Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• debd6e768c [XPU][MoE Refactor] Refactor xpu mxfp4 support into oracle (#37784) Kunshang Ji 2026-03-23 19:10:41 +08:00
• 9ace378a63 [Frontend][Responses API] Fix arrival_time recording for TTFT on initial request (#37498) Andrew Xia 2026-03-23 02:58:08 -07:00
• 27d5ee3e6f [FP8]add FP8 WoQ kernel abstraction. (#32929) Kunshang Ji 2026-03-23 17:47:47 +08:00
• 35141a7eed [Misc]Update gitignore (#37863) wangxiyuan 2026-03-23 16:14:10 +08:00
• e99fb98867 [ROCm] Fix fused_moe_fake signature mismatch and other AITER bugs (#36100) Chuan (Richard) Li 2026-03-23 00:48:31 -07:00
• a16133a0f1 [Perf] [Bugfix] Fix Triton autotuning in inference for Qwen3.5 (#37338) Artem Perevedentsev 2026-03-23 09:37:58 +02:00
• 54ab804e87 [Bugfix] Store Qwen3Next A_log in fp32 (#37810) Hojin YangandRoger Wang 2026-03-23 16:36:57 +09:00
• 02e6efe56d [Bugfix] JAIS: Only apply ALiBi when position_embedding_type='alibi' (#37820) r266-techandr266-tech 2026-03-23 15:36:34 +08:00
• 410d300893 [ROCm][Refactor] Enable AWQMarlinConfig on ROCm to use choose_mp_linear_kernel (#36505) Matthias GehreandMichael Goin 2026-03-23 08:36:08 +01:00
• d3fe857135 update doc for online fp8 quantization (#37851) Yan Ma 2026-03-23 13:19:03 +08:00
• f85e479e66 [Feature] ViT Full CUDA Graph (#35963) Baorun (Lauren) Mu 2026-03-23 01:01:10 -04:00
• 1f0d210641 [CI/Build][LoRA] Update Qwen35 LoRA testing (#37816) Jee Jee Li 2026-03-23 12:55:49 +08:00
• 3bbe2e1e6e [Test] Consolidate tool parser unit tests to tests/tool_parsers (#37834) Ben Browning 2026-03-23 00:24:25 -04:00
• 6e04e79326 always use embed&token_classify for bge-m3 (#37632) Augusto Yaoandwang.yuqi 2026-03-23 11:10:57 +08:00
• e7767eccae Fix AudioFlamingo3/MusicFlamingo HF parity and RoTE handling (#37643) Lasha Koroshinadze 2026-03-22 22:29:07 -04:00
• 43877a620b [MRV2] Enable PP CUDA graph test (#37830) Woosuk Kwon 2026-03-22 16:30:25 -07:00
• 63f49b8bd4 [Model Runner V2] Enable piecewise CUDA graphs for pipeline parallelism (#35162) zhanqiuhuandWoosuk Kwon 2026-03-22 16:48:25 -04:00
• a5e9d511de [MRV2] Use FP64 for Gumbel noise (#37798) Woosuk Kwon 2026-03-22 12:28:10 -07:00
• c058ff44d4 [Bigfix]fix lora test by pass padded size back to the layer (#37811) Yongye Zhu 2026-03-22 15:20:13 -04:00
• ce9b1d76cf [MRV2] Skip hidden states allocation for PW CUDA graphs (#37818) Woosuk Kwon 2026-03-22 11:47:21 -07:00
• e74c17e153 Enable NemotronHPuzzle + NemotronHMTP (#37803) Netanel Haber 2026-03-22 17:13:58 +02:00
• eaf4978621 [Test] Only Run MLA model when user explicitly set for batch invariance (#37719) Wentao Ye 2026-03-22 09:09:12 -04:00
• 77d24c4bfe [Bug] Fix fp8 deepgemm batch invariant (#37718) Wentao Ye 2026-03-22 08:57:20 -04:00
• b3e846017d [Model Runner V2] Support multi-modal embeddings for spec decode model (#36097) Giancarlo DelfinandWoosuk Kwon 2026-03-22 02:48:43 -07:00
• cd1242d82a [ROCm][CI] Stabilize ROCm speech-to-text translation test with lower min acc threshold (#37723) Andreas Karatzas 2026-03-22 04:32:08 -05:00
• 4383f1532e [MoE] Move PF Methods to Folder (#35927) Robert Shaw 2026-03-22 04:42:59 -04:00
