Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 06c4873d95 Rewrite C++ meta funcs to Python (#28595) Jane (Yuan) Xu 2025-11-13 11:52:50 -05:00
• d3387750f1 [Misc] Turn off encoder torch compile by default (#28634) Roger Wang 2025-11-13 08:38:08 -08:00
• b230286fbc Fix get_num_experts when config sets it explicitly to None (#28652) Harry Mellorandbruceszchen 2025-11-13 16:02:42 +00:00
• 3035d1a166 [BugFix] DeepSeek-OCR: apply NoRepeatNGramLogitsProcessor to greedy path (#28617) Yuanping Song 2025-11-13 10:24:35 -05:00
• 07a606aa7e [CI Failure] Fix backend selection for encoder-only models (#28534) Huamin Li 2025-11-13 07:11:27 -08:00
• a7791eac9d [CI/Build] Install uv for AMD MI300: Language Models Tests (Hybrid) %N (#28142) amdfaaandzhewenli 2025-11-13 09:34:55 -05:00
• 8da2f28f53 [ROCm][BugFix]Fix get_cu_count in rocm_aiter_fa.py (#28618) Pleaplusone 2025-11-13 22:18:20 +08:00
• 86d15bfd8d [Hardware][PowerPC] Fix fp16 compilation error for Power in cpu attention backend and bump oneDNN version (#28535) Akash kaothalkarandAkash Kaothalkar 2025-11-13 19:02:21 +05:30
• c9fe6abe7c [Bugfix] Fix FPS value type for Qwen2.5-Omni video processing (#28630) Fanli Lin 2025-11-13 21:06:06 +08:00
• c47b6c85ac [XPU] add sym params to IPEXConfig (#28611) zofia 2025-11-13 19:35:04 +08:00
• c428e8d80b Fix io processor pooling #28273 (#28484) baonudesifeizhai 2025-11-13 06:34:14 -05:00
• 5e973209aa [BugFix] Fix type error when assign a trition kernel tensor to a torch.nn.Parameter (#28603) Zijing Liu 2025-11-13 03:30:04 -08:00
• e63fd44560 Fix: Correctly filter special tokens in benchmark_prefix_caching (#28615) Di Wu 2025-11-13 18:57:44 +08:00
• 11ac9ddd03 Support all interleaved layer types (#28485) Yong Hoon Shin 2025-11-12 22:57:20 -10:00
• 5c9ad138d5 [Frontend] supports interleaved thinking (#28531) Chauncey 2025-11-13 16:14:13 +08:00
• fa183e9271 [Bugfix] fix kimi-linear crash (#28445) Jiangyun Zhu 2025-11-13 15:59:58 +08:00
• 4ab34f6ef1 Add NUMA node validation for CPU thread binding (#28555) usberkeley 2025-11-13 15:03:52 +08:00
• c33b87e777 Use official xformers-0.0.33 built for PT 2.9 (#28600) Huy DoandRoger Wang 2025-11-12 22:48:53 -08:00
• 4504e8029b [Bugfix] Prevent crash on empty grammar string (#28210) tjandy98 2025-11-13 14:42:29 +08:00
• ca00b1bfc6 [ROCm][BugFix] Remove the usage of device_info from aiter (#28383) Pleaplusone 2025-11-13 13:43:42 +08:00
• d44fbbab0e [build][cmake]: Bundle static ACL and torch libgomp for CPU extension builds (#28059) Radu Salavat 2025-11-12 21:43:08 -08:00
• 7e082bc14e Support DeepEP for Kimi-k2-thinking through enabling gemm selection for compressed-tensor marlin wna16 (#28574) Lucia Fang 2025-11-12 21:40:45 -08:00
• dbbe0c756a [XPU] Support Triton path for LoRA operations on XPU (#28511) Fanli Lin 2025-11-13 13:31:42 +08:00
• 7dca0c90cb [BugFix][ROCm] Fix get_cu_count missing variable error (#28608) Pleaplusone 2025-11-13 13:18:56 +08:00
