Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 9dd2e72828 fix flaky multi example connector consistency (#48206) 3 people 2026-07-15 02:20:34 -05:00
• d119beb1b9 [ROCm] Add tuned selective_state_update config for AMD MI350 (#48159) Giuseppe Grossi 2026-07-15 03:18:09 -04:00
• 12a8057bfe [CI/Build] Split release artifact annotations by type (#48600) Kevin H. LuuandOpenAI Codex 2026-07-15 00:00:52 -07:00
• e281ac663a [Rust Frontend] Integrate MM audio support (#48554) 3 people 2026-07-15 15:00:17 +08:00
• adce068118 [ROCm][CI] fix test_common.py (#48676) Charlie FuandAndreas Karatzas 2026-07-15 01:42:06 -05:00
• b6770d7b54 [ROCm] Run init test engine in-process to avoid KV-cache OOM (#48527) djramicandAndreas Karatzas 2026-07-15 08:39:31 +02:00
• 3b39fd284a [Bugfix][Spec Decode] Support heterogeneous QK fusion geometry (#48671) aoshen02andOpenAI Codex 2026-07-15 13:37:10 +08:00
• 6472131298 [Bugfix] Set kv_quant_mode on the generic MLA KV-cache spec (#48379) drakoshaandMatthew Bonanni 2026-07-15 06:36:52 +03:00
• 37aa52821d Build with ABI stable FlashMLA (#48174) Jane (Yuan) XuandShengqi Chen 2026-07-14 23:29:28 -04:00
• 96d2ceda4b [Security] Replace diskcache to eliminate pickle deserialization (#44549) Russell BryantandClaude 2026-07-14 23:29:24 -04:00
• fdf2cf66d3 [LoRA][1/N] Integrate flashinfer MoE LoRA for BF16 model (#48632) Jee Jee Li 2026-07-15 10:54:00 +08:00
• 9b2be4e9a5 [Quant] Enable humming w[2-7]a[4,8] inference with compressed-tensors (#46390) HDCharles 2026-07-14 22:22:31 -04:00
• 3ad85e0de4 [CI][AMD] Configure MI300 tests for native execution without DinD (#48387) Andreas Karatzas 2026-07-14 21:14:16 -05:00
• 4f7fffb92f [Core][LoRA] Support fp32 lm_head (head_dtype) on the LoRA path (#48525) Karthik KothuriandClaude 2026-07-14 18:51:09 -07:00
• 6e073440b1 [ROCm][CI] Remove mxfp4 test skips after amd-quark 0.12 release (#47330) 4 people 2026-07-14 20:25:05 -05:00
• f7aadae5e5 add pad-aware reduce path (#48385) gnovack 2026-07-14 18:05:50 -07:00
• 442c421e79 [Perf] Remove redundant repeat and copy for dsv4, 1.8% E2E TPOT improvement. (#48137) Wentao Yeandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-14 20:48:10 -04:00
• 0bd6b85a1f [Bugfix] Preserve unloaded non-persistent buffers during layerwise reload (#44371) JoanandDakai An 2026-07-15 01:46:29 +01:00
• 3ca242d1b6 [Bugfix][R3] Exclude draft routers from expert capture (#48622) aoshen02 2026-07-15 08:45:35 +08:00
• 7e950521b3 fix: size FlashInfer prefill workspace to batch head footprint (#48428) Joe Rowell 2026-07-15 02:18:41 +02:00
• 0f0f28b537 [Bugfix][CI] Fix test_head_dtype quant_method test on ROCm (#48654) Micah Williamson 2026-07-14 18:31:36 -05:00
• 520a20ba4e [Bugfix] MoRIIO toy P/D proxy: add /health (#45222) 4 people 2026-07-15 07:56:46 +09:00
• 9182e86971 Log fully resolved pooling config at startup (#48030) Taneem Ibrahimandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-14 18:00:46 -04:00
• 313d01f507 [CI][Bugfix] Fix FlashAttention reported MLA dimension support (#48631) Matthew Bonanni 2026-07-14 17:33:02 -04:00
• 05d4f8bba3 [ROCm][CI] fix flashinfer import check (#48647) Divakar Verma 2026-07-14 16:54:19 -04:00
• 0b54201a04 [CI] Build macOS arm64 CPU wheel natively on the macmini queue (#48289) Michael Goin 2026-07-14 15:40:26 -04:00
