Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 05c2dee7e9 [DeepSeek + LMCache Multiprocess] handle MLA for deepseek model + LMCache Multiprocess connector (#29039) Kuntai Du 2025-11-20 09:40:49 +08:00
• 1d642872a2 [torchao] fix safetensors for sharding (#28169) liangel-02 2025-11-19 19:39:45 -05:00
• 9ccef8e333 [Misc] Colorize logs (#29017) Nick Hill 2025-11-19 16:26:04 -08:00
• 537cc635c7 [GC Debugger] Simply and improve GC Debugger Utils (#29029) Jialin Ouyang 2025-11-19 16:10:22 -08:00
• 5031cd5d55 [Refactor] Optimize select_experts (#28069) Wentao Ye 2025-11-19 18:53:15 -05:00
• 3aaa94ac99 [Performance] Reduce DeepGEMM N dim restriction from 128 to 64 multiplier (#28687) Alexander Matveevandmgoin 2025-11-19 18:47:13 -05:00
• 8e38e99829 [Feature] EPLB on Qwen3VLMoe and CompressedTensorsWNA16MoEMethod (#28849) JartX 2025-11-20 00:30:08 +01:00
• 0075bfffd4 [CI] Fix precommit rope_theta issue (#29040) Wentao Ye 2025-11-19 17:22:43 -05:00
• cb0a7b4bea [Bugfix] Move flashinfer kernel check into ``__init__` function of `FusedMoE`` (#29018) Max Hu 2025-11-19 16:54:15 -05:00
• 8f4f77a727 [BugFix] Fix false assertion with spec-decode=[2,4,..] and TP>2 (#29036) Lucas Wilkinson 2025-11-19 16:43:54 -05:00
• 22e44ad589 [ROCm][CI] Fix Weight Loading With Multiple GPU Tests on ROCm (#28984) Micah Williamson 2025-11-19 15:31:33 -06:00
• 88f5b19f0b [DeepSeek] Fix DeepSeek V3.2 Rope Embedding (#28968) Yongye Zhu 2025-11-19 16:30:04 -05:00
• 613abb50d5 [MoE] Nvfp4 Masked Gemm: Add flashinfer grouped_gemm_nt_masked (#25990) Shu WangandMichael Goin 2025-11-19 15:29:06 -06:00
• cdeec2e606 [BugFix] Ray with multiple nodes (#28873) Julien Denize 2025-11-19 22:20:58 +01:00
• 1607e664f0 [Bug] Fix Batch Invariant MLA test (#28967) Wentao Ye 2025-11-19 16:18:32 -05:00
• 68d7231991 [CI/Build] Fix test_prefix_prefill for AMD (#28905) Ryan Rock 2025-11-19 15:04:36 -06:00
• 2fd893b4ce [Feature] Prefill Context Parallel (PCP) basic support (#28718) 6 people 2025-11-20 04:52:44 +08:00
• 02f5903b84 Eagle: MM Cuda Graphs with MRope (#28896) Izzy PuttermanandCyrus Leung 2025-11-19 12:01:05 -08:00
• ac10fd3c69 Upstreaming aiter triton attention backend as a new backend (#28701) Aleksandr MalyshevandAleksandr Malyshev 2025-11-19 11:59:30 -08:00
• 9d2d561257 [Bugfix] Fix precision corruption when shared_experts_stream=None (#28942) 杰兮andzhyajie 2025-11-20 03:30:57 +08:00
• fe69f331f8 [Kernels] Improve H200 Fused MoE Config (#28992) Robert Shaw 2025-11-19 14:23:54 -05:00
• 3319a493fc [Core] Reuse created spec tokens lists to mitigate GC cost (#28917) Jialin Ouyang 2025-11-19 11:20:22 -08:00
• 61728cd1df Re-enable FlashInfer for Llama4 on Blackwell in e2e fusion tests (#28966) 4 people 2025-11-19 13:32:19 -05:00
• 0c80efd94f GLM-V video segmentation solution adjustment (#28941) Yuxuan Zhang 2025-11-20 01:32:55 +08:00
• a8b70304d6 Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542) Harry Mellor 2025-11-19 18:06:36 +01:00