• 1a0b157a2e [Frontend][responsesAPI][1/n] convert responses API tool input to chat completions tool format (#28231) 3 people 2025-11-12 20:47:22 -08:00
• 7c38ed0f1c [Frontend] split append tool output (#28333) Andrew XiaandAndrew Xia 2025-11-12 20:03:23 -08:00
• a1d3866dda [n-gen] DO NOT repeatedly return finished child requests (#28591) Jialin Ouyang 2025-11-12 19:36:07 -08:00
• 97d1c99302 Rename clashing method names for vLLM model protocol (#27583) Harry Mellor 2025-11-13 03:14:33 +00:00
• 3226283461 [Docs] Add some details about what the MoE block needs for the Transformers backend (#28588) Harry Mellor 2025-11-13 03:12:14 +00:00
• 8832fff972 [BugFix] Fix mm_encoder_attn_backend arg type checking (#28599) Nick Hill 2025-11-12 19:06:03 -08:00
• a543e678b4 [Bugfix] Fix SM100 gpt-oss regression due to faulty attn sink support (#28561) Michael Goin 2025-11-12 21:40:59 -05:00
• 2dacd57394 [platform] Move get_cu_count to utils (#27005) wangxiyuan 2025-11-13 08:48:47 +08:00
• d75ad04818 [ROCm][Bugfix] Revert removing setuptools version restriction (#28592) Gregory Shtrasberg 2025-11-12 19:46:58 -05:00
• 52eadcec9e [Docs] Update meetups.md description (#28583) Michael Goin 2025-11-12 19:00:23 -05:00
• 51c599f0ec Skip models that cannot currently init on Transformers v5 (#28471) Harry Mellor 2025-11-12 23:43:57 +00:00
• 69d0e90313 [MoE][Kernel][Perf] Improve Shared Expert Stream Overlap (#28406) Alexander Matveev 2025-11-12 18:37:24 -05:00
• 4ca5cd5740 [Core][AMD] Migrate fully transparent sleep mode to ROCm platform (#12695) 3 people 2025-11-13 01:24:12 +02:00
• 10f01d5a3a [Bugfix] Adjust Marlin CUDA arch selection to 8.0+PTX;9.0+PTX (#28294) Michael Goin 2025-11-12 18:14:13 -05:00
• 3eb0c2673e [TPU] Support GCS path in VLLM_TORCH_PROFILER_DIR (#28487) QiliangCui 2025-11-12 14:31:14 -08:00
• d8140b9833 [ROCM] Fix ROCm warnings, environment flag access, and GEMM kernel naming for consistency in _aiter_ops.py (#28464) vllmellm 2025-11-13 05:46:57 +08:00
• 74a9a9faad [Performance][B200] Fix deepgemm prologue (#27897) Varun Sundar RabindranathandVarun Sundar Rabindranath 2025-11-12 16:13:03 -05:00
• 478ee511de [Misc]Fix typo in llm_engine.py (#28584) Wei Wei 2025-11-12 12:59:43 -08:00
• 58ce8d12b7 [BugFix] Priority scheduling and spec tokens preemption (#28558) Andy Lo 2025-11-12 20:29:21 +00:00
• 94a9ebcf31 [KV connector][WIP] KV cache proxy based on LMCache multi-process mode (#27902) Yihua Cheng 2025-11-12 12:25:43 -08:00
• a39dd7bb06 [CI] Skip "Multi-Modal Models Test (Extended) 3" test that's broken in current Transformers (#28559) Harry Mellor 2025-11-12 19:38:13 +00:00
• 64d57c3be7 [Model] [Config] Correctly identify granite-4.0-micro as non-hybrid model (#28563) Thomas Parnell 2025-11-12 19:17:55 +01:00
• a1e7fa362a [EPLB][ROCm]: support EPBL for ROCm backend (#27731) PerryZhang01andPerry Zhang 2025-11-13 02:16:35 +08:00
• bac904565f Implement ARC KV cache eviction policy for CPU offloader (#27039) albertoandOr Ozeri 2025-11-12 17:51:39 +00:00
• 304419576a [Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479) Benjamin Chislett 2025-11-12 11:56:40 -05:00
