Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 5d07e268b1 [Quantization][INC]Add MXFP8 Linear Support (#47514) 3 people 2026-07-27 14:26:31 +08:00
• d742856610 [3/N][Core][KV Connector] Support reliable partial-tail KV offload for sub-block prompts (#49502) Dao007foreverandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-26 23:22:17 -07:00
• 544cb724c8 [CPU][Spec Decode] Optimize GDN conv path for speculative decoding (#48577) 3 people 2026-07-26 23:20:14 -07:00
• c314af1abf [CPU][Perf] INT8 Fused MoE Kernel for Arm CPUs (#48637) Fadi ArafehandLi, Jiang <jiang1.li@intel.com> 2026-07-27 06:53:09 +01:00
• f19ee27e39 [Hardware][Power] Add FAST_EXP for Power (#49571) Akash kaothalkarandAkash kaothalkar 2026-07-27 11:17:29 +05:30
• 5f89a03dcb [CI] Explicitly tear down speculative decode runners (#49910) Andreas Karatzas 2026-07-27 00:32:46 -05:00
• 53397fbfac [Bugfix][KV Offload][P2P] Fix EngineCore crash reconnecting to a reaped peer (#49823) JasonandOr Ozeri 2026-07-27 01:07:00 -04:00
• 49f31d7cee [ROCm] Make vllm_c RMSNorm output contiguous (#49913) Andreas Karatzas 2026-07-26 23:59:14 -05:00
• 74d3b799e1 [Bugfix] Fix mHC block-M prenorm GEMM cross-row reduction carry-over (#49429) Nick Hill 2026-07-26 21:42:47 -07:00
• 29fdeab254 [XPU][CI] Add more test cases in Intel GPU CI (#49422) xiangdongandKunshang Ji 2026-07-27 12:40:22 +08:00
• ff6173997d [CI] Add kimi and k3 auto-labeling rules (#49895) 3 people 2026-07-26 21:00:27 -07:00
• 8de50e46d4 [Docs] Document NVFP4 GEMM kernel selection and Marlin weight-only fallback (#49376) Harjoth KharaandClaude Opus 5 2026-07-26 20:53:28 -07:00
• da99ffcc13 [ROCm][CI] Keep native datasets cache off shared NFS (#49516) Andreas Karatzas 2026-07-26 22:31:24 -05:00
• 8040ef2426 [Frontend] expose stream_interval as req sampling param (#49754) Walter Beller-Moralesandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-26 23:30:37 -04:00
• bf4f633b4c [XPU] Enable QK Norm + RoPE fusion pass on XPU (#49394) Chaojun ZhangandKunshang Ji 2026-07-27 11:22:29 +08:00
• 854c33f380 [CI][ROCm] Keep global GPU memory cleanup opt-in (#49911) Andreas Karatzas 2026-07-26 22:19:55 -05:00
• ac87549cbd [CI][ROCm] Reduce V1 attention test runtime (#49916) Andreas Karatzas 2026-07-26 22:12:45 -05:00
• 439f336212 [Core] Fix gpu<->cpu syncs in MRV2 mamba_hybrid.py (#49736) Nick HillandBenjamin Chislett 2026-07-27 03:41:27 +01:00
• ffc4f08c8e [Core][KV-transfer] MoRIIO: heterogeneous TP<->DP prefill/decode read routing (#46116) limeward 2026-07-26 19:10:00 -07:00
• 50aa830482 [BugFix][MRV2] Don't create dummy requests longer than max_model_len (#49751) Nick Hill 2026-07-27 03:04:20 +01:00
• f0553889c0 [Bugfix] Prevent NaN poisoning in xpu_mla_sparse for fully-masked index chunks (#48366) 3 people 2026-07-27 04:06:07 +03:00
• fdaa0d9e59 [ModelRunner V2] Support encoder-only attention (#49331) Nick HillandTaneem Ibrahim 2026-07-27 01:38:44 +01:00
• 0934b26790 [CI/Build] Refresh tags before building macOS wheel (#49901) Kevin H. LuuandOpenAI Codex 2026-07-26 13:07:32 -07:00
• 9e50e1037e [Bugfix][CuMem] Make KV-cache wake cleanup tag-safe (#49857) aoshen02andaoshen02 2026-07-27 03:09:13 +08:00
