemotron-3-Nano FP8+NVFP4 (#34725) Michael Goin 2026-02-18 12:39:22 -05:00
• 6b3166a7c7 [CI][Bugfix] Fix multinode test script (#34820) Ilya Markov 2026-02-18 17:45:10 +01:00
• 25e2e136ef [CI] temporarily disable multi-node tests (#34825) Robert ShawandRobert Shaw 2026-02-18 11:32:44 -05:00
• 6874638bc4 [Model Bash] DeepSeek R1 BF16 Min Latency QKV A GEMM (0.5% E2E Speedup) (#34758) Robert ShawandRobert Shaw 2026-02-18 10:42:36 -05:00
• e24663c5a9 Add unit tests for fp8 output fusion of triton_attn (#34228) Burkhard RingleinandLuka Govedič 2026-02-18 12:22:49 +01:00
• c50e105a88 [Model Runner V2] Avoid prepare prefill kernel launch overhead (#34780) Nick Hill 2026-02-18 00:49:21 -08:00
• a766b30349 [Renderer] Deprecate code paths for old input processing (#34775) Cyrus Leung 2026-02-18 16:35:04 +08:00
• 1faa8cb73c [Quantization] - Added uses_meta_device_weights to quant config (#34645) Asaf Joseph Gardin 2026-02-18 09:43:44 +02:00
• e89a91d927 [Bugfix] fix activation in cpu_fused_moe_torch call (#34696) Marek Michalowski 2026-02-18 07:39:46 +00:00
• 909b147197 [Bugfix] Fix prefix creation for Qwen3.5 (#34723) Michael Goin 2026-02-18 02:39:15 -05:00
• a88b3be7c4 [Bugfix] Fix quant RMS norm fusion for quantization with TMA-aligned scales (#33255) ElizaWszolaandLuka Govedič 2026-02-18 08:35:04 +01:00
• a49ea5a58f [Model Runner V2] A bit more PP simplification (#34766) Nick Hill 2026-02-17 21:39:07 -08:00
• 30ebe0dc3c [CI/Build] Remove use of skip_v1 (#34699) Cyrus Leung 2026-02-18 12:19:11 +08:00
• cef65f0715 [ROCm][CI] Removed hard-coded attn backend requirement for Qwen VL (#34753) Andreas Karatzas 2026-02-17 21:59:53 -06:00
• 6f3b2047ab [Core] Fix SSRF bypass via backslash-@ URL parsing inconsistency (#34743) Russell Bryantandisotr0py 2026-02-17 22:53:35 -05:00
• 02e8f26cea [torch.compile] Turn on silu+fp4 quant fusion by default for O1+ (#34718) Luka Govedič 2026-02-17 22:29:15 -05:00
• 4a00a511bb [BugFix] [Build] fix string literals comparison in indexer_k_quant_and_cache calling site (#34653) Hongxia YangandHongxia Yang 2026-02-17 22:19:41 -05:00
• a0d8d944e2 [Renderer] Move MM Hash parsing into Renderer (#34711) Cyrus Leung 2026-02-18 11:18:55 +08:00
• df3f537a66 [CI] Remove unused precompiled wheel args from image build (#34767) Amr Mahdi 2026-02-17 18:58:18 -08:00
• 7743152957 [Attention] Refactor check_and_update_config (#33600) Matthew Bonanni 2026-02-17 20:06:54 -05:00
• ab33d2a629 [Feature] Decode Context Parallel support for GPU model runner v2 (#34179) Wentao Ye 2026-02-17 19:27:15 -05:00
• be3af2d29e [Model Runner V2] Further simplification for PP (#34724) Woosuk Kwon 2026-02-17 15:18:18 -08:00
• c656ba3b4d [Kernel] Triton-based Top-k and Top-p sampler kernels (#33538) 3 people 2026-02-17 15:14:30 -08:00
• dc5fa77a4e [Bugfix][MTP][Sparse MLA] Allow sparse MLA with MTP to run with FULL cudagraphs (#34457) Matthew BonanniandJee Jee Li 2026-02-17 14:01:27 -05:00
• 1e4a084c8e [CI] Fix flaky test_parsable_context (#34717) Flora Feng 2026-02-17 13:42:52 -05:00
• 7967e854da [BugFix] Fix sp tests (#34716) Richard Zou 2026-02-17 12:07:56 -05:00
• 6bd6d0c3c1 Fixed whisper CPU test that does not spawn properly. (#34324) almayneandCyrus Leung 2026-02-17 14:46:23 +00:00
• 8e962fef5f [CI][Nixl] Add CrossLayer KV layout tests (#34615) Nicolò Lucchesi 2026-02-17 14:35:40 +01:00
• 574fe75245 [Renderer] Move InputPreprocessor into Renderer (2/2) (#34560) Cyrus Leung 2026-02-17 21:29:01 +08:00
• c61a98f529 [CI][BugFix] ShellCheck cleanup to remove baseline and preserve runtime behavior (#34514) junuxyz 2026-02-17 21:22:56 +09:00