• 6eedec6e36 [ROCm][CI] Make some duplicated tests optional so that they are only evaluated in our nightly (#37780) Andreas Karatzas 2026-03-22 03:03:18 -05:00
• ffc8531524 [ROCm][CI] Added missing resampy dependency for MM audio tests (#37778) Andreas Karatzas 2026-03-22 03:02:41 -05:00
• 6ecba840d7 [ROCm][CI] get_cu_count was renamed to num_compute_units in #35042 (#37764) Andreas Karatzas 2026-03-22 03:02:21 -05:00
• 3b06c55c78 [ROCm][CI] Fix MEGA_AOT_ARTIFACT fallback when PyTorch < 2.10.0 lacks AOT support (#37763) Andreas Karatzas 2026-03-22 03:02:03 -05:00
• b050700462 [Perf] Optimize glm4.xv VIT (#37779) Yang Liu 2026-03-22 02:12:34 -04:00
• 5dac719b2b [Bugfix] Handle libsndfile sf_error(NULL) race condition in audio fallback (#37782) Andreas Karatzas 2026-03-22 00:37:29 -05:00
• c862481c02 [CI] Skip ISAAC multimodal tests due to broken upstream HF model weights (#37781) Andreas Karatzas 2026-03-22 00:23:32 -05:00
• c86b17cfe6 [ROCm][CI] Add large_gpu_mark to test_max_tokens_none for ROCm (#37717) Andreas Karatzas 2026-03-21 23:25:16 -05:00
• 66f927f205 [Bugfix] Fix pooling non-determinism from pinned prompt_lens aliasing (#37775) Andreas Karatzas 2026-03-21 22:22:24 -05:00
• e78bc74268 [ROCm][CI] close missing quote in kernels/moe block in run-amd-test.sh (#37774) Andreas Karatzas 2026-03-21 20:42:34 -05:00
• 6b2fa3a762 [MoE] Move FlashInfer CuteDSL experts into fused_moe/experts/ (#37759) Robert Shaw 2026-03-21 19:15:16 -04:00
• eeee5b262d [Quantization][Deprecation] Remove PTPC FP8 (#32700) Robert ShawandRobert Shaw 2026-03-21 18:10:16 -04:00
• 5ad0446572 Revert "Consolidate AWQ quantization into single awq_marlin.py file" (#37768) Robert Shaw 2026-03-21 17:20:41 -04:00
• 8cc700dd6a Consolidate AWQ quantization into single awq_marlin.py file Robert ShawandClaude 2026-03-21 17:09:17 -04:00
• 80b70884eb Add tensor IPC transfer mechanism for multimodal data (#32104) Brandon PelfreyandNick Hill 2026-03-21 13:10:20 -07:00
• 61e381dcf0 [Perf] Add SM 10.3 (B300/GB300) all-reduce communicator tuning (#37756) Mohammad Miadh Angkad 2026-03-22 03:43:47 +08:00
• 88f1b374f5 [Core] Enable allreduce fusion by default for SM 10.3 (B300/GB300) (#37755) Mohammad Miadh Angkad 2026-03-22 03:40:37 +08:00
• 298e510848 [Hybrid] calling get_mamba_groups() once at MambaCopyBuffers.create() (#37318) Francesco Fusco 2026-03-21 10:29:43 +01:00
• 3982bc2cd0 [ROCm] Enable DeepEP ROCm as all2allbackend for AMD GPUs. (#34692) Chaitanya Sri Krishna LollaandTej Kiran 2026-03-21 13:02:31 +05:30
• 02eec7ecbe [ROCm][CI] Update GSM8K eval config to use fp8-and-mixed models list (MI355) (#37721) Andreas Karatzas 2026-03-21 02:27:12 -05:00
• 17ee641c45 [Responses API] Add kv_transfer_params for PD disaggregation (#37424) Bongwoo BakandChauncey 2026-03-21 14:48:54 +09:00
• 0d50fa1db6 [ROCm][CI] Mark gemma3 as large GPU test to avoid OOM on MI250 (#37610) Andreas Karatzas 2026-03-20 23:57:25 -05:00
• 1fa1e53a73 Revert "[compile] Initialize passes at VllmBackend init" (#37733) Simon Mo 2026-03-20 21:35:49 -07:00