• b5b61c622c [Core][Distributed] Add process-checkpoint lifecycle hooks for communicators (starting with Flashinfer) (#46877) Schwinn Saereesitthipitak 2026-07-26 11:47:50 -07:00
• b68d7ef262 [Bugfix][KV Offload] Namespace auto cache dtype by effective dtype (#49438) 3 people 2026-07-27 02:59:09 +09:00
• 7154856f3d [Bugfix] Fix handling 5D KV cache in kv_postprocess_layout_on_receive (#47791) Daniel SocekandKunshang Ji 2026-07-26 10:22:42 -04:00
• 3f1d40960f [KV Offload] Fix num_tokens_after_batch for different termination types (#49285) 3 people 2026-07-26 21:22:58 +08:00
• 0da6e7f3d6 [Bugfix] Reject contradictory custom-op directives (#49134) Taneem Ibrahim 2026-07-26 08:42:14 -04:00
• 5559679229 [Bugfix][KV Offload] Bound unaligned SWA loads by physical GPU blocks (#49052) 5 people 2026-07-26 05:53:32 -06:00
• da3a252fd1 [KVOffload][P2P] Generic P2P secondary tier: peer lookup and serving via ParentManager (#48021) 3 people 2026-07-26 11:45:47 +03:00
• 21fd9e85a0 [Model] Support top_k and top_p sampling for DiffusionGemma (#45429) Guan-Ming Chiu 2026-07-26 16:39:25 +08:00
• 8d28b48d01 [Perf] Isolate MM preprocessing on its own executor (#49524) Guan-Ming Chiu 2026-07-26 16:04:17 +08:00
• 0164022c90 [CI] Fix speech correctness check rejecting improved WER (#49853) Taneem Ibrahim 2026-07-26 02:24:07 -04:00
• 30b0714031 [Perf] DeepSeek-OCR-2 TTFT Optimize (#49531) REDandIsotr0py 2026-07-26 13:53:06 +08:00
• 2e860de498 [Doc] Add compile cache volume example to the Docker deployment page (#49782) Nils Matteson 2026-07-25 23:24:01 -06:00
• 7eca0e1a64 [KV Offload] Deduplicate replicated MLA KV in the shared CPU region (#48906) Chang GuoandOpenAI Codex 2026-07-25 22:22:30 -07:00
• 7a29a3c54c [Bugfix][KV Offload] Namespace persistent cache by model runner (#49440) Jonguk CheongandOr Ozeri 2026-07-26 14:21:52 +09:00
• 1240c74c0a [Bugfix] Respect declared attention contract for ColQwen3.5 retrievers (#49372) Athrael Soju 2026-07-26 06:08:07 +02:00
• 48ebd6f2f1 [Bugfix][KVConnector] Disable cross-layer KV blocks for per-token-head quant (#49226) achyuthan.sandOr Ozeri 2026-07-26 06:24:27 +04:00
• b153ae6089 [XPU][CI] add heterogeneous TP UT (#49651) liuzhenwei 2026-07-26 10:01:15 +08:00
• 7a6a5b3667 [CI] Compute speech WER directly with jiwer (#49773) Chang Guo 2026-07-25 17:51:21 -07:00
• 0111002323 [Kernel] TD operand loads for batched MoE GEMM (moe_mmk) on XPU (#46340) Lena OnyshchenkoandKunshang Ji 2026-07-26 02:50:07 +02:00
• d30b1ecd1b [Bugfix][KV Offloading] Defer request finalization until final store (#49671) Rui YinandOr Ozeri 2026-07-26 04:58:09 +08:00
• dbd80cc031 [UX] DCP Topology Validation (#49777) Taneem Ibrahim 2026-07-25 16:53:57 -04:00
• 70009fb934 [MM][CG] Support ViT CUDA Graph for Gemma-4 (#46837) anthonsuandLinkun Chen 2026-07-25 13:02:09 -07:00
• ee1d996367 [Build] Fix for DeepEP manylinux pidfd sycall usage (#49814) Tyler Michael Smith 2026-07-25 15:29:38 -04:00
• 6b0103d1c9 [CI] Stabilize Pooling Rerank Equivalence Test (#49822) Taneem Ibrahim 2026-07-25 14:36:50 -04:00
• 9321aff536 [Bugfix] Wait for the linear bias before layerwise online processing (#49805) Harry MellorandClaude Opus 5 2026-07-25 18:56:32 +01:00
• 26d725c334 [Model] Add VaultGemma via Transformers modeling backend (#49803) Harry Mellor 2026-07-25 17:54:15 +01:00