• a742134cc5 Remove deprecated fields from CompilationConfig (#27593) Harry Mellor 2025-11-12 16:10:28 +00:00
• 728a9eb70e [Misc] Refactor Attention kv transfer methods into decorator (#27816) Nicolò LucchesiandMark McLoughlin 2025-11-12 17:05:44 +01:00
• bc5bd45c7d [Refactor] Remove redundant TP gather/split in split_qkv in QwenVL (#28271) Canlin Guo 2025-11-12 23:56:47 +08:00
• f76e85c299 [Performance][Hopper] Avoid M dim padding to 4x for most cases (due to cuda graphs paddings) (#28492) Alexander Matveev 2025-11-12 10:51:43 -05:00
• 54aecd9ed5 Fix pre-commit (and XPU) on main (#28556) Harry Mellor 2025-11-12 14:13:41 +00:00
• 10138c92a5 [V0 deprecation] Deprecate use_v1 parameter (#28112) wangxiyuan 2025-11-12 22:03:52 +08:00
• a9d18b5107 [Bugfix] Fix gpt_oss packed_modules_mapping (#28536) Jee Jee Li 2025-11-12 21:02:06 +08:00
• edb59a9470 [ROCm] [Bugfix] Fix fused_qknorm_rope_kernel rocm compatibility (#28500) TJian 2025-11-12 05:01:14 -08:00
• c5f10cc139 add cpu option for p/d in nixl_connector (#28356) ZhengHongming888 2025-11-12 03:53:08 -08:00
• d143152308 [KVConnector] Enable get_block_ids_with_load_errors() in LMCache connector (#27978) 3 people 2025-11-12 18:44:58 +08:00
• a4730c1b4f [XPU]Fix crash due to removed VLLM_USE_V1 attribute (#28520) Chaojun Zhang 2025-11-12 18:20:55 +08:00
• d3ade61e42 [Model] fix glm4_moe_mtp load weights with GLM-4.6 checkpoint. (#27597) wuyaoxuehunandwuao.scotty 2025-11-12 17:14:00 +07:00
• 1761dea1a8 [BugFix]: --enable-lora with model granite-4.0-micro crash (#27733) yyzxw 2025-11-12 17:03:56 +08:00
• c748355e0d [CI] Introduce autorun_on_main feature (#27836) Huamin Li 2025-11-12 00:51:19 -08:00
• 91864b79b3 [CI/Build] Fix crash due to removed VLLM_USE_V1 attribute in EPD (#28521) 3 people 2025-11-12 15:09:33 +08:00
• ac0bb2c307 [Core] Cache vllm_is_batch_invariant (#28304) Lukas Geiger 2025-11-12 05:03:01 +00:00
• f31419ed8b [Benchmark] Add retry support to fix workload bias in multi-turn benchmark (#28493) ai-jz 2025-11-11 21:00:45 -08:00
• b9ce9a3013 [BugFix] Add fallback path in apply_rotary_pos_emb_flashattn for non-cuda platforms (#28447) Fanli Lin 2025-11-12 11:13:21 +08:00
• 4ccffe561f [Core] Encoder separation for Encode-Prefill-Decode Disaggregation (#25233) 7 people 2025-11-12 10:58:33 +08:00
• cbb799e314 [Model][Qwen3VL] Simplify get_mrope_input_positions using numpy (#28302) Lukas Geiger 2025-11-12 02:55:10 +00:00
• 9f0247cfa4 VLLM_USE_TRITON_FLASH_ATTN V0 variable deprecation (#27611) Andreas Karatzas 2025-11-11 20:34:36 -06:00
• 7f829be7d3 [CPU] Refactor CPU attention backend (#27954) Li, Jiang 2025-11-12 09:43:06 +08:00
• e1710393c4 [[V0 deprecation]]Remove VLLM_USE_V1 env (#28204) wangxiyuan 2025-11-12 09:22:16 +08:00
• 3f770f4427 [Performance] Cache loaded custom logitsprocs to avoid overheads (#28462) Isotr0py 2025-11-12 08:49:29 +08:00
• 48c879369f [Frontend] Change CompilationMode to a proper Enum (#28165) Yanan Cao 2025-11-11 16:46:18 -08:00