• 3ffa52009f [ROCm][CI] Guard CudaPlatform/RocmPlatform imports to fix test collection on cross-platform builds (#37617) Andreas Karatzas 2026-03-20 22:58:58 -05:00
• 87bd91892f [MoE Refactor] Mxfp4 oracle rebased (#37128) Yongye ZhuandClaude Opus 4.6 2026-03-20 22:37:04 -05:00
• c7f98b4d0a [Frontend] Remove librosa from audio dependency (#37058) Isotr0py 2026-03-21 11:36:15 +08:00
• 1c472f8fe1 Add get_device_uuid for rocm (#37694) tmm77 2026-03-20 23:33:16 -04:00
• c57d38d603 elastic_ep: Fix issues with repeated scale up/down cycles (#37131) Itay AlroyandRon Tourgeman 2026-03-21 01:13:02 +02:00
• e5ed6c6c13 [BugFix] Allow qk_nope_head_dim=192 in FlashInfer MLA backend checks (#37475) Kaihang Jiang 2026-03-20 18:14:55 -04:00
• b3d0b37908 [Refactor] Remove unused dead code (#36171) Wentao Ye 2026-03-20 18:12:51 -04:00
• 85f671b8e1 [Model Runner V2] Support Streaming Inputs (#37028) Santino Ramos 2026-03-20 13:42:25 -07:00
• 8bc6b5cdb0 [ROCm][CI] Setting some mi325_4 tests back to optional (in parity with upstream) (#37711) Andreas Karatzas 2026-03-20 14:25:08 -05:00
• 4f16ebbbd3 [Bugfix] Disable monolithic TRTLLM MoE for Renormalize routing (#37591) (#37605) Vadim Gimpelson 2026-03-20 23:19:26 +04:00
• 12fd17eb51 [compile] Initialize passes at VllmBackend init (#35216) Angela Yi 2026-03-20 11:40:33 -07:00
• 37aadf6237 [Model] Update Kimi-K25 and Isaac processors to fit HF-style (#37693) Cyrus Leung 2026-03-21 02:30:22 +08:00
• d7d2b5e405 [Bugfix] Disable --calculate-kv-scales for hybrid GDN/Mamba+Attention… (#37565) Le Yang 2026-03-21 02:28:34 +08:00
• 6ec5e9fd37 refactor: abstract deepgemm support into platform (#37519) SherryC41andsherryC41 2026-03-21 01:54:08 +08:00
• e1d85e5c24 [Attention] Support distinguishing between short extends and decodes (#37303) Lucas Wilkinson 2026-03-20 10:49:36 -07:00
• 79eb9369c5 fix CUDAGraph memory being counted twice (#37426) Peter PanandMatthew Bonanni 2026-03-21 01:36:32 +08:00
• e80cfe575d [MRV2] Avoid recompilation of _gather_block_tables_kernel (#37645) Woosuk Kwon 2026-03-20 10:31:45 -07:00
• d0532bf38d [Perf] Eliminate redundant SparseMatrix creation in gpt_oss_triton_kernels (#37683) Xin Yang 2026-03-20 10:28:41 -07:00
• fb4e8bf442 [ROCm][CI] Fix accuracy for llama-nemotron-vl pooling tests (#37613) Andreas Karatzas 2026-03-20 12:16:59 -05:00
• 6ade4bc5a5 Fix various config related issues for Transformers v5 (#37681) Harry Mellor 2026-03-20 16:30:12 +00:00
• 2e089b96a8 [compile] Add compiled artifact counter for VLLM_USE_MEGA_AOT_ARTIFACT=1. (#37589) Zhengxu Chen 2026-03-20 12:22:46 -04:00
• 880be2b1b8 [Metrics] Some small refactoring for better maintainability (#33898) Martin Hickey 2026-03-20 16:11:34 +00:00
• c0f5fae601 [compile] Fix aot test failures with torch 2.12. (#37604) Zhengxu Chen 2026-03-20 12:06:29 -04:00
• aa84e43ccb [Pixtral] Enable Pixtral language model support Eagle3 (#37182) Rémi Delacourt 2026-03-20 16:50:15 +01:00
• 5e806bcf54 [Bugfix] Fix ConchLinearKernel channelwise quantization (group_size=-1) (#37329) Matthias Gehre 2026-03-20 16:32:21 +01:00
• 56a62c310c [Bugfix] Reject channelwise quantization (group_size <= 0) in ExllamaLinearKernel (#37331) Matthias Gehre 2026-03-20 16:31:57 +01:00
