Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 41f9104fa6 [DSv4.1] Integrate Mega-Gate from DeepGEMM (#56266) 3 people 2026-09-18 03:06:36 +08:00
• a9a7e45f31 [ROCm] Restore VLLM_ROCM_USE_AITER_FP4_ASM_GEMM and default w4a4 ASM GEMM back to off (#57055) afriedri 2026-09-17 13:39:52 -05:00
• 6a344d8c85 [Bugfix][Metrics] Do not log a 0.0% prefix cache hit rate before any query (#54990) Michele Campi 2026-09-17 20:19:13 +02:00
• 09c379ab61 [Bugfix] Max-load throughput cliff when max_num_seqs is not a multiple of 8 (#57355) Andy LoandOpenAI Codex 2026-09-17 19:09:18 +01:00
• ff6b5808c4 [ROCm] Resolve the indexer fp8 cache dtype once at import (#53792) Sriram KumarandDouglas Lehr 2026-09-17 20:51:35 +03:00
• f3aa88d230 [Kimi K3 Bug] Fix kimi k3 reasoning parser (#57098) Wentao YeandFlora Feng 2026-09-17 13:05:52 -04:00
• b2f2bd71ba [CI][Bugfix] Add tp_shard_with_padding to padded MoE reload test mock (#57402) 3 people 2026-09-17 10:02:24 -07:00
• e30bf70c5d [ROCm][CI] Fix Entrypoints Integration (Pooling) tests on TheRock image (#57380) MattandClaude Opus 5 2026-09-17 09:22:39 -07:00
• fef55fd26d [ROCm][CI] Adapt MoE tests to the triton_kernels 3.8 API (#57385) MattandClaude Opus 5 2026-09-17 09:21:22 -07:00
• e960ead0aa [Mypy] Fix mypy typing for Transformers models (#54320) Taneem IbrahimandHarry Mellor 2026-09-17 11:11:31 -05:00
• b78fbab4be [Docs] Fix the typos in the document (#57190) YoungandYoung 2026-09-17 23:52:10 +08:00
• d7e755c230 [Bugfix] Fix wrong vLLM version reported by pip install (proto-v* tag collision) (#57295) 3 people 2026-09-17 23:50:14 +08:00
• 6a7de3ba3a [Bugfix][GDN] Fix CuteDSL BF16 KKT inversion divergence (#53864) Zoe923 2026-09-17 23:49:56 +08:00
• 493d4b3c04 [bugfix] Mark draft tokens to rebuilt their embeddings. (#57356) HieDeanandhiedean 2026-09-17 23:20:37 +08:00
• 1df336c3ba [Perf][ROCm] Enable AITER QuickReduce + RMSNorm fusion (#48249) 3 people 2026-09-17 18:05:48 +03:00
• 4f9a41ac94 [CI/Build][Hardware][NVIDIA] Add Rubin CUDA 13.4 nightly images (#55953) 5 people 2026-09-17 11:05:31 -04:00
• 3affd35f31 [Fix][ROCm] MXFP4 MoE round-up inflates TP-sharded expert weights on CDNA3, starving KV cache (#56359) vllmellmandTJian 2026-09-17 22:52:12 +08:00
• 438434b5b5 [ROCm][Perf] Insert MiniMax-M3 sparse-PA K/V without a contiguous copy (#56849) akii96 2026-09-17 17:44:54 +03:00
• 9e1cd262cb [Bugfix] Use DP index for dense DP weight updates and EC CPU region (#56950) 5 people 2026-09-17 10:44:36 -04:00
• 861a299adc [Bugfix][Model Runner V2] Route dummy tokens to MoE experts during profiling (#57270) Robert ShawandClaude Opus 5 2026-09-17 10:11:14 -04:00
• 78fdf4efb8 [Bugfix][MoE] Convert FlashInfer BF16 weights in place (#54699) 4 people 2026-09-17 07:00:04 -07:00
• 117cf43fb6 [HARDWARE][POWER] Enable W8A8 INT8 MoE on POWER (#55316) 4 people 2026-09-17 19:15:13 +05:30
• 0eae9acd4d [Bugfix][ROCm][MoE] Fall back instead of crashing when AITER MoE is requested for a non-gated (is_act_and_mul=False) model (#56590) Shantipriya Parida 2026-09-17 16:39:26 +03:00
• fe284ea17d [BugFxi] Fix DeepGEMM FP8 workspace over allocation (#53914) Lucas WilkinsonandRobert Shaw 2026-09-17 09:35:34 -04:00