• 7fe6d3c76b [Perf] Fix moe reduce_scatter perf regression by removing additional comm, 5% E2E throughput gain back. (#48763) Wentao Ye 2026-07-25 12:36:19 -04:00
• 2e0da24150 Mergify message not on cancelled (#45117) Harry Mellor 2026-07-25 17:20:24 +01:00
• 0b0bd2b5f6 [Feature] Add fault tolerance framework (simplified) for DP+EP external LB deployments (#44428) 10 people 2026-07-25 23:49:09 +08:00
• 33ef67e9fb [BugFix] Increase the max supported duration for MOSS-TD (#49403) Canlin Guo 2026-07-25 22:48:12 +08:00
• 3e74c60b9c [Docs] Use gen-files for generated docs content (#49587) Harry Mellor 2026-07-25 15:43:21 +01:00
• d1a8ba63d9 [Bugfix][MiniMax-M3] Fix token-major top-k buffer handling in Triton … (#49149) rongfu.leng 2026-07-25 21:45:27 +08:00
• 1423569ff5 [Bugfix][Tool Parser] Fix dropped streaming arguments in Jamba and InternLM2 parsers (#48852) mosya415andmosya415 2026-07-25 16:34:43 +03:00
• 9a50464698 [CI] Stop flaky test from downloading model every time (#49800) Harry Mellor 2026-07-25 14:29:36 +01:00
• b9b6306ebe feat[vLLM × v5]: Add audio support for the Transformers backend (#39330) 3 people 2026-07-25 15:20:58 +04:00
• ca0defa343 Make bare hugging_face imports forbidden (#49726) Harry MellorandAndreas Karatzas 2026-07-25 12:18:39 +01:00
• 0b1a8bb1f6 [Bugfix][CI] Fix stale Mooncake lookup expectation broken by a merge race (#49802) Harry MellorandClaude Opus 5 2026-07-25 11:10:26 +01:00
• fe5145765f [Core] Keep attention backends eligible for text-only serving of prefix-LM models (#48796) Tri VoandCyrus Leung 2026-07-25 17:06:14 +07:00
• dbcc1cdd0a [Model] Remove Ouro (#49786) Harry MellorandClaude Opus 5 2026-07-25 10:50:08 +01:00
• a82f1b388f [Perf][V1] Skip LRU hash-split in free_blocks when prefix caching is off (#48017) Agata DobrzyniewiczandClaude Opus 4.8 2026-07-25 11:06:20 +02:00
• 190be7dad2 [Docs] Fix confusing docstring indentation in nemotron_h.py (#49781) Johnny-Liou 2026-07-25 03:33:43 -04:00
• 94682b79f4 [multimodal] Make PyNvVideoCodec decoder concurrency configurable (#49753) Brandon PelfreyandOpenAI Codex 2026-07-24 23:31:41 -07:00
• 0ba2aa35a8 Stabilize GPU memory teardown between ROCm CI tests (#49242) 4 people 2026-07-24 23:21:36 -05:00
• aaaeda98dc [CI] fix compile test | refactor VLLM_DISABLE_COMPILE_CACHE for tests (#49770) Divakar Verma 2026-07-25 00:15:19 -04:00
• d9cd774198 [ROCm][CI] Force native compile caches onto local disk (#49763) Aarushi JainandAndreas Karatzas 2026-07-24 23:14:37 -05:00
• 70052fb924 [Bugfix][KV Connector][Mooncake] Keep TP-sharded Mamba state out of the KV-head dedup (#49499) Yifan QiaoandClaude 2026-07-24 21:02:22 -07:00
• 318b527cc2 [XPU] add warning for xpu graph limitations (#49419) liuzhenwei 2026-07-25 10:26:15 +08:00
• 6a1acac3fe [BUGFIX] Fix log capture in KV test (#49655) liuzhenweiandKunshang Ji 2026-07-25 08:36:53 +08:00
• 213f681f81 Revert "[Perf][GLM-5.2] Blackwell decode optimizations" (#49768) Woosuk Kwon 2026-07-24 16:04:59 -07:00
• 33c4f3551c [ROCm][CI] Wait for ROCm VRAM to settle between compiled and eager LL… (#49739) Aarushi Jain 2026-07-24 17:19:46 -05:00