• d44e9df7d4 [Model][Mamba] Add selector for mamba attention backend and make it pluggable for other device (#26487) Shanshan Shen 2025-11-20 00:24:55 +08:00
• 48fc8b1e59 [BugFix] Fix async-scheduling + FlashAttn MLA (#28990) Lucas Wilkinson 2025-11-19 10:04:07 -05:00
• 1ffe934c8a [torch.compile] caching of config fields should be opt-out by default (#26468) 5 people 2025-11-19 06:13:54 -08:00
• 2c8b9182b5 [CI] Reorganize compile tests so new tests are automatically included in CI (#28625) Yanan Cao 2025-11-19 06:13:50 -08:00
• 4f5299f717 Relax Transformers modeling backend MoE experts check (#28952) Harry Mellor 2025-11-19 14:50:30 +01:00
• 09540cd918 [Doc]: fix typos in various files (#29010) Didier Durand 2025-11-19 13:56:21 +01:00
• da2f6800e0 [Feat][Perf] Enable deepep-low-latency with round-robin expert placement. (#28449) Chen BruceandHarry Mellor 2025-11-19 20:46:24 +08:00
• ba558c029a [config] Expose get_total_num_hidden_layers() in ModelConfig (#28961) 3 people 2025-11-19 13:37:11 +02:00
• 97cfa99d59 [Docs] Take env var definition out of folded admonition (#29005) Harry Mellor 2025-11-19 12:32:04 +01:00
• bbc6c2f1e5 [CI/Build] Fix broken build on Apple M1 (#28999) j20120307 2025-11-19 03:07:22 -08:00
• 8151609583 refactor(cpu_types_scalar.hpp): Unify scalar loop implementations using unroll_loop (#28847) ihb2032andlyd1992 2025-11-19 19:05:44 +08:00
• fdf93486d6 [Docs] Clean up moe_kernel_features.md (#28530) Michael Yao 2025-11-19 18:35:29 +08:00
• d69062c67a add support for --fully-sharded-loras in fused_moe (#28761) gnovackandJee Jee Li 2025-11-19 00:32:00 -08:00
• ae4821a108 Add CPU support model (#28697) Louie Tsai 2025-11-18 23:47:57 -08:00
• 7ed27f3cb5 [Doc]: fix typos in various files (#28945) Didier Durand 2025-11-19 07:52:30 +01:00
• a4511e38db Speed up macOS smoke test (#28954) Michael Goin 2025-11-19 01:46:32 -05:00
• 71d0ae1c54 [Misc] Update embedding/cross encoder tests to use mteb v2 (#27329) 5 people 2025-11-19 09:28:40 +03:00
• 3d4e7d34be [Model][QwenVL] Simplify cos/sin rotary embedding indexing (#28962) Lukas Geiger 2025-11-19 05:43:01 +00:00
• 6a25ea5f0e [Docs] Update oneshot imports (#28188) Uranus 2025-11-19 13:30:08 +08:00
• 73ff872db0 [Bugfix] Fix typo in Qwen3 Next model executor (#28960) Gleb Kurchanov 2025-11-19 08:21:02 +03:00
• 468a8d72ba [Bugfix] Fix FusedMoEModularKernel for triton backend (#28913) Xin Yang 2025-11-18 21:05:22 -08:00
• 4c23690f43 [Attention] FlashAttention ViT support, make default backend (#28763) Matthew Bonanni 2025-11-18 23:06:21 -05:00
• 814843e021 Enable bitsandbytes quantization on AMD GPUs that use warp size 32 (#27307) Strahinja Stamenkovic 2025-11-19 04:12:31 +01:00
• 20852c8f4c [CPU] Refactor CPU WNA16 (#28826) Li, Jiang 2025-11-19 10:32:00 +08:00
• 40b6b38f2c [Core] Switch Flat logprob control from environment variable to SamplingParams (#28914) Jialin Ouyangand22quinn 2025-11-18 18:10:02 -08:00
• da94c7c0eb Move online quantization to model.load_weights (#26327) Jerry Zhang 2025-11-18 16:52:41 -08:00
