Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• e52407ef4c [ROCm][CI] Add MTP and other spec-decode acceptance coverage (#53399) Andreas KaratzasandOpenAI Codex 2026-09-02 03:27:05 -05:00
• 8052102c21 [Bugfix] Fix cross-batch buffer race corrupting DiskBackend loads (#51667) Fabian JoswigandYifan Qiao 2026-09-02 10:18:41 +02:00
• 01eeb798b6 [CI][AMD] Preserve diagnostics for unwritable checkouts (#53437) Andreas KaratzasandOpenAI Codex 2026-09-02 03:18:27 -05:00
• f5711fa138 [CI] Shard LoRA TP distributed tests (#52350) Kevin H. Luu 2026-09-02 01:16:44 -07:00
• b2558f8da3 [Bugfix] Fix launch render hanging on shutdown (#54913) Chauncey 2026-09-02 16:15:35 +08:00
• c23e15be9f [CI][Fix] Resolved the Ascend NPU test build image fail and add file dependencies (#50504) Zeyu YangandYikun Jiang 2026-09-02 16:14:00 +08:00
• 878ec4bfee [CI] Shard entrypoints API-server tests (#52344) Kevin H. Luu 2026-09-02 01:13:38 -07:00
• 62588e0592 [CI] Batch the swap_blocks verification instead of copying block by block (#54558) stefankoncarevic 2026-09-02 10:12:40 +02:00
• 1b4b2a18b8 [CI] Shard H100 MoE refactor integration tests (#52352) Kevin H. Luu 2026-09-02 01:10:14 -07:00
• df09c76735 [KV Connector][Offloading] Look through UniformTypeKVCacheSpecs in the canonical portability gate (#51690) 4 people 2026-09-02 10:43:26 +03:00
• 798b557e06 [Frontend] Add site-packages support for reasoning/tool parser plugins (#45241) Yusuke OdaandChauncey 2026-09-02 16:41:13 +09:00
• f4e6136146 [Kimi-K3] Bump FlashKDA to fix unstable inverse (#54859) Thien Tran 2026-09-02 15:16:29 +08:00
• c00091e026 [Frontend] Gate scale-out endpoints behind opt-in flag (#54579) Francisco Javier Arceo 2026-09-02 02:31:14 -04:00
• 300f688321 [Bugfix] Implicitly close DeepSeek DSML parameters (#54838) Flora Feng 2026-09-02 02:13:36 -04:00
• 3976eada8b [CI] Use PR head label for Buildkite branch to avoid main collision (#54895) Kevin H. Luu 2026-09-01 23:09:50 -07:00
• aae36577fb [XPU][UT] skip fp8_per_channel test on XPU (#54861) Qiming Zhang 2026-09-01 22:17:10 -07:00
• 56b5495377 [ROCm][AMD][Installation] Add mooncake build to rocm base image (#52650) Giuseppe GrossiandAndreas Karatzas 2026-09-02 01:14:58 -04:00
• ad76610c40 [XPU][CI] Move heavy jobs to nightly test in Intel GPU CI (#54863) xiangdong 2026-09-02 13:11:01 +08:00
• 73029d4244 [CI][ROCm] Calibrate AMD test timeouts from nightly runtimes (#54695) Andreas KaratzasandOpenAI Codex 2026-09-01 23:53:12 -05:00
• f870b92976 [Qwen3.8-Flash-Next] Fuse Qwen4Exp PLE kernels (#54517) Thien TranandCodex 2026-09-02 12:32:16 +08:00
• 05201d8ecf [CI] Shard basic model initialization tests (#54753) Kevin H. Luu 2026-09-01 21:31:57 -07:00
• 396c5a5632 [Bugfix] Lazy-import FlashInfer PCIe IPC all-reduce in kernel_warmup (#54869) Gabriel WuandKimi Code 2026-09-02 12:31:03 +08:00
• 1e300895ac [Qwen4] validate FP8 PLE weight scale after loading (#54722) Harry Huang 2026-09-02 12:30:58 +08:00
• ee3c00bbf4 [Performance] Register Triton W4A16 GEMM as a custom op (#51453) 3 people 2026-09-01 23:55:34 -04:00
• 1d8d7a3965 [Bugfix] Fix RoPE construction for deepseek-v4 sparse SWA layers (#54815) Isotr0py 2026-09-02 11:48:50 +08:00