• 1779c09898 [ROCm] Enable wvSplitK skinny GEMM kernel for RDNA4/gfx1x decode (#34709) L.B.R.andL.B.R. 2026-03-20 15:11:23 +00:00
• 44eea10f68 [ROCm][Quantization] make quark ocp mx dtype parser robust for weight-only quantization (#36232) xuebwang-amd 2026-03-20 23:10:03 +08:00
• 8b6c6b9505 [Model] Add LFM2-ColBERT-350M support (#37528) Ilya Boytsov 2026-03-20 15:57:57 +01:00
• 9f6d9dd371 Fix attribute error in isaac_patch_hf_runner (#37685) Harry Mellor 2026-03-20 14:49:40 +00:00
• dd20ee4e3e [UX] Enable torch_profiler_with_stack (#37571) Jee Jee Li 2026-03-20 19:17:26 +08:00
• 0523449c9c [Misc] Use logger.info_once for auto tool choice log message (#37661) Chauncey 2026-03-20 18:40:36 +08:00
• b4c1aef21c [Refactor] Relocate tests from tests/v1/entrypoints/ to tests/entrypoints/ (#37500) Flora Feng 2026-03-20 05:50:34 -04:00
• 6050b93bed [Refactor] Move serve entrypoint tests under tests/entrypoints/serve/ (#37595) Flora Feng 2026-03-20 05:10:47 -04:00
• 5a4a179591 [ROCm][CI] Fix granite_speech test for gfx90a by selecting compatible attention backend (#37611) Andreas Karatzas 2026-03-20 04:07:26 -05:00
• 37cd9fc107 [ROCm][CI] Remove deepep DBO tests on gfx90a (#37614) Andreas Karatzas 2026-03-20 04:07:07 -05:00
• 9cfd4ebb5e [ROCm][CI] Update GSM8K eval config to use fp8-and-mixed models list (#37619) Andreas Karatzas 2026-03-20 04:06:53 -05:00
• ed359c497a [Model] Deprecate the score task (this will not affect users). (#37537) wang.yuqi 2026-03-20 16:07:56 +08:00
• dcee9be95a [Model Runner V2] Fix draft logits not populated during cudagraph replay (#37639) Giancarlo Delfin 2026-03-20 00:43:47 -07:00
• bd8c4c0752 [CI] Removing deprecated rlhf examples reference (#37585) Andreas Karatzas 2026-03-20 02:20:33 -05:00
• 0140eafb15 [Bug] Fix FlashInfer allreduce fusion workspace uninitialized error (#37461) 3 people 2026-03-20 03:09:21 -04:00
• bdf6a0a57b [XPU] bump vllm-xpu-kernels to v0.1.4 (#37641) Kunshang Ji 2026-03-20 15:04:38 +08:00
• 0674d1fee7 [PluggableLayer][MM] Add PluggableLayer for CustomQwen2Decoder (#37293) Wangbei25andWangbei25 2026-03-20 14:24:07 +08:00
• 30108fc8b0 [Model] Refactor Step3-VL processor to HF style (#37579) Cyrus Leung 2026-03-20 14:05:08 +08:00
• e2d1c8b5e8 [Refactor] Relocate entrypoint tests to match serving code structure (#37593) Flora Feng 2026-03-20 01:31:23 -04:00
• 6951fcd44f [XPU] Automatically detect target platform as XPU in build. (#37634) Huanxing 2026-03-20 13:30:15 +08:00
• 39474513f6 [Model Runner V2] fix draft attention metadata generation (#37364) Giancarlo Delfin 2026-03-19 21:05:15 -07:00
• 638a872d77 fix(xpu): Re-compute compile ranges after platform-specific config updates (#37523) Yuxiang Liangandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2026-03-20 11:52:35 +08:00
• 9040151fe1 [V0 Deprecation] Deprecate --disable-frontend-multiprocessing (#37612) Flora Feng 2026-03-19 23:31:43 -04:00
• 8fbe3f303f [Bugfix][LoRA] Fix Qwen35 LoRA (#36976) Jee Jee Li 2026-03-20 11:09:32 +08:00
• ea2c148fa7 [compile][graph_partition]Add tensor size handling (#36038) Xiao 2026-03-19 19:55:25 -07:00