• 1395461f5f [Hybrid][torch.compile] Refactor mamba2 forward to avoid obscuring linear projections under custom op (#28587) tomeras91 2025-11-19 02:49:36 +02:00
• 9912b8ccb8 [Build] Add OpenAI triton_kernels (#28788) Varun Sundar RabindranathandVarun Sundar Rabindranath 2025-11-18 19:45:20 -05:00
• 49ef847aa8 [NVIDIA] Guard SM100 CUTLASS MoE macro to SM100 builds v2 (#28938) Johnny 2025-11-19 01:44:27 +01:00
• 67745d189f Supress verbose logs from model_hosting_container_standards (#28949) Michael Goin 2025-11-18 15:29:06 -05:00
• 2a2d5d2780 Replace torch.cuda.Event with torch.Event for better hardware compatibility (#26985) Kunshang Ji 2025-11-19 03:34:36 +08:00
• c3e2978620 [NIXL] fix cpu PD after physical <> logical block_size PR (#28904) Chendi.Xue 2025-11-18 13:03:23 -06:00
• e4bb2684bc [Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842) Isotr0py 2025-11-19 02:56:04 +08:00
• c64c0b78de [chore] Move the rest of wikimedia url to S3 (#28921) Kevin H. Luuandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2025-11-18 09:44:18 -08:00
• 0af3d4f0df [FEAT] [AITER] [ROCm] integrate aiter sampling ops (#26084) vllmellm 2025-11-19 01:28:34 +08:00
• da8dadf68b [Minor] Rename ec_producer field to is_ec_producer (#28884) Nick Hill 2025-11-18 09:26:07 -08:00
• f226a3f0c1 [CI][NIXL] Change default block_size for tests (#28927) Nicolò Lucchesi 2025-11-18 18:22:30 +01:00
• c2612371ad [Model] Add Gemma3 GGUF multimodal support (#27772) 3 people 2025-11-18 13:56:29 -03:00
• 49a986ecd4 [Benchmark] multi_turn: Report warmup-inclusive runtime (#28937) Ido Segev 2025-11-18 18:38:22 +02:00
• f6aa122698 [CI Sprint] Quantization CI Cleanup (#24130) Alex 2025-11-18 08:21:48 -06:00
• 184b12fdc6 [Bugfix][NIXL] Fix block_size_ratio when logical !=physical blocks (#28925) Nicolò LucchesiandCyrus Leung 2025-11-18 15:07:50 +01:00
• b9489f51e1 [Model][Perf] Use cos and sin cache in QwenVL (#28798) Canlin Guo 2025-11-18 19:51:54 +08:00
• 285eaa4285 [Bugfix] Safeguard against missing backend in AttentionBackendEnum (#28846) Song Zhixinandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2025-11-18 18:53:44 +08:00
• 439368496d [BugFix] Fix PP/async scheduling with pooling models (#28899) Nick HillandCyrus Leung 2025-11-18 00:20:45 -08:00
• 896e41ae04 [CI/Build] Replace wikipedia url with local server ones (#28908) Isotr0py 2025-11-18 16:10:55 +08:00
• 5bb1da5190 [MISC] Remove format.sh (#28906) Kuntai Du 2025-11-18 13:28:31 +08:00
• 5bdd155277 [CI] Fix async scheduling + spec decoding test flake (#28902) Nick Hill 2025-11-17 21:26:32 -08:00
• 0168f69e50 [Misc] Remove unnecessary parentheses from log statements (#28897) Ning Xie 2025-11-18 12:33:46 +08:00
• 083cf326dc [Doc]: fix typos in various files (#28863) Didier Durand 2025-11-18 05:32:14 +01:00
• bf9e1e8767 [Bugfix] Fix wrong CLI defaults for dynamic SchedulerConfig fields (#28872) Cyrus Leung 2025-11-18 12:30:29 +08:00