• a566ea7e8f [Perf] Avoid more h2d copies from non-pinned tensors (#54660) Nick Hillandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-09-01 20:37:09 -07:00
• 3ba9907a1d [Kimi-K3] Overlap low-M TP8 KDA projections (#54697) Yongye ZhuandOpenAI Codex 2026-09-01 20:35:28 -07:00
• 12b9573c98 [Bugfix][KV Offloading] Fix eager SimpleCPUOffload cache registration and final flush (#53532) 4 people 2026-09-02 12:23:09 +09:00
• dbf1a044ea [Bugfix] Fix ColQwen3.5 pooler projector initialization (#54847) Divakar VermaandAndreas Karatzas 2026-09-01 22:35:43 -04:00
• 46c8a161f4 [Bugfix] Handle padded routes in CUTLASS MoE permutations (#54747) 3 people 2026-09-01 19:08:17 -07:00
• 003e34341a [Qwen3.8-Flash-Next] Separate prefill and decode paths for QSA indexer (#54513) Thien Tran 2026-09-02 09:08:10 +08:00
• 80389cfedd [CI/Build] Gate PR title check on ready PRs & use slim runners (#54827) Bugen ZhaoandCodex 2026-09-02 09:12:00 +10:00
• e01d4acbb1 [ROCm][CI] Handle tied experts in softplus sqrt top-k test (#52679) Andreas Karatzas 2026-09-01 18:04:14 -05:00
• 3b6c0bdae9 [CI] Shard long kernel test groups (#54754) Kevin H. Luu 2026-09-01 15:58:27 -07:00
• 96031b8623 [CI] Shard distributed model jobs above the 24h P90 threshold (#54752) Kevin H. Luu 2026-09-01 15:32:11 -07:00
• e90b608c42 [CI] Split nightly MTP acceptance tests (#52353) Kevin H. Luu 2026-09-01 15:29:33 -07:00
• b911fe85c4 [Rust Frontend] Attribute decoded text to tokens (#52910) Bugen Zhao 2026-09-02 08:25:31 +10:00
• 259a209bfa [Kimi-K3][Perf] Prefetch ll_bf16 router weights for M=1 (#53524) mingg26andOpenAI Codex 2026-09-01 14:40:19 -07:00
• a56e74afd6 [CI] Remove MRV2-specific tests (#54823) Nick Hill 2026-09-01 14:15:12 -07:00
• fc72fc39ac [Kimi Bug] Fix cannot access local variable 'active_non_spec_mask_cpu' (#54781) Wentao Ye 2026-09-01 17:08:50 -04:00
• ce6a283c09 [Bugfix] Restore weight_dtype in QuarkW8A8Fp8MoEMethod to fix GPT-OSS FP8 MoE weight loading (#54824) Micah Williamson 2026-09-01 16:08:11 -05:00
• 18c53727ce [Bugfix][KV Connector] Fix DecodeBench DCP block selection (#54679) Jeff (Junze) MaandOpenAI Codex 2026-09-01 16:31:30 -04:00
• 16f16876fd [CI] Read the CRCR report token from a Buildkite secret (#54605) Andrey Talman 2026-09-01 16:29:23 -04:00
• 7a977c0699 [ModelOpt] Redesign the LinearMethod classes using the generic QuantKey-driven method (#49381) 3 people 2026-09-01 13:28:56 -07:00
• 4bf06be985 [CI] Disable CUDA graphs for GLM PCP evals (#54745) Kevin H. Luu 2026-09-01 13:24:46 -07:00
• dc5cf437cf [Rust Frontend] Enable Qwen4-exp multimodal support (#54813) Isotr0py 2026-09-02 04:23:55 +08:00
• 5cc32fbdfa [Bugfix][KV Connector] Fix Mooncake physical-block transfer length (#54272) Zhewen Liandzhewenl 2026-09-01 13:23:46 -07:00
• 7cb9a88f50 [Bugfix][Mooncake] Offload producer partial tails on request finish (#52832) 3 people 2026-09-01 13:23:35 -07:00
• 55178f2d09 [Bugfix][Profiler] Fix API server crash on double /stop_profile (#51678) Alberto MassiddaandAlberto Massidda 2026-09-01 22:05:04 +02:00
• 73723b707f [ROCm][MoE] Fix gfx950 block scale swizzle for AITER Triton MXFP4 W4A16 (#54773) stefankoncarevic 2026-09-01 21:48:45 +02:00