• 75c71390d5 [Bugfix] Escape control characters in xgrammar choice grammar (#48115) CaiJohnandjohn 2026-09-17 20:57:47 +08:00
• 0ef8366ba3 [Frontend] Only show the summary line of config docstrings in --help (#57357) Harry MellorandClaude Opus 5 2026-09-17 13:53:11 +01:00
• 8be3ca35f3 [ROCm][CI] Fix AMD CI pipeline upload rejected by an invalid block-step key (#57375) MattandClaude Opus 5 2026-09-17 05:52:02 -07:00
• f46968cb6e [CI] Make ci-clean-log.sh portable to macOS/BSD sed (#57367) Kevin H. Luu 2026-09-17 05:42:59 -07:00
• 2e50824766 [Bugfix] Fall back to native sampling when FlashInfer cannot target the GPU (#48956) Martín el CheikhandClaude 2026-09-17 09:27:54 -03:00
• a85dd55177 Fix Laguna patch mutating flat RoPE parameters (#57189) 3 people 2026-09-17 14:22:57 +02:00
• 91a4c40b45 [Compile] Fix compile warning #177-D (#56609) Wentao Ye 2026-09-17 08:04:56 -04:00
• 667b26e50b [Bugfix][ROCm][GLM-5.3-Flash] Apply deferred tilelang.jit already on attribute access (#57192) Simon Danielsson 2026-09-17 14:02:39 +02:00
• da3c07bf7d [XPU][CI] skip test_hybrid_prefix_cache_hit_rate (#57301) Yan MaandKunshang Ji 2026-09-17 19:57:22 +08:00
• 9ff08f9493 [Bugfix] Honor skip_reading_prefix_cache for KV connector hits (#57269) aoshen02andaoshen02 2026-09-17 18:46:28 +08:00
• 0d72627565 [Perf][EPD] Batch image requests per encoder (#57095) 3 people 2026-09-17 18:45:52 +08:00
• 99ea5f525d [CI] Raise DSv4-Flash disaggregated engine readiness timeout to 1800s (#57334) Kevin H. LuuandKimi 2026-09-17 03:34:00 -07:00
• 683f6c8edf [CI] Raise H200 LM Eval Large Models timeout to 120 min (#57335) Kevin H. LuuandKimi 2026-09-17 03:33:44 -07:00
• b6e7c1f1f0 [kv_offload] Skip scratch groups (#57145) Itay EtelisandOr Ozeri 2026-09-17 13:22:49 +03:00
• 40b40d1d39 [Bugfix][KV Offload] Register the offload region in chunks (#51081) 4 people 2026-09-17 13:19:56 +03:00
• 08633cb5cd [Qwen3.8-Flash-Next] Enable FP8 TP with FlashInfer TRTLLM MoE (#55867) 3 people 2026-09-17 16:56:28 +08:00
• f1c2f6ada8 [ROCm][CI] Stage H gating and MI355 test reallocation (#57080) 3 people 2026-09-17 03:44:11 -05:00
• 23cd0346ce [CPU] Align scheduler and NIXL CPU affinity per local rank (#53636) 3 people 2026-09-17 01:43:08 -07:00
• 88afb77700 [CPU][s390x] Pin protobuf to 7.36.1 and drop C++ extension removal workaround (#54978) coderfornowandLi, Jiang <jiang1.li@intel.com> 2026-09-17 14:05:17 +05:30
• de24e51908 [MM][V2] Enable encoder-only ViT CUDA graph capture (#56922) jiangkuaixue123andCodex 2026-09-17 16:12:17 +08:00
• 4e5ffda19d [Bugfix] Isolate supplemental FlashInfer BF16 autotuning (#57285) Jiahan Chang (Cyrus)andOpenAI Codex 2026-09-17 15:38:58 +08:00
• 528fa835fd [ROCm] Stage large pageable H2D copies instead of registering them (#56343) 3 people 2026-09-17 00:14:43 -07:00
• b3079e6e46 [Perf][DSpark] Stack DeepSeek V4 context WKV projections (#54674) leoandOpenAI Codex 2026-09-17 15:04:53 +08:00
• 7bbce752b8 [Perf][Model] Qwen4Exp QSA: sm_90 tuning table for _select_config (#57273) Shuolei WangandRoger Wang 2026-09-17 15:02:24 +08:00
• bd2d7ae7c1 [Perf] Reuse MoE workspace for DeepGEMM warmup (#57268) Lucas Wilkinson 2026-09-17 03:01:19 -04:00