• 32e632dfeb [Reasoning] Optimize TPOT for thinking budget when used with speculative decoding (#46662) rishitdholakia13andCursor 2026-07-14 14:55:40 -04:00
• 7ffb98e248 [ROCm] Retune MI355 selective_state_update float32 config on the unified effective_batch grid (#48373) vanshbhatia-amdandvanshbhatia-amd 2026-07-14 14:26:35 -04:00
• cdaa40d2a8 [KV Offload] Split cpu_cache_usage_perc into write/read usage gauges (#47666) 3 people 2026-07-14 12:13:41 -05:00
• ca3618bc69 [Doc] Sync four function docstrings with their signatures (#45437) 3 people 2026-07-14 10:10:13 -07:00
• b2f7d2560a [Bugfix] Make MLA+SWA check the layer's backend, not the model config (#48520) Michael Goin 2026-07-14 12:53:34 -04:00
• 1ff9429655 [CI Bug] Fully solve accuracy issue for DSv3.2 + MTP + Sequence Parallel (#48036) Wentao Ye 2026-07-14 10:00:24 -04:00
• af453e5647 [Bugfix] Gemma4 parser: classify channel-less output consistently in streaming and non-streaming (#48262) adhi29andBen Browning 2026-07-14 19:00:16 +05:30
• 32aef44388 [Bugfix] Include inline per-token-head scales in offloaded page transfer width (#48411) 3 people 2026-07-14 16:07:26 +03:00
• 7a74a9662b [NIXL] Avoid reading expired blocks in bidirectional turn-2 read (#47021) tomerg-nvidiaandNickLucche 2026-07-14 15:03:41 +02:00
• b6754f536e [Model] Enable LoRA support for tower and connector in LlavaNextVideo (#48594) karthik 2026-07-14 20:09:38 +08:00
• 793cf79c89 [Bugfix][Security] Fix concurrent sparse invariant race bypassing CVE remediation (#48583) Juan Pérez de Algaba 2026-07-14 13:08:24 +02:00
• 50ac1c7bab [Misc] Rename VLLM_TRITON_ATTN_USE_TD to VLLM_TRITON_USE_TD (#45781) Artur FierkaandClaude 2026-07-14 12:32:57 +02:00
• f04d3f640e [Test] Enable KV cache events for HMA models in CPU offloading test (#47754) Itay EtelisandOr Ozeri 2026-07-14 12:22:27 +03:00
• 0a9396a25e [XPU][CI] Add tests/v1/e2e/general/test_correctness_sliding_window.py in Intel GPU CI (#47231) xiangdong 2026-07-14 16:50:16 +08:00
• 038ec293b1 [Bugfix] Return 400 instead of 500 when multimodal data is sent to a text-only model (#48473) Hoang NguyenandClaude Opus 4.8 2026-07-14 15:15:43 +07:00
• 894ebb27f5 Add Cosmos3 Edge Reasoner model (#48291) adsridharandBartosz Stefaniak 2026-07-14 01:14:50 -07:00
• c9a788eedc fix(security): guard lm-format-enforcer regex compile with timeout (#47595) Juan Pérez de Algaba 2026-07-14 09:18:11 +02:00
• 0762f2afeb [Perf][Feat] Add generic cuteDSL LL BF16 router (GEMM) (#42562) Roberto L. CastroandLucas Wilkinson 2026-07-14 08:01:21 +02:00
• 31be872f55 [ROCm] Retune MI355 selective_state_update float16 config on the unified effective_batch grid (#48372) vanshbhatia-amdandvanshbhatia-amd 2026-07-14 01:16:29 -04:00
• 94c0ef3001 [Misc] Clean up "swap_space" (#48549) wangxiyuan 2026-07-14 12:43:45 +08:00
• af1f036a70 [Bugfix] Skip minimax_m3 tool parser tests when Rust extension is absent (#48523) Matt Woodson 2026-07-14 00:43:22 -04:00
• 95aab66e95 [ROCm][MiniMax-M3][Spec Decode] Support speculative decode with AITER sparse PA (#47984) 5 people 2026-07-14 12:12:53 +08:00
• dcf4072da9 [Perf][ROCm] Fix GDN KKT warmup regression on RDNA by avoiding fp32 tl.dot (#45000) nemanjaudovic 2026-07-14 05:48:54 +02:00