• 3439bad37e [Rust Frontend] Bound recursive argument parsers (#54303) Bugen Zhao 2026-09-02 05:45:50 +10:00
• ab54f5bd83 [Chore] Remove redundant _pack_topk_ids_weights_kernel in TrtLLM NvFP4 MoE (#46872) Julien Debache 2026-09-01 21:30:11 +02:00
• aa71f9bcc4 [Bugfix] Log platform plugin detection failures (#52285) Luyi XiaoandOpenAI Codex 2026-09-01 12:11:52 -07:00
• c0adee9231 [ROCm][CI] Add ROCm misc ops and env tests (#53279) Divakar VermaandAndreas Karatzas 2026-09-01 15:03:37 -04:00
• 6c58595c2d [Feature] Avoid flashinfer autotune each time when vllm source change (#54794) Wentao Ye 2026-09-01 15:03:30 -04:00
• d98bb2a879 [Bugfix][Frontend] Honor skip_decoder_start_token in async encoder-decoder rendering (#54799) waizuichougou 2026-09-02 03:00:28 +08:00
• 0ad5652a52 [Bugfix][Frontend] Restore the chat template content format mismatch warning (#54622) Pratham ShirbhateandCyrus Leung 2026-09-01 23:47:54 +05:30
• d1c15e589d [CI] Speed up quantization test group (#53291) fxmarty-amdandAndreas Karatzas 2026-09-01 20:15:24 +02:00
• e7cf4730d6 [Bugfix] Drop incomplete tool-call markup in non-streaming to match streaming (#47562) 3 people 2026-09-02 03:11:36 +09:00
• 2f01039666 [Misc] Share Buildkite CI failure skill across agents (#54806) Misha Goin 2026-09-01 13:22:09 -04:00
• c866ba9d11 [KV Connector] Support heterogeneous TP sharing in Mooncake Store Connector (#53129) 周峥andGuanyi Chen 2026-09-02 01:18:06 +08:00
• 2fe5cef35e [Fix] Fix FSE compatibility detection for Quark-produced models (#54573) 3 people 2026-09-01 19:12:20 +02:00
• 514c7314a0 [Perf][Kernel] Initialize NVFP4 padding in quant kernel (#53568) 3 people 2026-09-01 18:49:07 +02:00
• cdefd9d499 [Bugfix] Support Sentence Transformers 5.4+ serialized configs (#54533) Eli Nabro Yeongmin OhandCodex 2026-09-02 00:53:38 +09:00
• f1e5fdd7f2 [Transformers backend] Replace vocab embeddings in recursive_replace (#54760) Harry MellorandClaude Opus 5 2026-09-01 16:35:24 +01:00
• 82b7d49a6e [MoE] Generalize masked activation for padded layouts (#51217) Misha GoinandOpenAI Codex 2026-09-01 11:34:13 -04:00
• a232e29e9d [Bugfix] Gate sm_100-only kernel tests on the capability family, not >= (#54306) Bojiang Li 2026-09-01 08:29:15 -07:00
• 76f3249fbd [Mypy] Fix typing for M models (#54262) Taneem Ibrahim 2026-09-01 10:26:00 -05:00
• d9eb4e344f [Bugfix] Reject tokenless chat and audio streams (#54708) Taneem Ibrahim 2026-09-01 10:24:19 -05:00
• 339e16cbb6 [Bugfix] Support MCP SDK 2.x tool input schemas (#53870) Chauncey 2026-09-01 23:02:39 +08:00
• 0d4ad47981 [Kernel] Add B12X causal paged attention backend (#52017) Luke AlonsoandLucas Wilkinson 2026-09-01 07:42:02 -07:00
• adebc41b7e [Mamba] Add FlashInfer ReplaySSM backend (#52506) Andrii SkliarandAndrii Skliar 2026-09-01 16:26:43 +02:00
• 25efcfa788 [Attention] Enable adaptive verification for FLASHINFER_MLA_SPARSE_DSV4 (#52724) Ilya Markov 2026-09-01 15:43:28 +02:00
• 1f1f628859 [Feat][MM Hashing] include media_io_kwargs in multi-modal hashes (#54241) Jan Kwieciński 2026-09-01 15:17:01 +02:00
• 55aa766dc8 [Bugfix][Model] Fix GraniteMoeHybrid per-expert quantized weight loading (#54052) Priyansh Jain 2026-09-01 18:08:53 +05:30
• 40824284bc [Doc] Document FP8 GEMM kernel selection and Blackwell support (#49936) 3 people 2026-09-01 05:22:40 -07:00