• 3ddcf46011 [Refactor] Remove Unused Func in Batch Invariant (#28881) Wentao Ye 2025-11-17 23:29:29 -05:00
• d0a73620cc [ROCm][Quantization] add apply_vllm_mapper in quark config for models like gpt-oss (#28638) xuebwang-amdandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2025-11-18 11:16:45 +08:00
• 88ab591f0b Run macos smoke test workflow on main commit (#28752) Michael Goin 2025-11-17 22:16:03 -05:00
• b6e04390d3 [Bugfix] Fix Kimi-K2 tool parser concatenated tool calls parsing (#28831) 3 people 2025-11-18 03:13:25 +00:00
• 552cac95b5 [Misc] Fix wrong comment in scheduler (#28880) Zhuohan Li 2025-11-17 15:32:22 -08:00
• 61485844fc [BugFix] Corner case that could cause out-of-sync with external launcher mode and dp >1 (#28774) Bangsheng Tang 2025-11-17 15:22:11 -08:00
• f77bce001a [Model] Add Afmoe architecture implementation (#28332) PranavandMaziyar Panahi 2025-11-17 15:11:20 -08:00
• a289cc1dde [Test] Batch Invariant: Rename and organize tests (#27421) Wentao Ye 2025-11-17 18:09:47 -05:00
• 95ae50b7d1 [Quantization] [Eagle] Add complete quantization support to the draft model in Eagle (#28435) Shreyas Kulkarni 2025-11-17 18:01:34 -05:00
• 7765e5ba75 [BugFix] Fix PP performance and PP kv connector output regression (#28768) Nick Hill 2025-11-17 14:08:50 -08:00
• d8874c61a5 [Core] Async Scheduling X Spec Decoding Compatibility (#24799) 3 people 2025-11-18 04:16:20 +08:00
• f8b19c0ffd [Bugfix] Fix GPT-OSS on AMD after #28603 (#28816) Zhewen Li 2025-11-17 10:15:26 -08:00
• e42bd8c2e3 Cast return value to int64_t for cache size (#28814) tiehexue 2025-11-18 00:02:32 +08:00
• 7f064491f8 [Bugfix][Perf] Revert applying HF processor on text-only inputs for multimodal models (#28858) Roger Wang 2025-11-17 06:49:25 -08:00
• 64e39d667c [BugFix] Temporary fix for IMA with MTP = 2 and full-cg (#28315) Lucas Wilkinson 2025-11-17 09:41:22 -05:00
• 1b82fb0ad3 [XPU] work around for sp, avoid custom op import error (#28822) Kunshang Ji 2025-11-17 21:16:44 +08:00
• d4acf518d0 [Metrics] Fix KV cache usage percent metric multiproc (#28792) Jae-Won Chung 2025-11-17 04:54:15 -05:00
• ab01cd14e5 [BugFix] Fix glm4_moe_mtp load weights bug (#28805) wuyaoxuehun 2025-11-17 16:13:11 +07:00
• 577bb34fff [CPU][Bugfix] Fix _to_list in CPU model runner (#28824) Li, Jiang 2025-11-17 15:47:24 +08:00
• 3380ed5e11 [Doc] Add llama4 LoRA tag (#28825) Jee Jee Li 2025-11-17 14:08:48 +08:00
• 6f37419244 [Bugfix][Model] Prevent special token leakage in KimiK2ToolParser streaming mode (#28543) Jay Caldwell 2025-11-16 23:54:46 -06:00
• 60e089f0b9 [ROCm][Qwen3-32B] Fix AITER MHA accuracy issue cause by #25763 (#28670) Xiake Sun 2025-11-17 12:52:11 +08:00
• d64429bb36 [NIXL][XPU] update install script of NIXL (#28778) liuzhenwei 2025-11-17 11:01:33 +08:00
• 561253b37f [Performance][Fix] update nvfp4 code to support renorm routing (#28569) jiahancandMichael Goin 2025-11-16 18:02:42 -08:00
• 80b6080ddc [BugFix] Fix async scheduling + chunked prefill + preemption (#28787) Nick Hill 2025-11-16 14:46:46 -08:00