• 382bbd5144 [ROCm][Kernel] Add HybridW4A16LinearKernel: Triton prefill + HIP skinny decode (#40977) Matthias GehreandAndreas Karatzas 2026-07-14 05:22:00 +02:00
• b50ef9c6ed [ROCm][MiniMax-M2] Dispatch fused QK-norm + AllReduce via AITER (#44849) akii96andPawel Kowalski 2026-07-14 06:10:16 +03:00
• 9e289c553c up FI fp8 moe topk to 32 (#44462) Dan Blanaru 2026-07-14 04:58:16 +02:00
• c4f5cd60da [1/N] Add dense MHA path for sparse MLA short sequences (#47327) Matthew BonanniandClaude Opus 4.8 2026-07-13 20:29:56 -04:00
• 0b0ef8d7eb [Quantization][INC][ARK] Support INT2 XPU WOQ Linear (#47521) Zhenzhong XuandKunshang Ji 2026-07-14 08:29:45 +08:00
• 21472f32ea add pad-aware swiglu limit kernel (#48287) gnovackandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-13 16:48:19 -07:00
• fec64fea75 [BugFix] Correct OTEL span start time for Dynamo compilation (#40698) emricksini-handSimon Mo 2026-07-13 16:25:03 -07:00
• 8b8af2caf7 [Frontend] Expose logprob_token_ids on Python OpenAI endpoints (#43463) langzhao-netizenandClaude 2026-07-13 14:40:21 -07:00
• 7738ef35b8 [Feat] Add Support for BertForMaskedLM to vLLM (#48463) Snehlata 2026-07-14 02:26:25 +05:30
• 9a21f0d1a3 [BugFix] Initialize model_config for Qwen3-VL MoE (#44863) wenpengw-nvandRoger Wang 2026-07-14 04:43:53 +08:00
• 8ac8375270 [Core] Preserve Marconi caching with selective hybrid cache retention (#47782) Nick Hill 2026-07-13 21:24:20 +01:00
• 7dc447dda7 Added sliding window attention support for qwen-eagle3 architecture (#47568) shanjiaz 2026-07-13 16:20:44 -04:00
• 7fc97042c3 Add DCP + Eagle support for Tokenspeed MLA backends (#48180) Pavani MajetyandJingyi Yang 2026-07-13 11:46:02 -07:00
• 18c4067a54 [ROCm][CI] Unblock AMD: Language Models Test (Extended Pooling) (#48513) Micah Williamson 2026-07-13 13:38:10 -05:00
• 550218b136 [Bugfix][Frontend] Flush engine reasoning parser at engine-reasoning → tool streaming boundary (#47606) akii96andBen Browning 2026-07-13 21:06:10 +03:00
• 5c342876a6 [Doc] Add DeepseekV32ForCausalLM to supported_models.md (#48293) Gavin Morris 2026-07-13 12:43:59 -05:00
• 9427c45386 [ROCm][CI] Transformers: pass only one of input_ids/inputs_embeds (#48258) stefankoncarevicandHarry Mellor 2026-07-13 19:28:50 +02:00
• 43c8cbf79b [EC Connector] CPU Offloading EC Connector (#47423) omerpaz95andOr Ozeri 2026-07-13 20:09:41 +03:00
• 62286308c9 [Misc] Improve Matryoshka pooling dimensions validation (#48057) Taneem Ibrahimandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-07-13 12:57:36 -04:00
• 26587f9519 [BugFix][ModelRunner V2] Fix stale attn metadata in speculator prefill cudagraph capture (#48261) Nick Hill 2026-07-13 17:39:15 +01:00
• 93e3bc8f30 [XPU][CI]Adjust timeout_in_minutes in Intel GPU CI (#48418) xiangdongandKunshang Ji 2026-07-13 23:11:16 +08:00
• c2c9f7c5e2 remove force channels_last in Idefics3MultiModalProcessor (#48467) Yan Ma 2026-07-13 22:18:57 +08:00
• 1be6e937b2 lower memory required for capturing cudagraphs for large cudagraph sizes (#48483) Omer Ullman Argov 2026-07-13 17:14:25 +03:00
• b3cfca996c [Mypy Fix] Split mypy work (#48490) Wentao Ye 2026-07-13 08:42:42 -04:00
• 487dfb3418 [CI] Add SPDX license header to Rust/Protobuf sources (#48472) Bugen Zhao 2026-07-13 17:22:47 +08:00