• caa9cad31e [ROCm][Docker] Drop MORI_GPU_ARCHS so MoRI autodetects the device arch (#49737) Rohan PotdarandClaude Opus 4.8 2026-07-24 16:14:03 -05:00
• 7513d071bd [ROCm][CI] Fix XPASS(strict) on mixed audio embeds test (#49733) 3 people 2026-07-24 23:09:18 +02:00
• 89f6aa3a9e [KV Offload][CI] Fall back to buffered I/O without O_DIRECT; fix flaky api-server test (#49734) Harry Mellor 2026-07-24 21:56:57 +01:00
• 84d26b9ee3 [Model] Remove Plamo2 (#49729) Harry Mellor 2026-07-24 21:54:24 +01:00
• 9e6746b3c7 [CI] Stabilize memory-sensitive compile and structured output tests (#49749) Jiangyun ZhuandOpenAI Codex 2026-07-25 04:45:09 +08:00
• 972848f276 [Bugfix] Support non-uniform page sizes in KVBlockZeroer (#49704) Elvir CrnčevićandClaude Opus 4.6 2026-07-24 22:38:59 +02:00
• 2279575cd9 [AMD][Bugfix][EPLB] Fix elastic EP scaling accuracy on ROCm (#47206) 3 people 2026-07-24 14:35:27 -06:00
• 5d8e90a966 [WideEP] Update NCCL to 2.30.7 to enable DeepEPv2 in the vllm/vllm-openai image (#45321) 5 people 2026-07-24 16:00:02 -04:00
• e222c33f2f [Bugfix] Register axk1 config to fix A.X-K1 init (#49727) djramic 2026-07-24 21:57:49 +02:00
• c064fa52b6 Fix GLM-4.1V video placeholder token ID handling. (#49484) Aarushi JainandAndreas Karatzas 2026-07-24 14:35:16 -05:00
• 7e51939e25 [CI] Avoid unnecessary Hugging Face metadata requests (#49508) Andreas Karatzas 2026-07-24 14:33:53 -05:00
• 9863102ed9 [CI] Reuse loaded config for cached tokenizer (#49509) Andreas Karatzas 2026-07-24 14:32:40 -05:00
• 8c13ee5735 Add sm_107 for Rubin (#49387) Tyler Michael SmithandOpenAI Codex 2026-07-24 14:59:49 -04:00
• 41798069f3 [CI][AMD] Deprecate DinD for MI355 tests (#49257) Andreas Karatzas 2026-07-24 13:41:08 -05:00
• 866fea2b99 [Kernel] ReplaySSM: cache SSM inputs for faster Mamba2 standard decode (#48018) 5 people 2026-07-24 12:39:49 -04:00
• d02df748bf [Bugfix] Accept RFC 2397 parameters in base64 data URLs (#48973) Thomas FahrnerandClaude Fable 5 2026-07-24 08:23:08 -07:00
• 453f01783d [UX] Improve data-parallel launch validation (#49124) Taneem IbrahimandWentao Ye 2026-07-24 10:16:39 -04:00
• 7b40fb9645 [UX] Reject incompatible nested runtime overrides (#49247) Taneem Ibrahimandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-24 10:16:22 -04:00
• 8eac21a602 [ROCM] Fix AITER Fused AllReduce RMSNorm for Transformers Backend (#49673) BadrBasowid 2026-07-24 22:16:17 +08:00
• a454a1dd25 [Bugfix][Benchmarks] Restore --skip-tokenizer-init with custom dataset (#49180) mgazzandKevin H. Luu 2026-07-24 12:30:19 +01:00
• 833483f357 Encoder cache extension hooks (#48218) 3 people 2026-07-24 17:52:00 +08:00
• 163ecba377 [Bugfix] Skip linear bias in layerwise reload to avoid corruption (#49586) li-jinpengandxymli 2026-07-24 16:53:48 +08:00
• 589a5b884b [PD][NixlPush][Bugfix] Fix blocking handshake call on writer thread (#49221) Nicolò Lucchesiandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-24 10:41:37 +02:00
• 5c5434e2d8 Remove Quantization test parallelism (#49693) Kevin H. LuuandOpenAI Codex 2026-07-24 01:14:48 -07:00
• dd72658e7d [Perf][GLM-5.2] Blackwell decode optimizations (#48597) 7 people 2026-07-24 12:36:27 +08:00
• 0d77325b10 Bump Transformers version to 5.14.1 (#49223) Harry Mellor 2026-07-24 05:16:49 +01:00