• 191cecd51e [Kernel][Qwen] Add Hopper LL-GEMM tuning table for Qwen4Exp (#54560) Zheng Cai 2026-09-01 19:21:41 +08:00
• 481839ad9e [Feature][Spec] Support disabling trailing prefix-cache block dropping (#53388) Ziming HuangandOpenAI Codex 2026-09-01 19:12:36 +08:00
• 9e905f7450 [Bugfix] Account for client queue time in serve benchmarks (#54136) 3 people 2026-09-01 19:40:51 +09:00
• 92ccd2c306 [Bugifx][INC] Fix INC quantization method selection for non-quantized layers (#47237) 3 people 2026-09-01 18:40:47 +08:00
• 754d5e1f65 [CI/Build] Fix entrypoints coverage (#54750) Cyrus Leung 2026-09-01 18:33:02 +08:00
• 504bb8b0c3 [CI] Add repository-local OTel tracing helpers (#52851) Kevin H. LuuandCodex 2026-09-01 02:36:46 -07:00
• 8f03625b3d [CPU][Zen] Route Int8 MoE inference through zentorch on AMD (#44834) 3 people 2026-09-01 15:01:31 +05:30
• c35551f892 [Bugfix][KV Offload] Isolate tiering shutdown failures (#52290) AlexHuangandOr Ozeri 2026-09-01 17:24:54 +08:00
• 4707679cd2 [Bugfix][MiniCPM-V] Route video_embeds to the shared vision parser (#54633) Subhash Polisetti 2026-09-01 02:16:07 -07:00
• 5414b4e694 [XPU][TEST] Add entrypoints test in Intel GPU CI (#53980) xiangdongandKunshang Ji 2026-09-01 16:57:55 +08:00
• ce7391712b [Bugfix][Security] Bound embedding densification before to_dense() (#54632) lzhan011andCyrus Leung 2026-09-01 03:45:14 -05:00
• 8905633687 [Bugfix][Frontend] Preserve token offset origins after left text pre-trimming (#54692) waizuichougou 2026-09-01 16:02:56 +08:00
• b65af5e339 [CI][ROCm] Expand weight loading test coverage on AMD and cap its KV cache (#54037) stefankoncarevic 2026-09-01 09:57:22 +02:00
• 63988f3c2d [Quantization][Refactor][1/N] Adopt QuantKey in QuarkConfig and methods, relying on weight_quant_key, act_quant_key for quant method dispatch (#52958) fxmarty-amdandAndreas Karatzas 2026-09-01 09:55:54 +02:00
• d0e695a91b [Distributed] Support pre-shared ncclUniqueId rendezvous for weight transfer (#53784) dharak-cohereandCursor 2026-09-01 02:55:57 -04:00
• fa99a6fea6 [Bugfix][Security] Bound the validation-error response body (#54684) lzhan011 2026-09-01 01:51:16 -05:00
• 4ac452ad98 [Core] Release NCCL communicator memory in sleep mode (#51485) 3 people 2026-09-01 14:31:16 +08:00
• ec32f669bb [Feature][MM_UUIDs] Allow empty video URLs when using multi-modal UUIDs (#54220) Jan Kwieciński 2026-09-01 07:55:27 +02:00
• ff0c3cb03c [Bugfix][Frontend] Truncate the assistant tokens mask with the prompt (#54539) Hotragn Pettugani 2026-09-01 01:53:22 -04:00
• 30dd1a7954 [DecodeBenchConnector] Fix HMA cache-group mapping (#54647) Jeff (Junze) MaandOpenAI Codex 2026-09-01 01:52:50 -04:00
• 40b2f62061 [ROCm][CI] Stabilize the sqrt-softplus top-k tie oracle (#54403) Andreas Karatzas 2026-09-01 00:34:40 -05:00
• 7c5dc571cb [Attention][DSA] Enable W4A16 DSA (#51724) sychen52 2026-08-31 21:37:47 -07:00
• ce2e343be1 [ROCm] Keep GLM-5.2 on MRV1 and disable default breakable cudagraph (#53155) 3 people 2026-08-31 23:17:54 -05:00
• dc9114b201 [ROCm][MoE] Split AITER CK and Triton MXFP4 W4A16 into separate backends (#50622) afriedriandAndreas Karatzas 2026-08-31 23:14:14 -05:00