• 107a03ba63 [Core] Support fp32 lm_head for generation models via head_dtype (RFC #48305 §3.6) (#48390) 3 people 2026-07-13 01:43:34 -07:00
• 56a357ed33 [Bugfix][KV Cache] Don't route uniform-page-size MLA+SWA models into DeepseekV4 packing (#48256) Nicolò LucchesiandClaude 2026-07-13 10:16:24 +02:00
• bea70c7cfc [Attention] Make sliding-window support an explicit backend capability (#48011) 3 people 2026-07-13 10:07:56 +02:00
• 75fe92a316 [Distributed][Perf] Enable FlashInfer MNNVL allreduce RMS quant fusion (#48064) Mohammad Miadh Angkad 2026-07-13 15:02:59 +08:00
• b7b58d1eba [ROCm][CI] Cache Rust builds by source inputs (#46527) Andreas KaratzasandMatthew Wong 2026-07-13 01:14:08 -05:00
• 36484e464a [BugFix] Restore full tokens for Qwen MTP When MoE SP (#48429) Canlin Guo 2026-07-13 13:29:41 +08:00
• 9e57de7197 [CPU] Create Proper Numa topology for s390x (#40714) Rehan KhanandLi, Jiang <jiang1.li@intel.com> 2026-07-13 10:28:43 +05:30
• 8c5dafcd09 [Bugfix][UT]Fix EagleMiniCPMForCausalLM meet TypeError (#48452) Yejing Lai 2026-07-13 12:37:23 +08:00
• 05fa8183a6 [CPU][Spec Decode] Support DFlash speculative decoding for GDN models on CPU (#46090) 3 people 2026-07-13 07:16:18 +03:00
• d973cce3ca Re-disable CUDA graph memory profiling on ROCm (#48440) Rohan PotdarandClaude Opus 4.8 2026-07-12 23:59:20 -04:00
• 775c1589ea [Bugfix][ROCm] Keep TP all_gather on base-class collective (#48446) Fangzhou AiandCursor 2026-07-12 23:53:53 -04:00
• 2595d5cebc [Model] Optimize Qwen3.5 on H20 (#48350) zzt 2026-07-13 11:30:48 +08:00
• ee5a89f4d7 [ROCm][MiniMax-M3] Add AITER sparse paged attention (#47287) 5 people 2026-07-13 10:27:29 +08:00
• e26264f3ef [Kernel] Implement CUDA kernel for ReLUSquaredActivation (relu^2) (#39058) Tanish MalekarandClaude Sonnet 4.6 2026-07-12 19:18:03 -07:00
• 4c81772e8b [Bugfix][KV Offloading] Fix stale transfer_jobs after reset_cache + harden job completion (#48102) AlexHuang 2026-07-13 01:00:04 +08:00
• 27c3e579f0 [CI][Rust Frontend] Pin cargo tool versions (#48222) Bugen Zhao 2026-07-12 23:34:26 +08:00
• 8df14cfc8c [EC Connector] Add EC Transfer Params (#42433) omerpaz95andOr Ozeri 2026-07-12 14:35:33 +03:00
• 370b678a02 [CI][2/N] reduce CI time (#48394) Jiangyun Zhu 2026-07-12 19:16:55 +08:00
• 5c0c987c03 Make tiering offload region DP-replica aware (#47987) liranschourandOr Ozeri 2026-07-12 13:10:21 +03:00
• 5f8e73cb8b [Bugfix] Guard mixed-dtype allreduce RMSNorm quant fusions (#48330) Hugo Centeno 2026-07-12 11:39:27 +02:00
• 83762b77b0 [Frontend] Add /abort_requests to the RLHF dev API router (#47173) 3 people 2026-07-12 14:21:02 +08:00
• a02984ed47 [Perf][Qwen] Replace MOE all-reduce with reduce-scatter (#47006) Canlin Guoandyewentao256 2026-07-12 14:14:49 +08:00
• fc1c548093 Runtime Draft Weight Update for Speculative Decoding (#46725) 4 people 2026-07-12 13:51:53 +08:00
• 481e481be7 [2/N][Core] support partial prefix cache hit for hybrid model (#46384) Jiangyun ZhuandYifan Qiao 2026-07-12 13:37:51 +08:00
• 8e981630c9 [CI][CPU] Add Qwen2-VL multimodal tests for CPU backend and fix incompatibilities (#48072) zhao, zhenhui 2026-07-12 12:30:34 +08:00
• 9a48eef89a [Bugfix][LoRA] Support ark_linear base layer in _get_lora_device (#47690) Alejandro Paredes La Torre 2026-07-11 17:13:50 -07:00