• 1788aa1efb [BugFix] Graceful handling of torch symm mem errors. (#27671) Ilya MarkovandMichael Goin 2025-11-12 01:41:54 +01:00
• d23539549a Use FLASHINFER MLA backend when testing fp8_kv_scale_compile (#28491) Adrian Abeyta 2025-11-11 18:34:58 -06:00
• 412e153df5 [Feature] Allow configuring FlashInfer workspace size (#28269) Max Huandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2025-11-11 18:32:20 -05:00
• e5f599d4d1 [Bugfix] Disable shared expert overlap if Marlin MoE is used (#28410) Michael Goin 2025-11-11 18:16:12 -05:00
• 28534b92b9 Add Zurich vLLM Meetup (#28488) Michael Goin 2025-11-11 17:53:59 -05:00
• d4902ba56d [Misc] Cleanup Executor interface (#28441) wangxiyuan 2025-11-12 06:28:07 +08:00
• df4d3a44a8 [TPU] Rename path to tpu platform (#28452) Kyuyeun Kim 2025-11-11 11:16:47 -08:00
• 9d1c474704 [LoRA][1/N]Remove LoRA extra vocab (#28382) Jee Jee Li 2025-11-12 03:06:21 +08:00
• 8c32c6e4b4 [Misc] fix typo in DCP comment (#28389) Jie Luo 2025-11-12 02:59:16 +08:00
• de120bc94f [V0 deprecation] Clean up num_prefill_tokens logic for V0 (#28203) Canlin Guo 2025-11-12 02:57:12 +08:00
• 4228be7959 [Perf] Use np.ndarray instead of list[list[int]] to reduce GC overhead (#28245) Jialin Ouyang 2025-11-11 10:28:47 -08:00
• 76e4dcf225 [Misc] Remove unused attention prefix prefill ops functions (#26971) Lukas Geiger 2025-11-11 18:26:04 +00:00
• d5edcb8678 [BugFix] Fix Siglip2Attention on XPU (#28448) Fanli Lin 2025-11-12 02:18:02 +08:00
• 6c3c0f8235 [Kernel] Optimize rms_norm kernel (#27931) Xin Yang 2025-11-11 10:02:23 -08:00
• 684f254585 Prefer FlashAttention MLA as default over FlashMLA (#27363) Matthew Bonanni 2025-11-11 11:13:51 -06:00
• e553424919 [CI/Build] Refactor Attention backend for test_prefix_prefill from xformers to SDPA (#28424) Zhewen LiandRoger Wang 2025-11-11 09:09:47 -08:00
• 5a1271d83a [Quantization] fix attention quantization of gpt_oss model (#27334) xuebwang-amd 2025-11-12 01:06:00 +08:00
• 05576df85c [ROCm][Quantization] extend AMD Quark to support mixed-precision quantized model (#24239) 3 people 2025-11-12 01:05:22 +08:00
• 68c09efc37 [Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165) zhrrr 2025-11-12 01:00:31 +08:00
• a7ef3eb0cd [NIXL] Generalize block-first backend layouts (FlashInfer-like) (#28282) Nicolò Lucchesi 2025-11-11 17:57:43 +01:00
• f9a4087182 Remove weight_scale.T special case for SM90 Block FP8 CUTLASS kernel (#28431) Michael Goin 2025-11-11 09:46:04 -07:00
• 287bbbeb06 [Doc] Fix typo in serving docs (#28474) the-codeboy 2025-11-11 17:45:49 +01:00
• 3143eb23fc [BugFix] Add test_outputs.py to CI pipeline (#28466) usberkeleyandHarry Mellor 2025-11-12 00:01:30 +08:00
• b886068056 [BugFix] Fix RuntimeError in PixtralHFAttention on CPU/XPU (#28444) Fanli Lin 2025-11-11 23:29:33 +08:00
• a90ad7d838 Add @markmc to CODEOWNERS for Observability (#28457) Mark McLoughlin 2025-11-11 15:03:22 +00:00
• 533b018f72 [BugFix] Fix Failing Ruff Check (#28469) jvlunteren 2025-11-11 15:41:43 +01:00