• 28bffe9466 Fix docs build warning (#34686) Harry Mellor 2026-02-17 10:31:40 +00:00
• ad65177a19 [Bugfix] Fix 'remove_instance_endpoint' method logic in disagg_proxy_demo (#32922) ChenqianCaoandHarry Mellor 2026-02-17 18:06:53 +08:00
• d44a5b6c47 Remove dead bitsandbytes CxB code from 8-bit inference path (#34633) Tim DettmersandClaude Opus 4.6 2026-02-17 04:49:14 -05:00
• 1d65283e95 Revert "[Models] Fuse Qwen3.5 GDN's qkvz_proj and ba_proj" (#34683) Jiangyun Zhu 2026-02-17 17:29:27 +08:00
• c464b57374 [Ray] Propagate third-party env vars to Ray workers via prefix matching (#34383) kourosh hakhamaneshiandCursor 2026-02-17 01:08:42 -08:00
• c5c38e152a [CI] Fix bake config artifact path for AMI rebuild pipeline (#34656) Amr Mahdi 2026-02-16 22:39:44 -08:00
• d00df624f3 [Model Runner V2] Minor refactoring for penalties (#34662) Woosuk Kwon 2026-02-16 21:43:00 -08:00
• 9752da9d9c [Model Runner V2] Minor simplification for BadWordsState (#34669) Woosuk Kwon 2026-02-16 21:27:24 -08:00
• 04925b2202 [Model Runner V2] Minor cleanup for PP (#34666) Woosuk Kwon 2026-02-16 19:15:31 -08:00
• d74278fb67 [Model Runner V2] Fix unintended CPU-GPU sync in make_dummy (#34667) Woosuk Kwon 2026-02-16 19:00:29 -08:00
• b68fd899d1 [Bugfix] Fix fused MoE int32 overflow in stride*offset without perf regression (#34507) haosdentandMichael Goin 2026-02-17 09:58:49 +08:00
• 0b5f9b7204 [CI] Enable mypy import following for vllm/v1/kv_offload (#34639) Aneesh Puttur 2026-02-16 20:58:15 -05:00
• 9a8853f781 [Core] Pipeline Parallel support for Model Runner V2 (#33960) zhanqiuhu 2026-02-16 20:48:16 -05:00
• 387a1898d9 [Model Runner V2] support bad_words sampling param (#33433) zhrrrandWoosuk Kwon 2026-02-17 08:36:06 +08:00
• 3b30e61507 [NemotronH] Do not force router to run in fp32 (#34582) roikoren755 2026-02-16 20:15:32 +02:00
• 824f9e8f3c Targeting the MI355 agent pool with all existing tests (#34629) Alexei-V-Ivanov-AMD 2026-02-16 11:02:27 -06:00
• 6cc403e67d [Bugfix][CI] Fix flaky entrypoints/openai/test_response_api_with_harmony.py::test_function_calling[openai/gpt-oss-20b] (#34624) Nicolò Lucchesi 2026-02-16 17:11:07 +01:00
• 72d5951d02 [Bugfix] Treat generation_config max_tokens as default not ceiling (#34063) Almog Tavor 2026-02-16 17:58:24 +02:00
• a3205beffb [CI] Enable mypy coverage for individual excluded files (#34292) Lucas KabelaandHarry Mellor 2026-02-16 07:34:29 -08:00
• 6930becd45 (bugfix): Fixed encode in LLM entrypoint for IOProcessr plugin prompts (#34618) Christian Pinto 2026-02-16 15:33:55 +00:00
• 03a8770a6d [ROCm][CI] Fix plugins test group; updating terratorch and dependencies (#34589) Andreas Karatzas 2026-02-16 09:33:42 -06:00
• bc56a1d56e [Bugfix] Fix ARC touch KeyError for non-ready T1 blocks in kv offload (#34576) Yiqi Xue 2026-02-16 07:33:19 -08:00
• ec7d9e6745 Fix call to moe_mk in modelopt MoE modules (required for LoRA) (#34575) danisereb 2026-02-16 17:33:09 +02:00
• 3bb4e4311c [Models] Fuse Qwen3.5 GDN's qkvz_proj and ba_proj (#34492) Isotr0py 2026-02-16 23:32:51 +08:00
• 08f8c198ae [CI] Disable precompiled wheel path in CI image builds (#34606) Amr Mahdi 2026-02-16 07:14:43 -08:00
• a21cedf4ff Bump lm-eval version for Transformers v5 compatibility (#33994) Harry Mellor 2026-02-16 14:24:35 +01:00
• 3ef74cde5d [CI][Tracing] Fix race condition by adding server readiness check (#34364) emricksini-h 2026-02-16 13:57:39 +01:00
• cd81cdb399 [Scheduler][ASR] Fix CrossAttn blocks per-request for Variable length encoder inputs (#31058) Ekagra RanjanandNicolò Lucchesi 2026-02-16 06:08:44 -05:00