• fdcd42350b [ROCm][Perf] Enable HCA dual-stream overlap for DeepSeek-V4 (#56853) Shanshan Shen 2026-09-17 14:46:00 +08:00
• b978df3299 Revert "[Frontend] Omit absent fields from /inference/v1/generate stream chunks" (#57302) aoshen02 2026-09-17 14:31:59 +08:00
• 1eacb41e0f Merge remote-tracking branch 'origin/main' into sync/lesj/qwen4exp-bugfixes lesj0610 2026-09-17 15:01:29 +09:00
• 8538017f4b [Frontend] Omit absent fields from /inference/v1/generate stream chunks (#57264) aoshen02andClaude Fable 5.1 2026-09-17 13:58:56 +08:00
• e6b1a5e5e3 [XPU] Enable XPU eplb (#44987) Qiming Zhang 2026-09-16 22:19:13 -07:00
• 2743dc4f92 [Feature][Multimodal] Add cross-encoder caching to Mooncake P2P (#56242) 4 people 2026-09-17 13:16:41 +08:00
• 2d8a5c5741 [ROCm][Bugfix] Reduce CUDA graph divergences (#57229) Matt 2026-09-16 22:12:13 -07:00
• 0bfc7a15d0 [CI] Fix ruff docstring violations in EC connector files (#57283) Kevin H. Luu 2026-09-16 21:13:15 -07:00
• c38f92aae7 [A2A] Allow EPLB + SharedExpert Overlap + DeepEPv2 (#57236) Robert ShawandRobert Shaw 2026-09-17 00:02:39 -04:00
• 68f2c7e74f [ROCm][CI] Fix Nixl+Offloading PD edge cases on TheRock image (#57249) MattandClaude Code 2026-09-16 20:56:13 -07:00
• d2a2a82954 [Rust Frontend] Bump vllm-proto to 0.3.0 (#57233) Alec 2026-09-16 20:41:55 -07:00
• b8adc7710e Merge remote-tracking branch 'origin/main' into sync/lesj/qwen4exp-bugfixes lesj0610 2026-09-17 12:32:30 +09:00
• 9854b580df [Bugfix][KV Offload] Compact canonical MLA rows (#56799) Itay EtelisandOr Ozeri 2026-09-17 06:24:06 +03:00
• 5efbf0886a [CI] Support FORCE_COLOR env var; enable colors in CPU CI output (#56913) Nick Hill 2026-09-16 19:40:21 -07:00
• fa3622a4a1 [EC Connector] Add Metrics Collection (#54960) omerpaz95 2026-09-17 05:36:23 +03:00
• 2307a1d16c [Core] Don't issue a blocking collective RPC during engine handshake (#57226) Oxana KorzhandAndreas Karatzas 2026-09-16 20:30:59 -06:00
• 77b8783b25 Merge remote-tracking branch 'origin/main' into sync/lesj/qwen4exp-bugfixes lesj0610 2026-09-17 11:29:18 +09:00
• f730a93d2b [Scheduler] Add --max-num-active-seqs to cap RUNNING admission (#56758) Chuan (Richard) LiandCursor 2026-09-16 19:22:12 -07:00
• b1f9ebc0c3 [Test][Core] Add hybrid model prefix cache hit-rate coverage (#57127) Jiangyun ZhuandCodex 2026-09-17 10:17:45 +08:00
• 062f81d8ef Use skip-redaction option for crcr-report (#56532) 3 people 2026-09-16 22:01:54 -04:00
• 28ec03bc6e [Bugfix][Rust Frontend] Accept appended EngineCoreOutput fields (#56533) AlecandCodex 2026-09-16 18:54:43 -07:00
• 19b6ff62f2 [ROCm][Bugfix] Ignore descales for unquantized AITER caches (#56726) dongluwandAndreas Karatzas 2026-09-16 21:29:50 -04:00
• ed769b4a45 [A2A] Add DeepEPv2 to SP Supported List (#57210) Robert ShawandRobert Shaw 2026-09-16 21:29:10 -04:00
• e52da038d1 [Rust Frontend] Normalize native renderer reasoning controls (#56998) Bugen Zhao 2026-09-17 09:15:05 +08:00
• 3e267bae70 [ROCm] triton+triton_kernels 3.8 mxfp4 MoE support (gpt-oss + DeepSeek-V4) (#55934) 3 people 2026-09-16 20:03:34 -05:00
• 95f4925c3a [ROCm][CI] Enable AITER FP8/unquantized cases in modular-kernel sweep + fix MoRI per-tensor FP8 dispatch (#57074) Divakar VermaandAndreas Karatzas 2026-09-16 20:53:06 -04:00
• a615f53364 [Bugfix][Metrics][MFU] Size activation traffic from the model dtype (#57070) Thillai Chithambaram 2026-09-16 20:23:28 -04:00
• c58ff535f4 [Bugfix][Spec Decode] Fix EAGLE and dense draft startup with EP (#56930) 3 people 2026-09-16 17:03:36 -07:00
• 91b96a533c [ROCm][CI] Shard MI300 Multimodal Processor (#57056) Aarushi JainandAndreas Karatzas 2026-09-16 18:55:14 -05:00
• 6c5552776b Merge remote-tracking branch 'origin/main' into lesj/qwen4exp-bugfixes lesj0610 2026-09-17 08:16:59 +09:00
• 62f9482583 [Tests] Delete deprecate torchao tests for v1 configs (#52956) andrewor14 2026-09-16 19:11:27 -04:00
• fbf2c5e8be [Frontend] Add per-request metrics to Responses API (#55084) xinnywinneandBenjamin Chislett 2026-09-16 15:52:05 -07:00
• 0983aef8da [Frontend] Return prompt metadata from /inference/v1/generate (#53187) aoshen02 2026-09-17 06:29:07 +08:00
• 6c3d9816af [BugFix][KV Connector] Fix Deadlock with KVConnector + MTP under KV Pressure (#57104) Robert ShawandClaude 2026-09-16 18:22:07 -04:00
• 5a6ccc5892 [Bugfix] Add Responses cache_write_tokens for API compat and CC parity (#57222) yzong-rhandFlora Feng 2026-09-16 17:52:34 -04:00
• 929aa2a7ed [Build] Bump DeepGEMM pin to a6bbb80 (#57218) Yongye Zhu 2026-09-16 14:49:35 -07:00
• 24a63b3340 [CI][Bugfix] Initialize _transfer_layer_group_ids in region_pull_worker fixture (#57211) 3 people 2026-09-16 14:20:28 -07:00
• e0705731be [Refactor] Remove unused interface methods (#57194) Flora Feng 2026-09-16 17:00:48 -04:00
• 2bdbbc8080 [BugFix] Fix is_supported of cutlass FP8 linear (selected and fails on A100) (#55884) danisereb 2026-09-16 22:31:43 +03:00
• 30b847c1fa [Perf][DSV4.1] Remove MegaMoE padding and shared padding workaround (#57204) Woosuk Kwon 2026-09-16 12:09:30 -07:00
• bc0f47cd03 [ROCm][Bugfix] Revert #56433 + #51692 to fix accuracy breakdown for DeepSeek-V4 (#57132) Shanshan Shen 2026-09-17 02:58:23 +08:00
• 403f182c02 [MM] Keep raw pixels through dp-sharded ViT path (#56872) cjackal 2026-09-17 03:47:27 +09:00
• 6b6bf6b3c3 [Bugfix] Remove unsupported comma-separated detailed trace values (#55702) jxjandHarry Mellor 2026-09-17 02:44:48 +08:00
• 9639cbde04 [Rust Frontend] Expose local DP size in gRPC Control metadata (#57116) AlecandCodex 2026-09-16 11:28:08 -07:00
• fc8132a5e5 [Bugfix] Fix np.float64 leaking into the KV transfer metrics log (#57068) Nicolò Lucchesi 2026-09-16 20:09:05 +02:00
• 9f9e1dac26 [Bugfix][Model] Restore causal image SWA for DeepSeek V4.1 (#57152) 3 people 2026-09-16 11:09:00 -07:00
• 651a88c09a [CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212) Harry MellorandClaude Opus 5 2026-09-16 18:57:24 +01:00
• 1085b64425 [Benchmark] Retire stale benchmarks and consolidate RMSNorm (#57083) Misha Goin 2026-09-16 13:21:43 -04:00
• 0b9e018fb9 [MM] Further cleanup _apply_hf_processor_main (#53610) Cyrus Leung 2026-09-17 01:21:00 +08:00
• 6ca2b23e22 [ROCm] Expose kFp8DynamicTokenSym on AITER PTPC linears (#54248) Rebecca LeeandCursor 2026-09-16 09:48:30 -07:00
• 8c1557a79c Add pydocstyle to the ruff rules (#52136) Harry MellorandClaude Opus 5 2026-09-16 17:44:29 +01:00